We Are Metallurgists

We are Metallurgists, Not Magicians
Landmark Papers by Practising Metallurgists
Edited by D Pollard, G Dunlop and J Herzig
Spectrum 23
We are Metallurgists,
Not Magicians
Landmark papers by practising metallurgists
Spectrum 23
Edited by D Pollard, G Dunlop and J Herzig
Published by:
The Australasian Institute of Mining and Metallurgy,
Ground Floor, 204 Lygon Street, Carlton Victoria 3053, Australia
© The Australasian Institute of Mining and Metallurgy 2017
No part of this publication may be reproduced, stored in a retrieval system or transmitted

in any form by any means without permission in writing from the publisher.
All papers published in this volume were peer reviewed before publication.
The Institute is not responsible as a body for the facts and opinions advanced in any of its publications.
ISBN 978 1 925100 62 4
Desktop published by Belle Doley, Ross Palmer, Kelly Steele, Alexandra Talbot and Mia Wotherspoon
Printed by:
Focus Print Group
2 McIntyre Street, Burwood Victoria 3125, Australia
Front cover image (top):

Construction at Newmont Boddington Gold, photo courtesy of Newmont Australia.
Back cover image:

Just before sunrise at the Phu Kham Copper-Gold Operation process plant, photo courtesy of PanAust Limited.
introduction
The majority of papers included in this volume have been selected from AusIMM MetPlant and MillOps conferences from 2008 to 2016. The
compilation was inspired by the realisation that the Institute’s metallurgy conferences deliver papers of enormous value to practising metallurgists
and, unless they happen to be conference delegates, the information contained in them is otherwise lost to the wider metallurgical community.
They provide a unique resource for professionals at all career stages, whether working in design, construction or the operation of metallurgical
processing plants.
The editors reviewed over 40 AusIMM conferences identifying metallurgy papers relevant to the objectives of the present publication, viz:
• they are relevant to good design, construction and operation of metallurgical processing plants
• they do not have an ‘expiry date’
• they pass on relevant experience and inspiration to the next generation of plant designers and operators.
The papers have been updated, a few are new for the volume, and some have been re-written to combine the core narrative of two or more
conference papers. This compilation presents them in themed chapters, grouped with relevance to the challenges of the future as well as the
present.
Most authors expressed how privileged they were to have their papers included and how the initiative to have this wisdom distilled in one
place was a noble one. Many responded with additional comments, some of which are incorporated overleaf as testimonials.
Through this publication the editors hope that fewer wheels will be re-invented, and the accounts of successful ideas and innovative thinking
will inspire pathways to improved plant designs, operating efficiencies and profitability. Metallurgists in planning, plant design, construction,
operation, improvement and management will find the volume essential reading.
ACKNOWLEDGEMENTS
The editors are grateful for the suggestions and feedback from members and ex-members of the Metallurgical Society Committee, including Dean
David, Aidan Giblett, Andrew Newell, Philip Stewart and Peter Tilyard.
Janine Herzig joined the original editors as the paper selections were being finalised. The three editors worked together, contacting and
liaising with authors and securing the generous sponsorship from industry necessary to ensuring a hardcopy as well as electronic version. Janine
edited all of the revised papers, making an invaluable contribution.
David Pollard, Geoff Dunlop and Janine Herzig, Editors
testimonials
“A splendid outcome and a worthwhile contribution to the industry.”

E McLean
“My fellow authors and I are honoured that our papers are
considered for inclusion. The list of papers is very impressive.”
D Bennett (PanAust)
“I’m honoured to have some of my papers considered as ‘landmark papers’ and to be included as
a ‘Practical Metallurgist’! A very worthy project.”
T Napier-Munn
“I would be honoured … I see it as a great idea.”

P Thwaites (XPS)
“It is a great honour to be included in a such august company

and I must congratulate you on this wonderful outcome
and legacy to your vision, drive and management skills.”
A Newell
“Honoured to be included and love the title.”

P Bartsch
about the editors
David Pollard is a consultant metallurgist with an interest in education and professional development. His employment after graduation was
with the Port Kembla steelworks, and later as a metallurgy lecturer at University of Melbourne and the South Australian Institute of Technology
(now UniSA). For over a decade he managed the professional development activities of the Australian Mineral Foundation (AMF), providing short
courses and conferences across Australia and internationally, across a range of high level technical and management topics. Many of these courses
were offered in association with AusIMM conferences.
David brought the Metallurgical Plant Design (MetPlant) conference series to the AusIMM when AMF closed, and has convened the series
with Geoff Dunlop since 2002, and from 2017 with Janine Herzig. He has twice been Chair of the Adelaide Branch, and was Chair of the AusIMM
Metallurgical Society (MetSoc) from 2008 to 2013.During this time MetSoc initiated the groundwork for AusIMM to publish the third edition of
the Sir Maurice Mawby volumes, Australasian Mining and Metallurgical Operating Practices.
We are Metallurgists, Not Magicians was initially undertaken by David and Geoff, with the support of MetSoc, and will add to the professional
development resources for practising metallurgists and processing plant designers.
Geoff Dunlop is a metallurgical engineer who began his career at Mt Isa, rose to Concentrator Superintendent, and moved to a site research
role where he carried out early investigations on the geometallurgy of the Mt Isa orebodies. He characterised the ‘3000’ and Lead/Zinc orebodies.
He moved to AMDEL in 1970 as a project metallurgist, and developed process flow sheets and systems to provide financial information about
performance. In the early 1990s he worked with AMDEL collaborating with MIM Process Control and JKTech to develop instrumentation for
Indian, Burmese and Iranian processing plants. He continued to consult for some years following his retirement, including for BOC gases regarding
automated equipment for plants, as well as for international clients.
In 1998 Geoff edited, with Norton Jackson and Peter Cameron, the AMF Conference proceedings for ’Mineral Processing and Hydrometallurgy
Plant Design‘, a conference convened by David Pollard. This collaboration led into the partnership between David and Geoff on the AusIMM
MetPlant conference series from 2002.
Janine Herzig began her graduate career in Mount Isa as a metallurgical engineer with MIM in the Lead/Zinc Concentrator. She then moved into
the mineral sands sector with Iluka in various locations across Queensland and Western Australia, before being appointed Principal Metallurgist
for the Murray Basin Operations feasibility work and execution. After ten years with Iluka, Janine accepted the role of Mineral Processing Manager
with AMDEL and ultimately as the General Manager of the Minerals and Industrial Division, where she oversaw programs spanning multiple
commodities and mineralisation styles. Key projects included the first major ODX geometallurgy program under WMC then BHP, which involved
establishment of a major new metallurgical and geochemical testing facility, the Prominent Hill feasibility study, and management of a major
expansion of the on-site laboratories for the Whyalla steelworks and processing plants. Janine established a consulting business after leaving
AMDEL in 2009 conducting due diligence, desktop reviews and project management across a range of commodities including industrial minerals,
graphite, precious and base metals.
She has served on the AusIMM Adelaide Branch Committee since 2005, holding the positions of Chair, Secretary, newsletter editor,
and member on the South Australian scholarship interview panel. She continues to serve on a range of AusIMM conference organising
committees and other specialist taskforces. She was honoured to be invited to assist David Pollard and Geoff Dunlop with this publication,
and believes that it will become the essential reference for all metallurgists, plant designers, consultants, students and new professionals
for many years to come.
Sponsors
Platinum Sponsors
Gold Sponsor
Silver Sponsor
Bronze Sponsors
Our purpose is to create value and
improve lives through sustainable
and responsible mining
Follow us on: www.newmont.com

PanAust is an Australian-headquartered copper
and gold producer with producing assets in Laos
and a portfolio of organic growth projects in
Laos, Papua New Guinea, Myanmar and Chile.
PanAust’s producing assets are the Phu Kham

Copper-Gold Operation and the Ban Houayxai
Gold-Silver Operation, both located in the
Company’s 2,600 square-kilometre Phu Bia
Contract Area in Laos.
In Papua New Guinea, PanAust holds an 80 per
cent interest in the Frieda River Copper-Gold
Project, with the remaining 20 per cent held by
joint venture partner, Highlands Pacific.
In Myanmar, PanAust holds an 80 per cent interest
in Wuntho Resources Company Limited (WRCL)
and has established a joint venture with Myanmar
Energy Resources Group International Company
Limited (MERG), a Myanmar-based company which
holds the remaining 20 per cent of WRCL. WRCL
owns 100 per cent of the rights to areas covered
by three Exploration Licences in the Sagaing
Division of Myanmar.
At PanAust, operational excellence and
sustainable business outcomes are pursued in a
complementary manner. This strategy, which sees
excellence in one area underpin, encourage and
facilitate excellence in the other, has been the
driving force behind PanAust’s success. PanAust is
determined to remain Way Ahead.
For more information, visit www.panaust.com.au.

gold SPONSOR – Ausenco
Ausenco is a world leading engineering company providing consulting, project delivery and asset management and optimisation solutions to the
resources, energy and infrastructure sectors. We add value to our clients’ projects and deliver results even in the most challenging environments.
Ausenco has a global footprint with offices and current projects in Australia, the Pacific, Asia, Africa and North and South America. Our experts
thrive on their ability to create ingenious solutions that improve our clients’ return on investment. We’re proud to be a Gold Sponsor of this
publication and support the continued development of excellence in Minerals Processing.
CONTENTS
Overview
‘We’re metallurgists, not magicians!’ 3
E McLean
Presented at MetPlant 2011, this paper has been updated and edited for this compilation.
Back to the future – why change doesn’t necessarily mean progress 11

P D Munro and P A Tilyard
Presented at MillOps 2009, this paper has been updated and edited for this compilation.
Back to the future – still on the dark side 19

P D Munro
Presented at MillOps 2016 as a follow-on to the paper: ‘Back to the future - why change doesn’t necessarily mean progress’, this paper has been updated and edited
for this compilation.
Undermining productivity – when good key performance indicators go bad 27

J Pease
Presented at MillOps 2014 as a PowerPoint presentation, this paper has been written specifically for this publication.
Geometallurgy
Geometallurgy – what do you really need to know from exploration through to production? 33
K Ehrig
Presented at MetPlant 2013, this paper has undergone minor edits for this compilation.
Integrating geometallurgy with copper concentrator design and operation 37

G Harbort, K Jones, D Morgan and C Sola
This paper was presented in various forms at MetPlant 2011, MillOps 2014 and MillOps 2016. It has been updated with a new geometallurgy model, undergone
significant edits and is the integration of the papers presented at the three events outlined above.
Integrated mining and metallurgical planning and operation 55

P L McCarthy
This is a compilation of three papers presented and published for MetPlant 2011, MetPlant 2013 and MetPlant 2015.
Project economics
Guidelines for economic evaluation of projects 65
P Card
Presented and published at MetPlant 2011, this paper has been significantly reviewed and updated for this compilation.
Sensible cost cutting for resource projects 69

D Connelly
Presented at MetPlant 2011, this paper has undergone significant edits and updates for this compilation.
When does further processing at the mine site make sense? 75

C Fountain, S La Brooy and G Lane
Presented at MetPlant 2008, this paper has had minor edits for this compilation.
The ABC of Mine-to-Mill and metal price cycles 85

P Cameron, D Drinkwater and J D Pease
Presented at MillOps 2016, this paper has been reviewed and abridged for this compilation.
Base metals concentrate sales contracts – change Pavlov and the dog 91
P D Munro and S E Munro
Project design
Karouni Gold Project from drill core to commissioning 101
K Nilsson and D Connelly
Upgrades, modernisations, automation and expansions … where will the expertise, capability and skills come from in the future? 107
R Coleman, J King and T Hunter
Presented at MillOps 2016, this paper has undergone minor edits for this compilation.
x
Is an 80th percentile design point logical? 115
D David
Presented at MetPlant 2013, this paper has undergone significant edits for this compilation.
Measuring and taking notice of orebody variability – an essential ingredient for reliable plant design 121
D David
Getting optimum value from ore characterisation programs in design and geometallurgical projects associated with comminution circuits 131
S Morrell
Cost-effective concentrator design 137

G Lane, P Dakin and D Elwin
Project management and delivery

Fatal flaws in technical due diligences 145
A J H Newell
Presented at MetPlant 2015, this paper has undergone significant updates and edits for this compilation.
Guidelines for mineral process plant development studies 155

P R Whincup
Project delivery 163

G Lane and E Skinner
Mineral project management – a perspective from four decades in the industry 173
J S Dunlop
Keeping projects on the rails 183

J Canterford
Operations versus projects – how do people think and what are the implications? 187
G Lane and B Clements
Performance testing – when, what and how? 193

G Lane, M Davis, E McLean and J Fleay
Presented at the Project Evaluation 2007 conference, this paper has undergone minor edits for this compilation.
Unit design and development

Process development and throughput forecasting at the Phu Kham copper-gold operation, Laos PDR 201
D Bennett, I Crnkovic, P Walker, A Hoyle, A Tordoir, D La Rosa, W Valery and K Duffy
The subject matter for this paper was written, published and presented at three AusIMM conferences: MillOps 2012, MillOps 2014 and MetPlant 2015.
Significant effort has gone into combining, updating and editing this combination of the three papers for this compilation.
What can go wrong in comminution circuit design? 235

C Bailey, G Lane, S Morrell and P Staples
Presented at MillOps 2009, this paper has undergone significant edits for this compilation.
Do we need a gravity circuit or not? A case study in applying best practice 243
A Giblett and K Afewu
Solvent extraction of uranium – towards good practice in design, operation and management 249
P Bartsch and S Hall
Design of copper-cobalt hydrometallurgical circuits 257

G Miller
A review of best practice in gravity circuit design and operation 263

A Giblett, A Bax, G Wardell-Johnson and W Staunton
xi
Leach residue and pregnant liquor separation – process and capital comparison of counter-current decantation and counter-current
washing with vacuum filtration 273
R Klepper and P McCurdie
Filtration test work – extracting the whole story for studies and design 285
G Bickert and B Länger
Considerations for effective gold process development 297

A Giblett, D Appelhans and R Dunne
Energy efficient ball mill circuit – equipment sizing considerations 307

A Jankovic and W Valery
Advances in dense medium cyclone plant design 315

T J Napier-Munn, G Gibson and B Bessen
Sampling, metallurgical accounting and control

Using metallurgical data to drive continuous improvement 327
R Dunne
The importance of sampling in the mineral industry 333

R J Holmes
A progressive, iterative approach leading to reliable metallurgical accounting sampling systems 343
F F Pitard
This paper was written for MetPlant 2015 but not presented at the conference. It has undergone significant edits for this compilation.
Sampling, corporate governance and risk analysis 349

G J Lyman and F S Bourgeois
Presented at MetPlant 2015, this paper has been significantly edited and updated for this compilation.
Metallurgical management tools for continuous improvement 355

W McCallum and R Dunne
Measuring the influence of sample size on the precision and accuracy of gravity gold estimation 361
A Giblett and T J Napier-Munn
Defining practical metallurgical accounting discrepancy limits for gold operations 365
A Giblett, R Dunne and K McCaffery
Manual control, process automation or operational performance excellence – what is the difference? 373
P Thwaites
Presented at MillOps 2014, this paper has been significantly updated with new information for this compilation.
Unlocking processing potential by empowering our operators 401

X Li, M S Powell and W McKeague
Current developments in the operation and control of autogenous and semi-autogenous grinding mills in Australia 411
J Karageorgos, Y Atasoy and D Baas
Environmental management and sustainability

Mine waste risk minimisation by integrated waste management and process optimisation 419
D Brett
Sustainability, made easy – a business improvement case study 427

G Corder, N Currey and G Becker
xii
Continuous improvement – how changes in metallurgy can have a substantive impact on environmental impact and closure: lessons
learnt from environmental auditing 437
P Mulvey and G McMillan
Process improvements and case studies

Improving fines recovery by grinding finer 445
J D Pease, M F Young and N W Johnson
Presented at MetPlant 2004, this paper has had significant updates and edits for this compilation.
Cunning solutions to process improvement 455

T J Napier-Munn
How to prioritise process improvements 461

D J Hill
Presented at MillOps 2014, this paper has had a significant update for this compilation.
Some practical problems in running statistically valid plant trials and their solution 469
T J Napier-Munn
The optimisation of semi-autogenous grinding and ball mill based circuits for mineral processing by means of versatile and efficient
high pressure grinding roll technology 475
S W Kirsch and M J Daniel
Presented at MillOps 2009, this paper has been significantly updated for this compilation.
Optimisation opportunities for high pressure grinding rolls circuits 483

M S Powell, M M Hilden, C M Evertsson, G Asbjörnsson, A H Benzer, A N Mainza, L M Tavares, B Davis, N Plint and C Rule
Presented at MillOps 2012, this paper has been significantly updated for this compilation.
An update on applications of high frequency screens in closed grinding circuits 499

S B Valine, J E Wheeler, B Packer and A N Cavendor
This paper is a compilation of two papers from MetPlant 2011 and MillOps 2012. It has been significantly updated with additional case studies and information
obtained for the Alban Lynch publication Comminution Handbook.
The influence of liner wear on milling efficiency 509

P Toor, T Perkins, M S Powell and J Franke
Presented at MetPlant 2011, this paper has been significantly updated for this compilation.
Development of the Fosterville gold mine heated leach process 523

M Binks and P Wemyss
Presented at MetPlant 2011, this paper has been significantly updated for this compilation.
Collector – addition point and consumption 529

C J Greet, W J Bruckard and D MacKay
Presented at MetPlant 2008, this paper was revised several years later for Transactions of the Institutions of Mining and Metallurgy – Mineral Processing and
Extractive Metallurgy, and has been further updated and edited for this compilation.
The Ausenco carbon reactivation kiln initiative 537

J K Claflin and S R La Brooy
This paper was specifically written for this compilation as an author foreword to the following two papers.
Lessons learnt and performance – installing and commissioning an Ausenco carbon reactivation kiln in Africa 541
J K Claflin, S R La Brooy and D Preedy
This paper was presented at MetPlant 2015. Context for this paper and the paper immediately below has been provided via an author foreword (above), which also
provides significant new insights and updates.
Carbon management in a high gold price environment 549

S R La Brooy and J K Claflin
This paper was presented at MetPlant 2013. Context for this paper and the paper immediately above has been provided via an author foreword (above), which also
provides significant new insights and updates.
The impact of gravity gold recovery at Kalgoorlie Consolidated Gold Mines 559
A Giblett, D Hillier, K Parker and V Ramsell
Presented at MillOps 2012, this paper has undergone very minor edits for this compilation at KCGM’s request.
Agglomeration – the key to success for the Murrin Murrin heap leach 569
D Readett and J Fox
Author index 575
xiii
Overview
Contents
‘We’re metallurgists, not magicians!’

E McLean1
ABSTRACT
The decision-making checklists and the hierarchy of information for process and
plant design are generally wide-ranging and complex.
Grades, mineralisation and ore types in an orebody, the mine development
sequence and results from metallurgical and physical test work present complex
and competing influences for flow sheet development, values for design, equipment
selection, extraction/recovery performance and operating strategy.
Challenges are presented on several fronts: how to manage these variables;
understand their impacts on performance; and best provide a cost-effective plant
design and robust process that meets daily, weekly and monthly operating objectives.
Throughput, production and grade or product specifications need to be met. The
orebody is not homogeneous; its range of metallurgical and physical characteristics
may be moderately to highly variable. Average values derived from test work results
are unlikely to be suitable for design purposes.
This paper assesses some of the competing influences and conflicting issues in design
and equipment selection, and their effect on production and performance expectations.
THE DESIGN SEQUENCE

Each item of process equipment, the flow sheet, process logic and the operating
strategy for the treatment plant are there by design and purpose. This is because a
sequence of process engineering activities has led to their implementation, installation
and construction in the project.
The development activities for the project follow the staged sequences shown in
Table 1. Although not exhaustive, this table simplifies a complex procedure and
identifies the main process and engineering decision-making activities. In practice,
iterations following reviews, revisions at each stage and the linked sequence of stages
increase the complexity in project delivery.
Reviewers of the performance of project delivery, plant start-up and operation
(Agarwal, Brown and Katrak, 1983; Performance Associates International Inc, 2017)
identified fundamental issues that contributed to financial shock, whether it was due
to attaining nameplate capacity at a slower rate than anticipated, or not achieving
this at all. They noted that, notwithstanding experienced and competent mining,
design and planning personnel on the project, the chief technical problems, start-up
difficulties and poor performance for new mineral processing operations, occurred
across all stages of the project from the following items:
•• representativity of samples – high heterogeneity of orebodies, difficulties with
adequate sampling, extent of testing was inadequate (stage I in Table 1)
•• definition of the process – complexities of the process route, influence of process
water (stage II)
•• engineering – design deficiencies with mechanical equipment, equipment
design and/or installation is inappropriate (stage III)
•• operation – operators lack knowledge and skills to properly operate the plant,
management is unable to cope with the problems experienced during initial
operation, maintenance is not executed properly, particularly during the initial
operation (stage IV).
The authors stated that although problems with mechanical equipment were
expected, these risks can be reduced through efficient design and scale-up. They
attributed a number of these problems to basic design deficiencies such as incorrect
specifications, unit capacity and duty.
The activities in Table 1 match elements in the schematic project flow chart (Lane
et al, 2008) in Figure 1, which shows the relationship between design inputs and
1. FAusIMM, Manager Minerals Consulting,
Ausenco Minerals & Metals, South Brisbane
project outcomes:
Qld 4101. Email: eddie.mclean@ausenco.com •• stage I input = design inputs (green boxes)
3
E McLean
Table 1
Development sequence for principal process design and engineering activities.
Stage Development activities Description of key activities in development stage

I – Input Geology, mineralisation, resource model Discuss deposit(s) with exploration geologists
Review long- and cross-sections showing lithology and structures
Geological interpretation, reconcile metallurgical classification for zones or domains
Sample selection and sampling protocols Representative samples and the basis for selection
Characterisation – metallurgical and physical, of the major ore types, ore zones or domains
Variability – metallurgical and physical, of parameters that influence design and production
Production composites, as required, for selected periods during the initial two to three years of operation
Test work, by laboratory and vendor Preparation of samples, ensuring integrity of sample, preventing oxidation
Batch or semi-continuous small or large scale (pilot) for flow sheet development and to obtain data for process design
Reproducibility and robustness of tests; simulation of operating conditions; check/repeat and diagnostic tests
II – Definition Analysis of data Collate and compare, interrogate for trends and relationships, investigate outlier data
Ore sequence in likely mine plan and weighting by ore characterisation
Processing strategy and process alternatives, flow sheet development
Design value Recommend value/specification for unit process and for process stages
Define duty point in range of operating conditions, considers variability
III – Engineering Calculation – size and select equipment Text book or standard methodology, with efficiency factors
Include equipment factors – industry applicable, vendor recommended
Include operating factors – experience, reconciliation of operation with previous design and performance
Include agreed or nominated engineering margin, owner margin, allowance for engineering risk
Tender and select for procurement Vendor receives engineering specification, criteria, data sheets, drawings
Vendor uses proprietary modelling, calculations and database to size and recommend
Vendor submits equipment that complies with throughput and duty required
Vendor selection also considers commercial aspects: performance warranties, penalties for non-performance,
reputation, competition (cost)
New or second-hand equipment may be offered
Technical and commercial evaluation by engineer – recommend to purchase
IV – Deliver Install, start-up and ramp-up operations Implementation plan to coordinate and schedule activities
Project Install and commission in accordance with standards and procedures
Maintenance – planned and in accordance with recommendations
Operator experience and training
FIG 1 – Relationship between design inputs and project outcomes (from Lane et al, 2008).
4 we are metallurgists, not magicians

•• stage II definition = project design concept (orange oval) (month, quarter, half-year) to support production and
•• stage III engineering = project outcomes: engineering cash flow forecasting.
design equipment selection and quantities and take-offs Resource composites which are blended by ore types and
(blue boxes) grade to represent samples such as whole-of-pit, life-of-mine
•• stage IV delivery = project execution: implementation, and annual averages do not necessarily provide the basic
procurement, construction and commissioning (yellow information and ‘building blocks’ for flow sheet development
boxes). and design that characterisation and variability samples do.
Plant operation is unlikely, if ever, to treat such samples
The project design concept and definition stage are pivotal:
on a sustained basis. Specific ore properties that may cause
all information channels through here for interpretation and
metallurgical and operational difficulties are diluted in a
analysis, and then forms the basis for subsequent engineering
composite sample and any impact these have on performance
design and planning for the execution phases of the project.
is dampened, or perhaps not identified in the test results.
Activities and decisions in this stage are able to impact on the
project cost, schedule and plant performance. For a design based on average values plus allowances for
mechanical and vendor margins, there is a high likelihood that
Although all items are integral to a successful project outcome the expected performance and corresponding production will
and some are critically important such as ore representativity not meet planned targets when treating ores over sustained
and sampling, this paper reviews the following activities operating periods with characteristics significantly above
described in Table 1 that have a high ability to influence the the average value. Furthermore, the opportunity to catch
final project outcome: up production when ore with characteristic values below
•• analysis of the data to develop a viable processing strategy the average are treated may be constrained by physical and
•• values for design to specify the duties for the unit volumetric limits in the plant and equipment; for example,
processes pump capacity, launder size, screen area, thickener area,
filter rate.
•• calculations to size and select the process plant and
equipment.
Data analysis
In the early stages of project development when drilling and
THE DATA SEQUENCE geological interpretations are ongoing and resource definition
Practical and robust interrogation, and interpretation and is in its formative stages, the body of test work data is limited.
analysis of the data from the body of metallurgical information Data is generally in small sets of information and the data
available are fundamental to successful design. The quality selected is typically the maximum or the average of the range.
of data at various stages of project development is also of Incremental increases in input data values may be made to
consequence with respect to its validity and applicability for assess trends and to assess sizes of plant and equipment.
design. As well, benchmarking of the values may be made with
comparable industry data or in-house data from similar ore
Samples and data types and similar applications to improve confidence at this
Samples for metallurgical and physical testing are usually early stage.
domain-orientated and can be categorised as characterisation As the project develops, not only does the volume of data
samples and variability samples. available from several stages and iterations of metallurgical
Characterisation samples normally represent a lithology, test work programs increase, but so too does the complexity
oxidation, alteration, mineralogy, grade, or spatial property of information from the amount of detail available with an
of the ore, as appropriate, for which metallurgical or physical improved understanding of ore characteristics and variability.
unit parameters are determined by the test work for that Superimposed on these are a number of ore feed schedules
specific ore classification. Variability samples are usually to the plant with variable tonnage, grade, mineralisation and
from each domain or ore classification in which one property lithology, and different ore exploitation sequences depending
varies with, for example, metal grade, sulfur grade, depth, on the mining strategy, mine development plan, pit financial
location in the geological structure. modelling and resource constraints.
Data generated from tests on both these types of samples Analysis of the characterisation and variability data by ore
can be associated with blocks of ore in a resource model; some type to determine values for design includes:
of the more common applications are described as follows: •• a statistical analysis which identifies the average, the
•• hardness and competency values populate blocks of ore standard deviation, minimum and maximum values,
in which the dominant lithology or alteration type is values at nominated percentiles (default is typically 75th
identified (characterisation) or 80th) of the data population
•• leach recovery values populate blocks of free-milling •• an assessment of outlier low/high data, analysis and
gold ore in which the extent of oxidation is identified, inclusion with basis and comments noted, or discard
ie for wholly oxidised, partially oxidised or transition, with justification
or fresh ore (characterisation) •• a plot of values in an ordered sequence which graphically
•• a relationship of recovery with a variable (or variables) shows the distribution, the location of the bulk of the
can be described by an algorithm to populate blocks values, and where the outliers or sparsely populated
of ore; for example, in gold ores, relationships may values lie
be established between leach extraction and feed •• identification of values, which characterise the major ore
grade, leach extraction and sulfur grade, and extent of types, unit process performance, process streams and
refractory gold and arsenic grade (variability) water quality
•• mill feed from a mine plan and ore schedule can be •• determination of a weighted value appropriate for
interrogated to provide information over a period principal blends of ore in the feed to the plant.
we are metallurgists, not magicians 5

E McLean
Sources of data The design value does not include factors such as catch-
Input data and information for studies from concept to final up capacity, any additional design margins, factors or other
feasibility, for basic engineer packages (BEP) or front-end allowances. Design factors and margins should be assessed
engineering design (FEED) are from several sources. These and identified in the calculation where necessary to achieve
can be grouped as follows: the design objective, specific equipment duty or operating
condition. Factors include allowances for items such as wear,
•• Project, operating and production parameters, which
screen pegging, minor surges, emergency relief, certain start-
are set by the client. These are based on financial and
up or shutdown conditions or short-circuiting. Operating
organisation parameters. These are usually agreed
margins are allowed for materials handling and in selected
and fixed, and form part of the design basis, such as
process equipment such as conveyors, cyclone feed pumps,
operating schedule and availabilities.
other slurry pumps, reagent pumps, water pumps and air
•• Test work (by laboratories, vendors, and specialists), services.
resource models and consultants’ reports. These are
the main sources of data for the project as they contain Equipment sizing and selection
detailed test work and analyses for the resource, its
geology, mine plan, all unit processes, the water quality Calculations for equipment sizing show the design value, the
and environmental controls. This body of work also input flow(s) and additional factors where these materially
contains reports from client’s specialists, processing affect the equipment size and basis of selection. The input
specialists, other consultants and third-party audits. flow(s) for the calculation are normally from a nominated
(usually at, or near maximum) operating case or from a short-
•• Operating practice and industry standards. Sources
term event which may include a step-change due to a batch
of data are in-house experience, operating history
flow stream addition. The calculation using the average mass
on similar treatment plants, published or confirmed
balance condition is also normally carried out to provide a
operating information.
measure of the range of operating performance for the unit.
•• Public domain information such as vendor data,
Selection and recommendation of the unit process
handbooks, regulatory and environmental standards,
industry codes, service and performance catalogues equipment or plant should meet two main criteria:
on like-for-like applications, published engineering or 1. function efficiently in accordance with the parameters
discipline text books, and published regulations. and specifications nominated
•• Engineers’ database, operating and commissioning 2. be cost-effective from both initial capital and life cycle
experience. This would apply in instances, typically as assessments.
follows: if data from previous sources are not available; Selecting the unit process equipment or plant achieves the
if the number of test samples and testing is considered desired balance between the competing demands of:
limited and/or incomplete; if the test work does not
•• undersizing equipment to minimise costs, which may
replicate the unit process performance well enough; if
increase the risk of underperformance, and
test data is from batch operation and needs adjustment
for a continuous basis; if scale-up factors need to be •• oversizing equipment which, whilst providing a
allowed for based on previous experience. measure of confidence to meet the operating duty, not
As a guideline, the hierarchy of sources of information and only over-commits initial capital for the equipment item
data for process and plant design is usually in the order listed (or unit process) but also can increase costs for concrete
(above). The client parameters (first group) normally form works and steel structure to install this item.
the financial, operational and production basis for the project.
Test work data (second group) are ore and deposit specific CASE STUDIES
and consequently form the prime source of information for Case studies for selection of design values for key process
flow sheet development and unit process design. Data from criteria and for the design basis of process areas in gold plants
the remaining three groups of information would be used in are described in the following sections.
preference where these could be demonstrated to be more
applicable and more representative than those in the first
Selection of design values
two groups.
The selection of a design value from their population of data
Design value is described for comminution, gravity and feed grade cases.
An example of variability data analysis for gold recovery
The design value is the input value used in calculations to
with sulfur is given. Data sets are from different projects and
size and select the processing equipment item (eg crusher,
provide examples of the basis used to select a value for design.
grinding mill, hydrocyclone, thickener, float cell, filter) or a
unit process (eg carbon adsorption circuit, counter current
wash circuit). The value can be a grade or an ore/mineral
Comminution – competency
characteristic, an attainable unit parameter, a rate or capacity The drop weight test parameters ‘A’ and ‘b’ are measures of
for the nominated duty and operating conditions for a unit of the competency of an ore used in grinding calculations for
process equipment or a specific circuit. semi-autogenous and autogenous grinding mills. Figure 2
shows the distributions of Axb values with samples tested for
The design value does not necessarily relate to production
schedules or integrate to a mass balance. The mass balance for three ore types in a supergene/hypogene type orebody with
the plant represents an operating condition for a continuous skarn mineralisation.
circuit in steady state in which inputs equal outputs for all The Axb value selected for design from the distribution
items (ie solids, water, metals and elements). Plant mass plots for each ore type was based on the 25th percentile in
balances are prepared for an average ore condition and for each ore type (lower Axb values represent more competent
a series of cases which represent specific feed parameters or ore). The design value for each ore type and the characteristic
operating conditions. blend of ore types in the mine plan are shown in Table 2.

FIG 2 – Correlation Axb values with samples tested for three ore types.
Table 2
Distribution of ore types by the principal mine plan blends. FIG 3 – Bond ball mill and rod mill work indices (WI)
for samples from the dominant type.
Lithology Axb Work index Mine plan Mine plan
(ore ball (wt %) (wt %) Gravity recovery
type) kWh/t years 1–3 years 4–8 Gravity recovery tests on ground samples (5 to 10 kg
Type 1 77 12.8 50 0 per sample) using laboratory scale centrifugal bowls
together with shaking tables or vanners measure the gravity
Type 2 76 11.5 20 20
recoverable gold from steams in a ball mill circuit closed
Type 3 38 15.9 30 80 with cyclones. The grind size for these types of gravity tests
Total -- -- 100 100 is usually at the design P80 grind size, or it may be up to
two to three times this value which recognises that the bleed
The (weighted) Axb design value for each mine plan blend stream from the grinding circuit is coarser than the cyclone
is 65 for the 1–3 year period and 46 for the 4–8 year period. The overflow final product size. The weight recovered to the final
corresponding Bond ball mill work indices are 13.5 kWh/t gravity concentrate in this type of test is normally in the range
and 15.0 kWh/t, respectively. 0.03–0.09 per cent by weight from feed. Figure 4 shows the
results from standard gravity recovery tests on three ore types
The ore blend is less competent (higher Axb value) and with auriferous quartz vein stockwork mineralisation.
less hard (lower work index) in the initial three years, and
consequently the unit power requirements for the semi- Gravity recovery was moderately high and variable; no
autogenous grinding (SAG) mill and ball mill are lower trend was observed by ore type or by feed grade. Recoveries
during that period compared to the latter five years. Operating at the 90th percentile, 15th percentile and average were
options allow either: assessed. The laboratory recovery at the 90th percentile was
49 per cent; this was rounded up to 50 per cent for design
1. ability to treat higher tonnages during the initial years
of metal recovery in downstream processing of the gravity
based on the installed power and nominated throughput
concentrate.
for years 4–8
Gravity recoveries for the average and 15th percentile
2. expansion of facilities and increase grinding capacity to
from the laboratory tests were 39 per cent and 29 per cent,
maintain the nominated throughput for years 4–8.
respectively. These results from laboratory tests, which
In this project, option (1) was determined to provide more
value to the project.
Comminution – hardness
Standard Bond work index rod mill and ball mill tests are
measures of the hardness of an ore and are used in grinding
calculations for rod and ball mills; they also form an integral
component of SAG mill grinding calculations. Figure 3 shows
the distribution of work indices with samples tested for a
volcanic, volcaniclastic, co-magmatic intrusive type orebody.
The work index values selected for design from the
distribution plots, based on the 75th percentile, were rod
mill work index 28.9 kWh/t and ball mill work index
21.9 kWh/t. These values provide sufficient flexibility for
the mill to accommodate the majority of ore types and ore
feed conditions (with blending). Work index values at higher
percentiles would add capital cost and an additional margin FIG 4 – Gravity recoverable gold from three ore
over the required feed rate case. types in quartz-veined gold deposit.

E McLean
were carried out in ideal, batch conditions, were discounted

to account for the lower mass pull by weight per cent to
concentrate in the plant, the type of gravity circuit installed,
the recovery effort, continuous operation and ore variability
in the plant. The discount factor for low-sulfide and low
specific gravity gangue is typically in the range 0.60 to 0.80.
The adjusted average gravity recovery for the mass balance
was 30 per cent. For downstream leaching, carbon-in-leach
(CIL) and carbon in pulp, the carbon circuit and metal
recovery were designed for periods of lower gravity recovery
and a gravity recovery of 20 per cent was adopted.
The silver in feed and to gravity was tracked to assess its
deportment and accountability. As in many gold and gold/
silver ores, silver recovery to the gravity concentrate was very
low, and significantly less than that for gold.
Front-end engineering design grades FIG 6 – Gold residue and leach feed grade relationship by sulfur increments.
The cumulative distribution of copper feed grades from
the mine plan and ore schedule for a long-life copper-gold A relationship of residue gold grade with feed grade for
porphyry mine is shown in Figure 5. sulfur grade levels was obtained. Further statistical analysis
The copper feed grade distribution in years 4–12 was of this database was required to develop a model which
selected as the basis for design for the flotation and concentrate estimated the leach residue gold grade for the range of sulfur
handling areas. The 75th percentile was used for the design grades expected in the ore schedule, and thus to calculate
copper value. Although gold grades in this copper-gold ore gold recovery from the corresponding feed gold grade.
were higher in the initial three years than in the year 4–12
operating period, gold has an economic contribution but no Design basis
material influence on the sizing of the copper concentrator A case study for the design basis for areas in a conventional
equipment. gold plant is described. The circuit comprised the following
Although the copper feed grade was lower in the initial key process stages: crushing, semi-autogenous and ball mill
three years, it was preferred that higher unit capacity was (grinding) with pebble crusher (SABC), gravity, leach/CIL,
installed at the outset to avoid future disruptions to ongoing thickening, cyanide detoxification, desorption, regeneration,
operations from construction of additional flotation and gold room cathode, gravity preparation and smelting.
concentrate handling facilities. Additional capacity was
therefore available in the initial years of operation for Plant feed characteristics
catch-up and ability to cope with any short-term grade or FEED distributions by tonnage and by feed grades for ore to
mineralisation fluctuations. the mill, sourced from three open pits are shown in Table 3.
The yearly intervals correspond to optimised pit development
Gold recovery variability to maximise gold and silver grades, as well as revenue. The
An apparent scatter of gold recovery results from a number of majority of the ore comes from Pit 1, with ore from Pit 2
routine, standard grind and cyanide leach tests on the gravity boosting gold and silver grades in the first two years. Pit 3
tail for a large open pit gold resource, was resolved (Smith, is the lowest grade and is developed towards the end of the
2005) by assessing the response by sulfur grade increments. project life.
The leach residue versus leach feed grade trends for four
sulfur increments were plotted, shown in Figure 6. Each Characterisation – comminution and recovery
sulfur grade represented approximately equal proportions of Ores from all pits were characterised by lithology and
the sulfur distribution in the mine model for the ore reserves. alteration types and tested for their comminution competency,
hardness and abrasion properties. Three categories were
identified and grouped as ‘hard’, ‘medium’, ‘soft’ according to
unconfined compressive strength (UCS), Axb, rod mill work
index and ball mill work index values. Ore type distribution
in the mill feed tonnages by period is shown in Table 4. The
corresponding recoveries of gold and silver, based on the
characteristic gravity and leach/CIL recovery performance
Table 3
Mill feed from open pits – distribution by weight and average grades.
Year Pit 1 Pit 2 Pit 3 Total Grade Grade

(wt %) (wt %) (wt %) (wt %) Au (g/t) Ag (g/t)
1–2 80 20 0 100 2.2 20
3–7 95 5 0 100 1.6 9
8–9 60 15 25 100 1.6 6
FIG 5 – Copper cumulative distribution from mine
and ore schedule (LOM: life-of-mine). 10–12 0 20 80 100 1.1 2

Table 4
Distribution ore competency/hardness and recoveries by year.
Year Hard (%) Medium (%) Soft (%) Total (%) Throughput index Recovery Au (%) Recovery Ag (%) Au-equivalent index
1–2 20 60 20 100 0.90 88 60 1.35
3–7 10 60 30 100 1.00 84 50 1.00
8–9 10 40 50 100 1.05 84 50 1.03
10–12 0 5 95 100 1.25 80 40 0.78
by ore type for the blend of ores in each mill feed category, are thus directly affected solution and carbon inventories. The
shown in this table. Comparative throughput and production maximum gold and silver inventories, and the resulting
performances are expressed as a throughput index and a carbon transfer, desorption and regeneration systems to
gold-equivalent index (converts silver value to equivalent manage these metal inventories occurred during the initial
gold) in the table. two years of operation (Table 4). Consequently, this period
was the basis for the design of the carbon adsorption, elution
Grinding and regeneration circuits.
The ore hardness variability in the mine resource block model As the soluble ratio of silver to gold during the initial two
showed that a significant number of blocks in the first two years was about 6:1, silver management was the main design
years of operation had a predicted work index above those consideration and determinant for the carbon circuit design
expected for the following five to seven years. The design and operating strategy.
values for comminution were based on treatment of ores
for years 3–7 in the schedule which comprised nearly three- Gold room
quarters of all ore scheduled for the initial seven years. The Design of the gold room and smelting activities was based on
comminution circuit power and mill sizes were selected based treating peak weekly inventories of gravity recoverable gold
on comminution parameters for this period and a throughput and electrowinning cathode clean-up.
index of 1.0 was assigned. Although the average throughput
during this initial two-year period was 90 per cent of the design
Volumetric capacity
throughput based on the installed power of the selected mills,
production for the first two years on an equivalent recovered The amount of ore classified as ‘soft’ increased with mine
gold basis (Table 4) was 35 per cent higher than at the design life and is the major to dominant proportion of ore feed to
throughput due to the higher grades and higher recoveries the mill towards the end of operations, in years 8–12. During
with the harder ore blend. this period the grinding circuit was able to treat above-design
throughputs and most likely up to treating a 25 per cent
This strategy avoided oversizing and over-capitalising the
higher feed rate. This helped to offset the loss in production
grinding circuit and underutilisation of grinding power in
due to falling grades and decrease in recovery.
subsequent years at the design throughput rate. This also
prevented a flow-on of higher capital costs if the downstream Slurry viscosity and rheology tests on all ore types,
process plant and equipment were increased in size to match particularly the characteristic moderate- to high-alteration
the grinding circuit throughput. ores in the soft category, provided comparative information on
pulp density, viscosity, slurry transportation properties and
Gravity hydraulic gradients in the circuit. The slurry characteristics
and flow properties directly affected duty and specifications
Gravity recovery tests showed that gold recovery was
for equipment such as slurry pumps, trash and safety
moderately low and highly variable for all ore types. Gold
screens, launders, intertank screens and the thickener. Hence,
recovery was about 20 per cent for the lower grade main pits
provision was made in the initial design for equipment, pipes
and approximately 10 per cent for ores from the (smaller) high-
and launders to handle larger flows for periods when poor or
grade pit. The impact of gold deportment to gravity ahead of
adverse slurry properties dictated, or allowances were made
the leach/CIL circuit was relatively minor, particularly in the
to subsequently add to or upgrade existing equipment.
initial years treating the higher grade ore.
For design of the leach/CIL circuit, it was assumed that the
leach circuit would treat all gold in feed, that is, no gravity CONCLUSIONS
gold recovery. This ensured that all gold was recovered in Data obtained from metallurgical and physical ore testing
leaching and adsorption when the gravity circuit was not programs is appropriate and applicable to the extent that the
operating and also provided a small operating margin for sample selection is representative and correct preparation
variability in gold grades to leaching. protocols are practiced.
Silver gravity recovery was consistently very low, The recommended methodology for sample selection as a
typically less than five per cent to a final gravity concentrate basis for flow sheet development and process design is by
in all tests on various ore types. For design purposes, it was characterisation and variability categories. Characterisation
assumed that all silver in the feed was available for leaching accounts for metallurgical and physical properties of the ore
and carbon adsorption. by domain, zone, ore type or specific geological classification.
Variability encompasses subsets of domains such as
Precious metal recovery mineralisation, range of grades of economic metal, range of
A carbon adsorption and desorption circuit which met grades of deleterious or penalty elements, ratio of grades of
the duty required for the initial two years was selected for metals, spatial location along strike and at depth.
this operation. Although the throughput rate was lower, Practical and robust interrogation, interpretation and
ore grades and recoveries were highest in this period and analysis of the complex database of information and detailed

E McLean
data available are fundamental to successful design. This ACKNOWLEDGEMENTS

is not possible without an understanding and an iterative
The inspiration for this paper comes from beleaguered
assessment in conjunction with the mine planning group
colleagues, many of whom have been in the firing line from
of ore feed schedules to the plant by tonnage, grade,
project managers and production superintendents to explain
mineralisation and lithology; of ore exploitation sequences,
‘why’, ‘why not’ and ‘when’.
mining strategy, mine development plan, pit financial
modelling and resource constraints.
The selection of the appropriate design value for each unit
REFERENCES
process or process stage is critical. Coupled with relevant Agarwal, J C, Brown, S R and Katrak, S E, 1983. Taking the sting out of
out of project start-up problems, adapted from a presentation at
design margins, calculation factors, scale-up, benchmarking, The American Mining Congress, September 1983, San Francisco.
operating experience and input flow conditions, the unit
process equipment or plant selected should meet two Lane, G, Staples, P, Dickie, M and Fleay, J, 2008. Engineering design
of concentrators in Australia, Asia and Africa – what drives the
main criteria: to function efficiently in accordance with the
capital cost?, in Proceedings Procemin 2008 V International Mineral
parameters and specifications nominated; and to be cost- Processing Seminar (ed: R Kuyvenhoven, C Gómez and A Casali),
effective from both initial capital and life cycle assessments. pp 30–38 (GECAMIN: Santiago).
The engineer, scientist, or metallurgist who can make Performance Associates International Inc, 2017. So you are investing
apparently impractical things happen and is able to in a mining project – what usually goes wrong [online]. Available
produce outcomes that achieve objectives shows initiative, from: <http://www.perfnet.com/wp/white-papers/investing-
is innovative, creative and resourceful. The inexplicable or mining-project-usually-goes-wrong/> [Accessed: 20 July 2017].
impossible events outside these parameters are magic.

Contents
Back to the future – why change

doesn’t necessarily mean progress
P D Munro1 and P A Tilyard2
ABSTRACT
There have been enormous changes in mineral processing in the past four decades.
For example, grinding mill power has increased by an order of magnitude, regrinding
is done to -10 µm and flotation machines are 100 times bigger. Operating staff have
unprecedented opportunities for online monitoring and performance control of
mineral processing plants. Sophisticated instruments can provide a plethora of data
characterising the mineralogy and surfaces of particles. Digital computers allow
complex calculations on huge amounts of data including modelling and simulation of
machine and plant performance.
However, all these changes have not necessarily led to better metallurgical results. An
analogy can be drawn with the thoroughbred racing industry in Australia. Significant
advances in scientific knowledge in animal genetics, physiology, biomechanics and
nutrition applied to the business have resulted in only a two per cent reduction in
winning times for the Melbourne Cup and Caulfield Cup since the 1920s.
A critical look at some mineral processing metrics suggests similar failures to
improve performance despite putting in more resources. In fact, certain parameters
such as operating times and plant start-up performance are considered to have
remained static or even deteriorated.
There has been an emphasis on ‘process’ at the expense of ‘outcomes’. The industry’s
strength has been in finding technical (or ‘hardware’) solutions while its weakness
has been at the people end of the business in maximising and consolidating the gains
from the technologies. Some trends in plant design over these years have exacerbated
the apparent deskilling of operating and technical staff. Despite unparalleled options
for communications, some staff are embarrassingly uninformed about technical
developments in their fields.
The ‘boom and bust’ cycles of the industry, together with trends in tertiary education
and the effects of fly-in, fly-out (FIFO) operations, raise serious questions about the
sustainability of human capital in the mineral processing sector.
This paper by two experienced mineral processing engineers, with contributions
from other senior practitioners, reviews these trends. While there may be an element
of ‘the older we are, the better we were’, it is an attempt to identify the issues and
propose solutions.
INTRODUCTION
The terms ‘mineral processing engineer’, ‘metallurgist’ and ‘graduate’ are used
interchangeably in this paper.
Most observations and examples have been drawn from base metals sulfide
concentrators using flotation as the separation method. However, the authors
have enough experience with other operations, such as gold leaching and iron ore
processing, to expect that these have similar issues.
The authors have attempted to contrast the mineral processing sector that they entered
as new graduates at the beginning of the 1970s with the current situation in 2009.
While there may be an element of ‘the older we are, the better we were’ in this
paper, it is an attempt to identify the issues and propose some solutions.
AROUND 40 YEARS AGO

1. FAusIMM, Senior Principal Consulting
An industry snapshot in 1970:
Engineer, Mineralurgy Pty Ltd, Taringa Qld
4068. Email: pdmunro@bigpond.com.au •• The typical starting salary for a graduate mineral processing engineer was
2. FAusIMM(CP), Former Group ~$3500/a; accounting for overhead makes the cost to the employer ~$5000/a,
Metallurgist, MMG (now retired). which after six weeks’ annual leave and 8 h/d = $2.72/h cost to the employer.
Email: tilyards@bigpond.net.au Using the Australian Consumer Price Index as a multiplier gives a cost in
11
2009 of $26/h. Another way of looking at this is that •• The design of flotation plants brought the operators and
at 1970 metal prices of US$1300/t for copper, US$35/t metallurgists close to the froth allowing observation
oz for gold, US$350/t for lead, US$2750/t for nickel, and giving an excellent ‘feel’ for the process. As the
US$1.88/t oz for silver and US$300/t for zinc with American baseball commentator Yogi Berra once said
the exchange rate of A$1 = US$1.12, the graduate’s ‘you can see a lot by looking’.
annual salary including overhead had the following •• Comminution was done by multistage crushing with
approximate metal equivalents: rod milling plus ball milling giving stable grinding
•• 4.3 t of copper throughput so the metallurgical ‘narrative’ focused on
•• 161 t oz of gold separation performance.
•• 16 t of lead •• ‘Running time’ was the only thing that mattered, with
•• 2 t of nickel both maintenance and operational groups zealously
guarding their reputation on minimising downtime.
•• 2988 t oz of silver
There was significant focus at a high level on areas for
•• 18.7 t of zinc. improvement.
•• Employment conditions relative to Australian norms •• The flotation section was operated under a strict
were generous and encouraged young people to seek a ‘theory x’ (command and control) set of guidelines with
professional career in the industry; McCarthy (2006) has operators allowed to adjust parameters (such as air, froth
commented on this issue. depth, reagent additions etc) only within limits set by the
•• Fly-in fly-out (FIFO) did not exist. People lived in mining metallurgist. This reflected the high level of accountability
towns where socialising and talking ‘shop’ gave you a the metallurgist had for separation performance.
good appreciation of other disciplines such as mining
•• Prototype on-stream analysers (OSA) started appearing
engineering and geology.
in flotation plants.
•• Joining the Australasian Institute of Mining and
•• Quantitative mineralogical data were collected by
Metallurgy (the AusIMM) was almost a condition of
manual point counting. It was not uncommon for a site
employment and contributing to local branch activities
to have a mineralogical laboratory.
was expected.
•• There were no personal computers. •• Computer models of mineral processes were empirical
and run on main frame machines, programmed on
•• The internet did not exist, with no time consumed punched cards. Conclusions from the models were
reading emails!
tentative because of their novelty.
•• Telephone calls were expensive and making an STD
•• Metallurgists routinely:
(subscriber trunk dialling) call required permission.
•• were held accountable for metallurgical performance
•• Fax machines did not exist.
daily and subjected to relentless queries; this may not
•• Copying was changing over from wet process be in accord with current ‘warm and fuzzy’ human
duplication to ‘Xeroxing’. relations approaches but it certainly provided focus
•• Process control computers had just arrived with 4–8 KB and was character-building – the authors observed
of memory, programmed in Assembler. that those metallurgists who survived and thrived
•• 4 KB of memory cost $4000 (these are 1970 dollars). in this trial by ordeal had the potential to become
•• There were a few electronic calculators around. One of the competent plant managers
paper’s authors was hugely impressed as a new graduate •• had to be very sceptical about the veracity of instrument
to find out that the mill clerk at Mount Isa Mines Limited readings as much of the technology was in its infancy.
(MIM) had one with a square root function! Pneumatic control with 3–15 psi air was the norm
•• Companies had on-site technical capabilities with the •• checked crusher gaps with the ‘leads’
larger ones such as CRA and MIM doing world-class •• checked mill power draw
research.
•• checked hydrocyclone spigots with callipers putting in
•• Information was provided through a company technical a maintenance work request when the wear exceeded
library. This housed all significant reports and circulated 6 mm (¼ inch)
journals. Graduates were expected to read the technical
literature in their professions. •• routinely inspected mill and flotation banks on plant
shutdowns, putting in maintenance work requests.
•• The assay function had not been ‘outsourced’ and the
Decisions on equipment repair issues possibly
company’s chief chemist was a source of wise counsel on
affecting throughput and metallurgical performance
all analytical and chemical matters.
were not abdicated to the maintenance department or
•• Large complex sulfide flotation concentrators could even the purchasing department
be competently run by three metallurgists: control
•• produced a detailed monthly metallurgical report,
metallurgist, metallurgist control plus one graduate. The
commenting on performance on a sized basis
mill clerk adequately coped with most data collecting
and reporting functions. •• produced detailed cost comments – costing systems
•• Flotation was controlled using the vanning plaque (or were logically divided up into ‘cost centres’ for unit
‘pan’) confirmed by wet chemical assays every two processes such as ‘crushing’ and ‘grinding’ comprising
hours. This was a skill that the metallurgist had to expense accounts (such as ‘electric power’, ‘50 mm
master to acquire any credibility with the operators. The diameter grinding balls’ etc).
‘pan’ had the advantage of providing real-time semi- •• The monthly metal balance was done manually and
quantitative mineralogy as well as an estimated ‘assay’. expeditiously, often by the mill clerk.

Back to the future – why change doesn’t necessarily mean progress
•• Meetings were relatively infrequent and brief. Senior is never further than the next rotation out. Support staff
staff generally exercised strict control to ensure they did are seen as a cost with harangues from accountants and
not become ‘talk fests’. managers about the ‘head count’ and cost of supporting
•• Despite the lack of the internet and without cheap people in the camp but neglecting to consider the value
telephone calls, staff at remote sites were well aware of they add.
technical developments in other locations. Publications •• There seems to be a lower level of participation in the
such as Jim Woodcock’s annual review Mineral activities of the AusIMM. FIFO does not encourage
Processing in Australia (Woodcock, 1978) were eagerly this, as attending technical meetings when a person is
anticipated to provide details on operating practices. rostered off is often viewed as ‘work’.
Similar enthusiasm greeted other review articles such
•• Everyone has access to a personal computer and the
those in the Mining Annual Review and World Mining.
internet with most of a person’s time spent in front of it
•• Graduates were expected to be literate and numerate. dealing with emails is a major preoccupation.
Memoranda and reports were carefully vetted by
•• Process control computers are ubiquitous with the cost
superiors for both their technical and English content.
of memory and data storage still falling in accordance
Senior staff were then competent enough in their
with Moore’s Law.
own literary skills to identify errors in grammar and
deficiencies in expression. Rewrites of offending sections •• Very few companies have significant technical capabilities,
were obligatory and frequent. and most rely on research institutes, consultants and less
frequently inhouse technical groups.
•• Reports were written in long hand and subsequently
typed by a secretary. Graphs were hand drawn in India •• The technical library was generally a victim of cost
Ink with the aid of ‘french curves’. cutting in the late 1980s and 1990s. Anecdotes tell of the
•• Future ore testing looked at the ore sources the plant few remaining professional information technologists
would process in future years. (ie ‘technical librarians’) falling victim to the downsizing
that occurred following the global financial crisis of
•• Graduate training at larger operations was structured 2008. Historical technical work is usually imperfectly
with around six months spent doing ‘hands on’ jobs in
filed and collated, often with important omissions. The
the plant including supervisory positions. The next step
unspoken belief appears to be that everything you need
was project work leading to time being responsible for
to know can be sourced from the internet. An unkind
metallurgical performance. There seemed to be more
comment from one senior engineer was that ‘the level of
emphasis on achieving competency in core skills though
inquiry was so basic that the answer is often found on
such jargon terms were not then used.
the internet!’ Conversations with metallurgists quickly
•• There were more opportunities for mentoring. It was reveal significant lacunae in their knowledge. Nothing
not uncommon to find metallurgists over the age of 50 more technical than the introductory textbook ‘Mineral
at a mine site. Processing Technology’ (Wills and Napier-Munn, 2006) is
found on most bookshelves. Those metallurgists with
IN 2009 an interest in reading the technical literature struggle
Observations on the situation in 2009: to convince senior management of the benefit of a
subscription to a data search/retrieval facility.
•• Typical salary for a second-year graduate mineral
processing engineer on a FIFO basis was ~$102 000/a; •• The assay function has been ‘outsourced’, sometimes to
assume 8/6 roster with 0.5 d travelling each way gives the ludicrous point where it is no longer possible to have
seven effective working days on-site. Subtracting four any ‘spot samples’ processed! Some sites don’t even
weeks’ holiday per annum gives 24 effective working have a rudimentary metallurgical laboratory.
weeks at 12 h/d = ~$50/h cost to the employer. Applying •• A large complex sulfide flotation concentrator requires
the same overhead as in the 1970 case gives ~$70/h. At six to eight metallurgists with no seeming improvement
2009 metal prices of US$4300/t for copper, US$900/t oz in the quality of operation. The extra numbers are
for gold, US$1350/t for lead, US$13 000/t for nickel, needed because of FIFO, as support staff were culled
US$12.50/t oz for silver and US$1500/t for zinc with the during the hard times and with metallurgists now often
exchange rate of A$1 = US$0.75, the graduate’s salary performing clerical functions.
including overhead had the following approximate
•• Flotation is controlled using OSA and (surprisingly)
metal equivalents:
still by two-hour assays in some plants. The current
•• 25.3 t of copper (~6 × the 1970 value) design vogue of open air flotation plants, having the
•• 120 t oz of gold (~0.8 × the 1970 value) cells packed together with walkways over the top, has
•• 80.6 t of lead (~5 × the 1970 value) distanced the operators and metallurgists from the
process. A dissenting opinion on the ergonomics of
•• 8.4 t of nickel (~4 × the 1970 value)
flotation plant design is contained in a paper on the
•• 8700 t oz of silver (~3 × the 1970 value) design of the Prominent Hill concentrator (Colbert,
•• 72.5 t of zinc (~4 × the 1970 value). Munro and Yeowart, 2009).
•• Employment conditions relative to Australian norms •• Comminution by autogenous grinding/semi-autogenous
are less generous, discouraging young people to seek a grinding (AG/SAG) mills gives varying grinding
professional career in the industry (McCarthy, 2006). throughput with the ‘narrative’ focused more on the
•• FIFO operations are increasingly the norm. grinding section to the neglect of evaluating separation
•• People do not tend to live in mining towns and do not performance.
socialise much after work. Even on-site the 12 hour day •• Previous simple classifications of plant downtime such
leaves little time for interdisciplinary interaction. An as ‘planned maintenance’, ‘unplanned maintenance/
unkind observation is that the solution to any problem breakdown’ and ‘operational’, ‘lack of ore, water or

power’ have become more complicated including the mineral liberation and beneficiation process. It seems
definitions such as ‘readiness’. to be for ornament rather than use.
•• The end result is that often no one is accountable for •• It is disappointing to observe that quantitative
actual running time, which used to be the mill manager’s mineralogical data are usually one of the items discarded
responsibility. Running times seem to have deteriorated as a response to low metals prices; however, this is
from values achieved in the 1970s and early 1980s. understandable if the metallurgists receiving such data
•• One nostalgic observation is that as soon as the can’t use it effectively.
industry went away from fine crushing and rod and/ •• Computer models of mineral processes are increasingly
or ball mills the ‘rot’ set in. Availability of the ball phenomenological and can be run on powerful laptop
mills at Bougainville Copper Limited (BCL) inched up computers
to 99.1 per cent in 1983, utilisation of available time •• Metallurgists less commonly:
was 99.7 per cent and total run time was 98.8 per cent,
•• are held accountable for metallurgical performance
equivalent to 8655 h/a. Compare this to the current
common SAG mill design run time of 8000 h/a daily and subjected to relentless queries
(91.3 per cent) and actual operating run time not much •• master the fundamentals of their profession, which
better in many cases. detracts from their later performance as plant managers
•• In 2006, one large mining company was designing •• disbelieve instrument readings
its iron ore plants for a run time of 74 per cent and •• check crusher gaps, mill power draw or hydrocyclone
probably still is. It seems to have learned nothing from spigot dimensions
the BCL crushing plant experience where line run time
•• routinely inspect grinding mills, hydrocyclones and
was 96 per cent and secondary crusher run time was
flotation banks on plant shutdowns, then submit
91 per cent. Even experienced iron ore metallurgists
maintenance work requests when required
found these numbers difficult to believe. The secondary
and tertiary crushers and the tertiary screens were fed •• make decisions on equipment repair issues possibly
from large surge bins which significantly contributed affecting throughput and metallurgical performance
to the operating efficiency of the BCL plant. and often abdicate these decisions to the maintenance
department
•• Flotation is now run under a laissez-faire regime where
seemingly everyone’s opinions are valued, regardless of •• produce a detailed monthly metallurgical report
skill and knowledge, to avoid upsetting individuals. commenting on performance on a sized basis
•• Operators are allowed wide scope to adjust parameters •• exploit the power of the spreadsheet which is ideal
(such as air, froth depth, reagent additions etc) and for manipulating mineral processing data to produce
even to make circuit changes. One hears statements like a thorough understanding of the performance of the
‘someone changed the collector addition last night’ at plant according to the axiom of ‘size-by-size mineral
the daily production meeting which doesn’t encourage particle behaviour by liberation class’
a culture of responsibility. •• are held to account for operating costs
•• Democracy has been substituted for the scientific method, •• read the technical literature of the profession.
eg the fact that in the 14th century most people believed •• The monthly metal balance is done in many cases
that the world was flat did not mean it was correct! on expensive computer packages with no seeming
•• There doesn’t seem to be much accountability for improvement in accuracy or speed. The computer can’t
performance. tell you that the head sampler is not in its correct rest
•• OSA and particle size monitors are supposed to be position and is getting constant splash into it.
standard equipment though there are some curious •• Meetings are much more frequent and last longer.
exceptions. At some sites the OSA and/or particle size •• Despite the so-called communications revolution from
measurement system is no longer operational through the availability of the internet, staff seem to be totally
systematic neglect or produces questionable data from unaware of past industry paradigms.
poor calibration procedures.
•• The authors have been astounded to find plants using
•• It is depressing to hear announcements trumpeting rubber or polyurethane spigots in hydrocyclones
the purchase of an OSA system or online particle taking SAG mill discharge. Hydrocyclone spigots were
measurement as a technical breakthrough when such supposedly ‘standardised’ to long-wearing ceramic units
items are assumed to be standard equipment in the 21st during the late 1960s (Munro, Eaton and Burton, 1982).
century. The old adage ‘if you can’t measure it, you can’t
control it’ surely applies. •• Even when company intranets are set-up with a
metallurgical site and discussion page, there is a
•• Quantitative mineralogical data can be rapidly produced
reluctance to seek advice and/or information from the
by automated X-ray methods such as QEMSCAN
wider pool of metallurgical expertise in the group.
(Quantitative Evaluation of Minerals by Scanning
Electron Microscopy) or mineral liberation analyser •• Graduates often lack literacy and numeracy, and when
(MLA; both, incidentally, are Australian innovations); memoranda and reports are written (which isn’t often
however, manual and automated point counting still enough in FIFO operations), they are often not well
survive, with the practitioners of these supposedly constructed and painful to read.
outdated techniques able to offer interpretation and •• With the exception of engineering design companies,
information as opposed to the data produced by the version control of reports and spreadsheets is almost
automated X-ray systems. unheard of.
•• Despite having the luxury of quantitative mineralogical •• A recurrent theme is the disregard of the statistics
data few metallurgists are able to use the data to manage of variation as evidenced by claims of observed

improvement well within the normal ‘noise band’ of •• In early 2006, a large copper concentrator did a thorough
plant and laboratory performance. maintenance overhaul of its flotation cells including
•• Verbal communication is resolutely qualitative rather replacing worn impellers and stators, refurbishing level
than quantitative with numbers replaced by ‘larger’ control equipment and cleaning cell lips. On restarting
versus ‘smaller’, ‘faster’ versus ‘slower’ etc. the plant the sulfide grade of the concentrate increased
from 78 per cent to 82 per cent with 2.5 per cent abs
•• It seems to be easier to get to the geologists’ office at higher copper recovery. Such a significant deterioration
the mine site from head office in a capital city or from in the condition of the flotation cells would and would
the consultant’s office than it does from the nearby have been noticed much earlier if the metallurgists had
concentrator office. ‘Future ore testing’ may have been looked at the froth surface on the cells. One of the authors
repackaged as ‘geometallurgy’ but you still have to talk had a similar experience in a copper zinc concentrator
to the geologists. where he suggested that the flotation cells be drained
•• Graduates spend around 25 per cent of the time doing and inspected on the next shutdown. The appalling state
‘hands on’ training compared with 40 years ago. of the impellers and stators (some had ceased to exist)
showed that no interest had been taken in the flotation
•• Many operations supposedly have detailed programs for
cells by metallurgical staff for years.
the development of graduates; however, there is often a
gap between rhetoric and reality when the authors have •• Daily data for mill products of ‘silver nitrate soluble
examined the knowledge and competencies of people copper’, ‘acetate soluble copper’, ‘cyanide soluble
coming out of these schemes. copper’ and ‘total copper’ were not used to explain
the performance of a large copper-gold concentrator.
•• The age spread of site metallurgists is narrow, rarely Management’s concern and displeasure at high rougher
finding one over the age of 45. tailings losses could have been mollified by pointing out
•• The head office support function also appears to be that the proportion of acid soluble copper in the feed
reducing with major mining companies cutting back on had increased. The copper minerals in this category are
experienced metallurgists in technical services functions. not recoverable by conventional sulfide copper flotation.
•• Mentoring does not seem to be valued as much, •• At a review meeting examining design options for a
and while there may be ‘motherhood’ statements major upgrade of a venerable concentrator, none of the
about it, along with a supposed commitment to staff senior operating staff present knew the current plant
development, the lack of any demonstrated commitment operating cost in $/t of ore treated or its components
to actually providing the resources for mentoring and according to activity.
staff development belies this point. •• At an overseas copper concentrator, a cost saving
initiative was suggested by a group of non-technical
IF THIS IS PROGRESS …? employees assigned the task of improving plant
performance and was accepted by the metallurgical staff.
Below are examples supporting the contention that mineral
The instrument air compressor was decommissioned.
processing outcomes are not uniformly improving:
In due course the wet air from the plant compressors
•• Throughput variations as ascribed to the ore being destroyed the level controllers in the flotation section.
‘harder’ or ‘softer’. It is a long time since we heard a •• Two very large overseas copper concentrators do not
metallurgist say something like ‘our current feed is own a cyclosiser and the metallurgical staff stoutly
from bench ‘AB’ where the predominant rock type is defend the fact that they have no data in the -37 µ size
‘andesite/granite etc’; grindability data for this shows region.
a bond ball mill work index of ‘A’ kWh/t, drop weight
•• At an overseas copper concentrator, a graduate
index a × b of ‘B’ etc’. Despite all the excellent tools for
metallurgist disputed the visiting consultant’s opinion
data collection, manipulation and display, people seem
that, in the absence of OSA (the equipment had fallen
resolved to be qualitative rather than quantitative. It was
into disuse), two hourly spot assays would be useful.
hoped that the Australian Minerals Industries Research
Association (AMIRA) P843 geometallurgy project •• Very few operations are able to present in a single
would change this, but it seemed to be driven more by document, the performance of the plant over the past
geologists than metallurgists when it ran its course. say five to ten years.
•• In 2004, a large copper producer in a foreign country

asked a respected Australian research institute to MINERAL PROCESSING BASICS
investigate why there was a difference in metallurgical The fundamental data requirements for managing a mineral
performance between its new plant with one type of processing plant are the following:
flotation machine for rougher duty and the older plant •• The target metallurgical performance (eg concentrates
which had another machine. After a detailed campaign grade, metals recoveries, what you want to reject and
including measuring parameters such as superficial the priority for these targets).
gas velocity and bubble surface area flux it was found •• An equipment list.
that the rougher flotation cells in the newer plant had
•• A mass balances for solids, elements, minerals and water
significantly lower effective volumes because of the
accumulation of tramp oversized ore particles from •• A simulation model of the comminution circuit with
‘upsets’ in the grinding section. After the cells were grindability data for current ores.
cleaned out copper recovery increased from 83 per cent •• Quantification of ore types (both grinding and
to the target of 90 per cent. It is a telling commentary on metallurgical performance) in the plant feed for future
how removed people now are from the flotation process years and relation to past performance.
that someone has to come from another country to tell •• Knowledge of the processing characteristics of the
you that your cells are full of rocks! orebody in a spatial sense. This should include

grindability, quantitative mineralogy, metallurgical making retrieval a laborious and daunting undertaking.
performance, distribution of precious and impurity Electronic files of plant operating parameters from process
elements and minerals (eg Au, Ag, As, F, Hg, organic measurement and control systems frequently have long periods
carbon, talc etc) where applicable. of missing or erroneous data. Sometimes complete archives of
•• Grinding and regrinding mill power consumptions. metallurgical reports and plant data have gone missing when
a computer system was upgraded. Are mineral processing
•• Element and mineral particle behaviour on a sized basis.
engineers condemned to be ‘children of lesser data’?
•• Mineral particle behaviour on a sized basis by liberation
class on a monthly basis.
THOROUGHBRED RACING Versus
•• Trade-off between concentrate grade versus recovery
and plant throughput versus separation section DENTISTRY – A CAUTIONARY TALE
(eg flotation) feed sizing. Quirk (2006) made some trenchant observations on the
•• Effect of concentrate grade and recovery on net smelter performance of the thoroughbred racing industry. Winning
returns (NSR). times for the Melbourne Cup, which is an open handicap
event and not a race of equals, have improved only
•• Detailed chemical analysis of concentrates and concentrate
three per cent since the 1920s. Similarly, the W S Cox Plate,
physical data such as transportable moisture limit.
which is weighted for age and supposedly a classic race for
•• A history of the metal balance/metal accounting with the best horses, has shown only a two per cent improvement
accompanying narrative on all adjustments made over the same time period. These data are shown in Figure 1.
to the ‘first pass’ numbers. This should also include
Similarly, analyses in the 1980s of the classic English races
reconciliation of data from dispatches and receipts of
the St Leger, the Derby and the Oaks showed that little or no
products to customers.
improvement in winning times had occurred in the previous
•• Regular analyses of process and effluent waters. 70 years.
•• Consumables usage, eg grinding media (also reported By comparison, human athletes in the Olympic Games over
as weight loss per kWh of mill power), reagents and the same time period had an eight per cent improvement in
including water. winning times for the 100 m, ten per cent for the 1500 m and
•• Operating cost data: 12 per cent for the 5000 m.
•• on Pareto graphs So, despite having an array of modern technologies in genetics,
•• by activity (eg crushing, grinding etc) nutrition, biomechanics, equine physiology etc available, the
thoroughbred racing industry, as measured by winning times,
•• by expense account.
has not improved its performance. The reason given for this is
•• A simple financial model of the plant which incorporates the limited equine gene pool available for development.
the technical drivers and their effect on NSR.
The authors are not currently suggesting that the gene pool
•• A metallurgical development plan which incorporates of metallurgists is the cause of our perceived dissatisfaction
the above trade-offs and outcome drivers etc. with the performance of mineral processing professionals;
•• An enthusiasm for capturing plant data on an ongoing however, the cautionary message from the thoroughbred
basis. Unfortunately, a compendium of the above data is racing industry is that merely applying modern scientific
rarely seen when visiting a plant. techniques does not ensure the desired outcome.
It is interesting to note that concentrator staff are likely This is sharply contrasted to advances in modern dentistry
to have compiled job safety analysis documents for many over the same time period where technology and innovation
physical tasks in the plant but have minimal cogent have clearly improved the patient’s physical (if not his/her
documentation on how to do the metal balance, production fiscal) well-being (O’Rourke, 1995).
forecasts or metallurgical planning.
THE CONTEMPORARY MINERAL PROCESSING SITUATION
GEOLOGISTS AS EXEMPLARS FOR DATA AND INFORMATION There have been significant improvements in mineral
We should take a lesson in data collection and information processing technologies over the past four decades as per the
management from our geological compatriots. following examples:
Mineral deposits can be discovered and mines can operate •• Machines have become much larger, eg 20 MW SAG mill
for decades going through multiple ownerships with the compared with a 3 MW (max) ball mill. This has allowed
geological database remaining intact and up to date. You
hardly ever encounter situations where there isn’t any data for
a drill hole because the assay sheet was shredded five years
ago when the project geologists were ‘outplaced’ during the
last cut-backs or the samples weren’t analysed to save money.
Chief geologists are not in the habit of throwing away drill
core to eliminate the cost of storing it.
Contrast this to the metallurgical test reports on those same
geological samples. You would be an optimist to expect that
they would still be found in the concentrator department.
Essential fittings for many metallurgists’ offices are a plethora
of seemingly important test work and survey reports covered
in grime and the circular marks of beverage containers
haphazardly scattered over flat surfaces. Electronic copies of
reports when archived on computer servers have cryptic and
sometimes unintelligible titles and are not filed systematically FIG 1 – Melbourne Cup and Cox Plate winning times (ten year averages).

a single unit to replace a multiplicity of very small ones The most significant contributor to this is the dominance
for both comminution and separation duties. of electronic information technology in the working life
•• Capital efficient AG and SAG mills instead of crushing of a metallurgist. This area, which without a doubt has
plants plus rod mills plus ball mills for comminution experienced the greatest advances over the last 40 years, is the
(though some would argue this has been at the expense prime cause of process overcoming outcome.
of rapid plant start-up, fluctuating metallurgical
performance and increased downtime). MANAGEMENT RESPONSE
•• High-pressure grinding rolls (HPGR). If management fails to comprehend and master the technical
•• Comminution technologies that can economically grind basics of plant operation it will not understand the outcome
down to -10 µ. ‘drivers’ and be less capable of directing the available
metallurgical talent pool.
•• Sorting technologies using sensing of multiple mineral
The authors rarely encounter a coherent plan of metallurgical
characteristics.
development for a mineral processing plant with the most
•• Very high field strength magnetic separators. precious resource on-site of ‘Competent Person hours’
•• ‘High g’ gravity separators. appropriately focused on the most productive opportunities.
•• Carbon-in-pulp (CIP) and carbon-in-leach (CIL) Management must recognise this ‘black hole’ of
replacing solid liquid separation plus Merrill-Crowe metallurgist’s time from:
process for gold extraction. •• the ‘bureaucratisation’ of work with ‘process’
•• Pressure oxidation and bioleaching for refractory gold overwhelming ‘outcome’
ores. •• producing abundant data but a paucity of information.
•• Solvent extraction plus electrowinning for copper leach What value does a metallurgist add to the understanding of
pregnant leach solution. a mineral processing operation by spending three hours per
•• Selective flocculation. day producing the daily report with a plethora of egregious
•• OSA; online measurement of both coarse and fine irrelevant details? Surely this is not an intellectually satisfying
particle size distributions; online measurement of outcome after spending four years at university?
chemical parameters such as redox potential, cyanide Why such atrocious wastes of professional time occur in this
concentration, dissolved oxygen level etc. supposed ‘information age’ and are tolerated, considering the
higher ‘metal equivalents’ of metallurgist’s time, are questions
•• Expert control systems.
for serious reflection.
•• Ability to collect detailed quantitative mineralogical
Corrective actions must include:
data including QEMSCAN plus MLA, infra-red (IR) for
alteration etc. •• Putting the ‘mineral processing basics’ in place.
•• Proven models for the simulation and control of mineral •• Freeing up metallurgists’ time; eg automatic collection
comminution and to a lesser extent separation processes. and processing of online measurements and hiring
sufficient clerical and support staff. With metallurgists
•• Computer aided design (CAD) for 3D visualisation
being relatively more expensive to employ than 40 years
of plant layouts with accompanying efficiencies in
ago, there is even less excuse to waste their time doing
fabrication and plant construction.
jobs that can be farmed out to others.
•• The capacity to collect, analyse, manipulate, store and •• Focusing the available professional time on a strictly
retrieve data plus information on a scale unimaginable prioritised list of metallurgical opportunities with
in 1969–1970. quantifiable outcomes.
We should be wary about confusing: Further development of the ‘Mineral Processing Toolbox’
•• movement with action section of the AusIMM website to include examples of good
•• activity with achievement practice for some of the mineral processing basics cited, may
be an appropriate method to focus on the real issues driving
•• change with progress.
mineral processing outcomes.
The contention is that all this extra hardware and software is
Such a web-based facilitation mechanism could lift industry
not necessarily giving uniformly better performances across
performance, especially if companies would be prepared to
the mineral processing sector.
contribute suitable examples in a collaborative spirit, similar
There is some truth in the retorts to the above complaints to an AMIRA project.
that:
•• some of the orebodies currently being treated are more CONCLUSIONS
refractory than those processed 40 years ago
By some criteria, the standard of operation of mineral
•• metallurgists are relatively more expensive to employ processing plants has not significantly improved in the last
than they were in 1969–1970 for most commodity 40 years despite the ‘information technology revolution’.
producers. Professional time for process monitoring and improvement
Yet instances of poor metallurgical outcomes seem to be has been eroded by the ‘bureaucratisation’ of the metallurgist’s
too frequent and too serious to be classed as aberrations. The work, despite the fact that, for most commodity producers,
suspicion is that the industry has some systemic problems process engineers are now relatively more expensive to
that have to be addressed. employ than in 1969–1970.
The authors contend that the decline in the productive Metallurgists need to get back to the basics of their profession
output of metallurgists is the erosion of the limited time to produce outcomes rather than being overwhelmed
available to do professional work. by processes. There is a risk that information technology

can ‘enslave’ rather than ‘liberate’ with data displacing McCarthy, P, 2006. Message from the Managing Director [online],
fundamental mineral processing information. Digging Deeper, AMC Consultants. Available from: <http://
www. amcconsultants.com.au> [Accessed: 5 June 2009].
The concentrator manager has to take the leading role in
turning around the current unsatisfactory state of affairs. Munro, P D, Eaton, R and Burton, E, 1982. Wear materials experience
in Mount Isa concentrators, in Proceedings Second AusIMM Mill
Operators’ Conference, pp 327–335 (The Australasian Institute of
ACKNOWLEDGEMENTS Mining and Metallurgy: Melbourne).
The authors thank Mineralurgy Pty Ltd and MMG Ltd for O’Rourke, P J, 1995. All the Trouble in the World: The Lighter Side of
permission to publish this paper. Overpopulation, Famine, Ecological Disaster, Ethnic Hatred, Plague,
and Poverty, 340 p (Atlantic Monthly Press: New York).
The following individuals are thanked for their observations
on these matters over the years: Gary Chilman, John Glen, Quirk, T, 2006. Correct weight [online]. Available from: <http://
Greg Lane, Rolly Nice, Joe Pease, Geoff Richmond, Peter www.onlineopinion.com.au/view.asp?article=5282&page=0>
Rohner, Stuart Smith and Michael Young. [Accessed: 26 July 2017].
The authors stress that the ideas, opinions and biases in this Wills, B A and Napier-Munn, T J, 2006. Wills’ Mineral Processing
Technology, Seventh Edition: An Introduction to the Practical Aspects
paper are their own.
of Ore Treatment and Mineral Recovery, 456 p (Butterworth-
Heinemann).
REFERENCES Woodcock, J T, 1978. Mineral processing in Australasia 1978,
Colbert, P J, Munro, P D and Yeowart, G, 2009. Prominent Hill Australian Mining, pp 16–90.
Concentrator – designed for operators and maintainers, in
Proceedings AusIMM Tenth Mill Operators’ Conference, pp 23–32 (The
Australasian Institute of Mining and Metallurgy: Melbourne).

Contents
Back to the future – still on the dark side

P D Munro1
ABSTRACT
This is a reprise of the paper ‘Back to the future – why change doesn’t necessarily
mean progress’ presented to the Tenth AusIMM Mill Operators’ Conference in 2009.
It was observed that the ‘boom and bust’ cycles of the industry, together with trends
in tertiary education and the effects of fly-in fly-out (FIFO) operations raised serious
questions about the sustainability of human capital in the mineral processing sector.
In 2016, the industry found itself in a ‘bust’ after enjoying a decade of the longest
commodity price ‘boom’ in a century. In the boom did the industry make any
technological breakthroughs, and/or put measures in place to improve its human
capital, especially in the area of ‘professional formation’?
This paper makes observations on the current performance of mineral processing
engineers and the organisational milieu in which they operate. It questions whether
mineral processing engineers are actually ‘adding value’ at some sites given their
organisational and operational practices.
Some technical trends are examined and what they might mean for mineral
processing engineers.
Given the above, and demographic trends for mineral industry professionals,
changes to the seemingly current laissez-faire model of professional development
are proposed.
INTRODUCTION
At the Tenth Mill Operators Conference held in Adelaide in 2009 Peter Tilyard and
this author looked at ‘human capital’ in the mineral processing sector contrasting the
situation with the industry that it had entered at the beginning of the 1970s (Munro
and Tilyard, 2009). After seven years, it felt timely to reconsider the position, and to
see if the industry had improved the situation, especially given the perception that
the ‘minerals boom’ driven by China’s massive demand for raw materials was now
over and the mining industry seemed be coupled with the word ‘crisis’ in the media.
This is a personal view derived from over 45 years of experience in the business.
In the paper, the author uses the term ‘metallurgist’ to generically cover the job
titles/descriptions of primary metallurgist, mineral processor and process engineer,
while ‘concentrator’ and ‘mineral processing plant’ are used interchangeably.
Concentrator references are to a sulfide flotation plant but are just as applicable to
any mineral processing operation.
HUMAN CAPITAL – LESS ‘KNOW-HOW’ IN 2016

The lost decade
From 2000 the minerals industry enjoyed the longest commodity boom in a century.
In Australia for the 12 years to 2012, the world price of its mining exports more than
tripled and investment spending by the mining industry increased from two per cent
of gross domestic product (GDP) to eight per cent. Mackenzie and Cusworth (2016)
stated that from 2005 to 2012 private new capital expenditure in the Australian mining
industry grew from A$14.2 B in 2005 to A$94.5 B in 2012 for a mean annual growth
rate of 31 per cent.
Did the industry:
•• make any real technological breakthroughs?
•• improve the quality of its human capital by strengthening educational
institutions and skill development schemes for graduates?
In the context of an industry that was intent on increasing production at any cost,
1. FAusIMM, Senior Principal Consulting the answer to both of the above questions is ‘no’. While other economies with strong
Engineer, Mineralurgy Pty Ltd, Taringa Qld mining industries in Africa and the Americas also enjoyed felicitous times, this author
4068. Email: pdmunro@bigpond.com.au believes that the industry’s processing knowledge base continued to erode.
19
P D Munro
Consider the following: that computer simulation studies are being substituted
•• Decline in technical competency – we complained about for laboratory time on subjects such as ‘unit operations’ to
this in 2009 and it has continued to be noticed by others save money in process engineering courses.
(McCaffery, Giblett and Dunne, 2014). The situation is now worse than in 1998, considering that
•• Decreased research and development capability: the demographics of the population of metallurgists will
result in imminent loss of technical expertise as older people
•• Closure of corporate research laboratories – Australian
leave the industry.
examples are BHP Billiton, MIM Holdings, Pasminco
and WMC with diminished capability at what remains.
The conversion of former corporate research and Professional formation and development
development facilities to commercial units offering The usual suspects of authors (Drinkwater et al, 2011;
services to external customers must emphasise Drinkwater and Bianco, 2013; Drinkwater and Napier-Munn,
investigating current problems at the expense of longer 2014; McCaffery, Giblett and Dunne, 2014) have discussed the
term ones. Major companies, who traditionally were need for professional formation and development.
major technical innovators are now very risk averse This author will examine, in more detail, the conversion of
and obsessive about intellectual property (IP) rights. chemical engineers into metallurgists. Quoting directly from
They have created bureaucracies in research and Drinkwater (2015):
development (R&D) making it unlikely that they will Data collected in Australia in 2011 (Lind, 2013) showed
ever achieve much by comparison with past efforts.
that most metallurgists in Australia took their first degrees
•• Reduced government activity for minerals research – in chemical engineering, chemistry, materials science
the United States Bureau of Mines closed in 1995 with or engineering and a variety of other disciplines; only
the loss of US$100 million/a in funding. Environmental 25 per cent studied in programs with the words ‘mineral
and sustainability issues are more likely to attract process’ or ‘metallurgy’ in the name. It may be surprising
governmental R&D dollars than minerals industry to some of our younger colleagues to learn that this is
ones. This is confirmed by looking at the respective not a new phenomenon. Many of our most prominent
numbers of Cooperative Research Centres (CRCs) in senior metallurgists started out 30 or more years ago as
each sector in Australia. analytical chemists, materials scientists, steel-works cadets,
•• One unsettling snapshot on the current status of mineral instrument technicians and (of course) chemical engineers.
processing is to compare the technical content on Chemical engineers are the largest and most reliable
processing of the third edition Australasian Mining and source of supply of new metallurgists. Chemical engineering
Metallurgical Operating Practices, third edition (Rankin, educators and practitioners point out that dealing with solids
2013) with that in the first two editions (Woodcock, 1980; is more difficult than gases and liquids (Merow, Phillips
Woodcock and Hamilton, 1993). While the third edition and Myers, 1981; Nelson, Davies and Jacob, 1995; Merow,
should still be in every metallurgist’s library, it doesn’t 2000). In the minerals business, the solid (ie ore) tends to be
have as much detail as the first two. The author wonders heterogeneous and to change over time. Metallurgists spend
if the lack of content from operations of some major more time than chemical engineers in their tertiary studies
companies indicates waning interest in, and commitment dealing with solids. The industry successfully converted
to technical excellence from senior management. chemical engineers into metallurgists through a combination
•• Drivel spouted by mining companies about ‘people of on-site training and formal courses. FIFO operations do not
being our greatest asset’ is just nonsense considering meet the industry’s requirements for the development of a
examples such as the lack of meaningful industry metallurgist’s technical skills.
funding put into education and training, the usual knee Formal continuing education programs such as Metskill
jerk responses in the economic downturn of not hiring
(Drinkwater and Bianco, 2013) are struggling in an economic
new graduates, not offering permanent jobs to those
environment that has put cost reduction ahead of developing
coming off graduate training schemes and stopping
its people.
technical skills programs for professional development.
Peter McCarthy of AMC Consultants was right when he In 2016, there are diminished employment and professional
said that real business improvement and value creation development opportunities for metallurgists.
comes from experts who have a deep understanding of
how a mining operation makes money (Anon, 2015). STILL GOING BAD
That is, the industry needs more technical people doing The author includes a list of some examples of egregious
meaningful things, not less. The human resources sections operating practices, malfeasances and misunderstandings
of companies don’t seem to make much noise about the to move the reader out of his/her comfort zone about how
long-term effects on the organisation’s human capital the industry is running plants and offer a quantum of solace
stock when these ‘downsizings’ happen. In 1998, the through Schadenfreude (ie ‘well at least we aren’t as bad as that’):
Minerals Council of Australia published Back from the
Brink: Reshaping Minerals Tertiary Education, the National •• The bizarre concept of ‘value engineering’ in plant design,
Tertiary Education Taskforce discussion paper. With where the norm is to leave 25 per cent ‘fat’ in the design
regard to the issues raised in that document short-term for later removal – another way to look at it is the art of
measures got us through the early 2000s but here we are rendering an economically marginal project technically
in 2016 with reports of mineral industry related courses at unviable by taking ‘bold’ decisions on equipment sizing,
universities still under continuing threat of closure. Why selection and layout. For an alternative view on plant
would a university run a mineral processing course for a design refer to a paper on the design of the Prominent
small number of students requiring expensive laboratory Hill concentrator (Colbert, Munro and Yeowart, 2009).
facilities when it could fill a lecture hall with full-fee •• An acceptance of mediocrity in equipment running
paying international students on a subject like ‘business times – Bougainville Copper showed what could be
studies’? The author is also concerned about anecdotes done by understanding the basics (Tilyard, 2009; Tilyard

and Clarke, 2010). Alternatively there are managements when it was actually a low quality one with less than
that expect Bougainville-type outcomes after reducing 50 per cent w/w copper sulfide minerals.
the maintenance workforce. •• The testing of a new supposedly stronger flotation
•• The existence of three large concentrators without collector, which actually had a shorter length carbon
ore:water ratio control on the primary grinding mill, chain than the existing collector – Sutherland and
something that had supposedly become standard in the Wark’s classic text Principles of Flotation (1955) states that
1960s. contact angle is dependent on the length of the carbon
•• The fact that at a recent international conference chain. Knowing the chemical composition and structure
a colleague underwent the torture of listening to of reagents would prevent time being wasted on
a presentation extolling the virtues of adding a meaningless testing of reagents. Two other irritants are
hydrocyclone to a grinding mill and operating in closed metallurgists trying to fix a mineral liberation problem
circuit. What will they think of next…gasp! with reagents and not understanding the statistics of
•• The installation of hydrocyclones without vortex finders, testing. With a supposed sound grounding in statistics
enduring a long run of bad metallurgical results, only to during tertiary education and the availability of Tim
find out that poor classification in the grinding section Napier-Munn’s seminal book Statistical Methods for
was the cause. Mineral Engineers: How to Design Experiments and Analyse
•• A new flotation circuit that experienced 99 flotation Data (Napier-Munn, 2014) there is no excuse for not
cell ‘sanding’ events in 120 days and even more ‘near doing tests rigorously. If the metallurgists can’t lever
sanding’ ones, causing poor metallurgy and resulting in off their education and training to add value, then as
some blocks of cells decreed to be inoperable and thus mentioned below, why have them on-site?
bypassed – the concentrator operating staff whinged
•• The act of doing locked cycle flotation tests to ‘save time’
about the cells ‘having problems’ and being ‘no good’
but couldn’t trouble themselves to quantify the problem. before establishing conditions using batch tests and
They did not share these facts with the engineers who compositing drill core over unmineable intervals.
were designing an expansion of the flotation section •• The installation of a new plate and frame filter press
which was proceeding based on using the same cells. at a cost of A$7.5 million to replace existing disc filters
This author expected that the operating staff would have for concentrate dewatering, only to find out that the
shown a modicum of interest and initiative in resolving latter could do the duty when the concentrate thickener
a problem that they would end up having to live with. underflow pump was upgraded, to allow delivery of
•• Incorrect maintenance decisions and practices affecting higher density slurry to the disc filters.
metallurgical performance. At times, it seems as if all
the good words spoken and written about the need BIG DATA – A DIRGE ON PLANT INSTRUMENTATION:
to stop people operating in ‘silos’, and the dangers of
inappropriate key performance indicators, have been LET’S DO GEOMETALLURGY INSTEAD
ignored. When this author sees lowering maintenance
costs by increasing equipment service intervals causing Instrumentation reality check
poor metallurgical performance, the author wonders if Mining company executives and the technical press extol
concentrators operate in some parallel universe. Peter the virtues of ‘big data’ and the wonders this concept will
Tilyard and the author commented on this in their 2009 supposedly achieve towards improving operating outcomes.
paper and as soon as metal prices dropped, on too many
There is a credibility chasm between the lofty concept of
occasions, it has been ‘back to the future again’!
‘big data’ and the reality in the current general unsatisfactory
•• Metallurgists willingly abdicating important decisions state of plant measurement instrumentation and process
affecting metallurgical performance to the maintenance/ control systems. These are too often characterised by low
reliability engineering section – a probable contributor
availabilities, inaccurate calibrations, discarded process
to such aberrant outcomes is that while operating costs
control strategies etc. The author distinguishes between
can be measured to A$0.01/t, metallurgical benefits are
‘embedded instrumentation’ in a piece of equipment
sometimes expressed more nebulously. This results from
a lack of data prescribed in the section ‘Mineral processing eg ammeter, voltage, electric motor rev/min versus
basics’ below and the reluctance of too many metallurgists instrumentation specific to a concentrator. Problems with
to tear themselves away from their computer screens and the measuring and sensing elements such as density gauges,
see what is happening in the concentrator. Two examples flowmeters, level sensors, on-stream analyser streams etc
suffice to demonstrate the problem: are the inevitable result of reducing maintenance input; the
1. A very large copper concentrator increased recovery author can’t see how you can expect high running times from
by three per cent after improved maintenance part-time instrumentation support on a FIFO basis.
restored the rougher flotation cells to the original Another concern in larger companies is the tyranny
installed condition. of enterprise data management systems typified by the
2. A maintenance section decided to cut the baffles out persistent encroachment of corporate information technology
of a flash flotation cell to reduce wear, resulting in no (IT) interests into the process control area. IT people seem to
feed going to the gravity concentration section. think that if it’s a computer then they want to own and manage
•• Achieving 50 per cent running time for regrind it. Too often the result is the imposition of inappropriate
mills where the copper sulfide minerals in the final software standards by people who may know something
concentrate were only 70 per cent liberated. The about computers but very little about process engineering.
predominant copper minerals were secondary sulfides, This ‘top down’ approach stifles innovation in process
so the operating staff comforted themselves that they control. This author has yet to hear of a mineral processing
were making a concentrate assaying 22–25 per cent Cu plant controlled in real-time by SAP.

P D Munro
Geometallurgy system. Claims that they are better supported than Excel are
just tosh; hasn’t anyone heard of Microsoft?
Consider one area where the industry actually does have data
but lag woefully in turning it into information resulting in This author fears that they further encourage metallurgists
actions for the plant operators. As mentioned in Munro and to spend even more time in the office rather than in the plant,
Tilyard (2009), geologists are better custodians of data than as they won’t tell you that:
metallurgists. They collect a plethora of data from drill core, •• the head sampler is stuck at the edge of the stream and
such as the following: is receiving splash
•• assays •• the flotation operator is decanting the slurry sample and
•• rock type and alteration from losing the higher grade fines before filtration.
•• geological logging We are missing the point if the industry is looking for an IT
solution to what has always been an equipment and people
•• petrological examination problem. A realistic treatment of measurement errors is often
•• photographs lacking in these systems, a concept that many accountants who
•• mineralogy and textural analysis from drive them can’t comprehend as they deal in money, which can
be counted; rather than tonnes or assays which, are measured.
•• hyperspectral logging
•• mineragraphy
CONCENTRATOR OPERATIONS
•• XRD (X-ray diffraction)
•• automated methods such as QEMSCAN and MLA Mineral processing basics
•• geotechnical data Nothing has changed from Munro and Tilyard (2009) in
the fundamental data requirements for managing a mineral
•• rock strength data eg UCS, PLI
processing plant.
•• competence RQD.
A particular peeve is the absence of a credible metallurgical
To this should be added the location in 3D mineralised development plan. Some of the documents that claim to be
space of every sample taken for mineralogical examination, one are laughably unfit for purpose. Below are common
comminution and flotation tests etc. Unfortunately, current examples of bad practice:
operations often make little use of the extensive metallurgical
•• A bewilderingly large number of ‘metallurgical
test work done for a feasibility study and fail to properly initiatives’ – some operations have hundreds of them.
integrate subsequent test work into this body of knowledge. The manager and technical superintendent should
They seem to be two separate ‘silos’ rather than being allocate the professional time required to complete
regarded as a continuum of knowledge. The author wonders each initiative then relate the total to the ‘discretionary
how interested people are in their profession when one hours’ available for the metallurgists from both internal
encounters operating metallurgists who either have not read and external resources. Going through such a list with
the feasibility study or, if done so, failed to understand it. ‘competent person hours’ as the limiting resource leads
People talk about geometallurgy, but when all is said to the realisation that the deposit will have been mined
and done there is a great deal more said than done. Some out before even a small fraction of the initiatives have
operations do use geometallurgy in managing plant been completed.
operations (Butler et al, 2016) but remarkably few have •• Competing priorities – this author has seen a set of
fully integrated it into their way of doing business. In 2016, metallurgical initiatives with 30 ‘priority 1’ items. The
the fact that too many geologists and metallurgists are still compilers of such a list must have lacked the basic mineral
operating in ‘silos’ is totally the fault of the latter. The author processing data listed above to appropriately allocate
hasn’t observed any significant improvement in this over priorities according to ‘pay-off’, ‘time to complete’ etc.
the last seven years with people continuing to be resolutely •• Vague targets – an example would be ‘increase copper
qualitative rather than quantitative. For example, throughput recovery by two per cent’ or slightly better, ‘increase
variations continue to be ascribed to the ore being ‘harder’ copper recovery in rougher flotation by two per cent’.
or ‘softer’ rather than hearing a metallurgist at the morning/ What this author wants to see is something like ‘increase
daily meeting say something like ‘our current feed is now recovery by four per cent of composite particles that
from bench/stope ‘AB’ where the predominant rock type is are 50 per cent chalcopyrite or more with non-sulfide
andesite; grindability data for this rock type shows a Bond gangue in the +75 μm -106 μm size fraction in rougher
Ball Mill Work Index of ‘A’ kWh/t, JK Drop Weight/SMC flotation’. The first two examples above are ‘motherhood’
a × b of ‘B’ etc so the predicted throughput from the grinding statements. They risk that if a metallurgical initiative
simulation model should be ‘X’ t/h’. is implemented to ‘increase copper recovery by
This author dreams of people at the daily meeting in every two per cent’, its benefit could be diminished by factors
concentrator reviewing a 3D representation of the orebody on such as a change in ore type or operating practices. It is
a screen that incorporates the data described, explaining both much easier to demonstrate to senior management that
yesterday’s performance and predicting future results. you actually did ‘increase recovery by four per cent of
composite particles that are 50 per cent chalcopyrite or
We have the technology to do this, but don’t use it. more with non-sulfide gangue in the +75 μm -106 μm
size fraction in rougher flotation’ as opposed to not
Bureaucratisation of the metal balance achieving the more general target. You retain credibility
Another gripe is the bureaucratisation of the metal balance; with senior management who are rightly sceptical of
it has become integrated into an enterprise data management previous promises about metallurgical improvements
system promoted by accountants and the IT department. which, if aggregated, often would give a metal recovery
There are very high ownership costs of these metallurgical over 100 per cent. This leads you to understand the
accounting add-ons to the enterprise data management process according to the axiom of size-by-size mineral

particle behaviour by liberation class but this a fundamental metallurgist, the author witnessed a senior manager fully
of the profession. This author wants a focused target comprehending the operations of two large concentrators and
instead of a general one where the result can be obscured two smelters in a little over an hour including travelling time
by the ‘noise’ of plant operations. between each plant. The emphasis was on information rather
If the concentrator manager doesn’t have a credible than data, the future rather than past, with metallurgists
metallurgical plan based on process fundamentals then held accountable for performance and subjected to ruthless
one will be imposed by senior management, generally with questioning about any deviations. We can learn from the
unsatisfactory results. ‘orders group’ concept used in military forces on how to run a
An excellent management book is Moneyball: The Art of successful meeting but if you want ‘participatory democracy
Winning an Unfair Game by Michael Lewis (2003), which then …’. This author has not sighted any evidence showing
describes how the Oakland Athletics baseball team managed that the workforce has superior skills in chemistry, physics
by Billy Beane successfully competed with a player roster that and mechanical aptitude compared to 45 years ago, excluding
lacked the stars fielded by other teams and had a budget only the ability to send text messages. Indeed, with respect to issues
one-third of that of the New York Yankees (read the book – like vaccination, ‘organic’ food and homeopathy etc, one
the Hollywood movie left out the mathematics, which was the could believe that society has become more technologically
really interesting part). illiterate!
Quoting directly from the book:
One cliché is the ‘five-tool player’, a player who can hit for SOLID LIQUID SEPARATIONS – NEGLECTED UNIT OPERATIONS
average and power, and run, field and throw like a demon.
You still hear it all the time. Solid liquid separations – uninteresting and unloved?
Mineral processing is simplistically described as ‘LSD’, that is:
Right. And everyone wants five-tool players. There’s just
very few of them on the planet! But virtually every player has •• liberation
one tool. So, we started saying, ‘Well, let’s investigate each •• separation
player’s strength. Is there a way to combine all these strengths •• disposal.
and cover up some of their deficiencies? We’re not going to be
able to do away with their deficiencies. But as a team, can we The ‘disposal’ part of the business for both product and
do it all?’ It was about piecing together players we had on our waste streams has been ancillary to the other two sectors of
roster and building a team that could do everything. You had ‘liberation’ viz comminution + classification and ‘separation’
to find value where it wasn’t readily apparent. as done by gravity concentration, flotation and other methods.
Taking the case of a sulfide flotation concentrator, previously
One lesson from Moneyball was that you can’t assume
mineral processing engineers have tended to focus their
that a collection of apparently talented individuals will
intellectual efforts on the ‘interesting’ activities of grinding
serendipitously give the required result. Hiring someone who
and flotation; the former is usually the highest cost unit
looks like the ‘complete metallurgist’, analogous to the ‘five-
operation and the latter the bit where it is easily discerned if
tool player’, and just leaving him/her to it under a ‘laissez-
money has been made or lost.
faire’ approach isn’t getting us the required outcomes listed
above under ‘Mineral processing basics’. An example of neglect in solid-liquid separations is the
indifference about the operation of clarifiers and thickeners
A second learning from Moneyball is that you can compete
(the two duties hereafter combined in the single unit operation
by focusing on the separate components that contribute to the
desired outcomes. If you accept that you need the items in of ‘thickening’) which are ubiquitous in concentrators and
the ‘Mineral processing basics’ list then you should prescribe extractive metallurgical plants. Now into my fifth decade in
that they have to be produced, it’s not some kind of an option; the business, this author has seen a considerable number of
then work out the processes and actions required from both them, and the operation of many could only be described as
internal and external resources to make it happen. ‘poor’ to ‘appalling’. Depressingly, it doesn’t appear to be
getting any better with the passage of time. Malfunctioning
Running under the premise that a ‘champion will somehow
solid-liquid separation unit operations threaten the viability
spontaneously emerge’ isn’t delivering the goods.
of hydrometallurgical process plants and recycled thickener
overflows containing fine solids can materially affect flotation
Meetings – still the practical alternative to work performance.
The limiting resource to getting anything done is ‘competent
Thickening is a good example of this neglect. Consider the
person hours’ so it’s a wonder that too many concentrator
following thickener fundamentals:
managers have a long and unfocused morning/daily meeting.
While such gatherings can be social successes, I deplore the •• In a thickener, there are two flows: simplistically
waste of time of having up to 20 people milling around for an solid going downwards and liquid going upwards.
hour. The morning/daily meeting should not be an exercise Calculation of the rise velocity of the liquid going
in participatory democracy if you have to endure a junior upwards gives a number that is usually higher than the
metallurgist drone on for ten minutes explaining 24 hours of settling velocity of the finest particles in the feed. Note
‘screen grabs’ of primary crusher operating variables in 15 that the settling rate of a ten micron quartz sphere in
pastel colours. All the people at the meeting need to know water is 0.6 cm/min at 25°C (Dorr and Bosqui, 1950).
about the crusher is the tonnes in the run-of-mine stockpile, Hence the importance of adequate flocculation (and
crusher operating hours and the tonnes in the crushed ore coagulation in water treatment) to create agglomerates
stockpile. with a higher settling velocity than the rise velocity.
In this author’s dream world, the geometallurgy section •• Hydrometallurgical plants often have multiple thickener
of the meeting should not be a long-winded discourse on operations eg counter-current decantation. Taking a
geology, but emphasise the nature of the future ore feed and desired 92.5 per cent operating time with a total of
what actions the operators should take in the next 24 hours five thickeners in series in the flow sheet means that each
to meet the metallurgical performance targets. As a young thickener has to be performing as specified 98.5 per cent

P D Munro
of the time ignoring all other causes of downtime and qualities. Water treatment could become common in
poor operation! concentrators.
In a mineral processing world of finer particle sizings, 2. Tailings – this is the ‘big one’. The mining industry has
due to the nature of ores treated, leading to more difficult a poor and increasingly indefensible record managing
solid-liquid separation duties, the following are some of the tailings dams; this author prefers the word ‘dams’ to the
depressing realities of thickener operations: euphemism ‘tailings storage facilities’. Doubters about
•• Flocculent systems malfunctioning with incorrect this should spend time on the internet using the search
addition and subsequent dilution; flocculents once term ‘tailings dam failures’. To dismal incidents such as
formed are ‘maltreated’ in their passage to, and through, Certej (Romania) 1971, Buffalo Creek (USA) 1972, Val
the thickener feedwell. di Stava (Italy) 1985, Omai (Guyana) 1994, Merriespruit
(South Africa) 1994, Los Frailes (Spain) 1998 and Baia
•• Feedwells in poor condition leading to excessive Mare (Romania) 2000 we can now add the recent events
turbulence damaging flocs, side feeding etc. of Mount Polley (Canada) 2014 and Samarco (Brazil)
•• Thickeners acting as a further stage of cleaner flotation 2015. Over the last century Azam and Li (2010) estimated
with voluminous froths on the surface exacerbated by the failure rate of tailing dams at 1.2 per cent compared
upwelling air bubbles; technologies to de-aerate flotation with 0.01 per cent for conventional water retention dams.
concentrate froths have been available for over 15 years What an indictment on our industry! This author sees
(Garraway and Kaboth, 2001) and should be standard a future where dry stacking of tailings will be enforced
for such duties. by regulation. An example is the proposed 72 000 t/d
•• Solid-laden concentrate thickener overflows reminding Rosemont copper-molybdenum-silver operation of
one of the Dutch farm adage ‘too wet to plough, too dry Hudbay Minerals near Tucson, Arizona, USA, which
to drink’. will only be permitted if tailings are dry stacked. If dry
stacking is mandated then the unit operation of filtering
•• Thickener overflow launders choked with deposited
tailings will become as important as grinding in terms
solids.
of plant throughput and operating cost. Metallurgists
•• Thickener underflow density reduced to cope with are the best qualified people to design and operate
inadequate pumping systems: this author has to such dewatering facilities. If this industry continues
suppress a tirade of vituperation when told by a process its current lack of interest in the ‘back end’, the civil/
engineer that the downstream filtration stage ‘works geotechnical engineers will ‘eat our lunch’ just as the
better at lower solids density’ when what he meant was mining/geotechnical engineers have got into the paste
that the underflow pumping system was undersized and fill business.
couldn’t deliver a high density slurry to the filters. How
could someone claim to be a professional metallurgist
after making such an asinine statement?
THE FUTURE FOR METALLURGISTS
•• Too often the suggested remedy is to install additional Metallurgists – are they needed?
instrumentation when the problems are caused by
McCaffery, Giblett and Dunne (2004) asked whether we still
flawed process fundamentals.
needed metallurgists. This author accepts that we still do, but
Process engineers are encouraged to get ‘back to basics’ would like to see more evidence that they are making a positive
(has anyone ever made a credible case for leaving them?) economic contribution as this author is currently unconvinced
on thickening. This includes learning from the excellent about the value of much of the ‘on-site professional hours’,
work done by CSIRO in the AMIRA P266 Project ‘Improving particularly in FIFO operations.
Thickener Operations’.
Consider a typical FIFO site with the following professional
staff:
Paradigm revolution in concentrator operations
•• concentrator manager
– driven from the ‘back end’
•• metallurgical superintendent
The author discerns a trend that processing aspects of mining
operations and projects will be increasingly driven from the •• two plant metallurgists
‘back end’ influenced by two main issues: •• two project metallurgists.
1. Water – generally, the mining industry will not be Annual cost to the employer would be at least A$1 million/a.
allowed access to water until competing demands for If the reality of the metallurgists on-site:
human, agricultural and pastoral uses have been met. •• have gaping deficiencies in the list of ‘Mineral processing
The result is that the industry has to maximise both basics’
water recovery from tailings and recycling of water. This
•• are being used for clerical rather than technical tasks.
is not a novelty for Australian mineral processing plants,
Managers and senior professionals are to blame for the
but it is for some of the jurisdictions where we work. The
allocation of metallurgists to routine and mundane tasks
move to dry stacking of tailings discussed below means
instead of process improvement ie ‘discretionary time’
operating the plant with a higher proportion of recycled
(Munro and Tilyard, 2009). Consider things such as not
water than the industry has considered the ‘norm’.
being able to manage a good plant survey and not doing
This reduces the ability to maintain water quality by
statistically valid test work. (This author has difficulty
bleeding out deleterious chemical species via the tailings
remembering when he last went to a concentrator and
stream. A build-up of certain chemical species will affect
the metallurgists could show him actual corrected and
the metallurgical performance of flotation which is a
reduced efficiency curves for the hydrocyclone classifiers)
separation process dependent on the quality of mineral
surfaces. Johnson (2013) has discussed treatment of •• are unable to sustain gains
recycled process water to achieve specific chemical •• lack process or corporate memory

•• have no awareness of current and past operating Metallurgists are too narrowly focused on liberation and
practices for the type of ore being processed and separation; solid-liquid separations are likely to be the future
seemingly little interest in finding out ‘drivers’ of concentrator design and performance.
•• are guilty of other technical malfeasances as mentioned A new model for the deployment of metallurgists could
has been mentioned. have less technical people on-site using experienced
If metallurgists on-site aren’t using the technical skills they external technical people in a scenario that provides genuine
supposedly acquired during four years of tertiary education, professional development of junior metallurgists.
then how could a company justify having them? Is their
presence some kind of ‘Potemkin village’ to reassure outsiders ACKNOWLEDGEMENTS
that metallurgy matters?
The author thanks Mineralurgy Pty Ltd for permission to
publish this paper.
A possible new model
A person more interested in outcomes than appearances The following individuals are thanked for their continuing
might consider how he/she might alternatively spend the observations on these matters over the years: Peter Colbert,
A$1 million/a outlaid for the metallurgical team. Here are John Glen, Rolly Nice, Joe Pease, Peter Rohner, Tom Shouldice,
some heresies to current practices: Stuart Smith, Jorma Tuppurainen and Michael Young.
•• the concentrator manager could be either a metallurgist The author stresses that the ideas, opinions and biases in
or a skilled production superintendent this paper are his own.
•• the metallurgical development plan is produced and
monitored by head office metallurgists and/or external REFERENCES
consultants Anon, 2015. Mining consultant rejects rivals’ advance, Mining
•• clerks perform many of the tasks of the plant metallurgists Magazine, November.
•• one can buy a lot of external ‘Competent Person hours’ Azam, S and Li, Q, 2010. Tailings dam failures: a review of the last one
for the yearly salary of a metallurgist on-site who is hundred years, Geotechnical News, December, pp 50–53.
effectively a clerk and should be quite demotivated Butler, C, Dale, R, Robinson, S and Turner, A, 2016. Geometallurgy
acting as one. – bridging the gap between mine and mill: a case study of the
Degrussa geometallurgy program, in Proceedings GeoMet 2016,
This new model has less technical people on-site and
pp 77–88 (The Australasian Institute of Mining and Metallurgy:
uses experienced external technical people in a scenario
Melbourne).
that provides genuine professional development of junior
metallurgists. There are analogies to the South African Colbert, P J, Munro, P D and Yeowart, G, 2009. Prominent Hill
‘mining house’ model where site managers had a reporting concentrator – designed for operators and maintainers, in
line on technical issues to consulting engineers and practices Proceedings Tenth AusIMM Mill Operators’ Conference, pp 23–32 (The
in the oil industry where much of the expertise comes from
specialist consultants. Dorr, J V N and Bosqui, F L, 1950. Cyanidation and Concentration of Gold
and Silver Ores, second edition, p 23 (McGraw-Hill: New York).
CONCLUSIONS Drinkwater, D, 2015. What should be in the Mineral Processing

curriculum?, The AusIMM Bulletin, 2:44–47.
The ‘mineral boom’ since 2000 has not led to any technological
breakthroughs nor improved the quality of our human Drinkwater, D and Bianco, N, 2013. Developing technical excellence
capital by strengthening educational institutions and skill in young Australian metallurgical professionals – a new graduate
development schemes for graduates. development program, in Minerals Industry: Education and
Training, IMPC Monograph (eds: J Cilliers, D Drinkwater and K
It is the author’s opinion that the situation is now worse than Heiskanen), pp 117–129 (Indian Institute of Minerals Engineers:
in 1998, considering that the demographics of the population Jamshedpur).
of metallurgists will result in imminent loss of technical
Drinkwater, D, Bradshaw, D, Tilyard, P and Munro, P, 2011.
expertise as older people leave the industry.
Professional development for metallurgists – improving technical
The industry is not behaving appropriately to both improve skills, in Proceedings Metallurgical Plant Design and Operating
the skills of its metallurgists and ‘convert’ chemical engineers Strategies (MetPlant 2011), pp 39–48 (The Australasian Institute of
to metallurgists. This is particularly so given the large number Mining and Metallurgy: Melbourne).
of FIFO operations. Drinkwater, D and Napier-Munn, T, 2014. Why good professional
A meaningful use of ‘big data’ would be to fully integrate development is key to profitability in the mining industry,
geological and metallurgical databases to make geometallurgy Chapter 20: Mineral economics and human capital, in Proceedings
a practical reality. XXVII International Mineral Processing Congress (IMPC 2014) (ed: J
Yianatos), pp 28–37, Chile.
Concentrators are still running with the usual glaring
examples of egregious operating practices, malfeasances Garraway, B and Kaboth, K, 2001. The Frothbuster – innovative
and misunderstandings. Poor metallurgical results due to ill- thickener froth reduction technology, in Proceedings 33rd Annual
Operator’s Conference Canadian Mineral Processors, pp 17–24,
informed maintenance cost savings make you wonder if the
(The Canadian Institute of Mining, Metallurgy and Petroleum:
industry will ever learn from previous mistakes.
Montreal).
Having the basics of mineral processing on-site to effectively
Johnson, N W, 2003. Issues in maximization of recycling of water in
operate a concentrator is still the exception rather than the rule.
a mineral processing plant, in Proceedings Water in Mining 2003,
Time management to ‘create Competent Person hours’ is pp 239–246 (The Australasian Institute of Mining and Metallurgy:
still a problem. Melbourne).
Many metallurgical development plans need to be trimmed Lewis, M, 2003. Moneyball: The Art of Winning an Unfair Game (W W
down to reflect the ‘possible’ and to have meaningful priorities. Norton & Company: New York).

P D Munro
Lind, G, 2013. Minerals industry engagement in metallurgical Napier-Munn, T J, 2014. Statistical Methods for Mineral Engineers, How
education in Australia’, in Minerals Industry: Education and to Design Experiments and Analyse Data, JKMRC Monographs
Training (eds: J Cilliers, D Drinkwater and K Heiskanen) No 5 (Julius Kruttschnitt Mineral Research Centre University of
pp 87– 104 (Institute of Minerals Engineers: Jamshedpur). Queensland: Indooroopilly).
Mackenzie, W R and Cusworth, N, 2016. The use and abuse of Nelson, R D, Davies, R and Jacob, K, 1995. Teach’em particle
feasibility studies – has anything changed, in Project Evaluation technology, Chemical Engineering Education, 29:12–16.
2016, pp 133–148 (The Australasian Institute of Mining and
Rankin, W J (ed), 2013. Australasian Mining and Metallurgical Operating
Metallurgy: Melbourne).
Practices, third edition (The Australasian Institute of Mining and
McCaffery, K M, Giblett, A and Dunne, R, 2014. Sustaining Metallurgy: Melbourne).
metallurgical competencies, Chapter 20: Mineral economics
Sutherland, K L and Wark, I W, 1955. Principles of Flotation,
and human capital, in Proceedings XXVII International Mineral
pp 98– 100 (The Australasian Institute of Mining and Metallurgy:
Processing Congress (ed: J Yianatos), pp 15–27, Chile.
Melbourne).
Merow, E W, 2000. Problems and progress in particle processing,
Tilyard, P J, 2009. The Bougainville Copper concentrator: a
Chemical Innovation, 30:35–41.
retrospective – part 1, The AusIMM Bulletin, 5:73–81.
Merow, E W, Phillips, K E and Myers, C W, 1981. Understanding Cost
Tilyard, P J and Clark, N, 2010. The Bougainville Copper concentrator:
Growth and Performance Shortfalls in Pioneer Process Plants (The
a retrospective – part 2, The AusIMM Bulletin, 6:69–74.
Rand Corporation: Santa Monica).
Woodcock, J T (ed), 1980. Mining and Metallurgical Practices in
Minerals Council of Australia, 1998. Back from the Brink: Reshaping
Australasia, third edition (The Australasian Institute of Mining
Minerals Tertiary Education, National Tertiary Education
and Metallurgy: Melbourne).
Taskforce Discussion Paper, February.
Woodcock, J T and Hamilton, J K (eds), 1993. Australasian Mining and
Munro, P D and Tilyard, P J, 2009. Back to the future – why change
Metallurgy, second edition (The Australasian Institute of Mining
doesn’t necessarily mean progress, in Proceedings Tenth AusIMM
Mill Operators’ Conference, pp 5–11 (The Australasian Institute of
Mining and Metallurgy: Melbourne).

Contents
Undermining productivity – when good

key performance indicators go bad
J D Pease1
ABSTRACT
All organisations, from businesses to sports teams, have key performance indicators
(KPIs). These are always well-intentioned. A small number are crucial – indeed, they
are ‘key’. But without constant vigilance KPIs grow like weeds and do more harm
than good. It takes strict discipline and a deep business understanding to maintain
truly ‘key’ performance indicators that benefit the organisation.
ON ARCHAEOLOGY, TREES AND HOME INSULATION

An archaeological team had uncovered the remains of an ancient settlement, and
needed to recover as much as it could before the next wet season. Luckily there
was a local population of poor, hardworking villagers who enthusiastically joined
the dig. As the work progressed, the organisers were initially excited by the large
amount of ancient pottery recovered. But gradually they became disappointed that
they rarely recovered a complete piece – almost everything was broken. As ever more
fragments were returned, this moved from disappointing to perplexing to statistically
inexplicable. The archaeologists checked their numbers, reviewed data from other
sites, considered all the technical factors. Finally – and sadly too late – someone asked
them how they paid the villagers. ‘Well, we pay them for each piece they bring us’.
The villagers may not have had much formal education, but they certainly
understood management KPIs.
A small Australian town had become a desirable destination for ‘tree changers’. A
new council was elected with a mandate to protect the town’s leafy green image. They
passed a by-law that all trees over a certain girth would be listed and protected by the
council. Special approval would be needed to cut or even prune them. That weekend
the sound of chainsaws rang out around the town. Some owners with trees above the
girth cut them down before they were listed, fearing the development restrictions of
a ‘listed’ tree would reduce their property value. Worse, many young smaller trees
were cut down too. These should have been the next generation of trees, but their
owners opted to remove them while it was legal.
The most well-intended laws had precisely the opposite effect.
Recent governments have excelled at this art of unintended consequences. From roof
insulation to solar feed-in tariffs to training schemes, programs have been introduced
that are as well-intended as they are wasteful. They seem to say ‘here is a bucket of
money, you can help yourselves, but we can trust you all to do the right thing …can’t
we?’ It is alarming how often well-intended KPIs cause disastrous outcomes. It is not
that they are simply inefficient or ineffective. Their careless design encourages the
worst behaviour. It would have been better not to have any KPIs at all.
Sometimes this is called the law of unintended consequences. I just call it lazy and
dumb. Thoughtless. Honestly, what did they think was going to happen?
OF BANKERS, MINERS AND FOUNDING FATHERS

We would like to think big business is smarter than that. But consider the Global
Financial Crisis (GFC). How many bankers paid back the bonuses they ‘earned’ by
writing the bad mortgages that led to the GFC?
Or consider miners. Though, here the story takes longer to unfold.
In the 1950s, the underground mine at Mount Isa was one of the richest in the world.
Work was hard – mining relied on hand-held drills and considerable manual labour.
It was tough work in a cramped space at over 40°C and high humidity. It was quite
Engineer, Mineralis Consultants Pty Ltd, reasonable that miners were paid relative to effort – metres drilled or tonnes mucked.
Brisbane Qld 4066. The miners worked hard and both they and the company were rewarded. This was
Email: jpease@mineralis.com.au capitalism at work, aided by an appropriate KPI and reward system. But technology
27
J Pease
changed, as it does. New drill rigs, new underground vehicles In a nearby village, an industrious ant managed the mill. It
were developed. Everyone knew that if the company bought strived to understand how all minerals behaved. It sampled
an expensive new drill rig to double productivity, then the plant and studied real mineralogical data. This took up
the operator’s rate per metre should halve. But people are most of the ant’s time, but it refused to trust the Big Bad
people so that was hard to negotiate. There was only one Data on the office monitor. One summer, after studying
manager to negotiate with maybe 50 miners each with a deep mineralogy and smelting thermodynamics, the ant installed
personal interest in capturing some of the productivity gain more equipment and improved concentrate quality. It became
for themselves. The manager was busy, had his own KPIs to even busier with more equipment mouths to feed, but the
meet and couldn’t afford industrial unrest. Even if some of smelter danced for joy. Overall far less energy and cost was
the benefits leaked to bonuses the company was still much needed to make metal. Yet the ant missed its KPIs because
better off and production wasn’t interrupted. unit costs increased and recovery didn’t. One day the ant
That seems okay – until it happens over several generations missed a process-improvement meeting because it was busy
doing a plant survey. This disappointed management and the
of technology change. After about 40 years of this ‘creep’
ant was quietly moved to the safety department.
underground miners in air-conditioned cabins were doing
less arduous work than road workers or truck drivers but
were paid three times as much. They also argued that they KENTROPY
shouldn’t lose money when the equipment broke down – it We all know the second law of thermodynamics: without the
wasn’t their fault the company couldn’t maintain it. So, they constant input of energy a system will move towards disorder
negotiated a formula to pay them the average rate of the last – entropy.
six shifts if the equipment wasn’t available. This encouraged Our organisations are the same with KPIs. Without
them to get a high ‘run rate’ for six consecutive shifts – even if significant energy and discipline our KPIs multiply and
it meant pushing the equipment so hard it broke down on the proliferate. They are weeds that get out of control. It isn’t
seventh shift; but that wasn’t their problem. because our organisations lack intelligent and hard-working
Perhaps the later generation of miners were simply lucky people. It is because they are filled with intelligent, hard-
to be in the right place at the right time. But the system was working people. As a group, that is what we do. Everyone
broken. The mine was older and ore was lower grade and wants to make a ‘process improvement’. We all know that
more complex. Some ore zones had to be excluded because others would work better if only they did things our way. The
they disrupted the plant. We asked the miners to avoid it or more people we have in an organisation the more layers we
tip it to waste. But at the same time, we paid them a handsome have and the more KPIs we have. And the less relevant they
bonus to mine it. What did we think was going to happen become to our essential tasks.
when no one was looking? How many readers can say they have a succinct set of KPIs
In this case the original KPIs – simple $/metre – worked that enable them to focus on the few critically important
well for a while. But they didn’t stand the test of time. The aspects of their job?
founding fathers of the bonus scheme were not visionary Our organisations aren’t stupid, but they end up doing
enough in their design. Once a bonus scheme is in place it stupid things because they don’t apply constant energy into
is really hard to change. You are trying to change the rules resisting the inexorable sprouting of well-intentioned KPI
during the game, in a way that costs people money. That is weeds. Without that constant energy, Kentropy prevails.
not going to happen.
Considered in this context, the founding fathers of HAS ANYONE SEEN MY KEY?
democratic Constitutions deserve enormous respect. They The K in KPI stands for key. It must be the most overused word
didn’t get everything right; no-one could foresee 200 years of in business. See if you can find a business memo not peppered
technological advances. But they had a wisdom about people with ‘keys’. People now even talk about ‘the key KPIs’.
and what drives behaviour. Constitutions have stood the test
Readers are encouraged to google ‘KPIs’. There are scores
of time much better than most management KPIs.
of titles such as ’20 000 KPIs used in practice’. Twenty
thousand KPIs? Seriously?
THE PRODUCTIVITY CHALLENGE Yes, business is more complex and we all need to manage
After every mining boom we focus again on productivity. We multiple things. But most of us have so many KPIs it feels like
cut costs where we can, we reduce staff and we look for new carrying a huge bunch of keys. Most of them just get in the
innovations. Yet most operating people know we can do a lot way. Without the right one it doesn’t matter how many others
better with the equipment we already have. We don’t need a you have in your back pocket.
big scientific breakthrough or lots of new equipment. Most
managers can list the improvements they could make. If only
they could focus on the things they know to be important
THE WISDEN OF CRICKET?
rather than all the other ‘bumph’ they have to do. In the 2000s, the Australian cricket team dominated the sport.
They had rare skills. With their coach they took the game to
So why can’t smart people do what they know they need to?
a new professional level and made a science out of analysing
opposition teams, field placements, nutrition and mental
THE ANT AND THE GRASSHOPPER disintegration of opponents. They didn’t need much coaching
Once a grasshopper managed a mill. Rather than understand on the basics of the game because they were excellent at that
minerals, it understood the company. Instead of surveying already. The performance focus was on other areas.
the plant, it studied the KPIs. Rather than understand the But over a couple of years the champions left the team.
client, the smelter, the grasshopper read the marketing Slowly it lost its aura. In response, management set more
contract. It increased tonnage, reduced unit cost and out- and more targets and goals, training and nutrition regimes.
negotiated the smelter. The grasshopper was promoted and This culminated in four of the best players being sent home
lived a prosperous life. before a crucial game because they didn’t submit written

Undermining productivity – when good key performance indicators go bad
homework. Gradually two factors became apparent. First, the The silos we bemoan are constructed from our very own
game wasn’t fun for them anymore. But more importantly, KPIs.
they simply weren’t good enough at batting, bowling and
fielding. Team management previously could take that for THE KEYS TO COPING WITH KEYS
granted. Now, without the basics, no amount of sports science
The devaluation of KPIs is caused by the good intentions of
and psychology and KPIs were going to help.
intelligent people. There is hope because it is well-intended.
Does this remind anyone of their own workplace? You are Your approach depends on your role.
probably overwhelmed by KPIs and standardisation. It wears
For researchers and suppliers: you can’t change your client’s
you down. You respond because you are queried about them
KPIs. But you need to understand them. If your business
and your performance review and bonus depends on them.
proposition requires the client to work across the silos, then
But when was the last time management asked you find out if their KPIs will encourage and reward people to do
about your grind size distribution, your cyclone efficiency this. If they don’t find another client. If an organisation can’t
curve, your mineral liberation, the source of diluents in work across silos for its own benefit then it certainly won’t do
your concentrate? These are the very basics of running a it for yours.
concentrator but if management doesn’t ask about them
For supervisors and middle management: maybe your own
people won’t focus on them. New graduates won’t embrace
KPIs are exhausting and you can’t change them. But you can
them. Gradually the skills fade away. We no longer know the
simplify them for your own people. Give them goals that
basics. We can’t win.
let them focus on the basics of the business – and then hold
Of course, other indicators are important and need your them accountable to that. Allow for enough subjectivity to
attention. But they are not ‘key’. They are not the handful of encourage the types of important and positive conduct and
things you absolutely must get right for the business. Just as performance that are hard to measure. Like a sports coach
cricket in its Wisden has volumes of statistics. But a batsman you know that if your team gets the basics right, your job will
at the crease only needs to know a few things: how many runs, be easier and more secure regardless of your own KPIs.
how many wickets, how many overs? Those true KPIs need
For senior management: you can’t be an expert at everything.
their full attention. If they get them right, the other statistics
You don’t have to have played elite sports to manage them.
look after themselves.
But you need to understand the basics of the game and how
This is the difference between big data and the right data. people behave and respond to incentives. You need to know
who to listen to in order to pick and develop a team who are
SOMETIMES IT IS BETTER TO BE SUBJECTIVE good at the basics. Then let your people focus on the basics
One of the reasons we have so many distracting KPIs is our and hold them accountable for delivering them.
beloved performance reviews. As rational and numerate people, You must have KPIs that encourage – or at least don’t
we eschew subjective measures. If something is important we discourage – people to work across the silos for the benefit of
must be able to measure it. So we can’t measure ore quality the whole business. This is so obvious that it is rarely done.
online but we can measure tonnes. We can’t measure value- The simple concept conceals devilish design and is resisted by
added at each step but we can measure $/tonne. grasshoppers who like the simplicity of being able to manage
What do we think will happen then? their own KPIs and bonus independent of the organisation.
The most important characteristics of employees can’t be Finally, you need constant vigilance against Kentropy.
measured numerically – their integrity; how they cooperate Like the second law of thermodynamics it is relentless
and mentor others; their tenacity and willingness to do and immutable. Your intelligent people (and our beloved
whatever it takes; to work across the silos abandoning their management consultants) will constantly propose helpful
own KPIs and bonus because a higher priority arose in another new KPIs and systems. They will ask you to replace subjective
department. We know these people when we see them, but measures with ‘hard numbers’ that they have developed.
we can’t put a number on it. They are the employees we Everyone has a compelling argument, but the cumulative
want; not the grasshoppers who argue about their score for effect will smother your organisation. You need constant
‘cooperation’ and assiduously manage their KPI spreadsheet energy and conviction not to adopt them.
to maximise their bonus. We don’t like the grasshoppers, but Just because you pulled out the weeds last year doesn’t
we reward them nevertheless. We all complain about the silos mean you don’t have to do it again this year.
in our organisation. Management says it wants to break them
down. It says that, but it pays us to do otherwise.

Geometallurgy
Contents
Geometallurgy – what do you really need to

know from exploration through to production?
K Ehrig1
ABSTRACT
You are a metallurgist, geologist, geometallurgist or processing engineer and are
fortunate enough to work for an organisation who asks you to design and execute
a geometallurgy program. What seems like a simple request is actually not a trivial
exercise. ‘Recipe’ style manuals are not readily available. However, many ‘experts’
are lurking around willing to deliver solutions they claim will solve all of your
problems. Even though similar ore deposit styles (eg porphyry Cu, Fe-oxide Cu-Au,
Ni-laterites, calcrete hosted U etc) have many similar characteristics, each ore deposit
is unique. Hence your geometallurgy program needs to be fit-for-purpose. This paper
offers some insights gained via designing and executing geometallurgy programs
from exploration through to production, and working at a mine where you are held
accountable for your metallurgical performance predictors.
INTRODUCTION
Metal grades, mineralogy, rock and mineral texture are variably distributed across all
‘world-class’ mineral deposits. Hence the metallurgical performance of ores mined
and processed throughout the production life of an ore deposit will also vary. Mining
industry experience over the past 100+ years has shown that:
•• ‘recovery’ = f (mineralogy, grade, ore texture, process conditions) (Bojcevski,
2004)
•• future performance ≠ f (historical performance).
‘Recovery’ in this paper refers to any metallurgical performance parameter (eg ore
hardness, mill throughput, flotation recovery, leach recovery, concentrate grade,
reagent consumption etc).
Mineralogy exerts the primary control on metallurgical performance. Recovery
assumptions, based on a lack of data, usually prove to be misleading and imprecise.
These points are obvious, but we (as an industry) seem to be constantly trying to
convince ourselves that the above do not apply to our deposits. Think about the
revenue impacts and opportunity losses when we get it wrong.
The failure (or under performance) rate for mining projects remains relatively high.
More importantly, the failure rates have remained relatively unchanged for decades
(McCarthy, 2003). Based on numerous surveys to identify the reasons behind new
mining projects and brownfield expansions not delivering to forecast, McCarthy (2003)
identified the following feasibility study problem areas which lead to commissioning
and operational under performance:
•• mine design and scheduling (32 per cent frequency)
•• geology, resource and reserve estimation (17 per cent)
•• metallurgical test work, sampling and scale-up (15 per cent)
•• process plant equipment design and selection (12 per cent)
•• geotechnical analysis (9 per cent)
•• cost estimation (7 per cent)
•• hydrology (4 per cent).
The driver behind geometallurgy at Olympic Dam is to reduce the technical risk
to the current operations and future expansions caused by unexpected mining and
process performance issues due to variable ore properties.
The primary objective of Olympic Dam geometallurgy is to develop metallurgical
performance predictors that reliably describe the process performance of different
1. MAusIMM, Principal Geometallurgist, ore types and spatially distribute these into the resource block model for use as a
BHP Billiton Olympic Dam, Adelaide SA 5000. fundamental input into mine planning. The geomet models are predictive mineralogy
Email: kathy.ehrig@bhpbilliton.com and recovery models which are applied to blocks in the mineral resource model to
33
K Ehrig
enable the estimation of mineralogy, metallurgical recovery, of economic metals and deleterious elements along with
and real value on a block-per-block basis. some minerals. The next level of ore characterisation beyond
The secondary objectives are to identify any ores which may assaying and geological/geotechnical/geophysical data is
be ‘problematic’ to the current plant, and provide variability the acquisition of quantitative mineralogy and metallurgical
data of suitable quality for process plant design and future performance parameters collected at a sufficient spatial
plant optimisations. And finally, the chemical, mineralogical, frequency to be incorporated into various block models. If
physical property, and ‘recovery models’, collectively called you are incredibly fortunate, the number of quantitative
the geometallurgical model, provide the data required to mineralogy and metallurgical samples is only one to two
support the JORC Code (2012) ‘modifying factors’. and two to three orders of magnitude less than the geological
assay samples, respectively.
The geostatistical techniques to estimate mineral
GEOMETALLURGY FUNDAMENTALS abundances into the space between the drill holes (typically
There has been a rapid increase in the volume of published 20–200 m spacing) or samples are straightforward because
geometallurgy literature over the past ten years. Numerous mineral abundances are additive. However, this may, or
papers are available which describe the definition, purpose, may not, be the case for metallurgical parameters. One way
benefits, and methodologies used in geometallurgy from the around the potential problem of non-additivity is to develop
perspectives of geologists, metallurgists, and process design mathematical relationships which express each metallurgical
engineers. A few of these papers are listed with the references. parameter as a function of additive properties such as metal
This section provides highlights regarding the links between grades and mineral abundances. This has been the approach
minerals and populating the resource block model and the taken to populate every block in the Olympic Dam resource
mine plan with geometallurgical data. model (~20 million blocks) with abundances of 15 minerals
and >50 metallurgical performance parameters, in addition to
Minerals 26 elements and metals and bulk density. This allows for the
Metals and elements occur in ore deposits as minerals. Minerals, assessment of a block’s real value to recover metal. To achieve
not elements or metals, are mined. Extractive metallurgy, maximum benefit from your valuable geometallurgical data,
in particular, mineral processing, hydrometallurgy, and it must be incorporated into the mineral resource block model
pyrometallurgy extracts metals and elements from minerals. because the resource block model is the primary input into
Grades or concentrations of metals and elements have the mine plan.
been extensively used as proxies for mineral abundances. Mine planning transforms the block model (eg 3D
Historically, it has been easier and more cost-effective to orebody information) into a time-based 1D ‘metallurgical or
assay a sample for a limited suite of elements than it is to process engineer friendly form’ suitable for process design
measure mineral abundances. However, over the past decade and production planning. A well designed mine plan is
relatively rapid mineralogy detection and measurement produced via an iterative process involving inputs from
methods such as automated scanning electron microscopy the geology, geometallurgy, geotechnical, hydrology, mine
(eg MLA, QEMSCAN etc) and short wave infrared (eg PIMA, design, metallurgy, infrastructure, environmental, business
HyLogger, CoreScan etc) methods have become more readily valuation, and financial teams.
available thus permitting the measurement of mineral
Once a mine plan (there will be a continual stream of mine
abundances at an unprecedented scale.
plans) is available, where each block in the resource model
In addition to the recovery relationship described above, is tagged with a mining period, cumulative distribution
another critical yet useful relationship is: curves can be generated for any variable in the block model.
The curves can be produced to represent any mining period
mineral (wt per cent) = f (sample composition) and can also be filtered based on any variable in the block
model. This is a very powerful tool and provides the process
The relatively low cost of obtaining multiple element assays design engineer with some of the data necessary to design the
via inductively coupled plasma optical emission (ICP-OES) processing plant.
and inductively coupled plasma mass spectroscopy (ICP-MS)
methods opens the path to estimate mineral abundances on a
THE FIT-FOR-PURPOSE GEOMETALLURGY PROGRAM
sample-by-sample basis based on multielement assays using
the above relationship. Obviously, the above relationship In describing the fundamental aspect of geometallurgy, Sola
needs to be calibrated against measured mineralogy. and Harbort (2012) quoted the following which was originally
sourced from Harry and Schroeder (2000):
If the recovery and mineral (wt per cent) relationships for
You don’t know what you don’t know
your deposit can be established and the distribution of the
process critical minerals can be quantified and mapped, then You don’t measure what you don’t value
a logical conclusion is that metallurgical performance can be You can’t value what you don’t measure
predicted across a deposit. If you can’t measure it you can’t control it
If you can’t control it you can’t improve it.
The geological database, mineral resource
Several questions need to be considered when scoping and
block model and mine plan then addressed in a geometallurgy program:
Our understanding of the mineral deposit comes from the
•• What, if any, are the technical confidence levels required
systematic sampling provided primarily via drilling. In
by your project funding institution? What are the
general, as the sampling density increases, the confidence
measures of the confidence levels?
level also increases. The geological database (includes
drilling, assaying, rock mass properties etc) is very likely to be •• Is the project time-constrained or resource-constrained?
the largest data set (~100 000 to >2 million assayed drill core •• What learnings (ie free lessons) are available from
samples) available to a project or operation. Typical resource similar deposit styles (or other deposit styles with
block models contain geostatistically estimated concentrations similar mineralogy) in the district, region, or globally?
34 We are metallurgists, not magicians

Geometallurgy – what do you really need to know from exploration through to production?
•• What are the market specifications for the final change with time. Waste today may become ore tomorrow.
products? What are the penalty elements/minerals and Metallurgical performance is almost always impacted by
their limits? variations in mineralogy and grade. The number of samples
•• What are the payable elements/minerals and and variety of ore types required for testing will likely
subeconomic elements/minerals? decrease as the project advances into production where the
focus shifts to better understanding and managing local
•• Which minerals and concentration ranges are deleterious
(production) scale variability.
to the process?
•• How do the process and deleterious minerals/elements Model validation
occur in the mineral deposit? ‘All models are wrong, but some are useful…’. (Box, 1979)
•• How does mineral and textural variability across Geometallurgy models should be treated with scepticism
the known extent of the mineral resource impact on until they are shown to be useful. Early in the geometallurgy
metallurgical performance? program, initially developed metallurgical performance
predictors can be used to predict the outcomes of future
Resource understanding and project evolution laboratory testing outcomes. Model validation and model
From exploration into various project development stages updates should evolve as testing continues during the various
through to the end of production, resource classification studies. The initial models can be used to predict the outcomes
terminology such as inferred, indicated, and measured is an of future laboratory testing. Full scale production is the
attempt to demonstrate the level of geological knowledge and real validation of the metallurgical performance predictors.
confidence in the resource. This knowledge and confidence During production, there will be repeated opportunities and
should be increasing as the drilling metres increase. An business requirements to improve the precision of the geomet
analogy in the engineering studies space is the increasing level models so that they are useful for short-term production
of cost accuracy required when progressing from conceptual planning!
to prefeasibility (PFS) and feasibility studies through to
construction, commissioning and finally into production. Data integrity and security
In an ideal world, resource confidence should be sufficient Spreadsheets should not be allowed for data storage solutions.
to support the next stage of project development. However, Spreadsheets can be used to analyse the data. Properly
in the real world, drilling to support geological (and designed and managed databases protect the integrity of the
geometallurgical) knowledge and confidence in the resource data and also allow for full integration and interrogation of all
to a level suitable for a PFS usually starts at the same time the technical data.
as the PFS for mine planning and process design. The same
is true for the feasibility study, and even into production. CONCLUSIONS
More simply put, the entire project team commences their
Back to the title of this paper, ‘Geometallurgy – what do you
individual studies at the same time. The impact on the project
really need to know from exploration through to production?’
is that the mine plan and cut-off grade strategy are likely to
continually change as the geologists continue to discover and •• mineralogy, mineralogy, mineralogy
improve the definition of ore zones within the deposit and the •• how the mineralogy responds to likely extraction
geomet team continues to identify more potential constraints processes across all grade ranges
or new revenue streams. Hence new ore types are likely to •• the concentration of process and final product deleterious
be discovered, and more importantly the perceived relative elements and minerals
proportions of the ore types will also continue to change. The •• how the mineralogy changes, hence how the
impact on process design is that the ‘goal-posts’ will continue metallurgical performance changes over the life of the
to move during the engineering studies and a process plant is asset.
often built which cannot meet project expectations.
Remember that the primary purpose of geometallurgy
is to develop metallurgical performance predictors and to
Geometallurgy sampling strategy
incorporate expected performance into the block model
Geological and resource ore types and domains may not and mine plan. Geometallurgy is not process flow sheet
necessarily equate to equivalent metallurgical ore types development, this still requires traditional metallurgical
and domains. The metallurgist or process engineer needs to testing programs.
understand that geological ore types are based on qualitative
visual estimates of mineral abundances, or as in the case of
alteration zones, simply the presence or absence of a specific REFERENCES
mineral or groups of minerals. Most geological descriptions Bojcevski, D, 2004. Metallurgical characterisation of George Fisher
of host lithology, alteration, and mineralisation zones are for mesotextures and microtextures, MSc thesis (unpublished),
University of Queensland, Brisbane.
the purposes of understanding the genesis of the ore deposit,
not defining metallurgical ore types. Resource domains are Box, G, 1979. Robustness in the strategy of scientific model building,
usually linked to geological boundaries, but may be further in Robustness in Statistics (eds: R L Launer and G N Wilkison),
312 p (Academic Press).
refined based on metal distribution patterns within an ore or
alteration type. However, when commencing sampling for Harry, M and Schroeder, R, 2000. Six Sigma: The Breakthrough Strategy
geometallurgical testing, the geology and resource ore types Revolutionizing the World’s Top Corporation (Doubleday: New
York).
and domains are all that is available.
JORC Code, 2012. Australasian Code for Reporting of Exploration
Initially geometallurgical sampling must be designed to gain
Results, Mineral Resources and Ore Reserves (The JORC Code)
a deposit scale understanding of the resource, so all mineral [online]. Available from: <http://www.jorc.org> (The Joint Ore
combinations across the grade spectra require sampling and Reserves Committee of The Australasian Institute of Mining and
testing. Remember that the classification of waste, marginal, Metallurgy, Australian Institute of Geoscientists and Minerals
low, and high-grade ore is based on economic criteria which Council of Australia).
We are metallurgists, not magicians 35

K Ehrig
McCarthy, P, 2003. Managing technical risk for mine feasibility Sola, C and Harbort, G, 2012. Geometallurgy – tricks, traps and
studies, in Proceedings Mining Risk Management Conference, treasures, in Proceedings 11th AusIMM Mill Operators’ Conference
pp 21–27 (The Australasian Institute and Mining and Metallurgy: 2012, pp 187–196 (The Australasian Institute and Mining and
Melbourne). Metallurgy: Melbourne).

Contents
Integrating geometallurgy with copper

concentrator design and operation
G Harbort1, K Jones2, D Morgan3 and C Sola4
ABSTRACT
The use of geometallurgical modelling with flotability component simulation provides
a design methodology with significantly less associated risk. The use of geological
data for optimisation of operating plants has become a significant part of the modern
process mindset. The underlying principle is to use spatial metallurgical information
to drive production planning, mine planning, blast design, blending strategies
and plant set-up. At a design stage the process designer can use geometallurgical
information to evaluate bottlenecks and potential design flaws and propose the best
investment strategy for the benefit of the project.
This paper provides a brief review of various copper sites that have implemented
geometallurgical studies. Two case studies are also presented. The first details the
geometallurgical characterisation of the Andash deposit and methodology used to
review the project’s detail design and projected production. The second discusses the
geometallurgical approach undertaken at the operating Northparkes mine, both for
circuit optimisation and life-of-mine planning.
INTRODUCTION
Early evolution of geometallurgy
Geometallurgy is considered a relatively recent development. It has been suggested
(Williams, 2013) that it evolved in the late 1980s or early 1990s. In reality, geometallurgy,
both as a practice and a science is far older. The practice of copper geometallurgy can
be dated to the first decade of the twentieth century. Dwindling stocks of high-grade
ore for direct smelting or gravity concentration, coupled with the discovery of massive,
low-grade Chilean porphyries were key drivers for its development (Yeatman, 1932).
This led to increased cooperation between geologists and metallurgists to achieve
economic extraction of copper from low-grade ores. Although practised by numerous
porphyry copper mines in an ad hoc manner, geometallurgy was not to gain favour
as a science for several decades.
Its initial academic development was to take the form of a collaborative research
project between the University of Melbourne and the University of Queensland
(O’Malley and McGhie, 1939). This investigation centred on the mineralogy of the
Black Star orebody at Mount Isa and its implications for mill practice. At the time it
was considered outstanding ‘…in focussing upon the mind of the reader, in a lucid
and emphatic manner, the metallurgical problems that result from complex sulphide
intergrowths …’ (Blanchard and Hall, 1939). Not only did the work correctly predict
metallurgical operation decades into the future, it also produced one of the first theses
on geometallurgy.
By the 1950s geometallurgy was becoming an established concept. The first
symposium on geometallurgy was held in Salt Lake City in 1955 (Anon, 1956).
Discussions were very similar to those held at modern geometallurgy conferences.
1. FAusIMM(CP), Technical Director, Amec Foster Key issues were considered to be the need for greater interdisciplinary communication
Wheeler Australia East, Brisbane Qld 4000. and research (McQuiston, 1956; Kirkland, 1956; Reno Sales, 1956).
Email: greg.harbort@amecfosterwheeler.com
In the same era geometallurgical concepts played a crucial role in the formation of
2. Senior Analytical and Evaluation Metallurgist, resource reporting in the Soviet Union. Raw commodities were considered strategic
Northparkes Mines, Parkes NSW 2870.
assets, with their treatment to provide maximum national benefit (Arden and Tverdov,
Email: kellie.jones@northparkes.com
2014). Soviet geometallurgy, or technico economicheskiye obosnovaniye (technical
3. MAusIMM, Technical Superintendent,
economic characterisation), included factors such as mineral complexity, treatment
Northparkes Mines, Parkes NSW 2870.
Email: dylan.morgan@northparkes.com
options, product quality and economic performance (Tverdov and Mikishichev, 2014).
4. Process Engineer – Geometallurgy, During this period detailed geometallurgical analysis at operating mine sites was
Amec Foster Wheeler Australia East, often constrained by a lack of resources. To quote Mount Isa Mines project metallurgist
Brisbane Qld 4000. A E O’Meara (1953):
37
G Harbort et al
The mill research metallurgical staff has endeavoured to processing behaviour and processing properties for stockpiling,
incorporate mineragraphy in solving mill problems and blending and block models (Fennel et al, 2005).
determining the effectiveness of laboratory tests, as well as In the last ten years, there has been a flurry of activity
examining minerals sent in by the geology department. For relating to geometallurgy. Excellent reviews of geometallurgy
many years the amount of equipment available was limited methodologies have been published by Dunham and Vann
and much of it locally made or improvised. (2007), Walters (2008), Coward et al (2009), Williams (2013),
This would change in 1950 when a fully equipped Dominy and O’Connor (2016) and McKay et al (2016). The
microscopy laboratory commenced operation at Mount Isa assigning of geometallurgical domains and their importance
Mines. Geologists and metallurgists shared offices, evaluating has also been detailed by a number of authors, including David
the effects of comminution on mineral liberation and flotation (2007), Johnson and Munro (2008), Hunt, Berry and Bradshaw
performance. Another example of geometallurgical success (2011) and Sola and Harbort (2012). A useful benchmarking
was the Brunswick concentrator program detailed by Petruk of geometallurgy programs by type and depth of usage was
and Schnarr (1981). discussed by Lund and Lamberg (2014) and Lishchuk et al
In 1968 Quiston and Beachaud published their (2015). Other areas of interest have included comminution
geometallurgy (Alruiz et al, 2009; Suazo, Kracht and Alruiz,
comprehensive review of sampling and testing a virgin
2010; Mwanga, Rosenkranz and Lamberg, 2015) and process
deposit from a metallurgical perspective. They proposed
mineralogy (Evans et al, 2011; Kuhar et al, 2013). Although
‘geometallurgy’ as a term to describe this perspective:
most of the recent publications focus on geometallurgical
… since geology is inextricably interwoven with metallurgy successes, a number also discuss potential problems (Kittler
in gaining an understanding of the complexities of a deposit, et al, 2011; Sola and Harbort, 2012).
eventually leading to a definition of mineable reserves, with
the development of a flowsheet and engineering criteria for Copper geometallurgy
the planning of a successful and profitable operation.
Further publications of note included the use of Australia
geometallurgy for comminution (MacPherson, 1976) and Within Australia, copper geometallurgy was initially
plant design (Young, 1983). centred at Mount Isa Mines. Hoffmann (1964) reviews the
The appearance of automated mineralogy and the vast development of flotation treatment of Mount Isa’s chalcopyrite
amount of information it supplied was to have a major ore between 1941 and 1963. Part of the development process
effect on geometallurgical development. In the 1970s various included microscopic examination of ore to reveal features
groups in Canada (Petruk, 1984), the United Kingdom (Jones, of metallurgical significance. Subsequent test programs
1984), South Africa (Oosthuyzen, 1983) and Australia (Reid considered these features in terms of reagent addition and
and Zuiderwyk, 1983; Reid et al, 1984; Reid and Wittenberg, flow sheet design. During operation, liberation studies were
1984) began to develop optical and electron beam methods also conducted to allow optimisation of concentrate grade
of automating the collection of point counting data. This and copper recovery.
led to the development of the QEM*SEM, later renamed Geometallurgical evaluations have been conducted on
QEMSCAN, and the Mineral Liberation Analyser (MLA), Australia’s iron oxide copper-gold (IOCG) deposits. Expansive
automated instruments that allowed the rapid identification geometallurgy programs were conducted both for the
and measurement of a desired number of mineral grains, drill design of the Ernest Henry copper concentrator (Strohmayr
core or ore sample sections. et al, 1998) and subsequent optimisation (Tew et al, 2003).
The former studies involved core mineralogy and flotation
The modern era of geometallurgy tests on 42 drill holes. Although the drill holes represented
The modern era of geometallurgy can be considered to have a good cross-section of the orebody, it was reported that
commenced circa 1985, where automated mineralogy and their compositing resulted in performance prediction issues
increasing computer availability laid the groundwork for due to underestimating the microvariability of the orebody.
it to become a common practice. QEMSCAN and later MLA Latter investigations during operation used monthly modal
units were installed in commercial laboratories around the composites to calibrate and predict performance. Each mineral
world in addition to research departments in major mining was given a predicted flotation response in terms of grade
companies. It has become commonplace not only to interpret and recovery. In addition, samples were composited from
the performance of operating plants by use of QEMSCAN or each mining block, with X-ray diffraction (XRD) conducted
MLA measurements but also to assess orebody composition, and correlated with plant performance. Investigation showed
including block models, and to predict plant performance from that feldspar and albite had a negative impact on throughput,
drill core and ore samples. In this process, not only the mineral while quartz and magnetite had a positive influence.
abundances, but the nature of the gangue minerals, the mineral For the Prominent Hill IOCG deposit, geometallurgy was
associations and the natural grain sizes are all important used for optimising mining block models, plant design
in predicting grind size, the design of flotation circuits and and operations trouble shooting. Khosrowshahi, Shaw and
the likely recoveries and losses. Relationships between ore McKevitt (2009) discuss the use of transfer functions to
characteristics and concentrate grade and recovery were assist in mine plan optimisation including copper speciation,
obtained for ‘best practice’ of the day (Jackson, Gottlieb and gangue mineralogy, contaminants, processing parameters
Sutherland, 1988). A further outcome of the QEMSCAN and and concentrate quality. Barns, Colbert and Munro (2009)
MLA interpretive capabilities was the reliable prediction from discuss the mineralogy approach used for geometallurgical
studies of feedstock for grind size and liberation (Gottlieb, domaining, test work and flow sheet development. Bradshaw
Adair and Wilkie, 1994). This type of study was used by the et al (2012) consider geometallurgy in their investigation of
Hellyer mine in Tasmania to build up a mine model related variable performance when treating different ore blends.
to plant performance and the creation of stockpiles based on Perhaps the most comprehensive IOCG geometallurgical
ore type (Lane and Richmond, 1993). Used at an early stage study has been that for Olympic Dam. The deposit has over
in the mining life cycle the resultant data allows prediction of 70 ore and gangue minerals, each with associated metallurgical
38 WE are metallurgists, not magicians

Integrating geometallurgy with copper concentrator design and operation
properties. In the initial comprehensive geometallurgy occurring during stockpiling of the supergene ores after
program 500 to 1000 samples were used for full metallurgical mining (Wong et al, 1991). The same situation reoccurred
characterisation, with approximately 10 000 samples for with treatment of the Reward deposit. Laboratory test work
mineralogical characterisation (Liebezeit et al, 2011; Boisvert, found that there was deterioration in flotation performance
Rossi and Ehrig, 2013). Olympic Dam Operations maintain a after only five days of ore aging and that this continued to
geometallurgy program to provide technical support to both adversely affect performance the longer the aging. The
the mining and processing departments. Metallurgical and strategy to reduce this oxidation and subsequent deleterious
mineralogical assessment is conducted on drill core within effects to plant performance was to mine at a rate consistent
five-year production plans. Predictive mineralogy algorithms with plant throughput, minimising stockpiles at both the mine
are used to estimate throughput, copper recovery and and plant. This strategy to minimise stockpiles impacted on
concentrate grade (Nzama and Kapo, 2014). plant performance, as blending of ore for mixing of ore types
Although the geometallurgy programs for the larger and to minimise head grade fluctuations was not possible.
producing Australian copper mines have received significant Recovery loss due to the higher Fe:Cu ratios was investigated
recognition, much work has also been done on the smaller by conducting mineralogy on flotation tails samples by
polymetallic copper operations. A number of these relate to X-ray diffraction and electron microscope techniques. The
mines within the Cobar Basin, New South Wales. The basin investigation found that there was no obvious mineralogical
is characterised by metal assemblages dominated by Pb-Zn- reason for the difference in flotation response between the
Ag ± Cu at the Endeavour Mine, Cu-Zn-Pb-Ag at the CSA Mine high Fe and low Fe samples (Kilgariff, 2003).
and Cu-Ag at the Peak Mine. Loidl (2012) published a thesis
on the geometallurgy of Endeavour. The comprehensive study South America
included analysis of lithology and alteration, mineralogy and Yeatman in 1932 discussed two early examples of copper
petrography, grain size distributions and liberation and their geometallurgy from South America. The first was at
implication for processing. The CSA mine has transitioned Chuquicamata in 1911–1912. Although drilling had identified
from being a lead and zinc producer to a copper producer. This a large deposit of copper ore the mineralisation in the form of
required extensive changes to the design of the grinding and brochantite was something entirely new in large commercial
flotation circuits. Ongoing optimisation has utilised scanning copper deposits and presented new problems in ore treatment.
electron microscope analysis (James and Scamardella, 2000). It was found that the oxidised ore could be leached by
The Peak mine treats ore blends from five different deposits, sulfuric acid, a treatment method held with suspicion at the
with significant challenges due to the inherent variability time. Laboratory tests were conducted on 100 t of ore taken
between each of the orebodies (Hartog et al, 2014; Taylor, 2011). from different sections of the deposit to determine solubility,
Other polymetallic copper geometallurgy programs of note amount of sulfuric acid available and percentages of chlorine,
included those of Hellyer mine (Richmond and Campbell, nitrate, iron etc. Construction of a 10 000 t/d copper leach
1992) and the Thalanga operations (Gregory, Hartley and / precipitation plant was started in 1913, with operation
Wills, 1987). At Hellyer core logging and mapping recorded commencing in May 1915. The second case was that of the
geological and geotechnical data which was considered of El Teniente deposit. The original 400 t/d gravity plant only
importance to mining, metallurgy and exploration from the achieved 65 per cent recovery with a feed grade of 2.5 per cent
onset. By selectively extracting various textures from the Cu, principally as chalcopyrite and, to a limited extent, bornite
mineralisation data set it was possible to construct plans and chalcocite. Samples were dispatched to London for testing
and sections showing the distribution of textural zones and with the new minerals separation flotation technology. The
thus predict the ore type for each stope. These aided mine initial tests were not successful with investigation showing
planning and production scheduling. Ore types which were that a considerable portion of the copper in the samples had
difficult to treat were stockpiled separately and then blended oxidised during transit. This led to one of the first protocols
with other ore types to optimise metallurgical performance for preservation of samples between collection and testing.
(Downs, 1990). Diamond drill core was subjected to extensive Additional fresh and protected samples of ore were sent to
mineralogy to determine implications for possible processing London, with tests achieving a copper recovery of 85 per cent
routes. The samples were further examined for evidence of to 90 per cent. This evaluation would lead to the installation
ore zones or mineralogical differences between different areas of flotation and form a basis for all large porphyry properties
of the orebody. This allowed the orebody to be classified into that would follow.
several categories on the basis of texture with various flow South American copper geometallurgy has typically focused
sheets tested for design purposes (Richmond and Lai, 1988). on porphyry deposits, with CODELCO having a significant
The Thalanga operations provide an example where even role. At the Chuquicamata processing plant recovery was
detailed geometallurgical analysis may not achieve targeted modelled as a function of geomining and metallurgical data
performance. The primary ore zone consisted of massive and ore characteristics obtained from a historical database
and semi-massive sulfides containing copper, lead, zinc and (Compan, Pizarro and Videla, 2015). Operational data on
silver, as well as gold mineralisation. Three supergene ore mill feed grades, ore hardness, particle size, mineralogy, pH
types were designated originally according to the proportions and reagents, representing several months of operation, was
of the valuable sulfide minerals. Liberation data and test collected and using multivariate regression techniques was
work allowed a flow sheet to be developed for the recovery used to predict recovery. The recovery model was validated
of separate copper, lead and zinc concentrates in a three- using monthly plant data between January and July 2014.
stage sequential flotation process. The major difficulty in The model prediction shows a correlation coefficient of
treating the Thalanga supergene ores was the separation of 89.7 per cent and a mean absolute error of 2.75 per cent. Other
copper sulfides and sphalerite. The poor flotation selectivity major geometallurgical studies by CODELCO were reviewed
existed in all three types of supergene. It appeared that in situ by Beniscelli (2011) and included El Teniente, Radomiro Tomic
activation of sphalerite has taken place during the process and Ministro Hales. The Radomiro Tomic study commenced in
of supergene alteration of the Central Thalanga deposit. 1996. Mineral samples were analysed for total Cu (CuT), acid-
This was confirmed by the presence of copper sulfate in the soluble Cu (CuS), total Fe, Cl, S, Mn, Mg, Al, P, Ca, Na, F and
supergene ores and evidence of copper sulfate precipitation SiO2. Also, pH leach and acid dosage tests carried out prior to
WE are metallurgists, not magicians 39

G Harbort et al
a bench scale column leaching characterised the metallurgical Baumgartner, Gomez and Escobar (2016) discuss the
behaviour of two geological units. At El Teniente, a study was mineralogical characterisation and implications at the Cerro
conducted based on geological mapping information, optical Corona Cu-Au porphyry mine. The operation uses a suite of
mineralogy and metallurgical data from 731 samples. This geometallurgical tools including quantitative mineralogical
allowed the generation of a model relating rock texture and analysis (modal mineralogy, grain size liberation and
grindability for the main igneous lithologies. For the Ministro association), semi-quantitative XRD and rougher laboratory
Hales project geometallurgical characterisation allowed the flotation tests to determine clay types and their effect on
identification of key process issues in terms of mineralogical performance.
species and major and minor chemical elements. These issues At Toquepala the orebody was characterised according
included low copper recovery in the mixed minerals zone and to lithology, mineralisation and alteration. The use of size
high final concentrate quality variability. In addition, Carrasco by mineral mineralogy allows ongoing optimisation of the
et al (2005) presented the results of several heterogeneity operating plant (Quiñones and Mattos, 2001).
and Ingamells’ tests done in CODELCO deposits including
During the design phase for the La Constancia project
Chuquicamata, Mina Sur, Radomiro Tomic, El Salvador,
(Klohn, Stephenson and Granados, 2016) a geometallurgical
Andina and El Teniente. Tests considering the natural
study was undertaken to evaluate comminution and flotation
variability were used to develop accurate sampling protocols
options. The study identified dendritic growths of sphalerite
for further geometallurgical analysis.
and chalcopyrite into pyrite, the presence of copper oxides
At the Minera Escondida concentrator metallurgical issues in the supergene zone with associated recovery implications
became apparent soon after start-up of the first phase of and the copper activation of sphalerite both in situ and during
expansion: blending of supergene and skarn ores (Greig et al, 2009). The
The orebody was proving to be more variable than expected initial geometallurgical model estimated copper recoveries
and the limitations of predicting behavior from the widely and concentrate grade and quality over two week periods
spaced original diamond drill holes were apparent. for the mine life to assist in mine planning and metallurgical
Additional drilling programs were initiated, and a joint strategies.
mining and metallurgy program of oreblock testing was Antamina is a complex, highly variable polymetallic
commenced, to determine mineralogical characteristics deposit. Ore is treated by campaign, where one of eight ore
and grinding and flotation response. This data was used types is treated over periods lasting from two to 30 days. Each
to schedule the mining sequence and allow blending, so as ore type has distinct mineralogical characteristics resulting
to avoid peaks in hardness and clay content, however ore in variable metallurgical performance and treatment. In
exposure was often limited which meant high variations in one geometallurgical study, an assessment of mineralogical
mill feed characteristics. (Kilgour, 1995) deportment and texture led to effective penalty element
The Sierra Gorda deposit was identified as having three strategies being developed by metallurgical staff (Kormos
main zones of sulfide mineralisation. The hypogene zone et al, 2010).
consists of chalcopyrite, pyrite, molybdenite and bornite. In Argentina, at the Minera Alumbrera operation,
Gold typically accompanies the copper sulfides. Atacamite, geometallurgical studies commenced approximately ten
brochantite, chrysocolla, vermicullite, lindgrenite, powellite, years before project development (Matar et al, 1986). They
ferrimolybdenite are minerals of note in the leached/oxide continued through plant design (Keran et al, 1998) and into
zone. The supergene zone contains chalcocite, covelite, operating philosophy (Harbort et al, 2000). The Instituto de
digenite, bornite and chalcopyrite. The original flotation Investigaciones Mineras conducted an integrated study on
circuit product recoveries and grades were developed by the deposit which included geological, geochemical and
Aminpro based on their proprietary simulation computer topographic modelling, geostatistical reserve estimation,
model. The final flotation circuit design criteria used in mineralogical behaviour and prefeasibility studies. During the
detailed engineering included modifications based on design phase mineralogy was conducted on both bench and
benchmarking and additional ore kinetic and variability data pilot plant samples to determine optimum design. Operation
(Comi et al, 2013). has used both monthly and survey specific modal mineralogy
The Candelaria mine implemented a geometallurgical and liberation data to optimise performance.
project to increase the plant throughput by optimising rock The Technological Characterisation Laboratory in the
breakage from blasting, crushing and grinding. The ore was Department of Mining and Petroleum Engineering –
characterised, and drilling, blasting, crushing and grinding University of Sao Paulo conducted a significant geometallurgy
processes audited to develop site-specific models for each program for the Sossego copper mine in Brazil (Kahn et al,
process. Simulations were conducted to identify integrated 2014). The study included classifying samples by X-ray cluster
operating strategies in the mine and plant to increase mill analysis and detailed automated mineralogy. The study by
throughput. Recommended changes were implemented, Bergerman et al (2008) evaluated SAG mill variability via
achieving throughput increases of ten to 20 per cent, ore characterisation, JKSimMet comminution modelling and
depending on the ore hardness (Muñoz et al, 2008). blast design methodology. Samples from both core and plant
A number of geometallurgical studies have been published feed were submitted for Drop Weight tests. Core representing
for Peruvian operations. At the Collahuasi operations studies two years of the mine plan, along with plant samples were
have been conducted on both comminution (Alruiz et al, 2009) taken as part of comminution surveys. The study allowed
and flotation (Suazo, Kracht and Alruiz, 2010). The flotation modifications to be progressively implemented into both
geometallurgy model evaluated flotation as a function of air blasting patterns and the comminution circuit, with good
dispersion properties, the feed particle size distribution and correspondence to simulated results.
introducing a parameter (Ф), which represented the inherent Fonseca and Sá (2005) published results from a
geometallurgical floatability of the ore. At industrial scale geometallurgical study on the Alemão IOCG deposit, Carajás,
the model was able to predict metallurgical results in a time Brazil. The basis of the program was a liberation study on
frame of several weeks with an average relative error of less coarse and fine particles correlated with the performance of a
than two per cent. mini pilot plant. The results were used to indicate the primary

target size for grinding and to predict copper recovery from operations was undertaken. An extensive laboratory program
the flotation rougher stage. The liberation by grade class was was developed to assess the metallurgical response of the
used to indicate the particle size for the regrinding stage, prior various ores produced from the underground mine. In some
to flotation cleaning stages. cases of extremely poor performance some ore types were
removed from ore reserves. In other cases, the process had to
North America be modified to optimise circuit response to the particular ore
The Colorado geometallurgy symposium gives an indication feed. A considerable amount of mineralogical and microscopic
of the significant focus on copper porphyry geometallurgy work was completed, indicating the need for better liberation.
in North America during the 1950s. A number of presenters In addition to the metallurgical laboratory work, pilot plant
stressed the importance of a geometallurgical approach from campaigns were undertaken to confirm the conclusions
initial deposit discovery through exploration and into mine reached from bench scale studies (Urbanoski, 1993).
planning and product (Clemmer, 1956; Richard, 1956). In Kidd mine conducted detailed geometallurgical analysis
addition, Kildale (1956) discussed the more complex Cu-Pb- in 2010–2011 to determine optimum plant performance. The
Zn deposits, providing examples of the effect of mineral grain analysis showed that the net revenue impact of copper recovery
size on both flotation and comminution performance. improvement was neutral while zinc revenue decreased due
The Cyprus Pima orebody in Nevada contained to the increased losses to the copper concentrate. The key
chalcopyrite, bornite, native copper, chrysocolla and cuprite. revenue driver for the project was found to be silver due to
In the first year of operation, a geometallurgical program was higher than anticipated recoveries and prices (Leggett and
run in an effort to determine the best operating conditions on Morin, 2013).
Cyprus Pima ore. Seven operating strategies were developed The complex Laronde ores require a series of grinding,
for the seven ore conditions that commonly occurred. Two copper/lead flotation and separation, zinc flotation, zinc
of these ore conditions had variations depending on the tails precious metals leaching, followed by a counter-
tarnishing of sulfide particles. The ore types corresponded
current decantation circuit and Merrill Crowe precipitation.
to specific areas of the pit, allowing selective mining and
The plant treats a range of ore types requiring multiple
treatment (Ramsey, 1976).
metallurgical processes and optimisations, as well as
Geometallurgy at the Lavender Hill concentrator was driven operational and metallurgical philosophies that have
by the complex copper mineralogy of the orebody. Chalcocite changed to adapt to custom centralised milling (Blatter,
occurred as rims around pyrite grains and as minute veinlets Cayouette and Cousin, 2011).
along fractures within the pyrite grains, many contained a
Metallurgical results indicated slight differences when
network of chalcocite veinlets. Microscopic examination of
ore from the Matagami Lake Cu-Zn mine was ground with
polished sections showed the chalcocite veinlets to be 10 µm
thick, or less. Oxide copper formed very thin coatings on the different grinding media. The grinds are compared on
chalcocite, barely perceptible under the microscope. Oxidation the basis of size analyses and of the free pyrite, sphalerite,
took place rapidly after the ore was broken, particularly when it chalcopyrite and galena grains in the flotation feeds. The
was hot and humid. There was little oxidation during crushing, sizes of the free grains in all flotation feeds were different for
but significant oxidation during grinding (Martin, 1957). each mineral, decreasing in the approximate order pyrite-
sphalerite-chalcopyrite-galena. This order corresponds to the
Geometallurgy at the Anaconda C E Weed concentrator was
order of decreasing hardness for the minerals. In addition the
instigated due to variability in the Berkeley Pit. The pit was
grain sizes for specific minerals were different when different
a very non-homogeneous mine with respect to ore type and
grinding media was used (Petruk and Hughson, 1977).
mineralisation. An understanding of the geology and mine
led to separate slimes and sands circuits and strategies for the The Highland Valley Copper geometallurgical study
treatment of ‘sweet’, ‘sour’ and ‘hard’ ores (Palagi and Stillar, (Mitchell and Holowachuk, 1996) provides a classic example
1976; Wraith and Fulmor, 1964). of throughput optimisation via geometallurgy. The study
involved analysis of historical data and the comparison of
A more recent geometallurgy study is that of Bingham
Canyon which was designed as a proof-of-concept test the different intensities of alterations and rock types with
to develop a method of quantifying key geometallurgical actual milling rates. Milling rates derived from these studies
properties. The deposit was divided into 33 ore domains were further defined by processing many of the ore types
by mine geologists and metallurgists on the basis of individually and in combination with one another. This
lithology, alteration assemblages, fracture density, copper data was supplemented by both historic Drop Weight tests
mineralisation and metallurgical properties. This allowed a and future ore tests. The study allowed an effective blending
combination of geological and mineralogical characteristics strategy to be developed with mine operations, engineering
to be determined that, when combined, impacted processing and geology personnel meeting daily to determine specific
(Ross et al, 2009). blending strategies for each 24 hour period.
The approach used to produce geometallurgical domains at At the Kemess Cu-Au porphyry operation, concentrate
the Pebble porphyry Cu-Au-Mo deposit in Alaska was to first regrinding requirements were evaluated using a
characterise representative samples from each geologically geometallurgical approach. Treatment towards the end of
distinct area of the deposit. The samples from each area mine life resulted in a 22 per cent increase in grinding energy,
were studied using a combination of optical petrography a seven per cent decrease in copper recovery, a 20 per cent
and automated mineral mapping to define the silicate and decrease in gold recovery and a three per cent Cu decrease
sulfide mineralogy and the copper and gold deportment. in concentrate grade. Mineralogical studies indicated poor
This allowed assessment of changes in gold deportment as a chalcopyrite/pyrite liberation, with finer grinding required
function of alteration across the deposit (Gregory et al, 2013). in concentrate regrind (Brissette and Roman, 2012).
Copper geometallurgy in Canada has a significantly higher The Mount Polley porphyry copper mine makes use of on-
focus on polymetallic deposits than the rest of the Americas. In site automated mineralogy, both for optimisation and strategic
order to achieve metallurgical improvements at the Brunswick planning. An initial study compiled daily mineralogy data for
Cu-Pb-Zn mine a thorough review of the concentrator several weeks to generate mineral balances around the circuit.

G Harbort et al
This allowed strategies for mine planning, forecasting and Chalcocite occurs as sooty coatings throughout the massive
operational tuning to be developed (Dobbe et al, 2014). sulfides, mainly in the interval immediately underlying the
The Bismark, Sabinas 2 and Tizapas concentrators current water table. This has resulted in complex association
in Mexico have produced copper concentrates from of sulfides in the supergene-enriched zones as well as in areas
polymetallic deposits. In the last 15 years, emphasis has been that are highly fractured and structurally exposed to water
placed on the use of modal mineralogy, mineral association penetration. A geometallurgical study involving mineralogy,
and liberation characteristics to optimise operations and hydrology, structural geology and extensive metallurgical
metallurgical performance (Magallanes-Hernández and tests was undertaken to evaluate the reasons for elevated zinc
Espinosa-Gómez, 2005). content in copper concentrate. This indicated the presence of
in situ zinc and pyrite activation by copper, within zones of the
Asia Pacific orebody. This resulted in a reagent scheme which exploited
The complexity of the Ok Tedi orebody resulted in numerous the varying preferences of copper and zinc in relation to
geometallurgical studies being conducted over its history. pulp potential, allowing significantly greater selectivity to be
Sulfide porphyry ore consisted of chalcocite, digenite and achieved (Umipig et al, 2012).
chalcopyrite with typical copper recoveries of 90 per cent. The King-king porphyry deposit copper mineralogy consists
Oxidation led to localised formation of malachite, azurite of chalcopyrite, bornite, chalcocite, digenite and covellite.
and chrysocolla with lower and more variable recovery. Copper silicates are the most abundant copper minerals in the
Skarn mineralogy is varied and includes magnetite skarns, oxide zone of the deposit. Gold is relatively abundant in the
massive sulfides, gossans and calc-silicate skarns. In massive oxide zone, with lesser abundance in the sulfides. Native gold
sulfides and magnetite skarns, pyrite is abundant with copper is occasionally observed on fractures and in quartz veinlets.
present as chalcopyrite, chalcocite, digenite and bornite. The The King-king metallurgical study was based on an extended
close association of copper and pyrite reduces both recovery sequential assay procedure to determine copper speciation,
and concentrate grade. Recovery of copper and gold from correlated with the results of extensive flotation tests. The
gossanous material was poorest of all the ore types. Specific model allowed prediction of copper and gold recoveries
reagent regimes are required to achieve optimum performance throughout the deposit, based on sequential assay results
for the various ore types. In addition, ore hardness varies with (Snider et al, 2013).
ore type and degree of weathering. A high level of cooperation
At the Sepon operations in Laos, geological logging
between geology, mining and metallurgy departments
uses a set of predetermined tables for lithology, structure,
was necessary in order to deliver blends of ore which gave
mineralisation, geotech, oxidation and alteration, in addition
satisfactory metallurgical performance (England, Kilgour and
to site developed metallurgical codes (metcode). These
Kanau, 1991). Initial electron microprobe analysis of porphyry
metcodes are based on the metallurgical behaviour in the
samples confirmed the presence of fresh chalcopyrite, bornite
Sepon processing plant. In addition the resource model is
and pyrite and also showed that weathering products
coded by geometallurgical type for scheduling the process
of chalcopyrite vary in composition between chalcocite,
plant and by lithology for determining mining costs and
covellite, digenite, geerite and blaubleiblender ‘synthetic’
geotechnical domains (Quigley, Hackhan and Broome, 2008).
covellite. Fresh pyrite was found to exhibit natural floatability
(Afenya and Mwaba, 1991). Later work focused on the At the Phu Kham concentrator, mineralogical examination
mineralogical and metallurgical examination of fluorosilicate of monthly composites has been conducted since operations
to define and implement strategy for the reduction of commenced in 2008. This information, together with targeted
fluorine in concentrates (Pangum et al, 1998). One landmark mineralogy, is routinely used for process understanding and
geometallurgy study was that of predicting ore aging within improvement. Evidence was provided of chalcocite-covellite
potential new block caves and its effect on metallurgical intergrowth with pyrite and rimming of pyrite. There was also
performance (Morey and Cantrell, 2011). indication of copper activation of pyrite from soluble copper
At the time of development Batu Hijau was the largest species in weathered and transition ores. The mineralogy
mining project that has ever been undertaken. Detailed data demonstrated that since primary chalcopyrite ores had
production planning for the concentrator commenced in the become the dominant source of plant feed, the major cause of
middle of 1997, two years before production commenced. loss of copper in-plant tailings had changed from slow floating
The planned production ramp-up was reviewed alongside finely liberated copper minerals to chalcopyrite locked in poor
actual production experience from more than 30 new mining quality coarse binary particles with non-sulfide gangue. This
projects dating back to the 1970s. The geometallurgical model knowledge, coupled with flotation modelling and simulation,
for copper recovery was based on lithology, mineralogy, led to an increase in total recovery of both copper and gold
and depth with reference to 500 bench scale flotation tests by six per cent into final concentrate by increasing mass
(DeMull, Spenceley and Hickey, 2001). Later work focused on recovery into rougher concentrate, and debottlenecking of
throughput prediction using a geometallurgy approach. The rougher concentrate regrind, cleaning, and final concentrate
deposit’s lithologies, their distributions and association with dewatering plants (Bennett, Crnkovic and Walker, 2011).
mineralisation and geotechnical measurements served as the Variability in concentrate grades (24.7–37.6 per cent Cu) and
basis for ongoing metallurgical studies. This allowed models copper recoveries (60–90 per cent) during treatment of ore from
of Bond Crushing Work, Ball Work, Rod Work and Abrasion the Malanjkhand porphyry deposit, India, reflect its geology.
indices, JKMRC impact breakage resistance and the JKMRC To obtain higher copper recoveries, a geometallurgical study
abrasion resistance to be developed. was undertaken to define the ore characteristics that affected
The Canatuan Cu-Zn deposit in the Philippines is comprised metallurgy and to assess applicable processes for recovering
of an iron oxide gossan underlain by massive sulfide. The the metals. It was found that chalcopyrite tarnished readily
sulfide occurs as either massive or banded sulfide. The massive and oxidised to chalcocite within several months. The
sulfide is composed of >50 per cent coarse-grained pyrite, presence of both the tarnished chalcopyrite and chalcocite
chalcopyrite sphalerite, galena and tennantite. The banded affected chalcopyrite flotation and reduced copper recoveries
sulfide has lower pyrite content with bornite and covellite. (Petruk and Sikka, 1987).

Europe – Middle East financial drivers to focus on geometallurgical issues within

the orebody. This led to a strategy of ensuring a consistent
In 2006 the Chelopech mine in Bulgaria commenced a
metal production profile, as a result of managed plant feed,
geometallurgical project to study the metallurgical complexity
in terms of grade and geometallurgical characteristics, with
of the ore and the key aspects of its treatment. A major aim of
a mine to stockpile to plant philosophy built into the life-of-
the project was justification of possible changes in the process
mine plans (Macfarlane and Williams, 2014).
flow sheet to achieve higher recoveries. The basis of the
program was samples collected during two full plant surveys Interestingly, there are similarities between Ruashi and
of the grinding, classification and flotation circuits. Sample the Kolwezi concentrator 50 years earlier. Kolwezi, a Cu-Co
analysis included elemental and mineral composition, size mine in the DRC also used early geometallurgy to stabilise
and nature of intergrowths, liberation, grindability and ore variability in its comminution and flotation plant. The
floatability (Baltov et al, 2008). variable crushing requirements led to both in-pit and stockpile
strategies. Understanding of gangue variability assisted in
Mikheevskoye in Russia undertook a study to improve
selecting and optimising flotation reagents (Saquet et al, 1962).
metallurgical performance through better understanding of
the variation in the porphyry orebody and mine planning. Another geometallurgy program in the DRC was conducted
Thirteen geometallurgical domains were established for the on the Kamoa Copper deposit. The expansive study initially
ore zones which had similar characteristics during processing. demonstrated that representative, true samples of drill core
The study predicted that a geometallurgical approach could had been extracted from the drill core inventory. Copper
potentially decrease the payback period for the project by mineralogical studies identified bornite, chalcopyrite,
1.5 years and significantly increase the net present value chalcocite, with lesser covellite and azurite. The variation
(Lishchuk et al, 2015). in copper speciation led to a mixed collector flotation
test program to optimise flotation of all of these species.
The Sarcheshmeh concentrator in Iran implemented a
Liberation studies were used to define both primary grinding
geometallurgical study to diagnose the reasons for fluctuating
and concentrate regrind requirements (Lotter et al, 2013).
copper recoveries. The study was instigated following reagent
and ball charge changes which were implemented as part of
a recovery improvement project. Five months of mineralogy
Methods of integrating of geometallurgy with plant design
data was reviewed, with specific focus on the liberation of Traditional engineering design for flotation circuits uses the
chalcopyrite, covellite and chalcocite (Banisi et al, 2003). ‘Rule of Thumb’ approach. Typically this uses a nominated
maximum head grade for design. A scale-up factor is applied
Africa to laboratory flotation tests for residence time. This scale-
up factor is usually based on a designer’s experience at
The Mufulira copper mine in Zambia represented an earlier
other flotation sites and may or may not be relevant to the
age geometallurgy endeavour. Its copper minerals included
circuit being designed, or the flotation equipment being
chalcocite, bornite, malachite, native copper, azurite, covellite
used. A small number of locked cycle tests are assumed to
and chrysocolla. Gold was also present throughout the
represent the orebody and the results of these are often used
orebody, as was graphitic carbon, which caused significant for financial analysis, independent of changes in mine plan
difficulties in the flotation circuit. Geometallurgical planning throughput and mineralogy. The engineering ‘rule of thumb’
led to extensive efforts at blending, both underground and approach is only strictly accurate where there is very little
in the crushing plant, to minimise the effect of ore variability variability in throughput, head grade and mineralogy. It may
(White and Adair, 1948). provide accurate design for mature established operations
At the Tsumeb polymetallic mine in Namibia, the where a brownfield expansion is being considered but is
presence of 40 economic minerals, some refractory, led to unlikely to provide accurate design for a new, greenfield
treatment difficulties, even with high grades present. They site. To overcome this Rule of Thumb approach to design,
included chalcocite, bornite, tennantite, malachite, native geometallurgical test programs and data review have been
copper, cuprite, galena, cerunnite, sphalerite, native silver, developed to provide realistic information into simulation
germanite, renierite, conichalcite, duftite, mottramite, packages to provide data for process evaluation and plant
miretite, anglenite, smithsonite and willemite. To quote design. This approach was applied to the Andash project to
Boyce, Venter and Adam (1970): aid final design and risk minimisation.
The distinct flotation characteristics of the different minerals A number of geometallurgical modelling techniques exist
in the ore, and the rapid changes in ore grade and mineral and can be divided into three approaches as discussed by
type, made the development of the current flow sheet a long Lishchuk et al (2015):
and at times, exasperating undertaking. 1. In the traditional approach the metallurgical response
Over the concentrator’s history, multiple geometallurgical of an ore in the mineral processing plant is calculated
studies were undertaken including in situ sphalerite activation, from the chemical assays using mathematical functions,
overgrinding of chalcocite, tennantite selectivity, galena which are often called recovery functions. The
flotation variability, intimate mineral locking and associations. functions are developed using variability testing and
The implementation of an on-site daily mineralogical statistical analysis to define the correlation between the
analysis at Nchanga in Zambia allowed sharing of information metallurgical response and feed properties.
within the processing department, and between departments. 2. The proxies approach uses geometallurgical tests for large
The geometallurgy program assisted in the diagnosis and numbers of samples. The geometallurgical test is a small-
solution of processing problems. In one instance the program scale test which indirectly measures the metallurgical
was credited with achieving a one per cent increase in sulfide response. Normally the geometallurgical test results
recovery (Siddiqui, 2000). must be converted with certain correction factors to give
The Ruashi mine in the Democratic Republic of Congo (DRC) estimates on the metallurgical results of plant.
has significant variability in both copper and cobalt grades 3. The mineralogical approach’s main characteristic is the
across the resource. The constraint on acid availability and cost, continuous and systematic collection of quantitative
coupled with the need to stabilise plant performance, were the mineralogical information.

G Harbort et al
The geometallurgical development discussed in this paper content. The ore is formed by carbonate-chlorite-
was conducted in a number of stages: sericite-quartz alterations. Hypergene minerals are
•• incorporation of mineral data, either from sequential significant and are represented by malachite, azurite,
copper assays or mineralogy, into the resource model covelline, cuprite, native copper, drogoethite, bornite
and mine plan and chalcosine.
•• development of floatability parameters based on the •• The lower ore zone is at depths of 60–175 m. This zone is
rougher performance in either locked cycle tests or mainly formed by chloritised and sericitised cataclasised
rougher batch tests granodiorite porphyry with increasing chalcopyrite and
•• generation of a floatability component model to simulate pyrite content.
the locked cycle tests or existing plant performance Numerous barren dykes of various compositions are present
•• calibration of the model so that model results equate to within the orebody.
either actual locked cycle test results or historical plant
performance Treatment
•• development of a JKSimFloat simulator incorporating The proposed Andash ore processing flow sheet, as shown
floatability data, proposed circuit design, flotation feed in Figures 1 and 2, was designed to recover gold bearing
rates and flotation machine characteristics copper concentrate through the following principal unit
•• determination of mineral recovery with ores of various processes:
copper mineralogies, based on variability tests and •• three-stage crushing, ore storage and reclaim
mine plans. •• single stage ball milling (expanded to primary and
secondary ball milling in Phase 2)
CASE STUDIES •• flash flotation (Phase 2)
Two previously published case studies are reviewed in this •• sequential rougher flotation (utilising split sulfide and
paper. The first is the Andash study (Harbort, Cordingley and
oxide circuits)
Phillips, 2011), a greenfields geometallurgical study.
•• combined cleaner, cleaner scavenger and recleaner
The second is Northparkes Mines (Clarke et al, 2014; Jones
flotation
and Morgan, 2016), an operating site with both brownfield
and greenfield components. The brownfield component •• sulfide concentrate regrind (Phase 2)
relates to optimising the plant operation, while the greenfield •• concentrate dewatering, filtration, storage and load-out
component relates to the prediction of future performance and tailings thickening (Phase 2)
with various circuit and mine plan options.
•• disposal and decant water return.
Andash Geometallurgical evaluation
Location During the development and design of the Andash
concentrator, numerous flotation tests were undertaken
The Andash Gold-Copper Project is located in the Kyrgyz
including variability tests. The variability tests confirmed that
Republic, approximately 260 km from the capital Bishkek in
the Talas region of Kyrgyzstan. the orebody exhibits significant variability in total copper
rougher recovery and minor variability in rougher gold
Geological description recovery. The wide copper variability was directly related to
copper mineral variability. The recovery of copper sulfides
The Andash deposit is formed by the stratified sediments of the
was relatively robust with 96 per cent rougher recovery when
Tohtonisai and Jangiturmush formations. These formations
no copper oxides were present, declining to 90 per cent at a
are characterised by bodies and dykes of the Taldysui
copper oxide to copper sulfide ratio of 0.8. At very high oxide
monzo-diorite, granite and Permian dyke complexes. The
Tohtonisai formations, which forms the south-west part of copper content the sulfide recovery decreased significantly.
the deposit with a thickness of 300 m, comprises of moderate The rougher copper oxide recovery was heavily dependent
alkaline basalts, trachy-andesites and associated tuffs, beds on the ratio of acetic acid-soluble copper, eg malachite to
of agglomerate basalts, tuff conglomerates and sandstone weak acid-soluble oxides such as tenorite. When copper
layers. The Jangiturmush Formation forms the majority carbonates were the only oxide copper minerals present, their
of the deposit’s area with a thickness of 450–500 m. The recovery approached 75 per cent and declined with higher
formation is formed by tuff conglomerates, tuff-gravelites, weak acid-soluble copper content. Due to the wide variability,
tuff-sandstone with interlayers of tuffite, tuff-andesites and it was determined that the locked cycle tests could not
agglomerate tuff of basalts. effectively be used as a measure of life-of-mine performance.
The Andash copper-gold mineralisation is structurally At this stage the test sample was approaching exhaustion,
confined to a hydrothermally silicified tectonic breccia with limited sample available to conduct further flotation
and quartz alteration, extending from and terminating into tests. A geometallurgical approach was adopted to evaluate
granodiorite. The orebody is zoned as follows: performance over the mine life.
•• The upper section of the ore zone spreads from surface
to a depth of 40 to 75 m and is composed mainly of Resource and mine plan development
cataclasised granodiorite porphyry. Hydrothermally Sequential assay results were input into the resource model.
imposed ore minerals include chalcopyrite, native gold Development of a mine plan based on this geometallurgical
and with hypergene minerals including cuprite, covelline, resource model allowed a mine plan and concentrator
bornite, native copper, chalcosine and hydrotite. schedule to be developed based on copper mineral species
•• The middle section of the ore zone stretches from (sequential copper extractions). The scheduled feed grade to
40–105 m with sharply increased gold and copper the concentrator on a copper species basis is shown in Figure 3.

FIG 1 – Andash Phase 1 circuit flow sheet.
FIG 2 – Three-dimensional view of the Andash circuit.

G Harbort et al
Key points in the schedule are: results to see if floatability was constant across the locked
•• At the commencement of mining, heavily oxidised cycle test. The comparison indicated that mineral flotabilities
material was stockpiled for end of mine life treatment. in the respective rougher concentrates were not constant
This resulted in an initial concentrator feed where the across their respective cleaner circuit feeds. Flotabilities in the
majority of copper was acetic acid-soluble, eg malachite. cleaner concentrate were, however, conserved into recleaner
feeds. The major variations in the cleaner feeds were:
•• During the second year of operation, the starter pit
penetrated a high-grade sulfide pod, with a significant •• a decrease in the floatability of tenorite and malachite in
increase in both feed grade and aqua regia-soluble copper, the sulfide cleaner feed
ie chalcopyrite. Year three represents a significant mine •• an increase in the floatability of chalcopyrite in the
expansion with lower grade copper carbonate material sulfide cleaner feed
feeding the concentrator. This period coincided with the •• a decrease in the floatability of tenorite in the oxide
planned expansion of the concentrator facilities. cleaner feed.
•• During the fourth year, increased amounts of copper A series of floatability transfer matrices were incorporated
oxides and copper carbonates are mined from the into the model to calibrate it against the locked cycle test. This
expanded pit and presented to the concentrator. The provided acceptable recoveries of copper and sulfur, with the
deepening pit also results in significant amounts of modelled final concentrate grade matching the locked cycle
chalcopyrite being mined and treated. test in terms of copper and sulfur. Modelled iron grades were
•• In the ensuing three years, the head grade to the lower in the final concentrate however and this resulted in
concentrator remains relatively constant, although over estimation of non-sulfide gangue. To correct the iron
the copper mineralogy changes significantly. Acetic imbalance non-sulfide gangue was adjusted in the model to
acid-soluble copper is progressively replaced with contain iron. (Later QEMSCAN – Quantitative Evaluation
chalcopyrite as the pit deepens. The amount of copper of Minerals by Scanning Electron Microscopy – analysis of
oxide material also decreases. the concentrate indicated the higher iron levels were caused
•• Year seven represents treatment of stockpiled material by the flotation of copper-goethite). The calibrated model
with substantial amounts of copper oxides, although was then compared against the second locked cycle test and
acetic acid-soluble copper still represents 50 per cent of showed good agreement. A further two locked cycle tests
the concentrator feed. were conducted with different configurations and simulated.
The simulated results agreed with actual results and it was
Floatability component modelling decided that the model could effectively be used to replace
further locked cycle tests and simulate full circuit operation.
Ore floatability components were calculated from the sulfide
dominant composite rougher kinetic locked cycle test results.
As no bubble size measurements were conducted, a bubble
Integration with plant design
size of 1.1 mm was assumed. It was also assumed that the Seventy-eight simulations were conducted on the Andash
batch cell froth recovery was 100 per cent and entrainment flotation circuit developed during the DFS. This circuit consisted
was zero. As the model utilised a mineral based system it was of sequential sulfide and oxide circuits and split sulfide and
assumed that acetic acid-soluble copper was malachite, weak oxide cleaner circuits. Each simulation was conducted on the
sulfuric acid copper was tenorite, cyanide soluble copper was basis of one month of the concentrator feed schedule.
chalcocite and the copper extracted in the final aqua regia was These simulations were initially conducted to determine the
chalcopyrite. variability of copper and gold recovery and concentrate grade
The floatability parameters indicated that chalcopyrite was throughout the planned operating life. They also provided
highly floatable in the sulfide rougher feed and was depressed detailed mass balances across each month of the mine plan
in the oxide flotation stage. Both malachite and tenorite had and were used as a cross-check of the DFS design. A number
low floatability in the sulfide rougher. Both became floatable in of significant problems soon became apparent from using a
the oxide rougher with the weighted floatability of malachite rule of thumb approach. Although the simulations showed
being approximately double that tenorite. only minor variation in recoveries, they indicated major
variation in stream pulp flow on a month-to-month basis. It
A simulation was the created to match the configuration
was expected that these variations would become worse over
of the locked cycle test, including cleaner stages. The model
shorter time frames. The sulfide cleaner and recleaner circuits
results were then compared to the measured locked cycle
typically operated within acceptable limits. This was not the
case with the oxide cleaner circuit where significant variations
were evident and circuit stability was expected to cause
operating difficulties and loss of metallurgical performance.
The oxide recleaner circuit would never have operated to
design without major modifications.
An initial solution evaluated was to combine sulfide rougher
concentrate and oxide rougher concentrate and treat them in a
combined cleaner circuit utilising the same amount of flotation
equipment specified in the DFS design. This provided a major
improvement in the ratio of anticipated residence time to the
nominated design residence time and hence circuit stability.
Simulations, however, failed to match those of the test work,
with any recovery benefit undermined by large decreases
in concentrate grade caused by the copper oxide flotation
chemicals generating high levels of gangue entrainment. A
modified combined cleaner circuit was simulated with staged
FIG 3 – Life-of-mine mineral grade in the concentrator feed schedule. addition of rougher sulfide concentrate and rougher oxide

concentrate to the cleaner circuit. This provided an increase in •• The cleaner Jameson cell tailing is treated by the cleaner
copper recovery of approximately 2.5 per cent when treating scavenger mechanical cells, where the concentrate
sulfide ore and little tangible benefit when treating oxide discharges to recleaner Jameson cell flotation. The tailing
ore. Considering the ore mineral split would represent an from the cleaner scavenger is pumped back to the feed of
overall improvement in copper recovery over the mine life of the rougher scavengers.
0.5 per cent. In addition, the revised circuit required two less •• Concentrate from the recleaner Jameson cell discharges
flotation cells and was expected to be overwhelmingly more to the final concentrate hopper. Tailing from the
stable than the DFS split-cleaner circuit design. The modified recleaner Jameson cell is returned to the feed of the
circuit was incorporated into the Andash detail design. In cleaner Jameson cell.
addition, the simulation mass balances were used to calculate
cleaner circuit make-up water requirements. This resulted in Geometallurgical evaluation
all water addition piping to the cleaner circuit being changed From June 2011 to November 2013 Northparkes conducted
in the detailed design to further maximise circuit stability. mineralogical evaluation of monthly composites to evaluate
plant performance. Conducted by the Rio Tinto Technology
Northparkes and Innovation group, these production investigations were
complemented by analysis of comminution samples, selected
Location flotation survey samples and drill core samples from potential
Northparkes Mines (Northparkes), an unincorporated future concentrator feed sources. Historical plant surveys
joint venture between the China Molybdenum Company resulted in the ongoing development of model simulations.
(80 per cent), Sumitomo Metal Mining Oceania (13.3 per cent) In 2013, the 2012 un-sized, copper only JKSimFloat simulator
and Sumitomo Corporation (6.75 per cent), operates block was expanded to include mineralogy and a floatability
cave and open cut mines and an ore processing plant located component and used to check the accuracy of performance
27 km north of Parkes in central New South Wales. estimates for the future. Simulations were conducted over
the period of 2011 to 2013 to confirm validity of the model.
Geology In March 2015, Northparkes installed a process mineralogy
The copper-gold porphyry mineralisation at Northparkes facility on-site. The laboratory included a TESCAN TIMA
is hosted by the Late Ordovician Goonumbla Volcanics scanning electron microscope (SEM) and was the first SEM
and occurs in stockwork quartz veins and disseminations facility to be commissioned on a mine site in Australia. The
associated with potassic alteration. Sulfides are zoned facility has not only continued with the analysis of monthly
laterally from the centres of mineralisation. The central plant composites, but has also supported: metallurgical test
portions are bornite-rich with minor chalcopyrite, zoning work programs; the evaluation of alternative ores and feed
outward through equal portions of bornite + chalcopyrite, material; plant and laboratory reagent trials; grinding media
to a chalcopyrite-rich zone. Pyrite increases outward at the trials; plant optimisation and improvement; changes to plant
expense of bornite. Beyond the chalcopyrite zone, pyrite is the configuration and provided laboratory support to external
main sulfide and copper sulfides are minor to absent. consultants working on NPM projects.
Treatment Optimisation
The Northparkes ore processing flow sheet is shown in Routine analysis of monthly plant composites has allowed the
Figure 4 and Figure 5. The concentrator consists of two measurement of plant performance over time and provided
modules. Each module has its own grinding circuit, flotation a baseline of feed characteristics versus performance for
circuit, concentrate thickening circuit and filtration circuit. comparison. Northparkes have been conducting quantitative
The tailings from the two modules are combined in a single mineralogical analysis of plant composites for more than
tailing thickener before being deposited in tailing facilities. 11 years. Important information obtained from this data
Flotation takes place in two distinct but similarly configured has included a measure of the variation in the mineral
modules, each fed by its own grinding circuit. The tertiary make-up, ratios and copper source minerals of the plant
cyclone overflows of each module feed the main flotation feed; individual mineral recoveries by size; liberation and
circuits with the following configuration: locking characteristics; grain size versus particle size; mineral
association and mineral maps. Monthly mineralogy is
•• Preflotation is performed in tank cells, with frother, routinely input into the circuit flotation simulations, together
sodium hydrosulfide and thiocarbamate promoter with other operating data to calibrate and validate the model
added. accuracy.
•• The preflotation tailing is further treated with xanthate The data has shown the reality of actual recoverable copper
collector and frother. The conditioned pulp flows versus the actual non-recoverable copper. This is an important
through a series of banks of mechanical rougher and distinction when evaluating performance and changes to the
scavenger cells, wherein down-the-bank xanthate is also circuit. Analysis of the concentrate has illustrated the nature
employed. of impurities and levels of penalty elements and facilitated
•• Further scavenging in a post-flotation stage is conducted the development of a test work program or strategy, to reduce
in large tank cells. The resulting post-flotation these. The microchemical analysis of gold particles by energy
concentrates are recirculated back to the tertiary dispersive X-ray (EDX) analysis has detected the presence of
grinding circuit, while the post-flotation tailings are gold and silver tellurides in the ore. This data has driven a
combined in a single thickener before being deposited in phase of laboratory method development to improve the fire-
the tailing facilities. assay process and quality of NPM gold analysis, particularly
•• Concentrates from preflotation, roughers and rougher in regard to shipment assays.
scavengers are combined to feed the cleaner Jameson cell. Both equipment changes and circuit improvement projects
Concentrate from the Jameson cleaner cell discharges to have incorporated geometallurgical data. This has driven a
the final concentrate hopper. reduction in the technical risk of specific projects with the

G Harbort et al
FIG 4 – Northparkes Module 1 circuit flow sheet (shaded areas represent potential future changes).
FIG 5 – Three-dimensional view of the Northparkes circuit.

data used to evaluate, justify, test and validate major plant CONCLUSIONS
upgrades. This has proven to be a major drawcard when
The use of geometallurgy with copper concentrator design
justifying funding for an expensive project. As an example, and operation has a long history. Recent years have seen it
installation of additional scavenger flotation cell capacity move from an ad hoc methodology at a few sites to a well-
demonstrated recovery improvements of around one per cent. established procedure.
Mineralogy information plus simulations using mineralogical
inputs were involved in all aspects of the project, with data The Andash case study details the use of geometallurgy in
used to successfully evaluate the risk, drive decision-making, greenfield plant design with limited information available.
Geometallurgical modelling using sequential copper assays
optimise and validate changes. Mineralogy was also used
and floatability component modelling provided a design
to critically review the post-commissioning performance of
methodology with significantly less associated risk than
the new cells. Additional implementations have included
traditional design.
installation of froth washing and subsequent optimisation
to decrease entrainment. The bypass of a primary cyclone to The Northparkes case study illustrates the use of
reduce overgrinding of already liberated copper minerals and geometallurgy, via process mineralogy and mineral
optimisation of reagents once the bypass was implemented. component modelling. The use of geometallurgy has
The Northparkes on-site laboratory has frequently been facilitated a common goal of mill optimisation, stability and
utilised to support work programs. Mineralogical analysis more realistic forecasting through orebody knowledge and
is used to evaluate change in reagents and grinding media the sharing of information between disciplines.
from bench tests and in-plant. Performing the mineralogy
and some of the assay components of the test work in-house ACKNOWLEDGEMENTS
has given the site control over quality assurance and project The authors would like to thank Northparkes Mines and
integrity. Amec Foster Wheeler for allowing this paper to be published.
In addition to optimisation in the concentrator, The case studies utilise three papers previously published by
geometallurgy has also been used for mine optimisation. the AusIMM and the authors are grateful for the permission
Northparkes are currently mining the Endeavour 48 (E48) to extract the content from these. Previous authors of those
orebody via a modified block cave technique. This orebody papers are also acknowledged, namely Guy Cordingley,
still has 15 years of mining life, yet up until recently there was Marius Phillips and Danica Clarke. Sections of the introduction
very little mineralogical information available. To remedy were extracted from an unpublished paper – ‘Technological
this, mineralogical analysis has commenced on material from advances in flotation’, written with the guidance of Professor
the current drawpoints to gain information about the samples Alban Lynch in 2011.
that make up the daily feed blend. The mineralogy and assay
combination is used to provide mineral abundance, elemental REFERENCES
deportment and textural data to map cave drawpoints. The Afenya, P M and Mwaba, C C, 1991. Flotation of Ok Tedi sulphide
data shows the mineralogical variability of the orebody minerals, in Proceedings PNG Geology, Exploration and Mining
which is currently mined according to fragmentation, caving Conference, pp 191–196 (The Australasian Institute of Mining and
characteristics and copper grade inputs. The long-term Metallurgy: Melbourne).
objective would be to add mineralogy data to that criteria. Alruiz, O M, Morrell, S, Suazo, C J and Naranjo, A, 2009. A novel
Some drawpoints contain very fine clay-bearing material approach to the geometallurgical modelling of the Collahuasi
that has the ability to move like a liquid. This is extremely grinding circuit, Minerals Engineering, 22:1060–1067.
dangerous in the cave environment as the material has Anon, 1956. Success depends on: geologists, mining engineers,
the potential to cause a fines rush. This fine material was metallurgists, presented to Symposium on Greater Cooperation
investigated using SEM imaging and EDX microchemical for Metallurgical Planning; Second Annual Rocky Mountain
analysis techniques to infer the size, illustrate the morphology Minerals Conference, Salt Lake City, October, Mining Engineering,
and measure chemical composition. Based on this data, July, p 698.
additional mining controls including exclusion zones have Arden, H and Tverdov, A, 2014. Resource and reserve valuation
been implemented where similar mineralogy indicators exist. practices in CIS countries, in Mineral Resource and Ore Reserve
Estimation: The AusIMM Guide to Good Practice, second edition
Life-of-mine planning pp 825–834 (The Australasian Institute of Mining and Metallurgy:
Melbourne).
The incorporation of mineralogical data and floatability
components into simulations to predict future performance Baltov, I, Hristov, N, Evtiminova, K, Dontchev, I and Kuzev, L, 2008.
for analysis has provided a more robust method of Metallurgical challenges in the recovery of copper and gold from
the Chelopech mine, Bulgaria, in Proceedings XXIVth International
prediction than the historical method of prediction using
Minerals Processing Congress (ed: W D Duo and S C Yao), pp 1785–
only a head grade and throughput relationship to predict 1792 (Science Press: Beijing).
recovery, based purely on laboratory flotation results.
Banisi, S, Iran-Manesh, H R, Shayestefar, M R and Shekarchyan,
This improved understanding allows for the development
H, 2013. Mineralogical tracing of metallurgical results – the
of circuit modifications and circuit options earlier in the
Sarcheshmeh copper mine case, in Proceedings 35th Annual
planning process as well as supporting the development of Meeting of the Canadian Mineral Processors, Ottawa, 21–23 January,
future concentrate specifications. The simulation of future pp 417–433 (Canadian Institute of Mining, Metallurgy and
performance provides an improved ability to manage future Petroleum: Montreal).
expectations and ore reserves, becoming a critical tool in Barns, K E, Colbert, P J and Munro, P D, 2009. Designing the optimal
optimising the mining sequence over the life-of-mine. Further flotation circuit – the Prominent Hill case, in Proceedings Tenth
expansion of the model will introduce the ability to predict AusIMM Mill Operators’ Conference, pp 173–182 (The Australasian
gold and impurity behaviour as well as the copper sulfides. Institute of Mining and Metallurgy: Melbourne).

G Harbort et al
Baumgartner, R, Gomez, P and Escobar, G, 2016. Comprehensive David, D, 2007. The importance of geometallurgical analysis in
mineralogical characterisation at the Cerro Corona Cu-Au plant study, design and operational phases, in Proceedings Ninth
porphyry mine – the fundamental key for geometallurgical AusIMM Mill Operators’ Conference, pp 241–248 (The Australasian
applications, in Proceedings The Third AusIMM International Institute of Mining and Metallurgy: Melbourne).
Geometallurgy Conference, pp 221–230 (The Australasian Institute
DeMull, T J, Spenceley, J and Hickey, P, 2001. Planning and teamwork
of Mining and Metallurgy: Melbourne).
lead to successful start-up at Batu Hijau, presented to SME
Beniscelli, J, 2011. Geometallurgy – fifteen years of developments Annual Meeting, Denver, 26–28 February, SME Preprint 01–159.
in Codelco: Pedro Carrasco contributions, in Proceedings The
Dobbe, R, Bajic, S, Kalstrom, M, Palko, A, Martin, C, Thorpe, R, Lamson,
First AusIMM International Geometallurgy Conference, pp 3–7 (The
G and Baron, J, 2014. On site automated mineralogical analysis at
the Mount Polley porphyry copper mine, Canada, presented to
Bennett, D, Crnkovic, I and Walker, P, 2011. Recent process XXVII International Mineral Processing Congress, Santiago.
developments at the Phu Kham copper-gold concentrator, Laos,
Dominy, S C and O’Connor, L, 2016. Geometallurgy – beyond
in Proceedings 11th Mill Operators’ Conference 2012, pp 257–272 (The
conception, in Proceedings Third AusIMM International
Geometallurgy Conference, pp 3–10 (The Australasian Institute of
Bergerman, M G, Morais, J, Castro Rosa, M A N and Delboni Jr, H, Mining and Metallurgy: Melbourne).
2008. Variability studies for the Sossego 41 ktpd grinding circuit,
Downs, R C, 1990. Geological mapping and logging at Hellyer, in
in Proceedings International Mineral Processing Conference (ed: W D
Proceedings Mine Geologists’ Conference, pp 1–7 (The Australasian
Duo and S C Yao) volume 1, pp 407–416 (Science Press: Beijing).
Institute of Mining and Metallurgy: Melbourne).
Blanchard, R and Hall, G, 1939. Discussion – the mineralogy of the
Dunham, D and Vann, J, 2007. Geometallurgy, geostatistics and
Black Star ore body and its relation to milling practice at Mount
Isa, The AusIMM Proceedings, 116:487–490. project value – does your block model tell you what you need
to know?, in Proceedings Project Evaluation 2007, pp 189–196 (The
Blatter, P, Cayouette, J and Cousin, P, 2011. Agnico-Eagle Laronde Australasian Institute of Mining and Metallurgy: Melbourne).
plant: metallurgical challenges present and future, in Proceedings
43rd Annual Canadian Mineral Processors Conference, pp 3–15 England, J K, Kilgour, I and Kanau, J L, 1991. Processing copper-gold
(Canadian Institute of Mining, Metallurgy and Petroleum: ore at Ok Tedi, in Proceedings PNG Geology, Exploration and Mining
Montreal). Conference, pp 183–190 (The Australasian Institute of Mining and
Boisvert, J, Rossi, M and Ehrig, K, 2013. Geometallurgical modelling
at Olympic Dam Mine, South Australia, Mathematical Geosciences, Evans, C L, Wightman, E M, Manlapig, E V and Coulter, B L, 2011.
45:901–925. Application of process mineralogy as a tool in sustainable
processing, Minerals Engineering, 24(12):1242–1248.
Boyce, J H, Venter, W J and Adam, J, 1970. Beneficiation practice at
Tsumeb concentrator, in Lead and Zinc (eds: D O Rausch and B Fennel, M, Guevara, J, Canchaya, S, Velarde, G, Baum, W and Gottlieb,
C Mariacher), vol 1, pp 542–570 (American Institute of Mining, P, 2005. QEMSCAN Mineral Analysis for Ore Characterisation
Metallurgical, and Petroleum Engineers: New York). and Plant Support at Cerro Verde, in XXVII Convencion Minera,
Arequipa, Peru, pp 1-11 (Instituto de Ingenieros de Minas del
Bradshaw, D J, Wightman, E M, Evans, C L and Triffett, B, 2012. Perú: Lima)
Characterising the reasons for the difference in performance of
two ore blends in early processing at OZ Minerals Prominent Hill Fonseca, E and Sá, L R B M, 2005. Methodological evaluation of a
operation, in Proceedings 11th AusIMM Mill Operators’ Conference liberation study in sulfide copper ore, in Proceedings Centenary
2012, pp 373–278 (The Australasian Institute of Mining and of Flotation Symposium, pp 953–958 (The Australasian Institute of
Metallurgy: Melbourne). Mining and Metallurgy: Melbourne).
Brissette, M and Roman, E, 2012. Concentrate grade and regrind size Gottlieb, P, Adair, B J and Wilkie, G J, 1994. QEM*SEM liberation
improvement on the east pit hypogene ore at Kemess Mine, in indices for grinding classification and flotation, in Proceedings Fifth
Proceedings 44th Annual Canadian Mineral Processors Conference, Mill Operators’ Conference, pp 5–14 (The Australasian Institute of
pp 27–38 (Canadian Institute of Mining, Metallurgy and Mining and Metallurgy: Melbourne).
Petroleum: Montreal). Gregory, M J, Lang, J R, Gilbert, S and Hoal, K O, 2013. Geometallurgy
Carrasco, P, Campos, M, Tapia, J and Menichetti, E, 2005. of the Pebble porphyry copper-gold-molybdenum deposit,
Heterogeneity and Ingamells’s tests of some Chilean porphyry Alaska: implications for gold distribution and paragenesis,
ores, in Proceedings Second World Conference on Sampling and Economic Geology, 108:463–482.
Blending, pp 139–150 (The Australasian Institute of Mining and Gregory, P W, Hartley, J S and Wills, J A, 1987. The Thalanga
Metallurgy: Melbourne). massive sulphide deposit, in Proceedings Pacific Rim Congress 87,
Clarke, D, Harbort, G, Morgan, D and Sola, C, 2014. Northparkes pp 573–577 (The Australasian Institute of Mining and Metallurgy:
Mines – flotation and planning for the future, in Proceedings 12th Melbourne).
AusIMM Mill Operators’ Conference, pp 203–208 (The Australasian Greig, D, Widenbar, L, Oliver, R, Harbort, G, Cuttriss, C, Kerr, T F
Institute of Mining and Metallurgy: Melbourne). and Cummings, R, 2009. Mineral processing and metallurgical
Clemmer, J B, 1956. Metallurgical planning for porphyry copper ores, testwork (Section 16), in Constancia Copper Project, Definitive
presented to Symposium on Greater Cooperation for Metallurgical Feasibility Study technical report NI 43–101, Rev 0, 28 September.
Planning; Second Annual Rocky Mountain Minerals Conference, Harbort, G, Cordingley, G and Phillips, M, 2011. The integration
Salt Lake City, October 1955, Mining Engineering, July, pp 701–702. of geometallurgy with plant design, in Proceedings MetPlant
Comi, T, Garcia, C, Kaleta, R and Kurzydło, P, 2013. Sierra Gorda 2011, pp 339–363 (The Australasian Institute of Mining and
– the most modern Cu-Mo mine in Chile, in Proceedings Mineral Metallurgy: Melbourne).
Engineering Conference MEC2013 (ed: P B Kowalczuk), Swieradow- Harbort, G, Lauder, D, Miranda, J and Murphy, A, 2000. Size by size
Zdroj, 16–19 September, pp 109–121 (Wroclaw University of analysis of operating characteristics of Jameson Cell cleaners at
Technology: Wroclaw). the Bajo de Alumbrera copper/gold concentrator, in Proceedings
Compan, G, Pizarro, E and Videla, A, 2015. Geometallurgical model Seventh Mill Operators’ Conference, pp 207–220 (The Australasian
of a copper sulphide mine for long-term planning, Journal of The Institute of Mining and Metallurgy: Melbourne).
Southern African Institute of Mining and Metallurgy, 115:549–556. Hartog, J, Beehan, V, Karageorgos, J and Beeson, D, 2014. Optimisation
Coward, S, Vann, J, Dunham, S and Stewart, M, 2009. The primary- of the Peak Gold Mine’s processing plant through advanced
response framework for geometallurgical variables, in Proceedings process control, in Proceedings 12th AusIMM Mill Operators’
Seventh International Mining Geology Conference, pp 109–113 (The Conference, pp 335–344 (The Australasian Institute of Mining and
Australasian Institute of Mining and Metallurgy: Melbourne). Metallurgy: Melbourne).

Hoffman, C W, 1964. A review of developments in flotation treatment Kuhar, L L, McFarlane, A J, Chapman, N M, Meakin, R L, Martin,
of Mount Isa chalcopyrite ores, AusIMM Proceedings, 103. R, Turner, N L and Robinson, D J, 2013. Calibration and testing
Hunt, J, Berry, R and Bradshaw, D, 2011. Characterising chalcopyrite of a geometallurgical leaching protocol for determining copper
liberation and flotation potential: examples from an IOCG mineralogical deportment, in Proceedings Second AusIMM
deposit, Minerals Engineering, 24(12):1271–1276. International Geometallurgy Conference, pp 177–186 (The
Jackson, B R, Gottlieb, P and Sutherland, D N, 1988. A method for
measuring and comparing the mineral grain sizes of ores from Lane, G S and Richmond, G D, 1993. Improving fine particle flotation
different origins, in Proceedings Third Mill Operators’ Conference, selectivity at Hellyer, in Proceedings XVIII International Mineral
p 61 (The Australasian Institute of Mining and Metallurgy: Processing Congress, pp 897–904 (The Australasian Institute of
Melbourne). Mining and Metallurgy: Melbourne).
James, B and Scamardella, M, 2000. The Re-opening of CSA Mine, Leggett, A R and Morin, G R, 2013. Maximising concentrate revenues
Cobar NSW, in Proceedings Seventh Mill Operators’ Conference, at the Xstrata Copper Kidd concentrator, in Proceedings 45th
pp 137-147 (The Australasian Institute of Mining and Metallurgy: Annual Canadian Mineral Processors Operators’ Conference, pp 465–
Melbourne). 478 (Canadian Institute of Mining, Metallurgy and Petroleum:
Montreal).
Johnson, N W and Munro, P D, 2008. Methods for assigning domains
in the primary sulphide zone of a sulphide orebody, in Proceedings Liebezeit, V, Smith, M, Kittler, P, Macmillan, E and Lower, C, 2011.
Ninth International Congress for Applied Mineralogy, pp 597–603 (The Geometallurgy data management – a significant consideration, in
Australasian Institute of Mining and Metallurgy: Melbourne). Proceedings First AusIMM International Geometallurgy Conference,
Jones, K and Morgan, D, 2016. On-site process mineralogy – linking
Melbourne).
departments and disciplines with the common goal of increasing
plant performance, in Proceedings 13th AusIMM Mill Operators’ Lishchuk, V, Koch, P-H, Lund, C and Lamberg, P, 2015. The
Conference, pp 149–156 (The Australasian Institute of Mining and geometallurgical framework: Malmberget and Mikheevskoye
Metallurgy: Melbourne). case studies, Mining Science, 22:57−66.
Jones, M P, 1984. Recent developments in the rapid collection of Loidl, G, 2012. The Elura orebody: a multidisciplinary study
quantitative mineral data, in Proceedings Second International investigating geochemical mineralogical and geometallurgical
Congress on Applied Mineralogy, pp 449–464 (Metallurgical Society ore characteristics, PhD thesis, University of Adelaide, School of
of AIME: Warrendale, Pa). Earth and Environmental Sciences.
Kahn, H, Antoniassi, J, Shimizu, V and Uliana, D, 2014. X-ray Lotter, N O, Oliveira, J F, Hannaford, A L, Amos, S R and Broughton,
diffraction cluster analysis and automated mineralogy of Sossego D W, 2013. Flowsheet development for the hypogene geomet
copper ore, Brazil, presented to XXVII International Mineral unit in the Ivanplats Kamoa copper project, in Proceedings
Processing Congress, Santiago. 45th Annual Canadian Mineral Processors Operators’ Conference,
Keran, V P, Zumwalt, F and Palmers, J, 1998. Design aspects of the pp 71–85 (Canadian Institute of Mining, Metallurgy and
Minera Alumbrera concentrator circuit, presented to SME Annual Petroleum: Montreal).
Meeting, Orlando, 9–11 March. Lund, C and Lamberg, P, 2014. Geometallurgy – a tool for better
Khosrowshahi, S, Shaw, W J and McKevitt, M B, 2009. Ore block resource efficiency, European Geologist, 37:39–43.
optimisation for open pit grade control, in Proceedings Seventh Macfarlane, A S and Williams, T P, 2014. Optimizing value on a copper
International Mining Geology Conference, pp 141–146 (The mine by adopting a geometallurgical solution, The Journal of The
Australasian Institute of Mining and Metallurgy: Melbourne). Southern African Institute of Mining and Metallurgy, 114:929–936.
Kildale, M B, 1956. Metallurgical planning in connection with lead- MacPherson, A R, 1976. A simple method to predict the autogenous
zinc or lead-zinc-copper mines, presented to Symposium on grinding mill requirements for processing ore from a new deposit,
Greater Cooperation for Metallurgical Planning; Second Annual Presented to SME-AIME Fall Meeting, Denver, September.
Rocky Mountain Minerals Conference, Salt Lake City, October
1955, in Mining Engineering, July, pp 706–707. Magallanes-Hernández, L and Espinosa-Gómez, R, 2005.
Mineralogical and metallurgical results at Peñoles’s mining
Kilgariff, B, 2003. Thalanga processing plant – as many lives as a cat
operations, Mexico, in Proceedings Centenary of Flotation
and just as curious, in Proceedings Eighth AusIMM Mill Operators’
Symposium, pp 887–894 (The Australasian Institute of Mining and
Conference, pp 129–141 (The Australasian Institute of Mining
Martin, H K, 1957. Milling practice at the Lavender pit concentrator,
Kilgour, I, 1995. Development of the Escondida Concentrator 1990–
Mining Engineering, November, pp 1229–1235.
1995, in Proceedings XIX IMPC, pp 145–148 (Society for Mining,
Metallurgy and Exploration: Littleton). Matar, J A, Perucca, J C, de Salmuni, N S, Berenguer, T and Carrizo, M,
Kirkland, A G, 1956. Mining engineer’s view of cooperation, presented 1986. Geological research as a basis to define metallurgical
to Symposium on Greater Cooperation for Metallurgical Planning; behaviour of the disseminated Cu-Au orebody of the Bajo la
Second Annual Rocky Mountain Minerals Conference, Salt Lake Alumbrera, presented to SME Annual Meeting, New Orleans,
City, October 1955, Mining Engineering, July, pp 707–708. 2–6 March.
Kittler, P, Liebezeit, V, Ehrig, K, Macmillan, E and Lower, C, 2011. McCaffery, K and Wirfiyata, F, 2011. Applied Geo-metallurgical
‘It Seemed like a good idea at the time …’ common mistakes characterisation for life of mine throughput prediction at Batu
in geometallurgy, in Proceedings First AusIMM International Hijau, presented to Fifth International Conference on Autogenous
Geometallurgy Conference, pp 133–138 (The Australasian Institute and Semiautogenous Grinding Technology, Vancouver,
of Mining and Metallurgy: Melbourne). 25–29 September.
Klohn, S, Stephenson, D and Granados, H, 2016. Constancia project McKay, N, Vann, J, Ware, W, Morley, C and Podkiewicz, P, 2016.
process plant design and start up, in Proceedings 48th Canadian Strategic and tactical geometallurgy – a systematic process to
Mineral Processors Operators’ Conference, Ottawa, January, add and sustain resource value, in Proceedings Third AusIMM
pp 100–112 (Canadian Institute of Mining, Metallurgy and International Geometallurgy Conference, pp 29–36 (The Australasian
Petroleum: Montreal). Institute of Mining and Metallurgy: Melbourne).
Kormos, L, Oliveira, J, Fragomeni, D, Whiteman, E and De la Cruz, McQuiston Jr, F W, 1968. Metallurgist’s view on milling base
J C, 2010. Process diagnosis using quantitative mineralogy, in metal ores, presented to Symposium on Greater Cooperation
Proceedings 42nd Meeting of Canadian Mineral Processors, pp 97– for Metallurgical Planning; Second Annual Rocky Mountain
119 (Canadian Institute of Mining, Metallurgy and Petroleum: Minerals Conference, Salt Lake City, October 1955, Mining
Montreal) Engineering, July, pp 698–700.

G Harbort et al
McQuiston Jr, F W and Bechaud Jr, J, 1956. Metallurgical sampling Reid, A F, Gottlieb, P, MacDonald, K J and Miller, P R, 1984. QEM*SEM
and testing, in Chapter 3.2 – Mine Development, in Surface Mining image analysis of ore minerals: volume fraction, liberation and
(ed: E P Pfeider), pp 103–121 (The American Institute of Mining, observational variances, in Proceedings Second International
Metallurgical, and Petroleum Engineers: New York). Congress on Applied Mineralogy in the Minerals Industry, pp 191–204
(Metallurgical Society of AIME: Warrendale, Pa).
Mitchell, J and Holowachuk, N V, 1996. Ore blending at Highland
Valley Copper, its necessity, effects, and benefits, in Proceedings Reid, A F and Wittenberg, J C, 1984. Technical note – rapid production
SAG 1996, vol 1, pp 1–24 (Canadian Institute of Mining, of high quality polished sections for automated image analysis of
Metallurgy and Petroleum: Montreal). minerals, The AusIMM Proceedings, 289:93–97.
Morey, M and Cantrell, R, 2011. Ore ageing test work for the Ok Tedi Reid, A F and Zuiderwyk, M A, 1983. QEM*SEM: automated image
skarns, in Proceedings MetPlant 2011, pp 615–629 (The Australasian analysis and stereological applications to mineral processing and
Institute of Mining and Metallurgy: Melbourne). ore characterization, Acta Sterol 1983; 2/Suppl I, in Proceedings
Sixth International Congress, Gainesville, 9–14 October, pp 205–208.
Muñoz, A, Álvarez, L, Colacioppo, J and Valery, W, 2008. Process
Reno Sales, R H, 1956. Full cooperation between departments is a
integration and optimisation at Freeport-McMoran Candelaria
must policy for mining enterprises, presented to Symposium on
mine, Copiapo, Chile, in Proceedings PROCEMIN 2008, Santiago,
Greater Cooperation for Metallurgical Planning; Second Annual
pp 303–315 (Gecamin Digital Publications: Santiago).
Rocky Mountain Minerals Conference, Salt Lake City, October
Mwanga, A, Rosenkranz, J and Lamberg, P, 2015. Testing of ore 1955, Mining Engineering, July, pp 709–711.
comminution behavior in the geometallurgical context – a review, Richard, K, 1956. Geologist-metallurgist cooperation in porphyry
Minerals, 5:276–297. copper exploration, presented to Symposium on Greater
Nzama, S C and Kapo, M K, 2014. Managing the giant – a mine Cooperation for Metallurgical Planning; Second Annual Rocky
geologist’s perspective to the mining of the Olympic Dam Mountain Minerals Conference, Salt Lake City, October 1955,
deposit, in Proceedings Ninth International Mining Geology Mining Engineering, July, pp 703–705.
Conference, pp 451–459 (The Australasian Institute of Mining Richmond, G D and Campbell, J J, 1992. Innovation at the Hellyer
and Metallurgy: Melbourne). Concentrator, in Proceedings AuslMM Annual Conference,
O’Malley, G B and McGhie, R R, 1939. The mineralogy of the Black pp 191–196 (The Australasian Institute of Mining and Metallurgy:
Star ore body and its relation to milling practice at Mount Isa, The Melbourne).
AusIMM Proceedings, 116:459–486. Richmond, G D and Lai, K F, 1988. Metallurgical development of
the Hellyer Ore, in Proceedings Third AusIMM Mill Operators’
O’Meara, A E, 1953. New microscope laboratory at Mount Isa, The
Conference, pp 9–14 (The Australasian Institute of Mining and
AusIMM Proceedings, 171:219–229.
Oosthuyzen, E J, 1983. The Application of Automatic Image Analysis
Ross, J, Appleby, S K, Hoal, K and Botha, P, 2009. Quantitative
to Mineralogy and Extractive Metallurgy, special publication 7, mineralogical study of ore domains at Bingham Canyon, Utah,
pp 449–464 (Geological Society of South Africa). USA, presented to SME Annual Meeting, Denver, 22–25 February.
Palagi, C G and Stillar, S S, 1976. The Anaconda C. E. Weed concentrator, Saquet, J J, Apraxine, V, Lakaye, J and Troch, P, 1962. Kolwezi mining
in Flotation A M Gaudin Memorial Volume (ed: M C Fuerstenau), and metallurgical operations of Union Miniere du Haut-Katanga,
second edition, pp 1029–1042 (Society for Mining, Metallurgy & Mining Engineering, December, pp 70–81.
Exploration – American Institute of Mining Engineers: New York).
Siddiqui, F, 2000. On-site mineralogy – SIDMIN program, in
Pangum, L S, Lauder, W, Manlapig, E V, Mackinnon, I R and Proceedings 32nd Annual Meeting of Canadian Mineral Processors,
Glatthaar, J W, 1998. Mineralogical and metallurgical examination Ottawa, 18–20 January, pp 459–475 (Canadian Institute of Mining,
of fluorosilicate mineral flotation in the Ok Tedi Circuit, in Metallurgy and Petroleum: Montreal).
Proceedings Mine to Mill 1998, pp 19–28 (The Australasian Institute
Snider, J W, Ibrado, A S, Hester, M G, Earnest, D, Aronson, J G,
Roman, R J, Rehn, C C and Harbort, G J, 2013. Mineral processing
Petruk, W, 1984. Image analysis measurements and data presentation and metallurgical testing, in King-king copper-gold project
for mineral dressing applications, in Proceedings Second preliminary feasibility study, NI 43-101 technical report, Rev 0,
International Congress on Applied Mineralogy, Los Angeles, 25 February, pp 90–104.
22–25 February, pp 127–140 (Metallurgical Society of AIME: Sola, C and Harbort, G, 2012. Geometallurgy – tricks, traps and
Warrendale, Pa). treasures, in Proceedings 11th AusIMM Mill Operators’ Conference,
Petruk, W and Hughson, M R, 1977. Image analysis evaluation of the pp 187–196 (The Australasian Institute of Mining and Metallurgy:
effect of grinding media on selective flotation of two zinc-lead- Melbourne).
copper ores, ClM Bulletin, June, pp 128–135. Strohmayr, S J, Barns, K E, Brindley, S K and Munro, P D, 1998.
Petruk, W and Schnarr, J R, 1981. An evaluation of the recovery of Mineralogy controlling metallurgy at Ernest Henry Mining, in
free and unliberated mineral grains, metals and trace elements Proceedings Mine to Mill 1988, pp 13–18 (The Australasian Institute
in the concentrator at Brunswick Mining and Smelting Corp Ltd,
CIM Bulletin, 74(833):132–159. Suazo, C J, Kracht, W and Alruiz, O M, 2010. Geometallurgical
modelling of the Collahuasi flotation circuit, Minerals Engineering,
Petruk, W and Sikka, D B, 1987. The formation of oxidized copper
23:137–142.
minerals at the Malanjkhand porphyry copper deposit in India,
and implications on metallurgy, in Process Mineralogy VII (eds: Taylor, B, 2011. The Perseverance gold copper deposit – a new Peak
A H Vassilou, D M Hausen and D J T Carson), pp 403–420 (The Mines challenge, in Proceedings Eighth International Mining Geology
Minerals, Metals and Materials Society: Warrendale). Conference, pp 367–378 (The Australasian Institute of Mining and
Quigley, P, Hackhan, D and Broome, J, 2008. Sepon Copper Mineral
Resource statement, in OZ Minerals Limited Mineral Resources Tew, A, Strohmayr, S, See, A and Harvey, R, 2003. Mineralogy
explanatory notes, 30 June, pp 8–26. controlling metallurgy at Ernest Henry Mining revisited, in
Proceedings Eighth AusIMM Mill Operators’ Conference, pp 211–
Quiñones, S and Mattos, R, 2001. Metallurgical control through 217 (The Australasian Institute of Mining and Metallurgy:
mineralogy in the Toquepala concentrating plant, XXV Convencion Melbourne).
de Ingenieros de Minas – Arequipa, Septembre.
Tverdov, A and Nikishichev, S, 2014. Mine design stages in Russia, in
Ramsey, T, 1976. The Cyprus Pima concentrator, in Proceedings A M Mineral Resource and Ore Reserve Estimation – The AusIMM Guide
Gaudin Memorial Symposium, pp 1079–1099 (American Institute of to Good Practice, second edition, pp 815–824 (The Australasian
Mining, Metallurgical, and Petroleum Engineers: New York). Institute of Mining and Metallurgy: Melbourne).

Umipig, C V, Israel, E E, Hutalle, G G and Williams, S R, 2012. Williams, S R, 2013. A historical perspective of the application and
Canatuan Cu/Zn flotation metallurgy – dealing with zinc pre- success of geometallurgical methodologies, in Proceedings The
activation, in Proceedings 44th Annual Canadian Mineral Processors Second AusIMM International Geometallurgy Conference, pp 37–47
Operators’ Conference, pp 155–168 (Canadian Institute of Mining, (The Australasian Institute of Mining and Metallurgy: Melbourne).
Metallurgy and Petroleum: Montreal).
Wong, K Y, Breem, A M, Doherty, R J and Phelan, J M, 1991. Flotation
Urbanoski, L S, 1993. Road to recovery – metallurgical improvements treatment of supergene complex sulphide ores at Thalanga, in
at Brunswick Mining and Smelting Corporation Limited, in Proceedings Fourth AusIMM Mill Operators’ Conference, pp 205–
Proceedings International Symposium World Zinc’93, pp 205–212 (The 212 (The Australasian Institute of Mining and Metallurgy:
Australasian Institute of Mining and Metallurgy: Melbourne). Melbourne).
Walters, S G, 2008. An overview of new integrated geometallurgical Wraith Jr, W and Fulmor, T G, 1964. Anaconda’s Butte Concentrator,
research, in Proceedings Ninth International Congress for Applied Mining Engineering, May, pp 55–73.
Mineralogy, pp 79–82 (The Australasian Institute of Mining and
Yeatman, P, 1932. Problems at the Chuquicamata and Braden copper
properties, in Choice of Methods in Mining and Metallurgy: a Record
White, J and Adair, R B, 1948. Mufulira Copper Mines Limited, of Experience in Making Engineering Decisions, pp 1–34 (McGraw-
Concentrator, Northern Rhodesia, presented to AIME New York Hill: New York; London: published for American Institute of
Meeting, February 1948, technical publication no 2250, Class B, Mining and Metallurgical Engineers).
Mining Technology, January (American Institute of Mining and
Young, K A, 1983, Metallurgical aspects of mining – project
Metallurgical Engineers: New York).
development, in Proceedings AusIMM Sydney Branch, Project
Development Symposium, pp 221–228 (The Australasian Institute
of Mining and Metallurgy: Sydney Branch).

Contents
Integrated mining and metallurgical

planning and operation
P L McCarthy1
ABSTRACT
To design and operate a successful mine the technical specialists must agree on what
is ore, what is not ore, and how quickly the orebody is to be mined and processed.
These seemingly simple questions defy easy analysis and can push mining engineers,
geologists, metallurgists and others to the limit of their capabilities. The questions
are rarely addressed adequately in feasibility studies or in operational planning. To
optimise the mine and processing plant requires proper determination of cut-off
grades or values, prediction of feed variability on an hourly, daily or longer basis,
right-sizing the mining rate for optimum product quality, understanding the cost of
capital and the expectations of investors, and many other considerations that involve
geology, geotechnical engineering, mining, metallurgy, environmental effects and
economics. This paper addresses the key matters that should be considered as inputs
to the practical issues of metallurgical plant design and operation.
INTRODUCTION
The mine planning process includes the mining, processing, infrastructure,
environmental and social aspects of an operation. It begins with a sequence of
increasingly detailed feasibility studies and continues throughout the life of the
mine through several long-term and short-term planning processes. These processes
all require the selection of a value descriptor that best reflects the value that can be
obtained from the material, which may be expressed as a metal grade, dollar value or
other measure. At various points in the sequence of mining and processing activities
there is a need to choose a particular value of the value descriptor, called a cut-off,
to enable decisions to be made about what to do next with raw or semi-processed
material. Table 1 of The Australasian Code for Reporting of Exploration Results,
Mineral Resources and Ore Reserves (‘the JORC Code’, 2012) provides a useful
checklist of the many considerations that determine whether in situ mineralisation
that is above the mining cut-off is ore (and should be processed) or is not.
Selection of appropriate cut-offs is an outcome of the mine planning and
optimisation process. While approximate cut-off grades may be used early in the
resource estimation process, these are superseded by better, value maximising cut-off
grades as planning progresses. The process is iterative and cut-offs are influenced,
among other things, by the chosen mining and processing rate. The two variables,
cut‑off and mining processing rate are the primary levers that can be used by planners
to optimise the mining and processing plan.
The selected mining rate determines how much of the above cut-off material can
practically be recovered, how much below cut-off material is included in the feed sent
to the processing plant, and the quality and variability of that feed on various time
scales. Slow mining allows careful selectivity; fast mining generally reduces recovery
and increases dilution.
MINING CUT-OFF
The optimum rate of mining is closely tied to the cut-off grade chosen. Hall and Hall
(2006) observe that:
…the major parameters that a mining company can make independent decisions about
are typically the mining method(s), mining sequencing, production rate, and cut-off
grade (or ‘cut-off’). Since the size and shape of the orebody and hence possible mining
methods and the range of feasible production rates may vary significantly with cut-off,
it is often the cut-off that is the key driver of value of the operation.
1. HonFAusIMM(CP), Chairman, Principal
Mining Consultant, AMC Consultants Pty Ltd, This is illustrated in Figure 1, which shows that a smaller operation optimised
Melbourne Vic 3000. using a higher cut-off grade is more robust than an operation optimised using high
Email: pmccarthy@amcconsultants.com price assumptions.
55
P L McCarthy
mine more carefully and selectively. While quick progress

can be made in a single year, longer term averages should be
considered when sizing the processing plant.
The floor of an open pit can be advanced quickly but, as
a rule of thumb, the open pit production limit is around
eight benches per annum, or 80 m vertical advance using
ten metre benches. Though some very large pits and small,
short-life gold pits can achieve over 100 m per annum within
the operating area. However, an average sink rate of 40–60 m
per annum is more likely for longer term planning in most
pits.
The vertical rate of advance in a pit is influenced by the:
•• number of activities that are included in the mining
FIG 1 – Risks and rewards of optimum cut-offs (Hall and Hall, 2006). cycle, eg grade control, presplitting, drill and blast,
loading
CONSIDERATIONS ON MINING AND PROCESSING RATE •• rate at which each activity can be carried out, as
A design rate of mining and processing is selected in every determined by the size and number of equipment
mine feasibility study, although any attempt to optimise that •• total available floor area
rate is rarely documented. To maximise return on investment, •• bench height, as this determines the vertical advance per
it has long been recognised that both the capital investment mining cycle
per unit of output and the operating cost per unit of output
•• need to develop a drop cut to establish the bench, or a
should be minimised. In general, both cost measures decrease
sump for dewatering, as is the case in mining in the base
as the scale of the project increases, so the initial temptation is
of pits compared to a mining a cut-back.
to ‘push the orebody to the limit’.
Once the vertical advance rate is established, the average
However, the technical and commercial risk both increase as
production rate can be determined from the available tonnes
the scale of the project increases. Hoover (1909) states that the
of ore per vertical metre (t/vm) within the pit design.
lower the production rate, the lower the required investment,
the longer the income stream and the lower the risk to the Similar considerations apply to underground mines. The
investor. While this was well before the advent of discounted production rate from an underground mine is not usually
cash flow (DCF) analysis, the point made by Hoover remains limited by the rate at which ramp or decline development can
a good one. be advanced, as a decline face can typically be advanced at
40 m per week on a 1-in-7 (14.3 per cent) grade, which is a
Until the last third of the twentieth century, most mine
vertical advance rate of 297 m/a. Rather, the production rate
developers did not have ready access to project capital and
is limited by the number of available working faces which
so they had to develop projects using mainly a combination
in turn depends on the amount of predevelopment, possible
of new shareholders’ funds and retained earnings. Available
rates of ongoing lateral development, infill drilling, stope
capital was a key consideration in sizing a new project. More
turnaround times, backfilling and so on, with interference
recently, and particularly during the ‘mining boom’, there
between these activities.
has been an assumption that any scale of project will attract
project finance if it satisfies the hurdles set by bankers. Capital In steeply dipping deposits underground mining can occur
constraints were not commonly included in the project on several levels simultaneously, but the mine production
optimisation process. Bigger projects were generally thought rate can still be related to the ore t/vm that will be mined. This
to be better. In reality, the interests of existing shareholder relationship can be expressed as the ‘effective vertical advance
owners of a mineral deposit may be best served by a modest rate’, or the relationship between actual mining rate and the
scale of development, with restricted use of external capital. t/vm available in the deposit. For example, a 1 000 000 t/a mine
The value of a smaller project as measured by net present with an average 20 000 t/vm would have an effective vertical
value (NPV) may be lower, but the risk-adjusted value to advance rate of 1 000 000/20 000 or 50 vertical metres per
existing shareholders may be greater. annum (vm/a).
The uncertainties in mining investment are many. Mineral Work by the author and others has shown that the risk of
prices are cyclical and to a large extent unpredictable. Over failure increases as the vm/a increases. A century ago, a rate
the life of a mine, these prices usually fall in real terms. The equivalent to one level per annum or approximately 30 vm/a
mineral resources being mined are finite and can be highly was considered a reasonable basis for planning. Tatman
variable in both size and quality, while ground conditions can (2001) was able to derive an empirical formula relating the
vary significantly with depth and location. Costs are difficult risk of failure to the geometry of the deposit and the rate of
mining and to conclude that for modern mines in steeply
to predict and subject to periods of rapid escalation. These
dipping tabular deposits thicker than 10 m, risks were:
uncertainties tend to favour a more modest capital investment
and hence exposure to risk, provided the expected product <30 vm/a low risk
cost falls within an acceptable range. 30–70 vm/a moderate risk
>70 vm/a high risk.
MINING RATE LIMITATIONS In a 2014 confidential study of 12 current Australian mines
There are physical limits to the rate that any orebody can using sublevel open stoping methods, the author found that
be mined. High rates of mining are associated with greater average vertical advance rates varied from 23–71 vm/a with
day-to-day or month to month production volatility, with a an average of 43 vm/a, while the maximum single-year rates
tendency for dilution of the ore to become excessive at high varied from 31–79 vm/a with an average of 56 vm/a. Only one
rates as the pressure of production reduces the ability to mine sustained a rate higher than 61 vm/a. As a generalisation,

Integrated mining and metallurgical planning and operation
special circumstances are required for any underground mine ORE STOCKPILES
to sustain a rate above 60 vm/a. Nineteenth-century mines used hand-pushed rail trucks,
These observations describe the limits that might apply typically of around 400 kg capacity, to move ore from chutes to
to mining rates for open pit and underground mines. the shaft and up to surface. Each truck could be marked with
The optimum rate can only be determined after detailed chalk as to its source. Stamp batteries were fed separately in
scheduling of alternative mining plans and the completion of groups of five stamp heads. It was possible to blend ore from
an optimisation study that balances revenue against capital different sources to achieve a steady quality in each stamper
and operating costs for the entire mining and processing box. Alternatively and more commonly, parts of the mill were
operation. tuned to run on ore from different stopes or orebodies. In a
gold mill, high-arsenic ores might be run over strakes and
MINING RATE AND HEAD GRADE gravity tables into barrels, while clean ores were first mercury
amalgamated on copper plates. Thus the mine and mill were
The assumption that ‘economies of scale’ will result from closely aligned, with feedback on ore grade being given to the
increasing throughput rates needs to be balanced by an miners for individual stopes or areas. The many ore trucks
awareness of the adverse effects of increasing the rate beyond formed a stockpile which decoupled the stopes from the
a level that is supportable by the resource. For each scale of hoisting and processing system. Old illustrations show trucks
operation considered, it is a reality that for any intended head lined up at the plat, waiting to be hoisted, with empties ready
grade, at the associated intended cut-off grade, the actual to be refilled (Figure 2).
head grade achieved will fall as the mining rate increases. This
By the mid-twentieth century orepasses acted as stockpiles
effect is known to people at operations but is not generally
at each stope to the haulage level, and from there, in parallel
recognised in current ore reserve estimation methodology.
with the hoisting shaft, down to a skip-loading station. This
Once recognised, this dependence of head grade on mining
allowed for a great deal of blending of ore from different
rate can be quantified and used to establish the economically
sources, while decoupling the mining activity from hoisting
optimum mining and processing rate (McCarthy, 2010).
and processing. The ability to reconcile stope grades and to
run parallel lines through the mill had been lost, but cost
MINING FLEXIBILITY savings and greater throughput were achieved.
It is unfortunately true that a mining operation presents Today many mines are designed without significant
challenges on every time scale from daily to annually. The stockpiles. Block caving mines in particular may have no
failure of many ‘mine to mill’ studies to deliver promised effective stockpile capacity between the drawpoint and
improvements is largely because the mining engineers the surface stockpile. Even the surface stockpile may be
are fully absorbed in meeting existing challenges and have eliminated in normal operation, with the inclined conveyor
limited capacity to vary the method and sequence of mining from the underground mine delivering crushed ore directly
to deliver a better or more predicable product to the process to the secondary crusher. In these circumstances the miners
plant. Monthly, quarterly or annual mining schedules are have no ability to manage product quality. In other operations
based on the ore reserve model, which is frequently found to such as sublevel open stoping or longhole retreat stoping
be deficient at those scales. Changes to the mining sequence without shaft hoisting, the old approach of using orepasses as
are made on the run, and the challenge for the mining engineer stockpiles has been eliminated and ore is hauled to the surface
is to deliver the scheduled tonnes, of any quality, above the in trucks. This provides an opportunity to manage ore quality
mining cut-off. by blending using two or more dump points at the run-of-
mine (ROM) stockpile (or dump at a low-grade stockpile for
Some mining methods, such as sublevel and block caving,
future processing). Management of ore quality then becomes
allow of no short-term control of product quality, although
the responsibility of the reclaim operator, but may require
differential draw may be possible on a timescale of years.
resampling to establish the variability and location of material
Provided it is above the mining cut-off, in the short-term the
within the stockpiles.
ore is delivered as it presents at the drawpoints. While other
methods in both open pit and underground mines can be more The more challenging the mining situation, the greater the
selective, the pressure to deliver the tonnes to schedule often stock levels need to be including developed (exposed) ore
precludes any management of product quality by scheduling stocks, drilled stocks, broken stocks and ROM pad stocks. If
a blend of material from several mining areas. Of course, the these stock levels are adequate then variability and volatility
higher the rate of mining, the less selective the mining can be.
There is evidence that more selective mining is possible if it
is designed into the process from the outset and the planned
mining rate is adjusted downward to allow it to take place.
A good example is the selective mining of acid-generating
and benign waste from a pit, with the benign waste used for
capping. This is a legislative requirement which cannot be
compromised, so it is given appropriate attention.
The best way to achieve predictable feed for the process
plant is to develop an accurate orebody model in which all
the variables of consequence are modelled faithfully, and to
set the mining rate low enough so that selective mining can
be practiced on every time scale. This requires competence in
the emerging specialty of geometallurgy, a healthy geological
budget for drilling and modelling, and an uncommon
appreciation of the benefits of mining at a rate lower than the
maximum possible. FIG 2 – Ore trucks at the plat (from Dicker’s Mining Record, 1867).

P L McCarthy
can be reduced to a minimum. The mine should be designed variability in ore quality or of the distribution in space of
so that all stockpiles, including orepasses in an underground valuable material, process contaminants, ore hardness and so
mine, have adequate capacity to smooth the short-term on. At best, the mining schedules are presented as monthly
surges to a level acceptable for the entire system, including averages for the initial few years and as quarterly or annual
ore processing. This is a commonly overlooked requirement. averages thereafter.
Some orebodies are amenable to visual grade control; others As initially constructed, the processing circuit must be
require assays on a short turnaround to allow ore selection designed to cope with or be adapted to the expected range
decisions to be made. The latter can suffer from dilution and of ore qualities, with an ability to respond quickly to any
high-grade volatility if the grade control program is not well changes. Alternatively, with a less flexible circuit, the cost
designed and given priority at the laboratory. and revenue impacts of possible variations in ore quality
The reliability of mining equipment has an effect on ore must be examined to ensure they fall within acceptable limits.
quality. Delays in mine development (accessing ore in an The impact may be greatest in the first year, when orebody
underground mine or prestripping in a pit) can lead to knowledge is weakest and cash flow is critical. As experience
periods when low-grade or high-impurity ore is all that is is gained during mining, and infill or grade-control drilling
available. Breakdowns in ore-production equipment can advances, the orebody model can be greatly improved and
lead to increased dilution because it is human nature to be the processing circuit adapted to reality.
less concerned about dilution when there is insufficient ore The planning process begins with a good 3D geological
available to feed the mill. Hence the old saying ‘waste plus (resource) model developed by an experienced geologist
ore equals more ore!’ familiar with the deposit type. Geological domains are
Proper mine design, planning, scheduling and maintenance identified, such that a common set of rules can be applied to
require good management. Ultimately, the capabilities of determine local variations in metallurgical responses within
the mine management team will determine the quality and each domain. The domain boundaries may be structural,
regularity of mill feed. mineralogical, alteration or lithological. Poor geological
modelling and domaining are the leading causes of failure
in geostatistical modelling for grade estimation and for
PROCESSING RATE modelling metallurgical parameters.
Unlike the mining rate, the processing rate is not physically
Domains should be defined beyond the ‘orebody’ to include
limited by geology. A plant of any capacity can be built,
all material that could find its way into the ore stream.
at a cost, although there are step changes in the capacity
Metallurgically, adjacent domains may have little or nothing in
of available components that make particular rates more
common. For example, the waste rock adjacent to the orebody
attractive. Sometimes environmental constraints limit the
across a sharp contact may be much harder and more abrasive
hours for activities such as stockpile reclaim and crushing
than the ore, and if the mining method will cause 25 per cent
but this can be addressed by using larger equipment. The
mining dilution of ore with waste rock then the crushing,
availability of services such as power or water may place an
grinding and wet plant performance of the waste rock must
absolute limit on the size of plant, or impose a large capital
be thoroughly understood. Similarly, dilution from nearby
cost burden for going beyond that point.
carbonaceous shales (such as a hanging wall zone) may be
Once a plant is operating, the processing rate may be limited preg-robbing or a talc-rich fault zone may impact on filtration,
by feed characteristics. For example, harder ore than expected although neither is considered part of the orebody proper.
may limit the milling rate while wet, clayey ore may limit the
Once the domains (both within and near the orebody)
crushing rate. The copper to sulfur ratio may limit throughput
have been described (as ‘wire frames’ or solid objects in a 3D
in a smelter. For these reasons plant components and the
computer model), representative samples from each domain
overall plant capacity may be oversized to some extent as
can be subjected to laboratory-scale test work to determine
compared to the selected mining rate. If this is recognised by
the rock’s response to each mineral processing operation.
management as an allowance for variable ore quality then no
Conventional geostatistics can then be used to model the
problem arises, but invariably when the ore quality is good
distribution of metallurgical responses throughout each
the higher capacity is pushed back to the mine as a demand
geological domain. This results in a model comprising a large
for a higher mining rate, with adverse consequences.
number of blocks in 3D space, each block being assigned
For some products the production rate may be limited all the geological, geotechnical, geometallurgical and other
by the market. The output of large individual producers of characteristics needed for mine planning and scheduling.
products such as iron ore may influence the market price, so The mine scheduling process can then produce from the
that the optimum production rate is less than it would be in geological block model not only a schedule of tonnes and
an unlimited market. In some circumstances there may be a grade but also a schedule of metallurgical performance and
limit to the road or rail haulage capacity for concentrate, or other characteristics such as ground support requirements or
the availability of ship loading facilities. even water inflows to the mine.
The geostatistical approach used to model metallurgical
OREBODY MODELLING performance need not be complex. Even the simplest
A resource model is initially based on exploration drilling only. approaches using the ‘rule of nearest neighbours’ (which says
Its accuracy is limited by the exploration budget, by difficulty each block in the model is likely to perform in a similar way
of access to the mineralisation, and by a lack of experience to the nearest sample) or the ‘rule of gradual change’ (which
with the deposit. This resource model is used by mining calculates a distance-weighted average of characteristics
engineers to design a mine at the feasibility study level, with based on nearby samples) is likely to provide a significant
their predictions of mining dilution and mining recovery used improvement in predictability of plant performance when
to estimate an ore reserve. The mining schedules produced compared with having no geometallurgical model. However,
in the feasibility study are used by the plant engineers and the more advanced geostatistical methods are not difficult to
metallurgists to design the plant. At this stage there may be apply and will further refine the result. Selection of the best
a very poor understanding of the hourly, daily or weekly techniques is the subject of ongoing research.

Samples for metallurgical testing are usually composited mined, stockpiled and processed separately. The premining
from diamond drill core. Hardness testing (for crushing and block grades were estimated from diamond drill assays by
grinding) typically requires 10–20 kg of sample, with some geostatistical methods using a 150 g/t Au top cut. Overall, Gill
tests requiring 100 kg, while flotation testing usually requires Reef was estimated from drilling to have a grade of 7.5 g/t Au
at least several kilograms of sample (Barratt and Doll, which reconciled (for comparable blocks) with a mine head
2008). Large metallurgical samples excavated from near the grade of 7.6 g/t Au. Individual block reconciliations, with a
surface of a deposit are unlikely to be representative of the linear best-fit line, are shown in Figure 3.
orebody at depth. Shafts sunk for the purpose of obtaining
large metallurgical samples may also yield unrepresentative It was possible to reconcile mill recovered grade against
samples, or samples that represent performance in only mine head grade for 45 blocks ranging from 2064–9799 t, with
one geological domain. Such exercises may be compared to a mean size of 4426 t. Head grades ranged from 2.2–12.6 g/t
searching for a lost wallet under a streetlight, because it is too with a weighted mean of 7.6 g/t. Reconciled metallurgical
dark to search elsewhere. recoveries ranged from 71.5–96.7 per cent with a weighted
In order to be useful in developing a geometallurgical mean of 89.1 per cent. Individual block reconciliations, with a
model, test results must satisfy the following (ibid): logarithmic best-fit line, are shown in Figure 4. A constant tail
grade model was not a good fit to the data.
•• results must reflect the properties of a ‘small’, identifiable
interval of drill core With sufficient experience in similar orebodies it should
•• the location of the interval must be identifiable in 3D have been possible to predict from the drill results that mining
space (to connect it to the block model) about 240 000 t of ore would yield a head grade of about
7.5 g/t Au and a metallurgical recovery of about 90 per cent.
•• the values being distributed through the orebody must
However, these long-term averages would be of little use
be reasonably additive, allowing unknown blocks in
in predicting or optimising process plant performance on
the model to be estimated by interpolating two or more
known samples. a daily or weekly basis at the actual mining rate from Gill
Reef of 2000–3000 t per week. Figure 5 shows the predicted
Sufficient sample material to achieve these aims may be
metallurgical recovery for each mining block (based on the
available from drill core for a large porphyry copper open pit
predicted head grade and the grade-recovery relationship)
which has a large selective mining unit (SMU or minimum
and the actual metallurgical recovery, with a weak linear best-
mining block size). For smaller, more complex deposits and
many underground mines where assays are obtained for fit relationship. It is clear that metallurgical performance was
each one-metre sample interval, the production of composite not predictable at a useful level.
samples of sufficient size for metallurgical testing may defeat While attempts could be made to improve the metallurgical
these aims and blur the modelling results. For example Barratt model by modelling other geological features such as
and Doll (2008) propose sample intervals of 15 m of HQ carbonaceous shales or associated sulfides, the innate
(63 mm) core for a JK SMC® (drop weight) test and 45–105 m geological variability makes success unlikely at the actual
of HQ core for a Bond test. It may be necessary to develop local scale and rate of mining. At much higher mining rates, or
correlations between the large-scale tests and other properties
such as point load strength, rock quality designation (RQD),
fracture frequency and mineralogy in order to obtain sufficient
data to create a meaningful 3D model.
The new techniques of geometallurgical modelling
are useful in improving process plant performance and
predictability in large orebodies such as porphyries, mined
with big equipment. In these situations zoning may allow
prediction of changes over a period of years. However, many
medium to small-scale mines may not be amenable to the
techniques being developed. The problem is that variability
of the factors affecting metallurgical performance occurs
at a scale smaller than can be sampled for metallurgical
testing, and at that scale the measured properties are not well
correlated with performance. The same problem exists for
geotechnical modelling, where very limited success has been FIG 3 – Predicted and actual block grades, Gill Reef.
achieved in predicting ground conditions and stability using
geostatistical methods.
CASE STUDY – PREDICTING THE UNPREDICTABLE IN GILL REEF

The Bendigo orebodies have been shown to suffer from an
extreme nugget effect, making prediction of grades from drill
data difficult. For this reason Gill Reef was mined, processed
and reconciled in small batches and provides a more detailed
picture of ore variability and plant performance than is
usually available. The author examined the data to see
whether metallurgical recovery could have been predicted
from the predicted head grade of each batch.
Gill Reef is a distinct quartz reef at the Kangaroo Flat
mine that was mined between 2009 and 2011. A total of
55 discrete ore blocks totalling 243 497 t from Gill Reef were FIG 4 – Head grade and metallurgical recovery, Gill Reef.

P L McCarthy
•• Surface mines suffer significantly higher variability in

ore production from budget than underground mines.
This is surprising, but reflects the ability to switch the
mining fleet from ore to low-grade to waste in reaction to
changed circumstances, including geological variations,
weather and equipment downtime. Another reason may
be that open pit schedules are less conservative than
underground schedules.
•• Despite the mining variability, processing rates in
surface mines run closer to budget, perhaps due to
having larger stockpiles.
•• Variability in mined and processed grades is similar,
suggesting that little use is made of stockpiles for
FIG 5 – Predicted and actual metallurgical recovery. blending and that grade control outcomes are similar in
the surface and underground mines studied.
when considered over longer periods, the volume-variance Another useful measure is volatility, the average percentage
relation would allow greatly improved predictability. change in a measure from one time period to the next. The
level of planned and actual volatility drives the stock
VOLATILITY AND VARIABILITY requirements and levels. When volatility is low, mining and
processing are efficient and capacity is being used effectively
Production volatility refers to the relative variation in a
with costs minimised.
parameter from one time period to the next, while variability
refers to the variation from budget or plan from one time Table 2 shows the month to month volatility from the
period to the next. The volatility of parameters such as feed benchmark study mentioned above. It can be concluded that:
tonnage, head grade, metallurgical recovery, throughput or •• ore tonnage mined is nearly twice as volatile in surface
product output can be measured hourly, daily, monthly etc. mines compared with underground mines
The more volatile the measure, the less use is being made of •• on other measures, underground mines are more volatile
the installed capacity and hence of the capital invested and than surface mines.
of the fixed component of operating cost. One of the key
Carter (2010) observes that volatility and variation in any
symptoms of a system that has been pushed beyond its stable
business process creates uncertainty, whether in determining
capacity is an increase in production volatility.
mining volumes and plant feed or in optimising maintenance
A mining project designed for a 1.0 Mt/a rate with schedules and supply chain management. The same holds for
five per cent volatility needs an installed capacity of the management of working relationships. It is this variation
1.05 Mt/a. If a decision to increase the rate by ten per cent and volatility which skews outcomes, and which can be
to 1.1 Mt/a leads to an increase in volatility to 15 per cent, reduced by implementing rigorous planning, scheduling,
then the installed capacity must be increased to 1.27 Mt/a, resourcing and execution processes, and most importantly,
an increase of 20 per cent. If the capacity is only increased by clarifying roles and accountabilities at each level.
by ten per cent, the increased volatility could lead to a slight
When volatility and variability are assessed on an hourly or
reduction in output when the mine and mill are considered
daily basis they are seen to be much greater than the monthly
together as a system.
measures in Tables 1 and 2. For example, Figure 6 shows
The specifications for the processing plant should reflect daily plant tonnes at Anglogold Ashanti’s Mponeng mine.
real hourly or daily mining outcomes, not a smoothed and The figure also shows the results of a business improvement
idealised schedule. Table 1 summarises the unpublished initiative which contributed to a 15 per cent increase in
results of benchmark studies of 44 underground mines throughput over the historical average. This is an especially
and 21 open pit mines conducted by AMC Consultants. significant result at the Mponeng plant, long regarded as the
The monthly variability shown is the average absolute flagship operation within the group. Before the initiative,
difference (as per cent of budget) between the mine budget ore from the Mponeng mine would regularly be trucked to
and production over a 12 month period. High variability neighbouring plants for processing, as the mill struggled to
indicates that a mine is not operating as intended, and that cope. After the initiative the plant had improved productivity
daily variability may be much greater than the steady hourly to the point that it now had spare capacity. Emphasis on
throughput for which the plant was designed. It can be stabilised processes also resulted in a 20 per cent reduction in
concluded that: sodium cyanide consumption.
Table 1 Table 2
Monthly variability from budget over a 12 month period. Month-to-month volatility over a 12 month period.
Underground mines Surface mines Underground mines Surface mines

Min Max Ave Min Max Ave Min Max Ave Min Max Ave
(%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%)
Ore mined (t) 5 43 14 7 68 29 Ore mined (t) 6 34 14 11 62 25
Head grade 7 33 13 3 38 11 Head grade 5 28 11 3 22 9
Ore processed (t) 5 54 12 4 39 9 Ore processed (t) 3 32 13 4 22 10
Processed grade 3 33 13 2 31 10 Processed grade 5 23 12 3 16 8

FIG 6 – Mponeng mine daily plant tonnes (over 12 months) (after Carter, 2010).
OPTIMISING THE MINE AND PROCESSING PLANT MANAGING THE DOWNSIDE RISK
The key parameters over which a mining company has The need to manage risk was well understood in the past. A
control are the size of mine (large or small, low or high-grade, small project was built, often with second-hand plant, and
as determined by the cut-off grade), the mining method, the then cash flow from the operation, or equity funding from the
production rate, the mining sequence, the processing method now-reassured investors, was used for a series of expansions
and the amount of money that will be spent on getting these and optimisations. If there was a problem with the initial ore
things right, which includes exploration drilling, geological reserve or cost estimates, the exposure of shareholders to this
modelling and bench and pilot plant testing. Other aspects problem was minimised and managed.
such as power supply, water supply, concentrate transport The author was involved in a large gold project that had
and logistics generally have a more obvious engineering high geological risk. It could be initially developed using a
solution and are ancillary to the optimisation process. nearby idle plant, either by purchasing the plant for a modest
The objectives of optimisation must be aligned with the sum or by toll milling, before building a full-scale plant.
This plan did not give a satisfactory project NPV because it
corporate objectives of the owner. Some stated corporate
delayed full-scale production by several years. A compromise
objectives, such as maximising annual ounces of gold
was found whereby a small plant would be built on-site
production or maximising mine life, cannot be optimised.
and later expanded. However, as the time for construction
Clearly, increasingly large subeconomic projects will satisfy
approached, deteriorating estimates again made the NPV
the former objective while decreasingly large subeconomic
unattractive. This was addressed by deleting the smaller stage
projects will satisfy the latter. Short-life projects carry the risk and building the larger plant immediately. This enhanced
that most of the production will be delivered into a trough the NPV substantially in spreadsheet models. When the
in the product price. Sensitivity analysis based on a range geological problem proved intractable the project failed, with
of price scenarios will identify the parameters that yield an capital losses around three times what they might have been
acceptable risk. with the original toll-milling proposal.
There is also the problem of capital allocation between This outcome might have been avoided using the value at
competing projects. If there is no restriction on the available risk (VaR) approach which is widely used in the financial
capital then corporate value is maximised by maximising the sector. VaR is the maximum loss not exceeded with a given
NPV of every available viable project and carrying all of them probability defined as the confidence level, over a given
through to production. In the real world, where available period of time. It is commonly used by security houses or
capital is restricted, the corporation must select projects investment banks to measure the market risk of their asset
for investment using some ranking technique. Economic portfolios (market VaR), over time periods of one day to a few
theory says that projects should be ranked using the present days. However VaR is a very general concept that has broad
value ratio (PVR), which is the ratio of NPV to initial capital applications.
investment. For simplicity, the capital investment is usually For example, a Monte Carlo approach to modelling net cash
taken to be the total of negative cash flows prior to achieving flow outcomes for a particular project development option
positive cash flows. If the perceived risks are similar, projects might show that 95 per cent of outcomes have a net cash
with higher PVRs are selected before those with lower PVRs. result better than minus $50 M. In other words, the cash loss is
A project with a high NPV but a low PVR may require more expected to be greater than $50 M only five per cent of the time.
capital than the corporation (or the investment community) This approach must have a constrained time period applied,
is able or willing to risk, or if developed it may displace such as the time to project payback or a fixed number of years.
alternatives which would have provided a better aggregate The various project development options can be modelled
return on investment. and a decision made based on both the expected NPV and the
From the above, the project should be designed to maximise VaR for each development option. Ultimately, the corporation
must be able to absorb and manage the ‘worst-case’ outcome.
the project NPV at the corporation’s agreed discount rate,
provided this leaves it with a PVR that will make it an
attractive investment. Arguably, the plan should be changed CONCLUSIONS
to improve the PVR, even at the expense of NPV, if this will The design of metallurgical plants should be undertaken
allow the project to proceed in competition with others. in the context of the broader optimisation of the mining

P L McCarthy
project. Often the appropriate metallurgical process can be ACKNOWLEDGEMENTS

selected early in the optimisation process and thereafter the
The author acknowledges AMC Consultants Pty Ltd for
plant design is simply a matter of good engineering, without
permission to publish benchmarking results and Unity
strategic options. By contrast the size of mine (large or small,
Mining Limited for permission to publish the Gill Reef data.
low or high-grade, as determined by the cut-off grade), the
mining method, the production rate and the mining sequence
are strategic decisions which form critical inputs to the REFERENCES
engineering design of the metallurgical plant. Barratt, D J and Doll, A G, 2008. Testwork programs that deliver
multiple data sets of comminution parameters for use in mine
These strategic decisions should be made by a team planning and project engineering, in Proceedings Procemin 2008,
which includes members who have an appreciation of the Santiago, Chile.
implications of those decisions to every part of the proposed
Carter, C, 2010. Project one holistic transformation plan to achieve
operation. Subsequent engineering design for the mine,
strategic goals, Anglogold Ashanti.
processing plant and infrastructure can then proceed within
appropriate constraints. Dicker’s Mining Record, 1867. Dicker’s Mining Record and Guide to
Gold Mines of Australia (newspaper), 28 November.
In recent years, many project investment decisions have
been made on the assumption that unlimited project finance Hall, B and Hall, A, 2006. Doing the right things right: identifying and
implementing the mine plan that delivers the corporate goals, in
is available. Due to global economic circumstances this is no
Proceedings International Mine Management Conference 2006, CD-
longer the case, and a more traditional approach to project ROM (The Australasian Institute of Mining and Metallurgy:
optimisation is called for. After consideration of risk, a Melbourne).
modest-sized, staged development may provide better
Hoover, H C, 1909. Principles of Mining, pp 153–160 (McGraw-Hill:
shareholder returns than the largest project that an orebody New York).
can theoretically support. Staged development may require
multiple parallel processing circuits and smaller, more JORC Code, 2012. Australasian Code for Reporting of Exploration
Results, Mineral Resources and Ore Reserves (The JORC Code)
selective mining machines operating at higher cut-off grades.
[online]. Available from: <http://www.jorc.org> (The Joint Ore
The VaR approach is one way of quantifying the downside Reserves Committee of The Australasian Institute of Mining and
when considering alternative project scales and development Metallurgy, Australian Institute of Geoscientists and Minerals
paths. Its use, in combination with tradition NPV analysis, Council of Australia).
provides a more complete picture of the options available to a McCarthy, P L, 2010. Setting plant capacity, Transactions of the
company when it sets out to develop an orebody. Institutions of Mining and Metallurgy, Mineral Processing and
Extractive Metallurgy, 119(4):C184–C190.
Tatman, C R, 2001. Production rate selection for steeply dipping
tabular deposits, Mining Engineering, October, pp 62–64.

Project economics
Contents
Guidelines for economic evaluation of projects

P Card1
ABSTRACT
‘Let’s get discipline and quality into easy to understand evaluations!’
Project managers and metallurgical plant operators accept poor quality evaluations
because they are generally unaware of what they should expect.
To provide guidance, the Australasian Institute of Mining and Metallurgy (the
AusIMM) established a subcommittee of practitioners that collated evaluation
systems and procedures regarded as best practice. This has been developed into a
free website for the greater mining industry: <www.economicevaluation.com.au>,
which is maintained by the author. It offers a range of modules and worked examples
that describe in plain language, the step by step sequence of conducting an economic
evaluation.
These practices apply directly to projects and metallurgical plant evaluations,
including those with a heavy technical basis. They can be applied to technical problem
solving where there are no monetary computations.
Project managers and metallurgical plant operators should demand that economic
models and evaluations follow the six principles. They should become working tools
which are easy to understand, fit for purpose, consistent, rigorous, record sources of
input data, have key graphs and are rapid to audit.
INTRODUCTION
Poor workmanship is common but usually accepted
If industry made an award for the worst performance in metallurgical plant project
design and plant operation then the odds-on favourite would be the economic
evaluation!
If an economic evaluation is sophisticated, complex and terribly clever with Excel so
that only one or two experts can use it, then it probably is worst practice. But if anyone
with only basic knowledge of evaluation can readily follow it, sees the correct data
being employed and feels it is easy to understand then it is on the way to best practice.
Project managers and plant operators need to take control and demand economic
evaluations they can quickly follow during their busy working day.
The world’s best practices are readily available on the internet and are in active use
in the mining industry. But for historical reasons most project managers and plant
operators will accept poor quality work in this arena. What is accepted for economic
evaluation would not be tolerated in its sister disciplines of geology, mining,
metallurgy, engineering and accounting. Fortunately this is rarely due to sloppy
management by project managers and plant operators, but due to lack of awareness
of what they should expect and demand.
As a horrible start, most professionals do not even call it by its correct name of
‘economic evaluation’ but talk of ‘financial modelling’ or even more incorrectly
‘financial analysis’ (more later). Project managers and plant operators would insist on
metallurgists doing the processing study work, insist on engineers doing the design
and Engineering, Procurement, Construction Management (EPCM), would want
accountants to do the books, but probably would accept almost anyone willing and
able to perform the economic evaluation. This person probably would be allowed to
do the evaluation, however they believed was best. This is because the discipline of
economic evaluation has evolved over recent decades to bridge between operations/
engineering and accounting without an academic or professional foundation.
Older style project managers and plant operators see economic evaluation as a back
room activity to be hurriedly completed, typically when the last of the cost estimates
1. MAusIMM, Consultant – Economic are finished the night before the submission is due to management. They see the
Evaluations, Aspendale Vic 3195. activity as a mathematical computation to fill in the paperwork with net present value
Email: mpcard@tpg.com.au (NPV), internal rate of return (IRR), payback etc. Fortunately, these are a dying breed.
65
P Card
These older style managers do not really understand •• combining their mathematics is very tricky, especially
they are designing or operating a business, but live in a adjusting the discount rate as debt is introduced (do
closed world of professional engineering or hands-on plant people still get fooled by false claims of improving
operating. They are very confident that they are ‘working on project returns by using debt?)
the important stuff’ and ‘getting things done!’ They do not •• the spreadsheets will become unnecessarily complex
realise that contemporary managers demand an economic to use and audit and so alienate all in the team except
evaluation up and running from Day One as a tool to steer the specialists in finance.
project or operating plant through the study process into the
The financing spreadsheets could be appended to the end
optimum state of business. The economic evaluation specialist
of the economic evaluation workbook providing the flow is
should be the second-best role in the team; after the leader.
one-way and nothing feeds back to the economic evaluation
worksheets.
Mantras of best practice
A few mantras of best practice in economic evaluation are: Evaluation, valuation, modelling
•• ‘If you do not readily understand and comprehend my ‘Modelling’ is a component of, but not all of ‘valuation’ which
evaluation then you do not have a problem, I do!’ is a component of, but not all of ‘evaluation’. They form
•• ‘Every worksheet should be as easy to read as a school a hierarchy with ‘modelling’ at the bottom providing the
text book!’ hands-on computations that feed results for various cases and
•• ‘Do not try to impress with sophisticated Excel functions, scenarios up into ‘valuation’. This ‘valuation’ quantifies and
but use your intelligence to convert complex interactions characterises the value of the project or metallurgical plant. In
into simple steps on the worksheet.’ turn this ‘valuation’ feeds up as one element in the intellectual
activity of ‘evaluation’ by fully understanding the project or
•• And definitely not ‘Trust me! I am the expert in
metallurgical plant as a business.
evaluation modelling!’
Anyone thinking that economic evaluation is all about
Financial analysis versus economic evaluation pouring numbers into a spreadsheet model to get NPV and
IRR is living in the past. Today it is all about having a working
Completion of a project requires two separate money-focused
knowledge of the whole project or metallurgical plant from
activities:
ore in the ground through all activities and influences to the
1. economic evaluation market. It encompasses nearly everything from engineering
2. financial modelling. to paying taxes. It is about understanding the key drivers
The first, economic evaluation, is all about understanding and key interactions of the business. It is getting a helicopter
the business health of the project. What cash will be required view of the total entity, deciding how it fits the existing
to establish, operate and pay taxes versus the cash generated business and helping to test ideas and create better projects
or saved by the project. This is the simple economics of cash- and metallurgical plants. It is all about putting the ‘E’ into
in and cash-out over time. Economic evaluation does not Evaluation. Yes, the economic evaluation specialist should
worry where the cash comes from (that is, financing) but have the second-best job in the team: after the leader.
rather it wants to understand the cash generating power
of the underlying project including what sort of prices and Best practice in economic evaluation
operating environment is required to cover costs and generate There are six principles at the top of the hierarchy.
an economic return (the discount rate is usually before
financing).
KEY PRINCIPLES
Everything is computed in cash in the year it is actually spent
or received. There is no accounting depreciation, no accounting The key principles to which one should adhere when
charges for future closure and no other non-cash items. performing spreadsheet modelling are simple but extremely
effective. They are:
There is no equity raisings nor company borrowings, and
so no financing charges on borrowings during construction. •• easy to follow
Taxes are included and computed before the apparent benefit •• tailored-to-purpose
of debt. •• transparent
The second, financial modelling, is all about sourcing the •• disciplined, rigorous and consistent
cash to establish and run the project until it becomes self-
•• recording sources of all data
sustaining. What mix of company internal cash flow, debt,
new equity, convertible notes, derivatives and hedging will •• rapid to audit.
the company use to progressively pay for owner’s costs,
EPCM, first fills, commissioning, ramp-up to commercial KEY PRACTICES
operations and interest on these borrowings? Flowing down from these six principles are key practices to be
How will each type of capital raising impact company incorporated in an economic model. They are:
profits, its balance sheet and share price? •• A non-expert should readily understand the function
Economic evaluation might be best lead by a person with of each worksheet, how it is arranged into component
a technical-operating background whereas financing might parts, how the data is entered, the computations and the
be best lead by an accountant. Both perform spreadsheet relative importance of the parts.
modelling of the future business but they are very different •• The visual flow down and across each worksheet should
in purpose and process. Both should be presented in simple be intuitive and logical.
language and easy-to-understand concepts.
•• Each worksheet should have a bold heading followed
The two activities should not be woven together because: by a brief outline of its purpose, and where helpful, its
•• each is a stand-alone decision important links with other worksheets.

Guidelines for economic evaluation of projects
•• Sections within each worksheet should be in discrete •• starts with a brief overview of results including multiple
work blocks, with obvious subsections and subheadings graphs of the four cash flows (see below) and all
using a cascading layout for subtitles. important inputs and outputs
•• Visually each work block should be self-contained with •• is audited by the specialist, by experts in their areas
an obvious step-by-step development toward a bold (example: metallurgist audited the processing section)
subtotal for that work block. and usually an external specialist
•• The separate work blocks should be in a logical •• is intuitive
sequence down and across the worksheet and their •• provides trust in what you see without drilling down
aggregation is obvious. into algorithms.
•• Complex and extended computations should be shown
in a series of small steps so that the logic is visible and USING EVALUATIONS IN FEASIBILITY STUDIES
the input parameters are obvious. There is no need to Project managers and metallurgical plant operators appear to
interrogate the algorithm. Explain the logic of complex use three phases of study for projects:
algorithms in a note.
1. Concept or scoping studies to assess if the project fits
•• Usually if a row of data that has already been presented company strategy and at least one alternative has a
above is needed again in a work block then the entire reasonable likelihood of being economically viable:
row should be repeated so that there is visual flow of
•• For these brief studies the modelling and evaluation
the logic. If referenced from another worksheet then it
usually can be relatively coarse and simple. It needs
should be coloured coded.
to explore the range of outcomes and key drivers of
•• Key inputs and results should be shown in graphs as a success and failure. It needs to weed out pet projects
self-check and for rapid understanding. and support only quality concepts.
•• The ‘Data Group and Outline’ facility can be used to define •• The evaluation person needs to work cooperatively
worksheet structure, collapse related groups of rows or between the project manager and the many experts
columns to reduce visual clutter, and aid navigation. inside and outside the company who have the
knowledge. The specialist may be a ‘back room’ type
WORKED EXAMPLE OF MODELLING from any background, but better if an active specialist
Worked examples of economic models in Excel for with an operating/engineering background.
concept studies, prefeasibility studies and final feasibility •• The project manager is likely to work closely with the
studies can be downloaded (free) from the website: evaluation specialist interacting every day or two.
<www.economicevaluation.com.au>. 2. Prefeasibility studies that assess the complete range of
project alternatives. The divergent thinking in this phase
READY-MADE EVALUATION MODELS generates the most value:
There was unanimity amongst the AusIMM practitioners that •• For these studies the modelling and valuation needs to
ready-made economic evaluation models, where users fill in be detailed enough to differentiate the economics and
the blanks, were too dangerous to use. Experience is that these character of each alternative. This phase excites those
black-box models transgress the six principles of best practice, who are creative but objective. It needs to confirm the
but more importantly have a bad history. Their computations attractiveness of the selected alternative and define its
cannot be audited, the models must be exceedingly complex business character.
(or deficient) to accept a wide variety of scenarios and they •• The evaluation person needs to frequently interact
simply cannot be trusted. with specialists inside the project team and outside,
drawing out the complete information, going back and
confirming it has been correctly modelled (audits) and
WHAT TO EXPECT FROM SPREADSHEET MODELLING being absolutely objective. The evaluation specialist
A project manager or metallurgical plant operator should must resist the temptation to be too clever with Excel
demand that every economic evaluation model under his/ functions but keep the model simple so everyone can
her management: readily understand, audit and feel it represents the
•• is absolutely rigorous in its construction alternative truly. The specialist should not be a ‘back
•• is in discrete blocks of simple steps with clear headings room’ type but probably from an operating/technical/
and obvious end results engineering background with a broad understanding
of the business and a work ethic that is energetic,
•• has input data coloured (example: blue) so that it is
collaborative, creative and accurate.
immediately visible and the project manager can skim
across it to quickly check validity of all inputs •• The project manager is likely to see the evaluation
model as an engine to assess alternatives and generate
•• has input data with a source (when, who and what) ideas. He/she is likely to rely heavily on the evaluation
typed in the row above so the project manager can specialist to understand where each of the alternatives
immediately see if it is the correct version is heading and to help steer the study activities to best
•• has data referenced across from another sheet coloured effect. Daily contact, with creative thinking, would be
(example: green) so it is immediately recognised common.
•• is absolutely consistent across rows, so that algorithms 3. Final feasibility studies that thoroughly assess and define
do not have hidden changes the alternative to be taken into execution or construction.
•• has every item of input data exposed in a row before This is convergent thinking with lots of detail:
being used and absolutely no fresh data entered as data •• For these studies the modelling and evaluation will
hidden in algorithms become quite detailed. The whole business needs to

P Card
be thoroughly defined. The modelling will evaluate DO BIGGER PROJECTS NEED BIGGER ECONOMIC MODELS?
different methods and different equipment within the
One of the world’s greatest iron ore mines was acquired in
selected alternative. It may need to work in nominal
the 1970s with the economic evaluation model provided to
terms in some areas and integrate with company’s
the company’s Board being just one page of very easy-to-
accounting and financing activities.
follow, manual computations. Contrast it with the volumes
•• The evaluation person needs to work in detail as part of modelling and synthesis that would be required today. In
of the enlarged study team. a strange way that one page from the 1970s was as potent as
•• By this time the project manager should know if the all the evaluation study work that we generate today. There
existing evaluation person is the right fit for this phase. in amazing simplicity for the Board members was the heart of
Contact may become periodic but the evaluation the acquisition decision, namely the risks in the forecasts of
model would continue as a vehicle to help steer the price, mineral resource, production and costs over the years.
Feasibility Study. Senior executives in the world’s biggest mining companies
have made it clear they would love simple one page models of
FOUR CASH FLOWS major investments, major acquisitions and life-of-mine plans.
Anyone opening a worst practice evaluation model is faced This is not to suggest that all evaluation models should
with a myriad of parameters and a convoluted array of be one page. Quite the reverse: a detailed evaluation model
computations that are understood only by the evaluation should be a centre-piece of a major project or metallurgical
specialist. Key outputs such as NPV are buried amongst the plant. It should draw together all the component parts as a
worksheets. Auditing would be tedious and take days. business so the team understands the relative importance of
each part, where to focus and how to optimise the overall
Best practice has simplified valuations into four streams of
design. It should be a tool used every day by members of the
cash flow:
project team/metallurgical plant to challenge and test their
1. revenue cash stream (production, stocks, sales, prices, ideas. Detailed evaluation models are needed to steer the
debtors) project toward the best configuration and to assess the risks.
2. capital cash stream (capital expenditure (CAPEX), Ironically, Management/Board need a simple, easy-to-
creditors, tax deductions) understand evaluation ‘one page’ model to help it understand
3. operating cost cash stream (operating expenditure the big decision of whether to invest in the project whereas
(OPEX), creditors) the project team might need a big, detailed, working model
4. taxes cash stream (royalties and income tax). to optimise the project’s configuration and the design of its
component parts.
Their sum represents the net cash flow each year and this can
be simply discounted to give NPV, or used to compute IRR.
Second and third level computations such as working HALL MARKS OF BEST PRACTICE ECONOMIC EVALUATION
stocks, debtors, creditors and tax deductions for CAPEX (‘tax From day one, project managers and metallurgical plant
depreciation’) are computed as high-level calculations within operators must set their expectations and demand that
these four cash streams. Their impact on project managers’ economic evaluation is:
and metallurgical plant operators’ decision-making will be •• easy to follow – it may not be simple but anyone in the
minimal so should be reduced to a few simplified rows. team should be able to follow it like a school text book
•• fit-for-purpose in that detail matches importance – begin
THE TWO BOOKENDS OF ECONOMIC VALUE as simply as possible and add complexity only when
In most mining industry businesses there are two ‘bookends’ warranted
which dominate the economics: the ore resources in the •• rigorous, transparent and intuitive – trust what you see
ground and the market. One end determines how big and •• fully documented – readily see where every piece of
good the business can be and the other end determines how data was sourced
profitable it will be. In between are all the very important and
•• graphs – to find errors and to give quick visual
exciting projects and operations to make it happen and make
understanding
it improve. It has been joked that a whole team of engineers
and metallurgists slave for ages on production, CAPEX and •• audited – both the mechanical computations and the
OPEX while somewhere in a back office a few people generate results.
price forecasts that swamp the valuation. None-the-less
the production, CAPEX and OPEX are critical and must be
forecast with appropriate quality.

Contents
Sensible cost cutting for resource projects

D Connelly1
ABSTRACT
As the resources industry becomes tougher with rising costs and cyclical metal prices,
mining companies need to continue to reduce their costs and do more with less.
Many new resource projects fail to come in under or on budget based on feasibility
studies. In particular, capital expenditure (CAPEX) proves difficult to achieve
for resource projects. Many resource projects have failed because of aggressive
plant CAPEX cost cutting, which results in projects with no surge capacity and
an inability to achieve design throughput. In addition, the plants are not operable
because of the omissions. In recent times, many businesses have also experienced
the ‘costs’ of simply implementing aggressive operating cost-cutting measures as a
strategy for solving business performance problems. This paper looks at technology-
driven, employment-related luxury, department and restructuring cuts. In the past,
companies used an incremental approach based on the use of past budget information
as an integral part of the budget construction process. The use of performance reports
and management information systems (MIS) is examined, along with the role of
continuous improvement in achieving sensible cost cutting.
INTRODUCTION
All companies look for ways to reduce costs and increase profits, and resource
companies are no different. With the scale of mining projects, a cost saving of a
few per cent could mean millions of dollars for the company. Budgets are prepared to
determine where money is spent and where it could possibly be saved. This applies
to all resource companies, whether an exploration company looking to build a plant,
an engineering design company or a mining company in production.
Historically, the resources industry has always gone through peaks and troughs.
Understandably, mining companies want to take advantage of the good times by
maximising output of existing mines or constructing new ones. Designing a plant
to these favourable market conditions during a mining boom can be a dangerous
exercise. If the metal price suddenly drops, a low ore grade or high processing cost
prospect that was previously viable, could result in a mine closure. To avoid this,
companies must look at cost-cutting measures to increase the viability of their project
for a range of commodity prices.
A plant in the design stage has a much larger scope for cutting costs from the
CAPEX and operating expenditure (OPEX) than an existing operation through circuit
optimisation or process changes; however, overly aggressive process cuts to reduce
the CAPEX may cause the process to run inefficiently during operation and could
reduce the overall availability of the plant, causing extra costs and less revenue.
Process changes and their operability implications must be thoroughly considered
by the design company. Mining companies should not simply cut costs to the design
process as this may cause greater costs to the project in the future (for example, Murrin
Murrin, Browns project, Ravensthorpe nickel project, Port Hedland hot briquetted
iron project, Rapu Rapu copper zinc project).
An existing operation that is struggling to be profitable may need to make some
significant operating cuts. Rather than one large cut, which could have serious negative
implications to the personnel or process, several small cuts should be implemented
to achieve an overall significant saving. These cuts could be related to employment,
luxury, department or restructuring, and are applicable to both engineering design
companies and production companies. Production companies also have scope to
refine the operating costs through power conservation, process automation and
1. MAusIMM(CP), Director/Principal Consulting
consumables optimisation. It is important that prior to implementation, all cuts be
Engineer, Mineral Engineering Technical
Services Pty Ltd (METS), Midas Engineering assessed to ensure that the negative impacts are minimised. Continuous improvement
Group, Perth WA 6000. Email: (CI) plays an important role in the cost-cutting procedure. Streamlining information
damian.connelly@metsengineering.com communication using MIS is just one aspect that could be included for CI.
69
D Connelly
BUDGETING AND BUDGET CUTS when travelling instead of business. Additionally, many
Creating a budget should be the first stage of any project. organisations often spend a good deal of money on office
It is a process of predicting and controlling the expenditure spaces or other items to impress their clients and competitors
over the life of any given project. Budgets are the foundation (for example, the use of hire cars, video conferencing instead
of an organisation’s financial success. The importance of travel, mobile phone calls, alcohol, dining expenses).
of creating a budget is that it forces an organisation to
consider the expectation for its products and services with Employment status related cuts
the required resources to meet that expectation. In addition, By employing consultants and independent contractors,
budgets can transform an organisation’s higher priorities organisations will be able to minimise the overhead salaries
into the appropriate resources required to achieve those spent on permanent employees. Although independent
aforementioned priorities. The potential problems could contractor arrangements can make a significant saving to an
be highlighted in a sufficient time to acquire the corrective organisation, there are some downsides that need to be taken
actions to be performed. A baseline can be created against into consideration. An unintentional effect of converting
which the actual results can be compared. several former employees into independent status may create
Due to the variability of the resources industry and the unfortunate tax complications in an organisation. Some
difficulty in predicting future costs, the budget is a document organisations offer year-end bonuses instead of pay rises to
which should be continually monitored. When a planned avoid a fixed commitment. There is a growing trend to utilise
budget is overshot, the consequences can vary from mere temporary workers to support some parts of an organisation’s
frustration to anger or even litigation, if it involves a new operations such as during the shutdown period of an operating
engineering design and construction project with other plant. Certainly, temporary workers will never be as fully
companies. Overspending an allocated budget will result invested in the company as would permanent employees and
in the reputation of the company being marred and good this could result in less efficiency and productivity.
business relationships being severed. (See Figure 1, which
highlights impact of changing budget). Technology driven cuts
With the advancement of technology, organisations have
Strategic budget cuts started to find ways to reduce the current workload. The
Strategic budget cutting is a common organisational policy. application of technology will help to reduce the number
Applying the right combination of budget cutting and of employees needed and other associated costs to perform
strategic growth is a fundamental input for the long-term the tasks. It may even allow employees time to be in part
success of any resource organisation. It is vital for managers redirected to improving the efficiency of other areas of the
to approach budget cuts practically so as not to affect the operations. Although technology can be a good way to reduce
organisation’s capability. the operating budget, it is costly to establish and requires
There are many ways to perform a budget cut, yet extreme time to implement new technologies. The initial investment
cut-backs can eventually affect the growth of an organisation. may discourage some companies from implementing
Many organisations today have come to understand how an the technology. Technology is a capital investment and
overly enthusiastic implementation of budget cuts can have consideration of the depreciation, ongoing maintenance
the unintended effect of inhibiting their revenues. Profits and replacement should be assessed adjacent to the realistic
were increased in the short-term due to the severe budget savings. New technology introduces risks and is not applicable
cut, but the overall revenues were then decreased due to the for new projects unless piloted first.
lack of growth. In some cases, budget cuts have caused the
subsequent declines in customer service and product quality. Department cuts
There are several types of budget cuts and these include Department cuts usually involve a participatory process where
luxury cuts, employment status related cuts, technology managers of each operation unit will identify the prospects
driven cuts, department cuts and restructuring cuts. to reduce their budgets without affecting the operation of
an organisation. Managers can start by checking the surplus
Luxury cuts funding in the past that may have been given but is not
entirely essential to maintain the level of services. When there
An approach to reduce the budget is to focus on the costs of
supplies and services along with employee related expenses. is a vacant position available, managers will decide whether
Travel and other related benefits spending can be minimised; it can be held without affecting the everyday operation of the
for instance, by employees only being offered economy class organisation. Managers should make sure that the budget cut
does not affect the organisation’s core strategies or the key
clients, customers and constituencies’ interest.
Restructuring cuts
When an organisation undergoes a major restructure, this
generally entails a shift from budgeting to strategic planning.
This change occurs as the demand from the market has
changed. Restructuring an organisation is not an easy process
and it involves the re-examination of the services it offers,
re-evaluating the departments and managers it should keep,
and the determination of the employees it needs depending
on their role in delivering new services and the application
of technology as an alternative to additional employees in
its operations. The result of the restructuring cut is usually
a significant cost reduction. The organisational changes in
FIG 1 – The cost of not getting the budget right. the 1990s where whole layers of middle management were

retrenched including specialists with long-term business reducing the cost spent on design. Mining companies may
knowledge is a specific example. be reluctant to spend large amounts of money on the design
stage of a project because until production commences they
Continuous improvement will see no revenue for all their expenditure. Cutting corners
Continuous improvement can be described in several ways. at this early stage, although saving in the short-term could
In the simplest terms, CI is the development of ongoing prove costly in the future. A high quality plant design can
improvement in quality and efficiency within a company. The ultimately save on plant CAPEX and reduce unexpected costs
objectives of CI are to: in the future.
•• provide a more disciplined approach to CI projects and Typically, in the early stages of design for a mineral
initiatives processing operation a number of options will be investigated
to determine the best possible design for the plant. These
•• allow for greater consistency across global sites with a options will be compared on the basis of process performance,
standard methodology and tools CAPEX and OPEX.
•• allow for a quicker response to CI opportunities
by facilitating the organisation of teams and the Unit design
development of both teams and individuals It can be tempting for mining companies to only design a plant
•• improve safety, cost, production and productivity to the nominal operating conditions in order to save on CAPEX.
performance Ultimately this could be detrimental to plant performance and
•• help all employees understand the importance of, and result in less availability or throughput. Increased flow rates
the ways to, improve processes and organisational due to surges need to be accounted for around the plant and it
relationships is important that the equipment is able to handle this. Pumps
are a typical example of this and the design should incorporate
•• help capitalise on global best-practices and share
a 10–15 per cent allowance for surges.
knowledge across sites.
Another trap is designing equipment to handle the average
The steps of CI are:
rather than the maximum ore characteristics. Ore hardness
1. identify the opportunity that will lead to continuous is one characteristic where the maximum value should be
improvement taken when designing equipment. Although sizing to these
2. study the opportunity with respect to the key business specifications may result in larger, more expensive pieces
needs for the organisation of equipment being required, designing the plant to handle
3. define the current state by researching and understanding average ore hardness can result in lower crusher throughputs,
the current process, system or organisation higher recirculating loads and less overall plant utilisation.
4. develop the future state then develop and test an Similarly, individual equipment availability needs to be
implementation plan considered. In a processing system with little surge capacity,
the overall availability is equal to the lowest availability,
5. implement the solution and maintain high levels of not the average. If a critical piece of equipment is offline the
communication to monitor the status of scheduled entire process stops. A high maintenance piece of equipment
activities like belt filters may have an availability of 65–70 per cent
6. follow up and document the new procedures that have due to cloth replacement and cleaning. If an overall plant
been implements and assess the effectiveness. availability of 90 per cent is desired, standby filters would be
Together these points seek to improve existing operating needed to achieve this. Removing these standby units to save
procedures, quality and efficiency. on CAPEX will ultimately result in lower plant throughputs
and smaller revenue.
Management Information System Alternatively if there is equipment which is known to have
Management Information System is the practice of a low availability, but the capital required makes the use of
managing data so that information can be delivered with multiple units unviable, instead of expecting unreasonable
insight, understanding and value for the employees in an availabilities these units could be decoupled from the process.
organisation. It is the process of organising an information That is, placing adequate surge capacity before or after the
database which is easily accessible, well defined and flexible. units to allow for the increased downtime. These units will
Although techniques and technologies will change, these require a higher throughput than the rest of the processing
principles will remain in the core information management plant to maintain the overall plant throughput with less up-
model. Through the MIS, information is primarily delivered time. A common example of this in a minerals processing
to the right person in the right structure at the right time with operation will be a crushing circuit which has been decoupled
a cost that adds net value to the organisation. For those that from the processing plant through the use of a crushed ore
apply their knowledge gained through experience it will stockpile. A crushing circuit will generally have an availability
produce a result with more effective outcomes. of around 80 per cent.
MIS is an important tool to streamline communication
Plant layout
throughout an organisation and serves to reduce time
generally used to search for information. This time saving Plant layout can affect the CAPEX for a project. Optimising
ultimately makes an organisation more efficient and hence the plant location and plant layout can produce significant
more profitable. saving when considering access and plant roads, material
handling distances, electrical and communications
infrastructure, and plumbing. The design must be optimised
Design Stage Cost Cutting with these factors in mind without having the equipment
Optimisation during the design stage of a project is an so close that access for maintenance is compromised.
important measure in reducing the CAPEX of a new project. Computer-aided design packages (for example, SolidWorks
However, reducing the CAPEX of the project does not mean by Dassault Systèmes) can be used to virtually construct

D Connelly
the plant and be used to produce quantities. Optimising the or uncontrollable factors are common, some level of error is
design in the virtual world can enable direct savings in the allowed.
real world. A more compact plant but with room for later When you hire an engineer you purchase service, not
expansion flows through to savings on civil, piping, electrical insurance, so you are not justified in expecting perfection
and construction costs in both the short and long-terms. or infallibility, only reasonable care and competence. An
engineer who makes a mistake causing injury or damage is
Contingency not sufficient reason to lead to professional liability on the
Mineral processing projects in development will always part of the engineer. For there to be professional liability,
have a contingency added to the price. Although this may it must be proven that the services were professionally
seem like an ambiguous addition to the price it is in fact an negligent, that is, they fell beneath the standard of care of the
important item in the project cost. This cost represents the profession. When hiring an engineer, there is an expectation
project unknowns which experience tells us will eventuate as of risk acceptance that this professional engineer potentially
the project progresses. Contingency is directly related to the may make a mistake whilst using reasonable diligence and
risk of the project whether that is political, environmental or best judgement.
otherwise. The more risk associated with a project the higher The standard of care is not what an engineer should have
the contingency will need to be. For example, if a project is done in a particular instance; it is not what others believe
to be constructed in an area with political instability or a an engineer should do, or how others say they would have
tropical area that is prone to cyclones, an allowance will need done. It is what competent engineers have actually done in
to be made in the event that the project becomes delayed or similar circumstances.
requires extra capital to complete. As the project develops and
gets closer to production, the chances of one of these events Operability
occurring suddenly decreases, as does the contingency. Operability is the ability to keep equipment, a system
The contingency for a project in the prefeasibility stage may or a whole industrial installation in a safe and reliable
be between 20–30 per cent of the direct costs. However, as the functioning condition, according to predefined operational
project gets to the detailed design stage, it would be expected requirements such as:
that this number will drop to 10–20 per cent. This contingency •• consideration of the operator
must be determined on a case-by-case basis and these •• distributed control
numbers can vary. Estimating contingencies can be a black •• slopes on sump floors – too flat is a problem
art. To reduce the CAPEX of the project it may be tempting
•• access to power, air and water points
for some companies to reduce or remove the contingency.
Although there is a possibility that the contingency will •• walkways and access
not be needed, it is there for a reason and cannot simply be •• flat launders, which lead to sanding
dismissed. Reducing or removing this could prove costly if •• skirting – spillage
the contingency is required (see Figure 2).
•• dust control
•• clean up in crushing plants
Engineering Shortcuts
•• surge between unit operations
Engineers have a duty to provide their services in a manner
consistent with the standard of care of their profession. •• low head height
A good working definition of the standard of care of a •• clear signage, pipes labelled
professional is that level or quality of service ordinarily •• attention to particular areas such as reagent mixing and
provided by other normally competent practitioners of lime mixing
good standing in that field, contemporaneously providing •• floor space around mills
similar services in the same locality and under the same
•• absence of or insufficient bunding height
circumstances. An engineer’s service need not be perfect. As
the engineer is using judgement gained from experience and •• insufficient process water storage
learning when providing professional services, and is usually •• ball charging
doing so in situations where a certain amount of unknown •• telemetry etc.
Maintainability
In engineering, maintainability is the ease with which a
product can be maintained to:
•• correct defects
•• meet new requirements
•• make future maintenance or expansion easier
•• cope with a change to maintain the plant
•• access to pumps for repairs or change out
•• gain access for mobile equipment and personnel for
maintenance, clean up etc
•• prevent lubrication of equipment under spillage or
placement well outside the area of operation
•• provide spare cyclones
FIG 2 – Value-adding through project life. •• have common equipment

•• install cranes in grinding area, flotation area hand sales, sourcing of appropriate equipment is generally
•• install bypass facilities (thickeners, trash screens) done during the definitive feasibility study (DFS), the last of
the project before construction.
•• separate the acid wash column or hopper
There are four factors that should be considered when
•• avoid the obvious (for example, vent exhaust next to a
determining the viability of using second-hand equipment,
high voltage switchyard etc).
and these are:
1. equipment costs
CONSTRUCTION COST CUTTING
2. the condition of equipment
Labour costs 3. suitability for process requirements
Labour represents a significant percentage of the construction 4. spares.
costs. When looking to reducing these costs the ‘per hour’
labour rate should not be the determining factor; rather the Equipment costs
efficiency of the workforce. It may be tempting for some The equipment cost does not only include the purchase of
companies to hire cheaper, unskilled or foreign workers for the equipment but also the dismantling, re-conditioning,
plant construction to reduce the labour rate. However, if one transportation and re-assembling at site. This should be
of these workers is half the price but takes three times as long considered carefully if the equipment is located in a remote
to do the work, it is not a cost-effective decision. mine site or is difficult to dismantle and transport, it may be
Examples exist of projects where cheap, foreign electricians cheaper to purchase new equipment.
have been used for African projects and poor supervision
resulted in extended commissioning problems, the need for The condition of equipment
rewiring and damage to equipment. Although second-hand equipment may be cheaper than new,
its condition needs to be considered. If substantial refurbishing
Plant modularisation is going to be required, the initial saving on CAPEX will be
Plant accommodation and site buildings have used modular outweighed by the cost required for the plant to become
design for some time now to save on construction and operational. In addition, used equipment may require extra
installation costs. The same benefits are now being seen maintenance during operation and this possibility must be
with plant and equipment modularisation. Companies taken into account.
that produce modular equipment have a standard design
and modify it as per the process requirements, reducing Suitability for process requirements
engineering costs. The equipment is commissioned off-site Second-hand equipment is unlikely to meet the exact
before being dismantled and packed into shipping containers. specifications as detailed by the proposed equipment list. Some
Once at site the equipment can simply be ‘bolted together’ concessions may need to be made and it must be determined
and it is ready for use. Using equipment modularisation can whether the equipment can handle the process requirements.
significantly reduce construction times and hence capital cost.
The concept is fine; however there are numerous examples
where the execution has been poorly handled resulting in
Innovative construction techniques the savings being less than originally considered. Other
With construction costs representing a significant portion of issues include a lack of drawings, lack of vendor support or
the CAPEX, there are companies that are marketing innovative unknown historical problems with the equipment.
techniques and materials to reduce the cost of the project.
There are several products and construction methods that Spare parts
look to reduce the amount of concrete used in construction.
One method uses earth retaining walls constructed from steel This is necessary to avoid the cost of having to specifically
mesh. Another is the use of steel arches overlaid with earth manufacture one-off parts.
to form bridges, underpasses and stockpile tunnels. Both
save time and money which is associated with the concrete OPERATIONAL COST CUTTING
structures that would otherwise be required. It can be difficult to achieve any one cost cut on an existing
operation that is going to make a great deal of impact on the
Reducing construction capital through unit hire overall operating cost. Also, a large cut will inevitably impact
It is possible to reduce the CAPEX of a processing operation negatively on the process or personnel of the organisation.
through the hire of equipment and infrastructure. However, Therefore, the most sensible way of achieving any significant
the reduced initial cost comes at the price of an increased saving is to implement several small cuts. With regards
OPEX. This could be an option investigated by junior miners to resources, the sector can be broken up into two distinct
looking to enter production but are finding it difficult to raise industries; these are mineral consulting companies and the
the initial capital. The mining fleet, mobile plant and the production companies. Consulting companies would use the
accommodation camp are typical examples of units that can strategic budget cuts as detailed previously to achieve costs
be hired for this purpose. cuts in the organisation. Production companies can also use
these strategic budget cuts; however they also have scope to
Second-hand plants reduce the operating costs of the processing plant.
The option of using second-hand equipment in the construction
of a processing plant can mean considerable savings to the Reagents and consumables
overall CAPEX of the project. However, this needs to be Optimising the operating costs with regards to plant reagents
considered on a case-by-case basis as it is possible for the costs and consumables needs to begin at the first contract tendering
to outweigh the savings. The use of second-hand equipment phase. The tenders need to be reviewed to obtain the best
can be considered at any stage of the project. However, due to product supplier and this may not always be the cheapest
the project timeline compared to the urgency of most second- option. The lead time for the products needs to be considered

D Connelly
and also the supplier’s production rate, particularly if they operation; however the capital to implement these changes
are supplying to an ever increasing market. It is necessary could be significant. These improvements could be related
that supply can be maintained to the plant, especially if that to automation of sampling, online analysis or automatic
product is critical to the process. reagent addition. These methods would be categorised under
This has been the case with cyanide when a global shortage ‘technology driven cuts’ as mentioned previously. Although
impacted on the smaller supply companies. The operations the improved efficiency could reduce the operating costs this
holding contracts with these smaller companies were paying needs to be weighed against the, often substantial, capital that
enormous margins at the time because of the difficulty to get must be outlaid for their installation. Flexibility to cope with
the product. The larger companies, on the other hand, were changing ore types and flow sheets is a must.
able to maintain supply to their clients but would not provide
product to the companies without existing contracts in place. CONCLUSIONS
It is also important to watch the usage of reagents and Sensible cost cutting is an important measure in ensuring
consumables, particularly with the more expensive ones such the viability of an organisation through a range of market
as cyanide in a gold plant. High usages should be investigated conditions. This is particularly important in the resources
and lowered if possible. In some cases it may be appropriate industry where the cyclical metal prices mean that boom and
to use automatic control to moderate reagent and consumable busts are inevitable. Cutting costs to remain profitable during
use. It should be investigated whether the capital outlay will the busy times is often necessary to avoid business or mine
be repaid through cost savings. closures. Careful budgeting, implementing an MIS and CI
There are numerous examples of reducing cyanide or are just a few of the measures that should be implemented to
flocculant consumption where considerable savings have keep costs down at all times. If budget cuts are required, large
been made because technical people were prepared to cuts should be avoided due to the detrimental effect upon
challenge what had been used historically and undertake test personnel or the process. Several small and well considered
work to support the case for reducing reagent usage. cuts should be implemented, as discussed in this paper. These
small savings can cumulatively produce an overall significant
Power saving saving to the CAPEX and/or OPEX costs.
Using power efficiently and saving energy where possible
can result in decent operating cost savings, especially if this ACKNOWLEDGEMENTS
power supply is diesel generated. In some cases an external The author would like to thank Robert Hanna from METS,
contractor can be brought in to determine the savings various companies, colleagues, engineers at various sites,
achievable through power saving. However, it is necessary METS staff and other consultants for their contribution. Also
to assess whether the cost of engaging an expert is worth the the management of METS for their permission to publish this
savings. paper and the constructive criticism of various drafts.
One way of saving power is to ensure that the operating
curves of process pumps are within the most efficient range. REFERENCES
Depending on the size of the pumps in the system this could Australian and New Zealand College of Anaesthetists (ANZCA),
equate to a significant cost saving. There are numerous n/d. How to carry out a continuous improvement project,
examples of cost savings for semi-autogenous grinding (SAG) Guidelines on continuous quality improvement [online].
mills and crushing circuits by using circuit surveys and Available from: <http://fpm.anzca.edu.au/resources/
simulation studies. educational-documents/guidelines-on-continuous-quality-
improvement> [Accessed: 2 March, 2011].
External contractors Mackenzie, W and Cusworth, N, 2007. The use and abuse of feasibility
It is important that costs for external contractors are monitored. studies, in Proceedings Project Evaluation Conference, pp 65–76 (The
If contractors are not managed closely, they may charge more
than budgeted for the job. One provision would be to never Maddox, D, 1999. Strategic budget cutting [online], The
hire a contractor for a large job on an hourly rate. Over the life Grantsmanship Center. Available from: <https://www.tgci.
of the project, the cost blowout could be substantial. com/sites/default/files/pdf/Strategic%20Budget%20Cutting_1.
pdf> [Accessed: 17 July 2017] (John Wiley & Sons).
Improved control Petty, J, n/d. Budgeting and one day reporting: developing and
managing a budget, towards one day monthly management
Improving the control systems of an existing processing reporting, Course notes.
plant does have the ability to increase the efficiency of the

Contents
When does further processing at

the mine site make sense?
C Fountain1, S La Brooy2 and G Lane3
ABSTRACT
A century ago, new mines were often accompanied by smelters. In Australia, smelters
were built at the Daydream mine near Broken Hill; at Kuridala and Mount Elliott near
Mount Isa; Broken Hill; Mount Lyell; Mount Morgan; Mount Isa; and more recently at
Tennant Creek; Kalgoorlie; and Olympic Dam. Now the big miners eschew site-based
processing in favour of shipping concentrate to smelters in Europe and Asia.
Yet there are times when on-site processing clearly makes sense. Few companies
ship gold ore or concentrate to China for gold extraction. New smelters are being
built in Zambia. Solvent extraction and electrowinning plants frequently produce
copper from oxide ores and concentrates while pressure leaching is increasingly
being considered for copper sulfide ores. Some argue that on-site processing is best
because the waste products can be returned to the ground from whence they came,
avoiding potentially large disposal costs in more-populated areas. Treatment and
refining charges could once again turn in favour of the smelter operators.
This paper examines when it makes sense to use on-site pyro- or hydrometallurgical
processes in today’s environment and comments on how current trends might alter
the balance in future.
INTRODUCTION
In a keynote address to the Ecological Society of Australia’s annual conference in
November 2007, Paul Ehrlich said:
… Australia is still, in the 40-some years I’ve been coming here, striving to become
a third-world country, to be a place that just exports its natural capital as fast as
possible, unworked upon, until it becomes truly poverty stricken. (Ehrlich, 2007)
In the same year, Chip Goodyear was in the process of vacating his leadership role
at BHP Billiton Limited (BHP). He was quoted as saying that BHP should concentrate
on mining and ore, and leave ‘to others the skill set of processing that material’
(Roberts, 2007).
It was not always the case.
When the Broken Hill Proprietary Company Limited was floated in June 1885 to
develop the deposit that gave BHP its name, a smelter was not far behind, opening
in May 1886 (Blainey, 1968a). Other companies built their own, so there were at least
five smelters operating along Broken Hill’s line of lode by 1891. BHP’s first smelter
was preceded in the district by a smelter at the Daydream mine near Silverton, which
treated the small ore arisings from the district.
Other early smelters included copper smelters at Burra in South Australia; Mount
Morgan, Mount Elliott, and Mount Cuthbert in Queensland; and Mount Lyell in
Tasmania. Lead and copper smelters were later developed at Mount Isa; a nickel
smelter at Kalgoorlie; and copper smelters built at Tennant Creek and Olympic Dam.
A short-lived nickel-copper smelter was built at the Radio Hill mine near Karratha in
Western Australia.
Similar stories can be told in North America and other parts of the world. In 1978, a
1. MAusIMM, Operational Readiness Manager, survey of converter practices elicited responses from 16 copper smelters in the United
Nyrstar Port Pirie Smelter, Port Pirie SA 5540. States (Johnson, Themelis and Eltringham, 1979), most of them near the deposits that
Email: chrisjan@iinet.net.au
provided their feed. Today only three remain.
2. FAusIMM, Principal Process Consultant,
Ausenco, Perth WA 6000. Most mining companies today follow Goodyear’s injunction and sell concentrate
Email: stephen.labrooy@ausenco.com to others. However, there are some who are still building processing facilities at or
3. FAusIMM, Chief Technical Officer, Ausenco near mine sites: new smelters have recently been constructed in Zambia; pressure
Minerals & Metals, South Brisbane Qld 4101. leaching, and roast, leaching and electrowinning plants are being built at some sites;
Email: greg.lane@ausenco.com and hundreds of millions of dollars have been spent on nickel-laterite leaching plants.
75
This paper considers the drivers for site-based processing •• higher capital costs
and outlines guidelines for project developers to assist in their •• lower transport costs
choice between on- and off-site processing options.
•• falling product prices
•• higher labour costs
FURTHER PROCESSING
•• higher employee turnover
The developers of new mining projects need to decide
how much processing there will be at the mine site. The •• lack of infrastructure
options range from direct shipment of ore with virtually no •• changing orebodies
processing (for example, hematite lumps and nickel-laterite •• spare capacity in existing processing facilities.
ore) through to the production of a finished commercial
product (for example, magnesium alloy wheels at Solikamsk Technological change
in Russia). Table 1 lists examples of different degrees of mine Concentration techniques significantly improved over the
site processing. past 120 years. Remember that the froth flotation process was
Hardly any mine sites take processing further than refined developed to treat sulfide ore and tailings in Broken Hill.
metal to fabrication of finished products. The operation at Before then, practically all the copper mined in the United
Solikamsk in Russia, where magnesium alloy automotive States came from underground mines chasing high-grade
wheels were produced for export at a magnesium chloride veins, averaging 2.5 per cent copper (Wills and Atkinson,
mine, is an exception to the rule. 1991). The high cost of transport could cripple lower-grade
In most cases, the selection of the degree of further deposits. Smelting was often a means of concentrating the
processing will be made on economic grounds. In some cases, valuable metals to reduce transport costs.
it might be dictated by political considerations; the former The past century has seen the invention and improvement
Soviet Union is not the only place where economics might be of froth flotation, improved grinding technologies allowing
ignored for ‘social’ reasons. However, it is the economic case better liberation of minerals (the development of the McArthur
that is the focus of this paper. River deposit depended on the development of ultra-fine
grinding techniques such as the IsaMillTM), the development
of solvent extraction – electrowinning (SX–EW), carbon-in-
DISCUSSION pulp, pressure oxidation, and heap leaching technologies.
The economics of on-site processing changed dramatically
Improved concentration techniques have lessened the need for
during the twentieth century. Some of the changes favour
local smelters. The fine grain of the Mount Isa lead–zinc deposit
reduced on-site processing, while others have worked to
made it impossible to produce high-grade lead concentrate,
increase it.
so a lead smelter was built at the mine. The McArthur River
mine, developed following the invention of the IsaMillTM fine-
Factors limiting on-site processing grinding technology, ships concentrate to smelters.
The factors limiting on-site processing include:
On the other hand, SX–EW has allowed the economic
•• technological change production of copper metal at site from low-grade oxide
deposits.
Table 1
Examples of materials shipped from mine sites. Higher capital costs
Smelters were once cheap to build. The Daydream smelter
Category Type of Examples of Examples of
was little more than a blast furnace and a chimney. Small-
material materials locations scale smelters built to last only a short time could be viable.
1 Ore Iron ore Pilbara, However, they had their costs: Blainey (1986b) wrote of the
Nickel laterite New Caledonia, Indonesia, problems of lead poisoning in Broken Hill – once strong men
Bauxite Philippines, throwing lead fits, the difficulty of raising kittens or puppies
Weipa … or even children. Such conditions are unacceptable today
2 Concentrates Cu Cadia, in nearly every corner of the world.
(flotation, gravity, Ni Cosmos, When the Mount Isa lead smelter was commissioned in
dense medium Zn Century, 1931, it was fitted with a baghouse to limit lead emissions and
separation, Sn Renison (when operating), so the citizens of Mount Isa were spared the lead problems
magnetic, Mineral sands Bemax, encountered by the early residents of Broken Hill.
electrostatic etc) Iluka Eucla Basin Emission controls increase the capital costs of smelting
3 Intermediate Alumina Wagerup, Pinjarra, and have contributed to a trend to build larger smelters.
products Nickel sulfide matte Kalgoorlie nickel smelter, High production rates reduce the relative cost of emission
Nickel hydroxide Ravensthorpe, controls by spreading their unproductive imposts over larger
Nickel oxide Goro quantities of product. Ramachandran et al (2003) showed
4 Crude metal Lead Mount Isa, that the size distribution of smelters has increased in most
Gold bullion Kalgoorlie Cons gold mine, countries over the years.
Blister copper Mount Isa (before 1979), These higher capital costs have made it impractical to build
Anode copper Mount Isa smelters for small mines with short lives and encouraged the
5 Refined metal Ni Murrin Murrin, sharing of capital expenditure through toll smelting by large,
Cu Olympic Dam often significantly depreciated smelters. It would be difficult
for an MIM to justify a copper smelter with the initial capacity
6 Manufactured Mg alloy wheels Solikamsk, of 15 000 t/a of the Mount Isa smelter (Pritchard, 1980). AGIP
product Russia Australia attempted such a feat with its Radio Hill smelter near

When does further processing at the mine site make sense?
Karratha in 1990, with the IsasmeltTM plant there designed to and action at the expense of understanding and planning.
produce 1.5 t/h of nickel–copper matte (Bartsch et al, 1990), Processing plants with a stable workforce will be at an
but this plant was quickly closed by low nickel prices (Player, advantage over those with high-turnover and chaos.
Fountain and Tuppurainen, 1992).
Lack of infrastructure
Lower transport costs Mine sites in remote locations usually lack the infrastructure
Lower effective energy costs and larger, more efficient trains, that is commonly available in industrial areas of the world.
trucks and ships have reduced the cost of transport. They The need to build power stations or long power lines, water
make possible the shipment of lower-grade (and hence value) storage and treatment facilities, and perhaps oxygen and other
products from mines, particularly those near coasts and ports. industrial gas production facilities, raises the capital investment
Thus, it is possible for nickel mines in New Caledonia and the required for further processing. Many of these facilities are
Philippines to ship ore containing about 1.6 per cent nickel
available almost on tap in developed industrial areas.
to Townsville for processing (Fittock, 2006). Lower transport
costs have reduced the benefit of mine site processing. Government assistance to single companies is often
constrained by international trade agreements, but
Falling product prices governments are less constrained when it comes to providing
Until recently, there has been long-term downward pressure infrastructure that will service multiple customers. It is
on metal prices (Fountain, 2002). In the mid-1990s, WMC difficult for a state or national government to provide a
Limited allowed an average 1 per cent real fall in prices each connection for a remote project to an electricity grid, but it can
year (Morgan, 1995). do so more easily if it is setting up an industrial estate that will
attract many companies.
This downward pressure, combined with rising costs,
compressed margins, made it harder to justify new capital
and increased the need to use existing facilities that had Changing orebodies
already depreciated capital. It encouraged brownfield rather Mining companies normally mine the easy deposits and the
than greenfield smelter capacity expansion. easy parts of deposits first. The shallow Broken Hill ores
mined in the 1880s were oxidised and readily smelted in cheap
Higher labour costs blast furnaces. As the surface mineralisation was depleted,
Metal prices have fallen relative to labour costs. In 1900, it took the miners reached the deeper sulfide deposits from which
an American on the average 84 minutes to earn the equivalent, the zinc minerals had not been leached. The lead–zinc ores
before tax, of the price of a pound of copper; in 2002, it was just could not be smelted economically with existing technology
over three minutes (Fountain, 2002). The resulting increased and the ‘sulfide problem’ provided the incentive to develop
standard of living in the developed world has reduced the the flotation process.
incentive for people to work in remote, undesirable regions. Many of the orebodies discovered today are complex.
The consequent workforce shortages have been compounded Complex orebodies can require complex treatment processes
by the actions struggling mining companies took to survive: and complexity increases capital cost.
reducing their support of universities and apprentice-training
schemes, and reducing their hiring and training of graduates The McArthur River deposit did not become an orebody
and other young people. until the IsaMillTM fine-grinding technology was developed
(Fountain, 2002). Even with grinding to an 80 per cent
To attract and retain a skilled workforce in residential mine
passing size of 7 µm, McArthur River produced a bulk lead-
site processing plants a premium must be paid to them. This
zinc concentrate that needed to be treated by the Imperial
leads to a further incentive to ship the material to a location
Smelting Process (Nihill, Stewart and Bowen, 1998). MIM
where such a workforce is more prepared to live or is cheaper
to employ. Holdings (now Glencore) bought the existing Avonmouth
and Duisburg lead-zinc smelters to treat McArthur River
Higher employee turnover bulk concentrate rather than building a new smelter for the
task. It subsequently sold the Duisburg smelter and closed
An increase in career opportunities has resulted in higher
the Avonmouth smelter, and sold the bulk concentrate to
employee turnover. The workforce was very stable at the
Broken Hill mines throughout most of the twentieth century. smelters operated outside the original MIM Holdings group
The mines paid well and the Barrier Industrial Council’s of companies.
policies favoured employing local men and single women
over married women and those people ‘from away’. Spare capacity in existing facilities
These days, however, mining companies that do recruit There are many established processing facilities around
skilled people are struggling to retain them. It takes time for the world. Construction of a new processing facility must
people to learn how to run a complicated process well. Many be weighed against using an existing facility with largely
employees leave before providing employers with a return on depreciated capital. Some of these facilities were constructed
the time and money invested in their training and development. to process ores or concentrates from local sources that have
The situation is further complicated by a management been depleted or are too high cost to compete with cheaper
philosophy that assumes that plant managers do not need imported feedstocks. Others sought to reduce their unit
to understand how the plant works – they just need to be costs by importing concentrate to supplement local sources.
able to manage the people who do. Such managers, who Examples included the Queensland Nickel nickel production
rely on others for the technical knowledge, often cannot tell plant at Yabulu; north of Townsville in Queensland; and
when they are getting poor advice. To make matters worse, the New Boliden smelters at Skelleftehamn in Sweden and
managers often change jobs frequently. Harjavalta in Finland.
High employee turnover makes it difficult to optimise The existing facilities most likely to be able to accept
complicated processes and creates a culture of fire-fighting imported materials are those near a port.

Other considerations Cominco, 2008a) but Teck-Cominco (now Teck Resources)

could not justify its use at its Highland Valley mine, which
Smelting sulfide ores and concentrates these days usually
produces readily marketable, high-grade concentrate (Teck-
results in sulfuric acid by-product. Smelters with local markets
Cominco, 2007). Vale built a 10 000 t/a CESL demonstration
for sulfuric acid, such as an established chemical industry or
plant to process copper-gold concentrate at its Sossego mine in
an acid-consuming leaching process, have an advantage over
Brazil’s Carajás region (Teck-Cominco, 2008b) with a view to
those that must transport acid over long distances, particularly
applying it to the high-fluoride Salobo concentrate (Defreyne
over land. Local acid demand helped the profitability of
and Cabral, 2009). Vale’s website makes no mention of using
Japanese smelters in the 1980s and the production of acid for
the process in its current Salobo flow sheet.
heap leaching helped the Miami smelter survive when many
other American copper smelters were closing.
High transport costs
Smelters and other complex processing technologies can
There are still parts of the world where transport costs are
have long ramp-up times, disadvantaging them against
high. Zambia, for example, is a landlocked African country
existing plants. Ramp-up times can be decreased by using
with transport routes through other African countries. Mopani
well-established technologies with good training programs Copper Mines, Konkola Copper Mines, First Quantum Minerals,
(McNulty, 1998; Arthur and Hunt, 2005). and a joint venture between China Nonferrous Metal Mining
(Group) Company Limited and Yunnan Copper Industry
Factors promoting further on-site processing (Group) Company Limited have either refurbished existing
Despite these disadvantages, there are circumstances that smelters or built new ones this century. The Mopani smelter at
promote on-site processing. These include: Mufulira (Ross and de Vries, 2005), the First Quantum Minerals
•• difficulty producing a saleable product smelter (Glencore Technology, 2016) and the Chinese smelter
at Chambishi use Glencore’s Copper IsasmeltTM technology.
•• problems concentrating valuable minerals
The Konkola smelter at Nchanga is using an Outokumpu flash
•• high transport costs furnace (Konkola Copper Mines, 2006).
•• additional transport problems Another example might be the Kennecott smelter near the
•• government subsidies Bingham Canyon mine in Utah. Rio Tinto spent US$1.1 billion
•• high treatment and refining charges on the new smelter and refinery that was commissioned near
Salt Lake City in 1996 (West, 1999).
•• new technologies
•• simple, low-cost process flow sheets Improvements in recovery
•• environmental considerations Additional on-site processing might improve recoveries of
•• government decree. valuable minerals. Concentrator operators are sometimes
forced to sacrifice recovery to achieve a concentrate grade
Product saleability that is economic to transport. Additional on-site processing
It is sometimes not possible to produce a salable product steps could allow higher recoveries by providing the ability
from a mineral deposit without further processing. This is to treat lower-grade concentrates. The Mount Isa lead smelter,
particularly the case at times of surplus supply when smelters for example, was designed for lead concentrate grades lower
and refiners can be selective in their feedstock purchasing and than typically transported due to the fine nature of the Mount
Isa ore and the difficulties experienced producing high-grade
can afford to eschew concentrates containing minor element
concentrate. As a result, the blast furnace is fluxed using
assemblages that are costly to treat or handle. A zinc smelter
higher levels of lime than commonly used in lead smelting
might reject a high-manganese zinc concentrate when times
(McLoughlin, Riley and McKean, 1980).
are good but purchase and blend it with other low-manganese
concentrates when zinc concentrate is scarce.
Additional transport problems
Western Australian mineral sands miners adopted the
Materials with high uranium levels can be difficult to permit
Becher synthetic rutile process to convert almost unsalable
for transport. WMC (now BHP) reduced the transport issues
grades of ilmenite into a premium synthetic rutile product
for its copper product by building a processing plant at
(George, 1993).
Olympic Dam to remove the uranium from the concentrate
St. Joe Minerals Corporation installed a multiple-hearth and to produce cathode copper at the site.
roaster at its El Indio mine in the Chilean Andes to remove
arsenic from its copper concentrate to make it marketable Government subsidies
(Smith et al, 1985). Governments sometimes offer subsidies to persuade
companies to maximise local processing. These might include
Ability to produce a concentrate assistance with infrastructure development, royalty discounts,
Some ores are difficult to concentrate with economic tax-free holidays, or outright cash grants. The Queensland
recoveries. Fine gold does not respond well to conventional government, for example, offered a royalty discount for
concentration techniques. Consequently, many gold downstream processing of nickel, cobalt, copper lead and
operations use some form of leaching followed by fire-refining zinc ores in the past (Queensland Government, 2003). The
to produce gold bullion. Gold can be produced economically government of Trinidad and Tobago has offered tax-holidays
from gold grades as low as 0.5 parts per million (Wills and to companies building smelters and petrochemical plants
Atkinson, 1991). This would not be viable without on-site using local natural gas.
processing. Even at $1500 per ounce, it would be uneconomic
to transport ore containing such a low concentration of gold High treatment and refining charges
any significant distance. Treatment and refining charges vary, depending on the
The Teck CESL hydrometallurgy process might be suitable balance between concentrate supply and smelter demand. The
for treating low-grade bulk copper-nickel concentrates (Teck- construction and expansion of copper smelters in China has

placed downward pressure on these charges, to the benefit of gangue minerals half-way around the world, perhaps to a
concentrate sellers. In the June 2005 – June 2006 period, BHP less-efficient plant. However, imposing strict carbon emission
was paying US$115 per t for treatment of its copper concentrate caps on resource industries developed-world nations is likely
and US$0.115 per pound for refining. They were able to reduce to have the perverse effect of driving metals processing
these charges to US$60 and US$0.06 for the period June activities to countries without caps, potentially resulting in an
2006 – June 2007 (Carlisle, 2006). Spot sales have achieved increase in total emissions.
even lower prices and The Sydney Morning Herald reported a
settlement at US$45 and US$0.045 between ‘an Indian smelter Government decree
and a major Western miner’ (Reuters, 2007). Treatment charges In some instances, governments require further processing to
subsequently rose progressively to be over US$100 and refining be undertaken as a condition of mining rights. Development
charges over US$0.10 by 2015 (Outotec, 2015). of Voisey’s Bay nickel deposit was delayed over the insistence
Low treatment and refining charges make justification of by the provincial government of Newfoundland and Labrador
further on-site processing difficult, but high treatment charges that Inco build a processing plant in the province.
tilt the balance in favour of building additional capacity. It is In 2008, the Zambian government introduced a 15 per cent
easier to justify building a smelter at a mine site when the levy on the export of copper concentrates to encourage further
industry as a whole needs additional capacity, although
processing in Zambia (Montia, 2008).
greenfield capacity still needs to compete with cheaper
brownfield expansions. In such circumstances, the developer must determine
whether the deposit can withstand the additional capital
New technologies impost, and either comply or walk away.
New technologies can favour additional mine site processing.
The development of SX–EW technology led to increased
Comments on specific metals
production of cathode copper at mine sites, particularly from
low-grade oxidised copper orebodies. The development of
Aluminium
IsasmeltTM technology has lowered the cost of copper smelting. Aluminium is produced from bauxite ores. Some mines have
Carbon-in-pulp, BIOX® and other new technologies have associated refineries that produce alumina as an intermediate
resulted in more gold mines being developed, mines at which product (for example, Alcan’s Gove operations), while
on-site processing is a necessity. Relatively new technologies others ship bauxite to refineries (for example, Comalco’s
such as BioheapTM and variations on pressure oxidation are Weipa operations). Aluminium smelting consumes large
emerging as potential competitors to the established secondary amounts of electrical power (Australian smelters consumed
processing options, principally smelting, for copper sulfides. about 15 MWh for every tonne of aluminium produced in
2006 (Australian Aluminium Council, 2006)). Consequently,
One of the driving forces for the recent implementation
of pressure oxidation technology for copper concentrate smelters tend to be built where electricity prices were cheap at
treatment at Freeport McMoran’s Cyprus Bagdad operation the time of construction – near hydroelectricity in Washington
was the production of acid on-site for use in its oxide heap state, Quebec, Tasmania and Norway, and increasingly
leach operation. Similarly, the Sepon operation in Laos uses through natural gas production in Gulf states.
the pressure oxidation of pyrite to provide ferric ion and acid
to the copper leaching process. Copper
In the past few decades, most owners of new sulfide copper
Simple, low-cost process flow sheets mines have elected to install concentrators at mine sites and
Given the difficulty in attracting and retaining skilled ship concentrates to smelters located elsewhere. As discussed
operators and professionals to mine sites, it is an advantage earlier, smelters have become larger and more capital-
if the process is simple and robust. Simplicity tends to mean intensive with time, making them less likely to be built
low-capital flow sheets. Low-capital reduces the economic on-site.
risks of on-site processing. China’s late start to heavy industrialisation, its extreme
demand for materials, and a focus on economic growth ahead
Environmental considerations of environmental concerns has allowed smaller Chinese
Processing concentrate to extract valuable metals produces smelters to remain economically viable for longer. However,
waste products, whether the process is smelting or economic, governmental and environmental pressures are
hydrometallurgical. Waste disposal can be difficult in heavily driving the move to larger production plants with associated
populated areas. A mine site is often the best place to dispose oxygen and acid plants. Sulfur capture in Chinese smelters
of those wastes. The slags from the Mount Isa smelters can be was around 85 per cent in 2006, the same as the Chilean rate,
used as underground fill (their pozzolanic properties make but below the world average of 90 per cent, and 95–97 per cent
them a substitute for cement). level in developed countries (Diaz and Mackey, 2007).
Slags and other metallurgical wastes can also be deposited in Except for Zambia and possibly China, most recent increases
mine site tailings storage facilities. Titanium dioxide pigment in capacity have come through upgrades of existing smelters,
producers in developed countries prefer to use the chloride including the expansion of Southern Copper’s Ilo smelter
process with high-grade feed (rutile, synthetic rutile or high- and Vedanta’s Tuticorin smelter using new, large-capacity
grade titania slag) because the waste disposal costs of the older IsasmeltTM furnaces.
sulfate route are significantly higher. This has encouraged Hydrometallurgical processing options for copper minerals
the production of synthetic rutile in Western Australia and increased in importance following the development of solvent
upgraded slag at Rio Tinto Iron and Titanium Inc in Quebec. extraction, which largely eliminated the cementing process
Transporting ores or concentrates uses more energy than with its transport of scrap iron. Processing is still limited by the
transporting final product. If human emissions of carbon acid consumption of the ore and availability of acid, but new
dioxide are causing the world to warm, it would often be more technological developments, such as pressure leaching, might
sensible to produce metals at the mine site than transporting lead to additional on-site processing. One way of coping with

acid demand is integration of different processes so that acid Lead

generated in one is available for leaching in another, reducing
Most lead concentrates are exported to distant smelters.
net acid requirements and avoiding transport issues. BHP’s
However, mine site smelters might be warranted if the
Olympic Dam complex uses smelter acid to leach uranium.
concentrate grade is low, incurring high transport costs, or if
Another approach, recently implemented, is to use pressure
the ore is composed of oxidised lead minerals. In the latter
oxidation of sulfide concentrates to generate acid for leaching,
case, lead metal poses less risk to communities than moving
thus avoiding the cost of an acid plant; examples include Sepon
more-soluble forms of lead, such as lead carbonate.
in Laos (Sherrit, Pavlides and Weekes, 2005), Kansanshi in the
Congo, and Phelps Dodges’ Baghdad and Morenci mines in Oxidised lead minerals can be smelted with a relatively
Arizona (Dreisinger, 2006). low capital cost using technologies such as rotary furnaces,
which are commonly used for recycling lead sulfate battery
Current hydrometallurgical technologies offer a range
materials (Rao, 2006). The product is crude lead bullion that
of processes – biological, fine grinding, ferric leaching and
can be exported for refining.
pressure oxidation (Dreisinger, 2006) – that between them
cater for different feeds (ore or concentrate) and different
levels of acid demand from other on-site processes. No one
Nickel
process suits all applications. For sites with existing SX–EW Limonite nickel-laterite (low-grade iron ore with around
facilities that are running out of oxide ore, hydrometallurgical 1–1.5 per cent Ni) suitable for pressure acid-leaching (PAL) or
processing offers a way of continuing on-site processing using Caron processing have typically traded at 10 per cent of the
existing equipment. nickel value in the ore, while saprolite (magnesium silicate
ores with around 2–2.5 per cent Ni) suitable for smelting to
Gold ferronickel or matte have typically traded at 25 per cent of the
Gold metal is conventionally produced at the mine site; value of the nickel content (Australian Mining Journal, 2001).
payable ore grades are at the g/t level and the capital cost With development of PAL, the economics of laterite shipping
of cyanide leaching and associated gold recovery is low are becoming marginal.
unless the ore is refractory. Even when refractory ores
require pretreatment to make them amenable to conventional Guidelines for decision-making
cyanidation, on-site processing is still better than transporting The authors offer the following guidelines to help organisations
ores long distances. This is normally the case even when the with their decisions about the extent of on-site processing.
contained value in the concentrate exceeds that of base metal
concentrates. An exception occurs when such an ore can be Is there an existing facility that will buy an unprocessed product?
concentrated and transported to an existing facility close This is the first question that should always be asked. Using
enough that the capital saved offsets the transport costs. existing facilities limits the capital expenditure for a new
Oceana Gold’s Reefton (New Zealand) mine is an example: mining project. However, treatment charges and transport
Reefton produced a pyritic flotation concentrate containing costs might make new processing facilities attractive relative
highly-refractory, sulfide-encapsulated gold at about 30 g/t; to existing ones.
the concentrate was transported 600 km by rail to Oceana’s
Shortages in existing processing capacity will result in
plant at Macraes Flat for treatment by pressure oxidation in
higher treatment charges and a better case for building new
an existing autoclave. The Reefton operation was placed on
capacity. However, new capacity needs to come in at the
care and maintenance at the end of 2015 (OceanaGold, 2016).
lower end of the operating cost-curve so that the operator is
Payment for precious metal values in concentrates can be a not left with a high-cost asset when the shortage abates and
factor. Smelters normally pay for gold and silver recovered treatment charges fall.
during smelting or refining. However, treatment charges
Special issues, such as the presence of deleterious minor
and the effective costs of late payment can be avoided if the
elements (for example, fluorine, arsenic or radioactive
gold can be recovered by gravity treatment (Gray, Katsikaros
elements), might make a product difficult to market in an
and Fallon, 1999) and as practiced at Newcrest’s Cadia and
untreated form.
Telfer operations.
There is no payment of gold value in antimony concentrates, Undertake a value-chain analysis
making alternative on-site processing more favourable. New
It is important that mining companies understand where the
England Antimony Mines Limited used a thiourea leach to
value is added to their products. This means examining the
extract gold from its antimony concentrate (Hisshion and
potential net profit at each step of the processing chain and the
Waller, 1984).
total return on capital invested. It would include considering
Iron the cost of transport and relative energy costs at the mine site
and at potential alternative sites.
Extensive economies of scale enable long-distance rail and
maritime hematite ore transport. Between 1997 and 2000 Such an analysis will usually identify a logical point to hand
Australian free-on-board (FOB) costs fell from US$9.71– processing to another facility.
7.08/t of iron ore (AME, 2002). With the failure of BHP’s hot-
briquetted iron process, there is currently little appetite for Is the project capital-constrained?
further processing of hematite ore near the mine site. On the If a project is capital-constrained, it is often better to sell
other hand, upgrading magnetite ore to produce concentrate lower-grade products than spend additional capital to
pellets can reduce transport costs sufficiently to justify the upgrade them. The situation can change over time as a
investment (for example, the Savage River and Port Latta positive cash flow develops and provides financial resources
complex in Tasmania). for further investment.

Are there government incentives or mandates for further processing? Mount Isa over a period of six months and, as a result, set a
new benchmark for smelter commissioning (Arthur and Hunt,
Determine whether any government incentives or mandates tilt
2005). A new processing plant is more likely to be successful if
the balance in favour of further processing. Are the incentives
the plant’s owner can find a similar plant on which to train its
sufficient to provide a return on investment that justifies the
workforce before commissioning.
additional processing? If there are further processing mandates,
is the project robust enough to withstand any opportunity costs
created by this mandate? If not, the opportunity to lose money
Consider the complexity of the process
is probably better left to someone else. Simple processes are more likely to be successful at the mine
site. The greater the complexity, the more difficult it will be to
Care needs to be taken to ensure that government incentives
find a workforce capable of sustaining the process and plant
are incentives. Tax-holidays to encourage companies to a
performance used in the initial economic justification.
region might not be as valuable as first thought once the tax
effects of plant depreciation in the early years of operation It is difficult to achieve high plant availabilities with highly-
are considered. complex processing systems with minimal surge capacity,
and long ramp-up times reduce or destroy a project’s net
Consider the valuable metal content of the product present value.
Table 2 shows typical metal contents in various materials.
Avoid the economies-of-scale trap
Most materials transported have a valuable metal content
greater than 25 per cent. Materials containing less valuable When a project is uneconomic at the initial scale, the
metal than this would normally require further processing. temptation is to increase output to reduce unit costs through
economies of scale. This can produce significant benefits, but
increasing size to achieve a target rate of return carries risks.
Transport distances These include an increased financial loss if the project fails
Long inland transport routes are more likely to favour on- to achieve its cost, quality and throughput targets, and the
site processing. Economies of scale and dedicated railways possibility that increasing the scale of production will alter the
might mitigate this effect, as in the case of the Pilbara railways supply–demand balance, depressing the price of the product.
owned and operated by Rio Tinto and BHP. This latter effect increases the risk to the project.
Are there environmental reasons for further processing at site? What is the projected mine life?
As discussed, there might be environmental reasons to The mine life needs to be sufficient to allow a reasonable
upgrade a product at site. If there is an environmental driver, return on the investment in additional processing. The larger
site-based processing might make the difference between the capital expenditure and the more complex the process, the
receiving regulatory approval to proceed and a refusal. greater the mine life needs to be.
As with government mandates for further processing for
If the project is near a port, it is possible that the processing
social reasons, the project proponent will need to determine
plant could operate well beyond the life of the orebody by
whether the project is sufficiently robust to proceed under
treating imported ore or concentrate.
these circumstances.
Resist justifying on-site processing by selling potential by-products
Consider the ability to attract a workforce
The authors are aware of projects where the principals
If it is relatively easy to attract a high-quality workforce, were keen to undertake further processing. The proposed
additional processing options are more likely to be successful. processing plants were not economic in their own right, so
It is better to have an experienced core group to build the rest their ‘project champions’ sought to bolster the economics
of the workforce around than to try to train an entire staff with by selling by-products. In one case, this led to a scenario
no experience. Inexperienced employees will make mistakes, of a billion dollar (1990s dollars) industrial complex in the
extend ramp-up times, and perhaps develop a culture that Australian outback to develop an 8 Mt tonne mineral deposit.
prevents the project ever realising its potential. In another case, there was a short period of wild enthusiasm
when it was thought that by-products previously regarded as
Opportunities for training wastes (with associated disposal costs) had such a huge return
Good training, particularly hands-on training, is important that they converted a marginal project into a potential bonanza
when commissioning new processes. When the Yunnan … until it was realised that overly optimistic assumptions had
Copper Corporation chose to build an IsasmeltTM plant at its been made about the value of the by-products and the capital
smelter in Kunming, it sent metallurgists and operators to and operating costs of producing them.
Table 2
Relative metal contents.
Class of material Iron Aluminium Copper Nickel Lead

Ore Ore (55% Fe) Bauxite (26–30% Al) Limonite (1.5% Ni), Saprolite (2.5% Ni)
Concentrate Magnetite pellets Concentrate (25–37% Cu) Concentrate (12–27% Ni) Sulfide (50–70% Pb),
Carbonate (60–70% Pb)
Intermediate DRI or HBI (93% Fe), Alumina (52% Al) Matte (40–80% Cu) Precipitated mixed sulfides (23% Ni), Oxide
HIron™ (96% Fe) or hydroxide (40% Ni), Matte (45–72% Ni)
Crude metal Pig iron (93–95% Fe) Blister Cu (95–99.5% Cu) Lead bullion (97% Pb)
Refined metal Steel Refined Al (>99.7% Al) Refined Cu (>99.85% Cu) Refined Ni Refined Pb (>99.9% Pb)
Note: DRI – direct-reduced iron; HBI – hot-briquetted iron.

Miners can feel offended by lack of payment of ‘full value’ One of the key actions that a company developing a
for minor elements (for example, gold, silver, indium, new deposit need to undertake is a value-chain analysis
and cobalt). However, extracting these elements requires to understand the logical end point of processing. The end
additional processing steps that add to complexity (Fountain, point will depend on the type of mineral mined and the
2013). Careful consideration of the capital and operating costs location of the mine. It might be affected by government
of these additional steps is needed to justify their construction. incentives or subsidies.
By-products might provide a supplementary income stream, On-site processing facilities will be easier to justify in times
but projects that depend on this stream are often not robust. when demand for processing exceeds global capacity (as
Exceptions include the use of sulfuric acid for leaching. occurred during the decades of rapid growth after World War
II) than times with surplus processing capacity.
Consider the quality of the existing infrastructure at the mine site Advocates of additional mine site processing need to keep
Additional mine site processing will increase the demand for a clear focus on the economics of the process. Increasing
water and electricity in particular. The availability of plentiful economies of scale or earning additional income through selling
water and cheap electricity can make the difference between by-products might not be as attractive as at first glance. Low-
shipping a relatively low-grade material and upgrading cost and simple flow sheets are easier to operate and expose the
it at the mine site. The need to install additional electricity company to less risk than complex, high-cost flow sheets.
generating capacity or power lines can often kill upgrading
projects in remote areas. REFERENCES
AME Mineral Economics, 2002. Australia boasts lowest iron ore
Will further processing improve metal recoveries? production costs in 2001 [online], Reed Business Information.
If further processing results in higher recoveries, are they Available from: <http://www.ferret.com.au/> [Accessed: 11
sufficient to justify the additional capital and operating costs? March 2008].
Arthur, P S and Hunt, S P, 2005. IsasmeltTM – 25 years of continuous
What effect will complexity have on development schedule? evolution, in Sustainable Developments in Metals Processing
(eds: M Nilmani and W J Rankin), pp 73–94 (The Australasian
Increasing the complexity of the project will increase the
complexity of the permitting and ultimate financing of the
project, and increase the development duration. Does this fit Australian Aluminium Council, 2006. Sustainability report 2006
[online]. Available from: <http://www.aluminium.org.au/>
with the owner’s capability and expectation?
[Accessed: 28 March 2008].
When should novel processes be considered? Australian Mining Journal, 2001. Trade in nickel-cobalt laterite ores
[online]. Available from: <http://www.theajmonline.com/>
Novel processes are sometimes considered as a saviour for [Accessed: 9 March 2008].
processing difficulties with complex orebodies. When used in
Bartsch, P, Fountain, C and Anselmi, B, 1990. Radio Hill project, in
greenfield locations, the application of novel technologies can
Pyrosem WA (eds: E J Grimsey and N D Stockton), pp 232–250
suffer from over enthusiastic expectations of performance, (Murdoch University Press: Perth).
operating cost and capital cost. A recent example was the
Blainey, G, 1968a. The Rise of Broken Hill, pp 20, 24 and 63 (Macmillan
development of the PAL technology in Western Australia,
of Australia: Melbourne).
which resulted in underachievement of expectations for on-
site nickel production from laterite ores on three simultaneous Blainey, G, 1968b. The Rise of Broken Hill, pp 86–95 (Macmillan of
projects. Extensive pilot and demonstration plants are Australia: Melbourne).
required to establish the technical and cost drivers for novel Carlisle, C, 2006. BHP Billiton secures big drop in Japan copper
processes and applications, particularly where the project is smelter charges, Mineweb, 5 October 2006.
the only or major source of cash flow for the company. Defreyne, J and Cabral, T, 2009. Early copper production results from
Vale’s hydrometallurgical CESL refinery [online]. Available from:
<http://www.teck.com/media/CESL-Publication-Copper-Vale-
CONCLUSIONS Metallurigcal-Results-ALTA-2009.pdf> [Accessed: 18 July 2017].
While many lament the export of relatively unprocessed
Diaz, C and Mackey P, 2007. Discussion of copper smelting
products from mines, the decline in mine site processing in technology practice and innovation [online], in Proceedings
recent years has been driven by a variety of economic and The Carlos Diaz Symposium on Pyrometallurgy – Cu 2007,
social factors. These include increasing capital costs associated vol 3. Available from: <http://cim.v1.myvirtualpaper.com/
with emissions controls, the existence of large facilities that report/2007061901/?page=10> [Accessed: 9 March 2008].
toll-smelt or buy concentrates or ores, the rise of Chinese Dreisinger, D, 2006. Copper leaching from primary sulfides: Options
smelters over the past decade, poor mine site infrastructure, for biological and chemical extraction of copper, Hydrometallurgy,
and an increasing reluctance by skilled people to work in 83:10–20.
remote areas. Ehrlich, P R, 2007. Keynote address to the Ecological Society
There are still circumstances that favour mine site of Australia’s annual conference, Perth, 25–30 November
processing. These include: 2007 [online]. Available from: <http://www.abc.net.au/
rn/inconversation/stories/2007/2109181.htm#transcript>
•• difficulty of producing a marketable product without it [Accessed: 24 February 2008].
•• problems concentrating valuable minerals (such as fine, Fittock, J, 2006. Nickel and cobalt refining by QNI Pty Ltd, Yabulu,
disseminated gold) Qld, in AusIMM Monograph 19 Volume. Available from: <http://
•• high transport costs www.ausimm.com.au/Content/docs/mono19model_paper.
pdf> [Accessed: 24 February 2008].
•• simple, low-cost process flow sheets
Fountain, C, 2002. IsasmeltTM and IsaMillsTM—models of successful
•• environmental considerations R&D, in Proceedings AusIMM 2002 Young Leaders’ Conference (The
•• government sanction and approval. Australasian Institute of Mining and Metallurgy: Melbourne).

Fountain, C, 2013. The whys and wherefores of penalty elements Pritchard, J, 1980. Copper smelting at Mount Isa Mines Ltd, Mount
in copper concentrates, in Proceedings MetPlant 2013 (The Isa, Qld, in Mining and Metallurgy Practices in Australasia
Australasian Institute of Mining and Metallurgy: Melbourne). (ed: J T Woodcock), pp 340–344 (The Australasian Institute of
George, I W, 1993. Synthetic rutile operations of RGC Mineral Sands
Limited at Capel and Narngulu, WA, in Australasian Mining and Queensland Government, 2003. Explanatory notes for SL 2003
Metallurgy – The Sir Maurice Mawby Memorial Volume, Second No. 174, Mineral Resources Regulation 2003 [online]. Available
Edition (eds: J T Woodcock and J K Hamilton), pp 1301–1304 (The from: <http://www.legislation.qld.gov.au/legisltn/sls/ris_
Australasian Institute of Mining and Metallurgy: Melbourne). en/2003/03sl174e.pdf> [Accessed: 9 March 2008].
Glencore Technology, 2016. Installations – IsasmeltTM [online]. Ramachandran, V R, Diaz, C, Eltringham, T, Jiang, C Y, Lehner,
Available from: <http://www.isasmelt.com/EN/Installations/ T, Mackey, P J, Newman, C J and Tarasov, A, 2003. Primary
Pages/Installations.aspx> [Accessed: 20 November 2016]. copper production – a survey of world copper smelters [online],
in: Volume IV: Proceedings Pyrometallurgy of Copper, Herman
Gray, S, Katsikaros, N and Fallon, P, 1999. Gold recovery from copper
Schwarze Symposium, Book 1, pp 3–106 (Canadian Institute:
gold gravity concentrates using the Inline Leach reactor and weak
Montreal Canada). Available from: <http://www.hatch.
base resin [online], in Proceedings Oretest Copper Gold Symposium.
ca/non_ferrous/articles/primary_copper_production.pdf>
Available from: <http://www.gekkos.com/documents/17-gold
[Accessed: 9 March 2008].
recoveryfromcoppergoldgravityconcentratesusingtheinlineleach
reactorandweakbaseresin.pdf> [Accessed: 9 March 2008]. Smith, E H, Foster, J W, Minet, P H and Cauwe, P H, 1985.
Selective roasting to de-arsenify enargite/pyrite concentrate
Hisshion, R G and Waller, C G, 1984. Recovering gold with thiourea,
from St. Joe’s El Indio Mine – from pilot plant to commercial
Mining Magazine, September: 237–243.
operation, in Complex Sulfides: Processing of Ores, Concentrates
Johnson, R E, Themelis, N J and Eltringham, G A, 1979. A survey and By-products (eds: A D Zunkel, R S Boorman, A E Morris and
of worldwide copper converter practices, in Copper and Nickel R J Wesely), pp 421–440 (The Metallurgical Society: Warrendale,
Converters (ed: R E Johnson), pp 1–32 (The Metallurgical Society Pennsylvania).
of the American Institute of Mining, Metallurgical, and Petroleum
Rao, S R, 2006. Resource Recovery and Recycling from Metallurgical
Engineers: New York).
Wastes, pp 199–200 (Elsevier: Oxford).
Konkola Copper Mines Plc, 2006. Expansion projects [online].
Reuters, 2007. Japanese smelters, BHP discuss copper fees [online],
Available from: <http://www.kcm.co.zm/expansion_projects.
The Sydney Morning Herald, 6 December. Available from:
php> [Accessed: 28 March 2008].
<http://www.smh.com.au/news/Business/Japan-smelters-
McLoughlin, B C, Riley, J F and McKean, G R, 1980. Development BHP-discuss-copper-fees/2007/12/06/1196812912011.html>
of optimum fluxing procedures for Isamine lead concentrates, in [Accessed: 28 March 2008].
Lead-Zinc-Tin’80 (eds: J M Cigan, T S Mackey and T J O’Keefe)
Roberts, J, 2007. BHP denies delays at Olympic mine [online],
pp 309–320 (The Minerals, Metals & Materials Society:
The Australian, 8 September. Available from: <http://www.
Warrendale, Pennsylvania).
theaustralian.news.com.au/story/0,25197,22382128–5005200,00.
McNulty, T, 1998. Developing innovative technology, Mining html> [Accessed: 24 February 2008].
Engineering, 50(10):50–55.
Ross, J and de Vries, D, 2005. Mufulira smelter upgrade project
Montia, G, 2008. Zambia introduces 15% copper concentrates levy – ‘Industry’ smelting on the Zambian copper belt [online].
[online]. Available from: <http://www.metalmarkets.org. Available from: <http://www.isasmelt.com/downloads/
uk/2008/03/10/zambia-introduces-15-copper-concentrates- XTpaper_Mopani_MSUPyromet05.pdf> [Accessed: 9 March
levy/> [Accessed: 29 March 2008]. 2008].
Morgan, H, 1995. Interview on The Bottom Line, Australian Sherrit, R, Pavlides, A G and Weekes, B L, 2005. Design and
Broadcasting Corporation Television, 18 February 1995. commissioning of the Sepon copper pressure oxidation circuit,
Nihill, D, Stewart, C M and Bowen, P, 1998. The McArthur River in Proceedings First Extractive Metallurgy Operators Conference,
mine – the first years of operation, in Proceedings 1998 Annual pp 21–27 (The Australasian Institute of Mining and Metallurgy:
Conference, pp 73–82 (The Australasian Institute of Mining and Melbourne).
Metallurgy: Melbourne). Teck-Cominco, 2007. Annual Report 2006 [online]. Available from:
OceanaGold, 2016. OceanaGold operates mines in the three largest <http://www.teckcominco.com/DocumentViewer.aspx?eleme
goldfields in New Zealand [online]. Available from: <https:// ntId=115704&portalName=tc>, p 21 [Accessed: 31 March 2008].
www.oceanagold.com/our-business/new-zealand/> [Accessed: Teck-Cominco, 2008a. Annual Report 2007 [online]. Available from:
20 November 2016]. <http://www.teckcominco.com/DocumentViewer.aspx?eleme
Outotec, 2015. Copper markets and TC/RCs [online]. Available from: ntId=124227&portalName=tc>, p 26 [Accessed: 31 March 2008].
<http://www.outotec.com/en/Search-material/Customer- Teck-Cominco, 2008b. Applied technology [online]. Available from:
newsletter-Smelting-News/Issue-1-February-2015/Copper- <http://www.teckcominco.com/Generic.aspx?PAGE=Technolo
markets-and-TCRCs/> [Accessed: 20 November 2016]. gy+%26+Innovation+Pages%2FResearch+Centre+Pages%2FCES
Player, R L, Fountain, C R and Tuppurainen, J M I T, 1992. IsasmeltTM L&portalName=tc> [Accessed: 31 March 2008].
developments, in Mervyn Willis Symposium and Smelting & Refining West, G, 1999. Breakdown leaves giant copper furnace out of action,
Course, July, pp 20:1–20:13 (The University of Melbourne). The Australian Financial Review, 12 January, p 22.
Wills, B A and Atkinson, K, 1991. The development of minerals
engineering in the 20th Century, Minerals Engineering, 4:643–652.

Contents
The ABC of Mine-to-Mill and metal price cycles

P Cameron1, D Drinkwater2 and J D Pease3
ABSTRACT
In the 1990s metal prices were trending in a long-term decline and the usual cost-cutting
exercises were adopted. Cuts to research and development made innovation difficult
and the outlook was grim. But necessity is a strong motivator, and these conditions
spurred the Mine-to-Mill movement, which optimised across organisational silos and
utilised new technology tools, innovative software and increased computing power.
Many applications of Mine-to-Mill exploited the fact that comminution is usually
the site processing bottleneck, and that blasting is more efficient at breaking rock than
grinding. This approach sought to:
•• understand and characterise rock breakage from mining to the mill
•• develop models and simulators for blast design, fragmentation, crusher and
mill circuits
•• develop tools to measure in real-time the particle size distribution of rocks on
conveyors and run-of-mine (ROM) muck piles
•• ensure effective communication across the silos between geologists, blast design
engineers, mining engineers and metallurgists.
These methods and the technology tools available in the 1990s were widely adopted
at numerous sites around the world with significant benefits. Semi-autogenous
grinding (SAG) mill throughput increases of 10–20 per cent were common. This
was the advance the industry desperately needed. It was low capital cost; it was
obvious. It was here to stay.
Except, in too many cases, it didn’t. In the minerals price boom of the 2000s Mine-
to-Mill was no longer necessary to ‘survive’, even though it was still good business.
Some operations, including some of the early success stories, slipped back to old
habits of optimising within organisation silos instead of across them. Fortunately,
other operators still embraced Mine-to-Mill and developed new techniques. Mine-to-
Mill was still successful, but not as widely adopted as expected.
Now the boom has ended and operations are again under cost pressure. Since
these are the same circumstances that created Mine-to-Mill, it seems time to achieve
the wider adoption.
Since the 1990s many new or advanced technology tools are available for Mine-to-
Mill projects: blasthole sensors, new explosives formulations, new blasting techniques
and modelling, ore tracking devices, improved image analysis to determine size,
texture and colour of coarse ore, grade sensors, whole-stream simulation tools, even
more powerful computing hardware and data analysis software.
If we could achieve so much in the 1990s, how much more can we achieve today
when we have the same imperative, the same potential and a larger number of high
technology tools?
INTRODUCTION
The Mine-to-Mill methodology was developed during the commodity price downturn
of the 1990s and was widely adopted at numerous sites around the world. Mine-to-
Mill relies on the fact that comminution is usually the site processing bottleneck, and
that blasting is more efficient at breaking rock than grinding.
1. MAusIMM, General Manager Australia, This approach and the technology tools available in the 1990s were widely adopted
Split Engineering, Mission Beach Qld 4852. around the world with significant benefits. Semi-autogenous grinding mill throughput
Email: pcameron@spliteng.com
increases of 10–20 per cent were common. We thought this approach to managing
2. MAusIMM, Principal Consulting Engineer, operations was here to stay, since it appeared to be free money for operators; however
Mineralis Consultants Pty Ltd, Brisbane Qld
during the subsequent boom and focus on ‘production at any cost’, while some
4066. Email: ddrinkwater@mineralis.com.au
operators advanced the approach, many reverted to old habits of optimising within
organisation silos. Mine-to-Mill wasn’t crucial to their survival.
Engineer, Mineralis Consultants Pty Ltd,
Brisbane Qld 4066. Now that the commodity cycle is back in downturn, the simple productivity gains
Email: jpease@mineralis.com.au offered by Mine-to-Mill look attractive again. Since the 1990s many new or advanced
85
P Cameron, D Drinkwater and J Pease
technology tools have been developed to enhance the approach. TABLE 1

Now is the time to adopt the new technologies, and to embed Mine-to-Mill case studies 1996 to 2002.
them into a more universal and lasting industry change.
Mine site Metals produced Country Product increase
PAST PRACTICE AND ESSENTIAL REQUIREMENTS Highland Valley Cu Canada 10%
In its broadest sense, Mine-to-Mill integrates all aspects of Alumbrera Cu Au Argentina 13%
geometallurgy and production steps with processing and
marketing. In this paper, the authors confine their comments Porgera Au Papua New Guinea 15%
to the subset of Mine-to-Mill that focuses on the integration KCGM Fimiston Au Australia 18%
of blasting with comminution and separation. They refer to Cadia Au Cu Australia 14%
this as ‘Advanced Blasting for Comminution’ – the ABC of
Mine-to-Mill. Red Dog Zn USA 12%
An excellent summary of the steps involved in any Mine-to- BHP Iron Ore Fe Australia 3% increase in lump
Mill project groups them into three main areas (McKee, 2013): KCGM – Kalgoorlie Consolidated gold mine.
1. First and foremost, good data. Data needs to be collected
about the ore before it is mined and as it passes through and on the conveyors from the primary crusher to the
the production chain, about the equipment and processes SAG mill feed
used in production, and the process performance • JKSimMet, simulation software for comminution circuits.
and cost. There was minimal information about mineralogy, and
2. Good, robust analytical tools and models to evaluate work-arounds had to be devised to deal with complex,
options and identify optimum operating points for a multicomponent ore types and non-standard mining and
range of feed types and conditions. The best models also processing scenarios. Early Mine-to-Mill projects relied
account for economic factors such as metal price and heavily on the knowledge, experience and desire for
operating cost. cooperation of the project team. Refer to Figure 1 for a
3. Finally, any Mine-to-Mill project requires tools for schematic representation of the work process.
ongoing monitoring, assessment and evaluation. Importantly, the projects and the survival imperative of the
Findings need to be validated and optimisation kept on- times engendered dialogue, collaboration and cooperation
track to fully realise the benefits. between geologists, blast and mining engineers, metallurgists
Further, these non-technical factors are critical: and General Managers (Kanchibotla et al, 1998; Lam et al,
2001; Valery et al, 2001). In 2012, Karen McCaffery suggested:
•• sustained management support
Mine-to-Mill (and geometallurgy) is just code for making
•• availability of staff with specialist skills
the effort and putting the processes in place to record, in
•• an enabling organisation structure. an accessible format, an understanding of the orebody,
Early Mine-to-Mill projects included scheduling to smooth how changes in the orebody and operating practice drive
ore variations, building stockpiles to ‘campaign’ different productivity and production, and understanding the
ore types, or redesigning underground activities to eliminate operating parameters in the mine and mill which can be
‘cash negative’ ore while rescheduling surface operations to manipulated to improve productivity and operating cost.
eliminate the supposedly ‘fixed’ costs associated with them. It is what people in mining and processing at sites should
These examples show that Mine-to-Mill projects did be doing as a normal part of their day-to-day business.
whatever was required to improve overall mine site (personal communication, 2012)
performance. They were orebody and situation specific. Many
case studies reported significant gains, typically in the range NEW TOOLS AND NEXT GENERATION OF MINE-TO-MILL PROJECTS
10–20 per cent productivity improvement across the mine Since 2002, there have been marked developments in image
site, with little or no capital expenditure. These were dramatic analysis, GPS, simulation software, radio-frequency tracking
improvements by any measure compared with working in devices for ore, in-plant instrumentation to measure flows,
isolated silos (McKee, 2013). online particle size monitors, mineral liberation analysis,
geometallurgy and equipment monitoring instrumentation.
MINE-TO-MILL IN 1996 Though some early adopters lost their way, new adopters
By 1990 higher speed computational power enabled achieved outstanding improvements, such as Antamina’s
innovative software for mathematical modelling and 45 per cent increase in SAG mill throughput followed by a
simulation of industrial processes including mining and further 10 per cent later (Rybinsky et al, 2011; Valery et al, 2012).
mineral processing. The AMIRA Project ‘Optimisation of Other case studies have been discussed by Bennett et al (2014);
Fragmentation for Downstream Processing’ (1996 to 2002; Hart et al (2011); Dance et al (2007); Diaz et al (2015); Renner et al
Table 1) was a collaborative research project to exploit this (2006); McCaffery et al (2006) and Gomes et al (2010).
capability and develop operating strategies to enhance mining Now the productivity imperative has returned it is time for
and downstream processing activities. The history, concepts more operators to learn from the old and the recent successes
and case study projects are presented in McKee (2013). and adopt the ABC approach using the new technology tools,
In 1996 the technology tools available to Mine-to-Mill were which include:
limited to elementary versions of tools such as (or similar to): •• more complex software for blast design, analysis and
• JKSimBlast blast design software management
• Kuz–Ram fragmentation model to predict particle size •• advances in fragmentation modelling
distribution of the blast in the stockpile •• image analysis with comprehensive rock fragmentation
• Split-Online, or other image analysis of the particle software for automated particle sizing at truck dump
size distribution at the truck tip to the primary crusher, and conveyor belt locations

FIG 1 – Extract from the Julius Kruttschnitt Mineral Research Centre (JKMRC) Mine-to-Mill brochure 1999.
•• end-to-end modelling packages such as the Integrated •• high energy explosives to improve post blast
Extraction Simulator (IES) developed by CRC ORE fragmentation (Hawke and Dominguez, 2015)
(Cooperative Research Centre Optimising Resource •• electronic detonators programmable in one millisecond
Extraction) and provided by JKTech steps to provide blast flexibility and precision
•• tools such as Split FX® that processes 3D point clouds •• rock tracking devices such as Metso SmartTagTM
from LIDAR (light detection and ranging) scans and (La Rosa et al, 2007) to track ore from blasthole through
photogrammetry to automatically characterise fracture mining, handling, stockpiling and mill feed. Combined
attributes with mill performance and online size data, the effects
•• GPS digital drilling systems to guide, monitor and save of ore and blast changes can be correlated with their
drill hole patterns impact on processing, refer to Figure 2
•• measured-while-drilling (MWD) data that captures rock •• image analysis of particle size distribution in the
hardness measurements from blasthole drills to reliably stockpile
categorise rock types •• MineWare’s Argus shovel monitor to improve shovel
•• blast movement monitors accurately measure three- and operator performance, optimise truckloads and
dimensional blast movement to minimise ore loss and reduce costs
dilution and significantly increase ore yield (LaRosa and •• prompt gamma neutron activation analysis (PGNAA)
Thornton, 2011) like GeoScan or CB Omni for online measurement of

P Cameron, D Drinkwater and J Pease
FIG 2 – Using a SmartTagTM to track ore batches from blasting (Metso image, from La Rosa et al, 2007).
elements in rock streams on belt conveyors for grade The fatal flaw was that earlier designs didn’t lock the
engineering or stockpile blending new operating methodology into organisational and
•• safe, accurate and quick measurements of the Crusher management systems. When profits rose in the boom,
Closed Side Setting (CSS) from 6 to 220 mm through the attention was focused on expanding output and resources.
‘C-Gap’ digital measurement tool Blasting engineers were rewarded for reducing their cost
per tonne and miners were rewarded for increasing tonnage.
•• accurate and reliable online particle size monitors
Metallurgists were incentivised to increase tons and recovery
(PSM) and sampling stations to sample and measure the
and to reduce costs, but were rarely encouraged to increase
particle size distribution and per cent solids of the total
product quality beyond ‘good enough to sell’. Smelters
grinding circuit classifier product
remove impurities at much higher cost than the concentrator
•• acoustic monitoring software to monitor operating – but that was the smelter’s problem.
conditions in SAG mills and allow optimisation
If we are to truly succeed with Mine-to-Mill in future,
•• automated quantitative mineralogy provides mineral
implementation must be supported, not undermined, by our
liberation and association data (using area scan of
organisation systems. We need to design key performance
particles) and elemental distributions of ore and waste
indicators (KPIs) that encourage integration and work across
minerals
the silos to provide mine site targets.
•• geometallurgy – the integration of geological, mining,
metallurgical, environmental and economic information Mining operations are complex. We need as many good
to maximise the net present value (NPV) of an orebody measures as we can find; they are getting better but are still
while minimising technical and operational risk. imperfect. We need to distil the complexity to the simple
fundamental basics of what makes a good integrated
organisational team, and then set the minimum few KPIs
20 YEARS LATER – WHAT HAVE WE LEARNED? that let clever people get on with their own jobs while using
For many of the early adopters, the initial success with Mine- their initiative for the group benefit. This will keep the
to-Mill fell victim to organisation systems and personal shareholders onside.
incentives that ‘sprang back’ to default when survival was
no longer in doubt. Any reinvention of Mine-to-Mill needs to TABLE 2
recognise why we failed to hold the gains first time, then set
Mine-to-Mill case studies 2003 to 2016.
about to fix those flaws.
Antamina and Cerro Corona demonstrated what can be Mine site Metals produced Country Product increase
achieved by applying some of the new tools in a Mine-to-Mill
Batu Hiju Cu Indonesia 10 to 15%
project. Models of drilling and blasting, crushing and grinding
were combined with a knowledge of different ore domains, Antamina Cu Peru 45 to 60%
plant surveys and SmartTagTM tracking of blasting changes. Los Bronces Cu Chile 15 to 20%
The result was a significant increase in mill throughput
from 2750 to 4400 t/h and a 25 per cent reduction in specific Cerro Corona Au Peru 15% hard ore, 6% overall
energy (kWh/t) (Valery et al, 2012). While at Cerro Corona, Phu Kham Cu Laos 8%
throughput was increased by 6 per cent overall (as much as
15 per cent for harder ores) and the SAG specific energy was Ahafo Au Ghana 8%
reduced by over 9 per cent (Diaz et al, 2015). Morila Au Mali 10%
Case studies from 2003 to 2016 are listed in Table 2, and Oyo Tolgoi Cu Mongolia 25%
there are even further operations that have benefited from
Mine-to-Mill but chosen not to publish the results. Iduapriem Au Ghana 21 to 32%

CONCLUSION Diaz, R, Mamani, H, Valery, W, Jankovic, A, Valle, R, and Duffy, K,

2015. Diagnosis of process health, its treatment and improvement
We know Mine-to-Mill works. We made it work well with old to maximise plant throughput at Goldfields Cerro Corona, in
technology. We have more tools and more appropriate data SAG 2015 Conference Proceedings, Vancouver, Canada.
now. We have a business imperative, so we can make it work
Gomes, M P, Tavarez Jnr, L, Nunes, E, Colacioppo, J, Jankovic, A and
much better again. Valery, W, 2010. Optimization of the SAG Mill Circuit at Kinross
We have to recognise the complexity of sites and the Paracatu, Brazil, Comminution 10.
differences between sites. The principles of integration will work Hart, S, Rees, T, Tavani, S, Valery, W and Jankovic, A, 2011. Process
everywhere, but the crucial components will be site-specific. integration and optimisation of Boddington, in SAG 2011
Off-the-shelf solutions provided by external management Conference Proceedings, Vancouver, Canada, 25–28 September.
groups won’t work; the components of the solutions need to Hawke, S J and Dominguez, L A, 2015. A simple technique for using
be specifically assembled by multidisciplinary teams of people high energy, in Proceedings 11th International Symposium on Rock
who know the details and constraints of each site, and who Fragmentation by Blasting, Sydney, Australia, pp 321–326.
understand the difference between minimising costs in silos Kanchibotla, S S, Morrell, S, Valery Jnr, W and O’Laughlin, P, 1998.
and maximising mine site profits. Exploring the effect of blast design on SAG Mill throughput
This time we need to support Mine-to-Mill with an at KCGM, presented at the Mine to Mill Conference, Brisbane,
understanding of human and organisational behaviour, Australia.
and lock that into our incentive systems. We need to greatly La Rosa, D, Valery, W, Wortley, M and Ozkocak, T, 2007. The use
simplify the overwhelming list of performance targets and of radio frequency id tags to track ore in mining operations, in
distil them to the critical few that encourage the behaviour Proceedings 31st International Symposium on Computer Applications
needed to integrate the operation. Then we can allow our in the Mineral Industries (APCOM), Santiago, Chile, 23–25 April
2007.
clever people to get on with the job.
Lam, M, Jankovic, A, Valery, W and Kanchibotla, S, 2001. Increasing
SAG mill circuit throughput at Porgera gold mine by optimising
ACKNOWLEDGEMENTS blast fragmentation, in Proceedings SAG 2001 Conference,
This paper was inspired by observation and comments from Vancouver, Canada.
many people that Mine-to-Mill is often overlooked, yet the McCaffery, K, Mahon, J, Arif, J and Burger, B, 2006. Batu Hijau
same problems which drove the original AMIRA Project are – controlled mine blasting and blending to optimise process
now repeating themselves. production at Batu Hijau, in Proceedings SAG 2006 Conference,
The authors thank the contributions made by the many Vancouver, Canada.
suppliers of the technology tools that are now available for McKee, D J, 2013. Understanding Mine-to-Mill, 96 p (Cooperative
Mine-to-Mill projects. Valuable contributions were made by Research Centre for Optimising Resource Extraction: Brisbane).
Walter Valery and Sarma Kanchibotla, both of whom have Renner, D, La Rosa, D, DeKlerk, W, Valery, W, Sampson, P, Bonney
been associated with Mine-to-Mill projects almost since their Noi, S and Jankovic, A, 2006. Anglogold Ashanti Iduapriem
inception. mining and milling process integration and optimisation, in
Proceedings SAG 2006 Conference, pp 249–264, Vancouver.
REFERENCES Valery, W and Rybinski, E, 2012. Optimisation process at Antamina

boosts production and energy efficiency: helping a large copper/
Bennett, D, Tordoir, A, Walker, P, La Rosa, D, Valery, W and Duffy,
zinc mine meet the economic challenge of processing harder ore
K, 2014. Throughput forecasting and optimisation at the Phu
types, Engineering and Mining Journal, September 2012.
Kham Copper-Gold operation, in Proceedings 12th AusIMM Mill
Operators’ Conference 2014, (The Australasian Institute of Mining Valery, W, Morrell, S, Kovovic, T, Kanchibotla, S and Thornton,
and Metallurgy: Melbourne). D, 2001. Modelling and simulation techniques applied for
Dance, A, Valery, W, Jankovic, A, La Rosa, D and Esen, S, 2007. optimisation of Mine to Mill operations and case studies, in
Maintaining the benefit – how to ensure mine-to-mill continues Proceedings of VI Southern Hemisphere Conference on Minerals
to work for you, in Proceedings Ninth AusIMM Mill Operators’ Technology, Rio de Janeiro, Brazil, 27–30 May.
Conference, (The Australasian Institute of Mining and Metallurgy:
Melbourne).

Contents
Base metals concentrate sales contracts

– change Pavlov and the dog
P D Munro1 and S E Munro2
ABSTRACT
The commonly traded base metals concentrates of copper, lead, tin and zinc are sold by
miners to custom smelters using contracts whose terms were generally established at
the beginning of the 20th century or even earlier. Technical clauses in these concentrates
for items such as payable metals and penalty elements do not reflect current realities of
extractive metallurgical technologies and hygiene/environmental issues.
It is an axiom that the miners and smelters have a symbiotic relationship. While the
existing antique and arcane commercial arrangements serve the purpose of transferring
money from the buyer to the seller, they are not likely to result in the industry achieving
better overall technical and economic outcomes. By examining the cases for copper
concentrates and zinc concentrates in particular, it will be shown that the practical
application of these antiquated technical clauses does not drive the miners to produce
the better quality concentrates that should be desired by the smelters. This is despite
development of mineral beneficiation and hydrometallurgical technologies that would
allow the miners to make better products.
The adage that any extractive metallurgical problem in the smelter is cheaper to solve
in the concentrator still generally holds.
While nickel concentrates are only touched on lightly in this discussion because sales
terms are more opaque than those for the other base metals, it is expected that the same
conclusions will apply.
INTRODUCTION
The commonly traded base metals concentrates of copper, lead, tin and zinc are sold by
miners to custom smelters using contract terms generally established at the beginning
of the 20th century or even earlier. Technical clauses in these concentrates for items
such as payable metals and penalty elements do not reflect current realities of extractive
metallurgical technologies and hygiene/environmental issues. The discussion is
confined to the consequences of these technical clauses rather than commercial ones
such as treatment and refining charge, price participation, quotation period for metal
pricing and so on.
This paper focuses on the situation for the traditional base metals because of the
greater transparency of sales terms and the common use of flotation as the beneficiation
method (tin being the exception by using mostly gravity concentration). Nickel
concentrates are only touched on lightly in this discussion. The cases of other metals
such as molybdenum, tantalum and tungsten are not discussed except where it relates
to flotation performance.
Lewis and Streets (1979) provided a comprehensive and masterly review of base metal
smelter terms from the miner’s point of view which is a good primer for the novice in
the area. Details on the contracts for the individual metals are found in the Australasian
Institute of Mining and Metallurgy publications Cost Estimation Handbook for the
Australasian Mining Industry, Monograph No 20 (1993) and Cost Estimation Handbook, 2nd
Edition, Monograph No 27 (2012).
The reader is referred to the relevant sections for individual concentrates from these
publications cited in the references list of this paper as follows:
•• copper – Wilson and Chanroux (1993a); de Sousa (2012)
•• lead – Wilson and Chanroux (1993b); Watters (2012)
1. FAusIMM, Senior Principal Consulting •• tin – Lewis (1993); Kettle (2012)
Engineer, Mineralurgy Pty Ltd, Taringa Qld •• nickel – Cunningham (1993); Selby and White (2012)
4068. Email: pdmunro@bigpond.com.au
•• zinc – Wilson and Chanroux (1993c); Wise (2012).
2. MAusIMM, Senior Process Engineer,
Mineralurgy Pty Ltd, Taringa Qld 4068. On reviewing the sales terms cited in the above papers and those for current contracts
Email: semunro@bigpond.com sighted in the course of our work, the striking impression is the continuity ie the lack
91
of change over the last 20 years. The reader does not get the and lead levels. Molybdenum concentrates are often
impression that the base metals industry is now operating in ≥55 per cent Mo.
a situation where it has to deal with vastly heightened public •• Zinc concentrate – while sphalerites in volcanogenic
concerns about toxic elements such as arsenic, lead, and massive sulfide and Mississippi Valley-type deposits are
mercury plus the issue of safely disposing of smelting and relatively ‘pure’, others contain significant amounts of iron
refining residues and wastes etc. with the mineral being more appropriately described as
Taking a holistic approach, it seems wrong to make marmatite ([Zn, Fe]S). This is the case for Broken Hill ore.
concentrates containing excessive waste and toxic elements Manganese can also be present in the sphalerite as is the
thereby moving the associated disposal and containment issues case for Broken Hill, Dugald River and Gamsberg. Many
from a remote mine site to a smelter located near a population zinc concentrates contain >90 per cent w/w sphalerite or
centre. marmatite. A preceding flotation step for copper and/or
lead usually prevents other sulfide minerals reporting to
A significant proportion of base metals concentrates are low
the zinc concentrate.
quality as measured by ‘weight per cent valuable mineral’.
Quantitative mineralogy often shows both high amounts of •• Copper concentrate – copper concentrates can be
misplaced free diluent minerals and unliberated valuable surprisingly poor quality with some of the world’s largest
minerals in composite particles with both non-sulfide and producers having <60 per cent w/w copper sulfides in
sulfide gangue minerals. their products despite hubristically quoting grades over
30 per cent Cu. This is because of the presence of the
The base metals industry is inherently flawed if current
high copper value supergene copper minerals chalcocite-
commercial arrangements encourage the miners to make low digenite (79.9 per cent Cu), covellite (66.4 per cent Cu)
quality concentrates without appropriate incentives to use the and bornite (63.3 per cent Cu).
best available technologies to make better ones.
•• Lead concentrate – those from Mississippi Valley-type
and carbonate-hosted ores can be high quality but those
THE QUALITY OF BASE METALS CONCENTRATES produced from texturally complex ores such the sedex-
type deposits in the Carpentaria – Mount Isa Inlier of
Measuring up to an industry benchmark? Queensland and the Northern Territory, Australia plus
The industry usually measures the quality of base metals certain volcanogenic massive sulfide deposits such as
concentrates by chemical content of the valuable constituent Hellyer and Woodlawn struggle to reach 50 per cent Pb.
for example per cent Cu, per cent Zn etc. A more appropriate •• Nickel concentrate – a distinction is made between
measure of the efficiency of the mineral beneficiation process concentrates produced from ores with a significant
(flotation for sulfide minerals or mostly gravity in the case of amount of pyrrhotite gangue such at Sudbury in Canada
tin) is by per cent valuable mineral in the concentrate. and deposits common in Western Australia. The latter
Examining the particular case of sulfide mineral flotation mostly komatiite-hosted nickel deposits are sulfur-poor
we use the benchmark of Johnson (February 2015, personal and magnesium-rich with altered phyllosilicate minerals
communication) that an industrial flotation process should of concern being:
be able to make a concentrate with 85 per cent w/w valuable •• the end product ta
mineral from a feed with >80 per cent mineral liberation •• the end products of the serpentine subgroup such as
(liberation is defined as particles of composition ≥98 per cent chrysotile, antigorite, lizardite etc. Ores with a high
w/w of the target mineral). pyrrhotite content can have a significant amount of nickel
Table 1 shows the grade of base metal concentrates using a associated with the pyrrhotite in pentlandite ‘flames’
composition of 85 per cent w/w valuable mineral. and in solid solution so rejecting pyrrhotite decreases
Contrast the above base metal sulfide concentrate with a nickel recovery. Concentrate with high per cent Fe is
magnetite concentrate (magnetite 72.4 per cent Fe) assaying desired to maintain the required Fe:MgO ratio in the
68 per cent Fe which contains 94 per cent w/w magnetite. feed of flash smelting to matte which treats ~75 per cent
of nickel sulfide concentrate (Crundwell et al, 2011). The
Some comments on Table 1 are warranted:
high MgO concentrates come from those deposits with
•• Molybdenum concentrate – quality is emphasized altered phyllosilicate minerals. The mineral separation
probably from the element’s major use as an alloying problem is difficult under circumstances where the
element in steel and the situation of a relatively few non-ssulfide gangue can be hydrophobic in the case of
‘converters’ roasting the sulfide concentrate to molybdic talc and with serpentine minerals positively charged
oxide (MoO3). By-product molybdenum concentrates at alkaline pH while sulfide minerals are negatively
from porphyry copper mines are often subjected to charged causing agglomeration of the mineral species.
a chloride leach at the mine site (The Brenda Process – As a consequence of the mineralogical factors for
Jennings, Stanley and Ames, 1973) to reduce copper these two types of deposits concentrate grades above
TABLE 1
Base metal concentrate quality as per cent valuable mineral.
Concentrate Mineral % w/w Valuable mineral Grade

Molybdenum Molybdenite 59.9% Mo 85 50.9% Mo
Zinc Sphalerite 67% Zn 85 56.9% Zn
Copper Chalcopyrite 34.6% Cu 85 29.4% Cu
Lead Galena 86.6% Pb 85 73.3% Pb
Nickel Pentlandite 34.2% Ni 85 29.1% Ni

Base metals concentrate sales contracts – change Pavlov and the dog
20 per cent Ni are rare. The sulfide nickel segment of the above. Using copper concentrate as an example it is
industry is more integrated than copper, lead or zinc a fruitless exercise to attempt to optimise the content
with major smelters such as Falconbridge, Kalgoorlie, of copper, sulfur and silica for a given grade-recovery
Norilsk and Copper Cliff receiving most of their feed curve in a particular beneficiation system. Selecting the
from the companies’ own mines. target copper grade of the concentrates fixes the copper
recovery and the sulfur and silica grades.
NEED FOR DIALOGUE Miners may not appreciate the increasing environmental
and industrial hygiene pressures applying to the smelters.
Miners versus smelters These are genuine problems exacerbated by the proximity of
Many participants in the base metals sector do not understand many smelters to population centres. Some smelters operate
the technical situations facing both the miners and smelters; in jurisdictions that impose a yearly site limit for particular
‘technical’ in this context includes environmental factors. This elements/compounds, ie limiting the total amount of the
element or compound allowed on the site for the year.
is compounded by the lack of meaningful dialogue between the
participants. The tremendous growth in the custom base metals Fountain (2013) has reviewed the situation for elements of
smelting and refining sector in China over the last 20 years has concern in copper smelting and refining.
been accompanied by a retreat from the smelting and refining This is an evolving situation with a seemingly endless
business by ‘western’ or ‘first world’ mining companies. procession of new issues. Some years ago we were aware that
Companies with fully integrated mining + concentrating + a smelter was concerned about the chromium level in a zinc
smelting + refining assets seem to becoming rarer along with concentrate. This seemed strange because chromium is not
possession of the associated ‘in-house’ technical knowledge known to affect the performance of the conventional roast-
across the whole sector. leach-electrowinning process. However, the leach residue
Contact between miners and smelters is increasingly through of this zinc smelter was treated pyrometallurgically to fume
the marketing department of the former meeting the concentrate out remaining zinc and lead leaving a discard slag possibly
buying department of the latter where the ‘deliverable’ is the used as construction fill material. We surmised that because
concentrate sales contract. The participants in this dialogue/ hexavalent chromium (Cr6+) is a known carcinogen, it would
negotiation are naturally focused on commercial issues rather an easier for the smelter to state that there was negligible
than technical ones. chromium in the slag than to expect that the relevant
regulatory authorities could draw a distinction between Cr3+,
which is not a carcinogen, and Cr6+.
Myths, misconceptions and misunderstandings
Miners Smelters
Some common misconceptions that miners have about smelters The misunderstanding that smelters have about miners is
that we have encountered are as follows: ignorance that many of them could actually make better quality
concentrates given appropriate incentives.
•• Smelting and refining is a profitable business that
The miner has the choice of moving along a given grade-
makes money by ‘gouging’ the miner. This myth is
recovery curve in a particular beneficiation system to make a
demolished for copper concentrate by looking at the
higher grade concentrate at a lower recovery or conversely a
history of treatment charge (TC) + refining charge (RC)
lower grade concentrate at a higher recovery. Taking copper
from say 1987 to 2014. Up to 2006 the smelters received
as an example, for at least the past 50 years the economics have
15–25 per cent of the value of the contained copper but
always favoured recovery at the expenses of concentrate grade.
since 2007 coinciding with the disappearance of price
participation, their share has fallen to only ~5–8 per cent The miner can move to a new grade-recovery curve
due to a concentrate supply deficit caused by the huge (conventionally drawn with concentrate grade on the y-axis
expansion of smelting + refining capacity in China. Thus and recovery on the x-axis) pushing the curve up and to the
the miners have appropriated 50–80 per cent of the value right by changing the flow sheet and operating conditions. In
of the copper in the concentrate that formerly went to many cases the higher concentrate grade from operating on
the smelters. The economics of smelting lead, nickel, tin this new grade-recovery curve will not be accompanied by a
and zinc concentrates can be expected to be similarly less significant recovery loss. This is because the flow sheet and
attractive than mining the ores containing these elements. operating conditions changes can be confined to the cleaner
For all base metals the miner receives the largest share of flotation block. The major loss of the valuable mineral is usually
the value of the contained metal in the concentrate and from the rougher flotation block with the cleaner flotation block
operators of high-grade deposits get superior returns usually contributing less than a third of the total.
because of their higher comparative advantage. Compare
this to the position of the smelters who receive a minor CONCENTRATES IN PARTICULAR
and diminishing proportion of the value of the metal in
the concentrate and have limited comparative advantage Copper
as they use the same technologies for smelting and
refining. Payment disincentive for superior quality
•• Smelters ‘want sulfur in the concentrate because it is free Normal payment is for 96.5 per cent of the copper content
fuel’. It is reasonable to assume that if sulfur is a desirable subject to a minimum deduction of one unit (Lewis and Streets,
constituent in a concentrate then it would receive a 1979; Wilson and Chanroux, 1993a; de Sousa, 2012).
payment. We have never seen a payment made for sulfur Table 2 shows how payment terms vary with the copper
after reviewing hundreds of base metals concentrates content of the concentrate.
sales contracts over the last five decades. Miners making a low-grade concentrate at 15–20 per cent
•• Smelters ‘want silica because it slags out the iron thus Cu appear to be treated relatively generously compared with
removing the need to buy flux’. See comment for sulfur producers of higher grade materials. Lower grade concentrates

will produce more sulfuric acid and/or make more slag A 1.7 per cent absolute increase in metal payment seems
increasing copper loss. The higher slag volume increases inadequate recompense for the miner who makes a 75 per cent
disposal cost and requires more silica flux if the concentrate Pb concentrate compared with one at 45 per cent Pb that has
isn’t self-fluxing. only 60 per cent of the lead per tonne of concentrate. The lower
grade material probably leaves the smelter with more sulfur to
Arsenic problem be eliminated as sulfuric acid. So-called ‘black acid’ from lead
Current ‘world scale’ copper smelters produce ~300 000– and zinc smelters is disadvantaged in the market because of
500 000 t/a of copper (Schlesinger et al, 2011). Assuming a mean concerns about toxic elements such as mercury, cadmium etc
feed concentrate grade of 25 per cent Cu then the 500 000 t/a entering the food chain via phosphate fertiliser. Other usual
unit requires >2 000 000 t/a of concentrate feed. Using the diluent elements such as zinc and iron must be removed as
slag, increasing lead loss and subsequent disposal cost of what
limit of <0.5 per cent As content for concentrates imported
must be considered a ‘sensitive’ material as it contains elements
into China (Anon, 2006) means such a smelter has to deal with
such as lead, zinc, and cadmium.
up to 10 000 t/a of arsenic or 13 200 t/a as arsenic trioxide.
Halving the arsenic content of the feed to 0.25 per cent As still
produces an alarmingly high amount of arsenic that has to be
Technology changes
stored because of insufficient demand for it. The conventional lead smelting and refining route is as follows
(Sinclair, 2009):
Arsenic does not assist in the smelting of concentrate to
anode copper with its only technical function needing to be of •• sulfur elimination in a sinter plant
sufficient concentration in the electrolyte when electrorefining •• bullion production in a blast furnace
anodes to eliminate the formation of floating slimes and assist •• thermal refining:
the precipitation of antimony and bismuth in the slimes layer. •• drossing in two stages to remove copper and arsenic
Some very large copper deposits such as La Granja and •• oxygen softening or Harris process to remove arsenic,
Tampakan remain undeveloped because metallurgical test antimony and tin
work has shown unacceptably high levels of arsenic in the
•• desilvering by the Parkes process
copper concentrate.
•• vacuum dezincing
Arsenic penalties (Wilson and Chanroux, 1993a; de Sousa,
2012) are generally in the range US$2.50–5.00 per 0.1 per cent •• bismuth removal by the Kroll-Betterton process, and
As over 0.2 per cent As. Concentrate >0.5 per cent As cannot •• caustic refining.
be imported into China with the Japanese and South Korean New lead smelting technologies include bath smelting
smelters unwilling to take material above 0.15–0.2 per cent processes such as the popular Shui Kou Shan (SKS) process
As. High arsenic concentrates enter the market by incurring from China (15 plants operating and a further 15 under
a negotiated higher treatment charge reflecting the cost of construction as of 2013), Ausmelt, Isasmelt and less commonly
shipping to a destination and the cost of supplying ‘clean’ the Queneau-Schuhmann-Lurgi (QSL) process with minor
material to blend down the arsenic content. This is usually contributions from other technologies such the Kivcet and
done through a trading house. Boliden Kaldo processes.
Thus the industry’s tactics of dealing with the arsenic issue Electrolytic refining by the Betts process has never been as
are the somewhat passive responses of rejecting ‘high arsenic’ well accepted by ‘western’ countries as in Japan and China.
material and limited attempts at blending down the arsenic The enormous growth of the Chinese lead smelting industry
level. means that the Betts process with associated limited thermal
treatment is now probably the dominant refining process.
Lead For the conventional lead smelting process a certain amount
of zinc in the concentrate improves sinter quality (Sinclair, 2009)
Payment disincentive for superior quality as the formation of an appropriate proportion of zinc melilite
Normal payment is for 95 per cent of the lead content subject to (hardystonite) gives the desired microstructure for optimum
a minimum deduction of three units (Lewis and Streets, 1979; blast furnace operation with high softening temperature and
Wilson and Chanroux, 1993b; Watters, 2012). open porosity (McLoughlin, Riley and McKean, 1980).
Table 3 taken from Sinclair (2009) shows how payment terms New lead smelting technologies do not make sinter thus
vary with the lead content of the concentrate. removing the need for zinc in the concentrate. The now
TABLE 2
Copper concentrate metal payments.
Concentrate copper content (%) Copper in concentrate paid for (%) Copper content paid for (%)
50 48.25 96.5
45 43.425 96.5
40 38.6 96.5
35 33.775 96.5
30 28.95 96.5
25 24 96
20 19 95
15 14 93.3

TABLE 3
Lead concentrate metal payments (after Sinclair, 2009).
Concentrate lead content (%) Lead in concentrate paid for (%) Lead content paid for (%)
75 71.25 95
65 61.75 95
60 57 95
55 52 94.5
50 47 94
45 42 93.3
dominant SKS process treats the lead-rich slag from the first- While it would be preferable that this would be done at the
stage bath smelting process in a conventional blast furnace. mines, the small-scale of many operations may make it
J Tuppurainen (February 2015, personal communication) impractical.
advises that blast furnace practice is to maintain a zinc:iron ratio
of feed material less than 0.7:1 to effectively remove the zinc in Zinc
the discard slag. Zinc contained in the blast furnace slag can be
recovered by fuming but maintaining reducing conditions at Payment disincentive for superior quality
high temperatures is an expensive process. Zinc concentrate sales terms pay for 85 per cent of the contained
Zinc in lead concentrate has moved from being a ‘necessary zinc subject to a minimum deduction or free metal allowance of
evil’ in the conventional sinter plant + blast furnace process eight units. This deduction is an antique term, possibly related
route to being a nuisance in processing by the current dominant to the relatively low first pass recovery of zinc in the now
lead smelting technologies. This leads to the question whether obsolete horizontal retort process dating from the 18th and
zinc contained in lead concentrate is being penalised too lightly 19th centuries. Modern electrolytic plants which produce over
by the smelters. 90 per cent of the world’s zinc (Sinclair 2005), achieve well over
Conversely, the predominance of electrolytic refining over 95 per cent zinc recovery.
thermal refining calls into question the validity of a severe Table 4 taken from Sinclair (2005) shows how payment terms
penalty charge for an element such as bismuth which is readily vary with the zinc content of the concentrate.
removed by the Betts process. Bismuth and antimony are Thus the smelter provides little incentive for the miner to
usually intimately associated with the lead mineralisation and make a concentrate much over 52 per cent Zn except to save
not rejected by physical beneficiation. transport costs to the smelter. Lewis and Streets (1979) observed
a similar ‘flattening’ out of payable zinc for concentrate grades
Tin above 54–55 per cent Zn.
The elimination of by-product revenue has changed the
Tin concentrate contract terms affected by quality economics of the zinc smelting business. In the 1980s zinc
The deductions and charges for tin concentrates shown in smelters could expect to receive payment for products such as
Lewis and Streets (1979) and later papers (Lewis, 1993; Kettle, cadmium, cobalt, copper residue, and mercury, lead residue
2012) are generally more variable than for other base metals containing silver and sulfuric acid extracted during processing
concentrates. This reflects tin contents varying from 20 per cent of the concentrate (Hamdorf and Woodward, 1980). Few zinc
Sn to 70 per cent Sn; a range rarely encountered in other plants in the ‘western’ world are currently paid for these by-
concentrates. Charges for penalty elements such as US$120–150 products and it is reasonable to assume that the Chinese
per 1 per cent As and $300–500 per 1 per cent Cu are very high smelters will reach this position in future. Toxic elements
compared with other concentrates. associated with the extractive metallurgy of zinc such as
Tin concentrate sales contracts increase the base deduction cadmium, lead, mercury and thallium are other ‘hot-button’
with increasing iron content which is logical considering issues for the public.
the difficulties of recovering tin from the tin-iron alloy
called ‘hardhead’. Iron problem
Table 5 taken from Sinclair (2005) for a conventional roast-leach-
Quality change from hard rock source electrowinning plant shows the dependence of zinc extraction
Glen (February 2015, personal communication) notes that in primary leaching on the iron content of the zinc concentrate.
depletion of alluvial tin resources means that relatively more High iron levels in the concentrate increase both the amount
tin concentrates are being produced from ‘hard rock’ deposits of primary leach residue that has to be processed by secondary
with custom tin smelters concerned about elevated levels of leaching and the amount of leach residue for ultimate disposal.
arsenic levels (Glen, 2015). Arsenic is a real issue in tin smelting It is anomalous that payment terms shown in Table 4 level
as its removal from molten tin is by the addition of aluminium out for concentrates assaying 52 per cent Zn and above
which forms a dross containing the two impurities. Wright considering that disposal of iron-rich residue from acid
(1966) describes this dross as ‘… probably the most dangerous leaching of calcine and/or concentrate is a major problem
substance met in classical metallurgy’ because of its propensity for zinc smelters (the term ‘smelters’ includes all zinc
to evolve arsine on contact with water or moisture in the air. extractive metallurgical plants). This is particularly true
Arsine is 50 times more toxic than hydrogen cyanide. for zinc smelters located near population centres. Storage
Custom tin smelters are examining treatment of incoming of jarosite or goethite-type leach residue in ponds is under
tin concentrates at their plants by flotation to reduce arsenic regulatory pressure. Conversion of the leach residue to
(presumably contained in arsenopyrite) to acceptable levels. hematite or alternatively making a relatively benign slag by

TABLE 4
Zinc concentrate metal payments (after Sinclair, 2005).
Concentrate zinc content (%) Zinc content paid for (%) Zinc content as free metal to smelter (%) Zinc content paid for (%)
46 38 8 82.6
48 40 8 83.3
50 42 8 84.0
52 44 8 84.6
54 45.9 8.1 85.0
56 47.6 8.4 85.0
TABLE 5
Primary leach zinc recovery and residue composition (after Sinclair, 2005).
Zinc in concentrate % Iron in concentrate % Zinc recovery % Zinc in primary leach residue % Primary residue amount % of concentrate
48 12 82.8 22.8 36
50 10 85.6 22.8 32
52 8 88.3 22.5 28
54 6 90.7 21.7 24
56 4 92.9 20.9 19
pyrometallurgical treatment such as the Waelz process or bearing their full share of downstream processing costs. In a
fuming by smelting is expensive. situation where the industry has a finite capacity to deal with
The case of the Budel zinc smelter in The Netherlands is a input iron in concentrate, removing some iron from high-
portent for the future of leach residue disposal. Since 1999 grade concentrates allows more room for iron in low-grade
on-site storage of residue has been banned with the former concentrates albeit at lower returns for miners producing
leach residue ‘waste’ now mandated to be a ‘lead product’ such materials.
sent for smelting by another party. Budel has achieved this Ameliorating the iron issue does not resolve the problem of
by using low iron zinc concentrate from the Century Mine in elements such as cadmium, mercury and thallium which are
Queensland as its major feed source to minimise the amount usually too closely associated with the sphalerite/marmatite to
of this ‘lead product’ (Sinclair, 2007). be removed by physical beneficiation.
Given this concern about iron it seems surprising that the
penalty for iron in a zinc concentrate hasn’t changed in nearly MINERAL PROCESSING SOLUTIONS
20 years being US$1.50 per 1 per cent Fe above 8 per cent Fe
(Wilson and Chanroux, 1993c; Wise, 2012). Greater use of existing technologies and practices
Miners can often make higher grade base metals concentrates
A new zinc concentrate contract? from flotation by:
We suggest that the zinc industry move to a new zinc •• increasing mineral liberation of the target mineral feed
concentrate sales contract that more appropriately aligns the going to the cleaner flotation block; too often the
capabilities of the miners with the needs of the smelters. This benchmark condition mentioned above of >80 per cent
new contract would comprise the following: mineral liberation for the valuable mineral in the feed is
•• Treatment charge reflecting the actual cost of producing not achieved
zinc metal from concentrate. Sinclair (2005) estimated •• improving mineral separation; this is evidenced by the
operating costs for a 200 000 t/a electrolytic plant as of June presence of excessive amounts of free diluents such as
2000 to be US$383/t (or US$191.50/t of zinc concentrate iron sulfide minerals and non-sulfide gangue.
with 50 per cent recoverable metal content). Power in Johnson and Munro (2002) have reviewed the technologies
his calculation was US$0.04/kWh. Cost of jarosite leach and plant practices developed to treat complex sulfide ores
residue disposal into rubber-lined ponds holding up to such as the zinc-lead-silver ores of the Mount Isa Inlier (Young
five years of production was US$6.42/t of zinc. et al, 1997) and difficult volcanogenic massive sulfide ores such
•• Realistic payment for contained zinc eliminating the as Hellyer (Lane and Richmond, 1993).
‘hocus pocus’ of ‘eight units of free metal’ and adjustments Better understanding of pulp chemistry together with
to the treatment charge. using ultra-fine grinding and washed froth cleaning have
•• Iron penalty commensurate with the cost of dealing been successfully applied in concentrators such as Century
with the leach residue. A concentrate with 50 per cent Zinc (Barham and Kirby, 2001; Obeng et al, 2013), McArthur
recoverable zinc and 8 per cent Fe generates at least 0.3 t River (Rossberg and Pafumi, 2013), Mount Isa – George Fisher
of residue per tonne of zinc metal produced. (Pafumi et al, 2013), Phu Kham (Bennett, Crnkovic and Walker,
Obviously there are going to be ‘winners’ and ‘losers’ among 2012) and Prominent Hill (Barns, Colbert and Munro, 2009).
the miners from such a contract. Miners with deposits of high A seemingly forgotten powerful flotation practice is reverse
iron sphalerites will receive a lower return than miners able flotation as exemplified by the so-called Brunswick process
to make high-grade concentrates. However, this is required (McTavish, 1980) which was used at Brunswick Mining and
to achieve a better overall industry outcome. We could argue Smelting and later at Kidd Creek. Sphalerite was depressed
that currently producers of low-grade concentrates are not by heat and sulfur dioxide addition with pyrite removed

as the concentrate. The standard by-product molybdenum •• Separation of the copper concentrate into a high tonnage
separation from copper-molybdenum ores can be considered lower arsenic portion and a low tonnage higher arsenic
to be a type of reverse flotation process upgrading the copper portion by exploiting Eh-pH flotation separation
concentrate. ‘windows’ for the different copper mineral present.
If these existing technologies and plant practices have to be Ametov has demonstrated the possibility of a reasonable
applied to ‘difficult’ ores to achieve ‘ordinary’ metallurgical separation of chalcopyrite + pyrite versus chalcocite
outcomes then using them on ‘easy’ ores such as a porphyry + bornite (Ametov et al, 2014). Ma and Bruckard (2009)
copper deposit would achieve ‘extraordinary’ metallurgical show that similar ‘windows’ exist for the arsenic carriers.
outcomes compared with current ones. It is unreasonable to expect that there will be clean
separation with all the copper arsenic minerals reporting
One consequence of better quality concentrates will
to one product.
probably mean miners and smelters having to cope with
finer particle size distributions. For the miners this means •• Hydrometallurgical treatment of the high arsenic
froth pumping and concentrate dewatering. This shouldn’t concentrate by an emerging technology such as the
be alarming as the plants mentioned above are successfully Toowong process (Nakon and Way, 2014; Turner, 2015)
dealing with these issues. A far bigger potential problem is which uses an alkaline leach. This has the advantage
the solid-liquid separation issues with the movement to dry over roasting of leaving the main copper sulfide minerals
stacking of tailings if regulators ban the conventional ‘wet’ substantially unchanged. Arsenic is fixed using a range
tailings dam. For the smelters finer concentrate particle size of conventional methods and the stabilised product
distribution means more dust. can be disposed into a suitable engineered facility.
Alternatively all the copper concentrate is treated
The reason that these technologies and practices aren’t used
through the Toowong process reducing the arsenic
more widely by the miners to make better quality concentrates
content to 0.05 per cent As.
is because technical terms in the sales contracts do not provide
any monetary incentive to do so.
As mentioned above a magnetite concentrate at 68 per cent SUMMARY
Fe contains 94 per cent w/w magnetite which is much higher The technical terms in contracts for the sales of for base
quality than the base metal concentrates in Table 1. Higher metals concentrates show little change over the last 20 years
capital and operating costs are needed to produce magnetite or even the last 50 years. This is despite all sectors of the
concentrate compared with direct shipping ore (DSO) so primary base metals industry now operating in a situation
a high quality product has to be made to get the premium where they have to deal with vastly heightened public
offered over the DSO material. Blast furnace iron producers concerns about toxic elements such as arsenic, lead, and
are much more conscious of feed quality than base metals mercury plus the issue of safely disposing of smelting and
smelters. An argument against using iron ore sector practice refining residues and wastes etc.
as a parable for base metals is that iron ore is a low value For mineral beneficiation by sulfide mineral flotation, many
commodity mined from very large deposits where recovery miners including some of the world’s largest producers do
isn’t important. One tonne of iron ore at 30 per cent Fe with not achieve our benchmark that an industrial flotation process
iron ore at US$100/t has the same in situ value as one tonne should be able to make a concentrate with 85 per cent w/w
of copper ore at 0.5 per cent Cu with copper at US$6000/t. valuable mineral from a feed with >80 per cent mineral
Large porphyry copper deposits can be the same size as iron liberation.
ore deposits. Mineral processing developments over the past 20 years, such
Magnetite concentrates are high quality because they have as better understanding of pulp chemistry, ultra-fine grinding
to be, customers won’t accept poor quality. It exemplifies and washed froth cleaning have been successfully applied to
Sinclair’s reiteration (2007) of the business truism ‘that if you the treatment of complex or ‘difficult’ ores.
don’t have customers you really don’t have a viable project’. Using them on ‘easy’ ores such as porphyry copper deposits
would achieve ‘extraordinary’ metallurgical outcomes
A viable technical solution to the arsenic problem? compared with some current mediocre ones leading to a general
High arsenic levels in copper concentrate are usually caused improvement in the quality of base metals concentrates.
by the presence of the copper arsenic sulfide minerals enargite
The smelters, after appropriate dialogue with the miners,
(Cu3AsS4) and tennantite ([Cu, Fe]12As4S13 or Cu13As4S13).
should alter the technical clauses in base metals concentrate
These minerals behave in flotation like other copper sulfide
sales contracts appropriately affecting economic returns. This
minerals and report to the concentrate. Arsenic in the form
should elicit the appropriate Pavlovian response in terms of
of arsenopyrite should be readily dealt with as it behaves
concentrate quality.
like pyrite.
If the industry doesn’t hang together on this then its
The proven mine site solution to high arsenic concentrates
constituent components will hang separately!
is roasting to remove the element as arsenic trioxide. This
was done at Lepanto and El Indio and is currently practised
at Codelco’s Ministro Alejandro Hales operation (formerly ACKNOWLEDGEMENTS
known as Mansa Mina). A less appealing solution from both The authors would like to acknowledge Mineralurgy Pty Ltd
the point of view of the miner and custom smelter is for the for the permission to publish this paper.
former to produce copper metal on-site by smelting or the We thank Mr Craig Walter formerly of MIM Holdings
pressure oxidation + solvent extraction + electrowinning route. Limited and later of Trafigura Beheer BV for his acute
Examination of the technical literature suggests the interesting observations, wise guidance and patient mentoring in the areas
possibility of a mine site treatment route as follows: of concentrate sales and metals marketing over 30 years.
•• Product of a copper concentrate by conventional or Mr Jorma Tuppurainen Senior Principal Consulting Engineer
enhanced means. with Mineralurgy Pty Ltd kindly provided technical advice

from his considerable expertise in the extractive metallurgy of Ma, X and Bruckard, W J, 2009. Rejection of arsenic minerals in
copper, lead, nickel and zinc. sulphide flotation – A literature review, International Journal of
Mineral Processing, 93:9–94.
Mr Roderick Sinclair is thanked for his very useful
publications The Extractive Metallurgy of Lead (2009) and The McLoughlin, B C, Riley, J F and McKean, G R, 1980. Development
of optimum fluxing procedures for isamine lead concentrates, in
Extractive Metallurgy of Zinc (2005) from which we have drawn
Lead-Zinc-Tin ‘80 (eds: J M Cigan, T S Mackey and T J O’Keefe),
heavily in writing this paper. pp 309–320 (Metallurgical Society of AIME: Warrendale,
Pennsylvania).
REFERENCES McTavish, S, 1980. Flotation practice at Brunswick Mining, CIM
Ametov, I, Harmer, S, Grano, S and Alford, R, 2014. Selective flotation Bulletin, 115:115–120.
of chalcopyrite and pyrite from a mixed copper mineral concentrate
Nakon, D and Way, D M, 2014. Method for Treating Arsenic Containing
under controlled redox potential condition, in Proceedings 12th
Materials, US Patent 8771619 B2 (Xstrata Queensland Ltd).
AusIMM Mill Operators’ Conference, pp 131–140 (The Australasian
Institute of Mining and Metallurgy: Melbourne). Obeng, D, Twining, D, Milsom, J, Brown, M, Smith, M, Horton,
R, Ackroyd, T and Smith, W, 2013. Century Mine MMG, in
Anon, 2006. The specification for limit on harmful element content of
Australasian Mining and Metallurgical Operating Practices (ed: W J
the heavy metal concentrates products (English trans) National
Rankin), pp 1661–1682 (The Australasian Institute of Mining and
Standard of the People’s Republic of China GB 2044–2006.
Barns, K E, Colbert, P J and Munro, P D, 2009. Designing the optimal
Pafumi, J, Rossberg, F, Golling, J, Armstrong, C, Walker, R, Grobler, J,
flotation circuit – the Prominent Hill case, in Proceedings Tenth
Boyer, J, Meyer, R and Beer, A, 2013. George Fisher Mine Xstrata
AusIMM Mill Operators’ Conference, pp 173–182 (The Australasian
Zinc, in Australasian Mining and Metallurgical Operating Practices
(ed: W J Rankin), pp 1639–1660 (The Australasian Institute of
Bennett, D, Crnkovic, I and Walker, P, 2012. Recent process Mining and Metallurgy: Melbourne).
improvements at the Phu Kham Copper-Gold concentrator,
Rossberg, F and Pafumi, J, 2013. McArthur River mine xstrata zinc, in
Laos, in 11th AusIMM Mill Operators’ Conference, pp 257–272 (The
Australasian Mining and Metallurgical Operating Practices (ed: W J
Rankin), pp 1543–1668 (The Australasian Institute of Mining and
Crundwell, F K, Moats, M S, Ramachndaran, V, Robinson, T G and Metallurgy: Melbourne).
Davenport, W G, 2011. Extractive Metallurgy of Nickel, Cobalt and
Schlesinger, M E, King, M J, Sole, K C and Davenport, W G, 2011.
Platinum-Group Metals (Elsevier: Amsterdam).
Extractive Metallurgy of Copper, 5th edition (Elsevier: Amsterdam).
Cunningham, W H, 1993. Nickel, in Cost Estimation Handbook for the
Selby, M and White, D T, 2012. Nickel, in Cost Estimation Handbook,
Australian Mining Industry (eds: M Noakes and T Lanz) (Chapter 16)
Second Edition (Chapter 20), pp 502–507 (The Australasian
Melbourne).
Sinclair, D F, 2007. Project evaluation – the customer’s perspective, in
de Sousa, A, 2012. Copper, in Cost Estimation Handbook, Second
Proceedings Project Evaluation 2007, pp 203–208 (The Australasian
Edition (Chapter 20) pp 483–486 (The Australasian Institute of
Sinclair, R J, 2005. The Extractive Metallurgy of Zinc (The Australasian
Fountain, C, 2013. The whys and wherefores of penalty elements in
copper concentrates, in Proceedings MetPlant 2013, pp 502–518 (The
Australasian Institute of Mining and Metallurgy: Melbourne). Sinclair, R J, 2009. The Extractive Metallurgy of Lead (The Australasian
Hamdorf, C J and Woodward, O H, 1980. Zinc production by
electrolytic zinc company of Australasia Ltd, Risdon, Tas, in Turner, D, 2015. The Toowong Process, Core Resources Pty Limited,
Mining and Metallurgical Practices in Australia (ed: J T Woodcock), Albion, Queensland.
pp 254–261 (The Australasian Institute of Mining and Metallurgy: Watters, T, 2012. Lead, in Cost Estimation Handbook, Second Edition
Melbourne). (Chapter 20), pp 493–495 (The Australasian Institute of Mining and
Jennings, P H, Stanley, R W and Ames, H L, 1973. Development of Metallurgy: Melbourne).
a process for purifying molybdenite concentrates, in Proceedings Wilson, P C and Chanroux, C, 1993a. Copper, in Cost Estimation
International Symposium on Hydrometallurgy (eds: D J I Evans and Handbook for the Australian Mining Industry (eds: M Noakes and
R S Shoemaker), pp 868–883 (Society of Mining Engineers of AIME: T Lanz) (Chapter 16), pp 346–348 (The Australasian Institute of
New York). Mining and Metallurgy: Melbourne).
Johnson, N W and Munro, P D, 2002. Overview of flotation technology Wilson, P C and Chanroux, C, 1993b. Lead in Revenue Calculations and
and plant practice for complex sulfide ores, in Mineral Processing Marketing of Cost Estimation Handbook for the Australian Mining
Plant Design, Practice and Control Volume 1 (eds: A L Mular, Industry (eds: M Noakes and T Lanz), (Chapter 16) pp 337–339 (The
D N Halbe and D J Barratt), pp 1097–1123 (Society for Mining, Australasian Institute of Mining and Metallurgy: Melbourne).
Metallurgy and Exploration: Littleton).
Wilson, P C and Chanroux, C, 1993c. Zinc, in Cost Estimation Handbook
Kettle, P, 2012. Tin, in Cost Estimation Handbook, Second Edition for the Australian Mining Industry (eds: M Noakes and T Lanz)
(Chapter 20), pp 511–514, (The Australasian Institute of Mining (Chapter 16), pp 364–367 (The Australasian Institute of Mining and
and Metallurgy: Melbourne). Metallurgy: Melbourne).
Lane, G S and Richmond, G D, 1993. Improving fine particle flotation Wise, W, 2012. Zinc, in Commodity Marketing of Cost Estimation Handbook
selectivity at Hellyer, in Proceedings XVIII International Mineral Second Edition, (Chapter 20) pp 516–521 (The Australasian Institute
Processing Congress, vol 3, Flotation 1, pp 887–905 (The Australasian of Mining and Metallurgy: Melbourne).
Wright, P A, 1966. Extractive Metallurgy of Tin, pp 138–139 (Elsevier:
Lewis, P, 1993. Tin, in Cost Estimation Handbook for the Australian Amsterdam).
Mining Industry (eds: M Noakes and T Lanz) (Chapter 16) pp 360–
362, (The Australasian Institute of Mining and Metallurgy: Young, M F, Pease, J D, Johnson, N W and Munro, P D, 1997.
Melbourne). Developments in milling practice at the lead/zinc concentrator of
Mount Isa mines limited from 1990, in Proceedings Sixth AusIMM
Lewis, P K and Streets, G C, 1979. An analysis of base-metal Mill Operators’ Conference, pp 3–12 (The Australasian Institute of
smelter terms, in Proceedings Eleventh Commonwealth Mining and Mining and Metallurgy: Melbourne).
Metallurgical Congress (ed: M J Jones), pp 753–767 (The Institution
of Mining and Metallurgy: London).

Project design
Contents
Karouni Gold Project from drill

core to commissioning
K Nilsson1 and D Connelly2
ABSTRACT
Troy Resources acquired the Karouni Gold project in Guyana, South America in 2014
and immediately proceeded to a Pre-Feasibility Study and project implementation.
The study considered two open pit mines feeding a conventional carbon-in-leach
(CIL) gold plant with a capacity of 1 Mt/a. The study assumed a total of 2.61 Mt of ore
with an average grade of 3.84 g/t gold with a recovered gold production of 303 526 oz
over a three year mine life.
The project development involved drilling metallurgical test holes, securing
second-hand major pieces of equipment where suitable and arranging shipping to
Guyana. The remote location meant logistical considerations were paramount. The
plant consists of conventional two-stage crushing with single stage ball milling and a
gravity circuit. The tailings circuit includes cyanide detoxification (DETOX).
The extremely high rainfall presented challenges with water management and
the necessity to dispose of clean water to the environment. The DETOX was also
mandatory due to the highly sensitive environment and past history in the area. The
criticality of this meant that in the design, any cyanide in the tailings would precipitate
in a plant shutdown.
Lessons were learned and the project has been both profitable and successful for
Troy Resources in a new operating location. In addition, the project has created jobs,
delivered royalties and brought new skills into an area where previously there was
little current mining activity.
PROJECT HISTORY
The Karouni project was previously owned by Azimuth Resources. The project
transferred to Troy Resources following their acquisition of Azimuth Resources in July
2013. DDrilling by Azimuth established inferred resources of 1.6 Moz for the combined
Smarts and Hicks deposits, with Smarts being the higher grade of the two (see Figure 1).
Guyana is host to several active junior mining companies, and also has activity from
the major minors as well, including BHP, Newmont and Lamgold. In 2012, mining in
Guyana produced over 400 000 oz of gold.
A majority of the population of Guyana lives along the coastal areas, with a growing
amount of exploration occurring in the interior of the country. Resources within the
interior include diamonds, semi-precious stones and gold (see Figure 1).
GEOLOGY
The Guyana Shield has been correlated with the Leo-Man Shield of West Africa and
it is generally accepted that prior to the opening of the Atlantic during the Mesozoic,
the two shields formed a contiguous craton. The Archaean Imataca Complex can be
correlated with the Archaean Liberian Province, the Central Guyana Granulite Belt
with the Dimbroko Zone in Ivory Coast, the Barama-Mazaruni greenstones with
the Birimian greenstones and the Trans-Amazonian tectono-thermal event with the
Eburnean Orogeny.
METALLURGICAL TEST WORK

The test work program investigated three ore zones; primary, shear and oxide, over
1. Project Director, Troy Resources Ltd, a range of processing parameters. This included comminution, gravity, cyanide
West Perth WA 6005.
leaching, variability test work and ore process characterisation. Additionally, some
2. MAusIMM(CP), Director/Principal Consulting tests were conducted on material classified as sulfide. This material was subjected to
Engineer, Mineral Engineering Technical
preliminary comminution testing and checked for gold grades.
Services Pty Ltd (METS), Midas Engineering
Group, Perth WA 6000. Email: Samples were selected based on geological data, orebody cross-sections and a review
damian.connelly@metsengineering.com of the core logs and assays as provided by Troy Resources to METS. Composites were
101
FIG 1 – Karouni Gold project location.
formed based on this information including assay data for one of the sulfide samples, with the high sulfide result being
further testing, both for variability and master composites. classified as hard (see Table 2).
Comminution results SAG mill comminution

Comminution testing was conducted on all areas of the SAG mill comminution testing is conducted to further evaluate
orebody. ore hardness and competency for size reduction, with a
smaller starting size than that of the crushing tests. The shear
Unconfined compressive strength material was seen to be the least competent, being classified as
The oxide material was classified as very weak, while the soft–medium. The primary and sulfide materials were seen to
primary and shear materials were classed in the range of be more competent as hard to very hard. These rankings were
weak to medium-strong. The sulfide ore tested was seen to achieved on the basis of the A*b values (material hardness),
be the most competent material, being classified as medium- and the ta. values (resistance to abrasion) (see Table 3).
strong (see Table 1). Simulations indicated SAG milling was an option for the
process plant.
Crushing work index
The crushing work index results showed that all material was Bond ball mill work index
classified as ‘very soft’. There will be no identified issues as The Bond ball mill work index results indicated that all
a result of processing this through the crushing circuit. The samples were medium to hard, across all zones tested. The
results ranged from 0.4 kWh/t through to 17.9 kWh/t for range of values was between 12.57 kWh/t and 15.28 kWh/t,
Table 1
Unconfined compressive strength (UCS) results summary.
Hole Zone UCS ranking Average Max Min Std dev

99 Primary Medium-strong 25.65 45.62 12.93 14.29
100 Oxide Very weak 2.75 3.30 2.20 0.55
100 Primary Weak 8.85 14.35 3.50 4.43
101 Shear Medium-strong 29.53 80.57 8.37 27.08
102a Primary Weak 21.76 39.56 10.59 7.52
102a Shear Weak 16.67 45.50 4.48 14.60
103 Primary Medium-strong 28.14 32.52 20.75 5.26
103 Sulfide Medium-strong 34.62 133.31 6.55 31.66

Karouni Gold Project from drill core to commissioning
Table 2
Crushing work index (CWi) results summary.
Hole Zone #CWi CWi(kWh/t, average) Max Min Std dev Ranking
99 Primary 21 3.13 4.79 2.39 0.73 Very soft
100 Oxide 20 1.46 3.52 0.37 1.05 Very soft
101 Shear 20 4.64 9.62 3.77 1.80 Very soft
102a Primary 30 4.24 7.25 3.44 1.12 Very soft
102a Shear 15 5.14 11.21 2.41 2.18 Very soft
103 Sulfide 34 5.87 17.93 2.37 3.03 Very soft
103 Shear 19 5.12 7.69 2.36 1.43 Very soft
Table 3
SAG mill comminution (SMC) results summary.
Sample ID A b A*b Rank ta Rank

SDD102a – Primary 58.7 0.63 37.0 Hard 0.35 Hard/mod hard
SDD103 – Primary 54.6 0.53 28.9 Very hard 0.28 Very hard
SDD103 – Shear 51.9 0.95 49.3 Medium 0.48 Medium
SDD103 – Sulfide 69.9 0.36 25.2 Very hard 0.24 Hard
SDD102a – Shear 46.3 1.81 83.8 Soft 0.83 Soft
with the highest value being recorded by the sulfide ore, and The gravity recovery varied between the three composites,
the lowest by the shear ore. with the oxide displaying a gravity recovery of around
30 per cent, the shear 40 per cent and the primary 60 per cent.
Bond rod mill work index Given the high gravity performance, it was recommended
Rod mill work index testing was conducted on the primary that a gravity stage be included in the processing circuit flow
and shear master composites. Both results showed the material sheet.
to be hard, with results of 19.9 kWh/t and 17.5 kWh/t for the
primary and shear zones respectively. Quantitative Evaluation of Minerals by SCANning
Electron Microscopy (QEMSCAN) – mineralogy
Bond abrasion index A QEMSCAN analysis was carried out on heavy liquid
The primary and shear master composites were tested. The separation sinks and floats following desliming of a subsample
primary material returned a result of 0.13, which classifies this of the primary master composite material. The floats contained
ore as mildly abrasive. The shear material returned a value of a majority of quartz, feldspar and other silicates. Based on
0.09, which is non-abrasive. These results indicate that there work conducted in the test work program, the inclusion of the
will not be any undue wear experienced in a crushing circuit. feldspar and silicates shows negligible negative effect on the
rheological properties of the primary ore.
Master composite testing The analysis of the sinks showed that the sulfides,
Master composites of primary, shear and oxide ores were identified in the head assays, were present as pyrite. Other
formed. While not all tests were conducted on all composites, major minerals present in the sinks included iron oxides/
the primary, shear and oxide master composites were used hydroxides, and iron silicates. These results give basis to the
to assess optimal leaching conditions and determine process assumption that the unleached gold remaining in the leach
performance. residue may be locked in pyrite.
The QEMSCAN results showed no major cause for concern
Gravity – cyanidation leach optimisation from a processing point of view.
A series of tests were conducted to determine the optimal
grind size and cyanide dosage for the leaching process. It Oxygen uptake testing
was determined, after the first round of testing, that a 24 hour The oxygen consumption ranged between maximum values
leach time would be adopted due to negligible improvements of -0.024 mg/L/min for the oxide and -0.035 mg/L/min as
in leach recovery being achieved from longer leach times. a maximum value seen for both the shear and primary ore
Cyanide was very low and lime consumption as expected. types. As a result air sparging will suffice.
The test work program, conducted on the oxide, primary and
shear master composites, determined that a P80 of 63 microns Carbon loading testing
would be used for subsequent testing and throughout the Carbon loading tests were conducted on the primary and
process plant, with a cyanide dosage of 250 ppm. Under these shear master composites, with the slurry from a bulk leach
conditions all samples displayed fast leaching kinetics and on gravity tailings being used to generate a representative
high recoveries. carbon loading feed.

The primary material had a k value of 177 h-1, and an n value Table 5

of 0.74. Outotec thickener test results summary.
The k and n values for the shear test were 202 h-1 and 0.70
respectively, and therefore indicate no issues for processing. Parameter Ore Type
Oxide Shear Primary
Rheology Feed density (% w/w) 15 15 15
Rheology testing was conducted on all three master
composites. Based on the expected design of the process Flocculant dosage (g/t) 30 30 20
plant, the tests were conducted at a P80 of 63 micron, at 40, 50 U/F density range (% w/w) 47.4–53.2 49.0–63.4 55.9–63.5
and 60 per cent solids by weight. A range of shear rates was Overflow clarity (mg/l) ~100 ~100 ~350
applied to the slurries, with shear rates of 5 s-1 and 200 s-1 being
taken as low shear and high shear conditions respectively for
The cyanide consumption was low for all samples tested,
the analysis.
while the lime consumption was at levels below concern.
Based on the current process plant design, with correct
thickener control it is not expected that any viscosity issues
will be seen through the process plant. PROCESS PLANT DESIGN
The process plant design is based on two-stage crushing to
Thickening a covered crushed ore stockpile with a nominal 3800 t live.
The tests were conducted on RC chip samples (SRC730 and The primary jaw crusher is a Metso C120 and a Nordberg HP
SRC711) provided by Troy Resources, with three flocculants 500 secondary crusher. This is based on operating 6500 hr/a.
being tested. It was hence recommended to proceed with the The product will be a P80 of 10.3 mm from a double deck
Magnafloc 351 as the preferred flocculant in the design work screen. Crushing was selected based on low wear rates,
(see Table 4). energy efficiency and process certainty compared to a SAG
mill.
Outotec thickening test work A second-hand ball mill 5 m diameter by 7.32 m long was
Thickener design test work was conducted by Outotec in sourced out of Canada using a 40 per cent ball charge with
two stages; initially flocculant screening was conducted on a 3.1 MW motor. The mill is close circuited by cyclones with
five flocculants including the previously tested Magnafloc the cyclone overflow going to a 25 m preleach thickener.
351. From this testing it was determined that Magnafloc 155 The thickener underflow is pumped to the first CIL tank.
would be used for the dynamic settling tests. A preleach thickener was included to allow flexibility of
cyclone overflow and grind size, increased residence time in
The test work results showed that the optimal feed density the CIL and reduced cyanide and lime consumption. This also
for all ore types tested was 15 per cent solids. Other parameters assists in clean-up of spills inside the bunded areas of a high
from the test work are summarised in Table 5. rainfall area.
Based on the results of this work, there should be no issues Two Knelson XC30’s gravity concentrators are used to
in thickening the Karouni ore or cyclone overflow. recover coarse gold before being processed using intense
leach reactors. The high gravity gold recovery made this
Bulk leach and detoxification essential for higher overall recovery, faster cash flow and
In order to conduct the detoxification test work, a 10 kg amount improved gold security.
of sample from the primary and shear master composites was
Cyanide leaching is using six CIL tanks. The nominal leach
used as feed slurry.
residence time is 24 hours. The loaded carbon is stripped in a
The primary material showed strong performance at all pressure Zadra stripping column.
ratios, with the cyanide level, both free and WAD cyanide,
The gold is electrowon and smelted in the gold room. The
being below 1 ppm after two hours. The shear material was
carbon is regenerated after stripping on an as-needed basis.
unable to achieve free or WAD cyanide levels below 1 ppm
after eight hours at the lowest ratio, however at ratios of 3
and 4 both the free and WAD cyanide was seen to be below DETOX
1 ppm. Cyanide destruction testing was based on the INCO The sensitive environmental area required a DETOX to
sulfur dioxide system. minimise the potential impact on the environment.
The leached slurry in CIL tank 6 will pass from the final
VARIABILITY TESTING launder into the DETOX circuit, at DETOX tank 1 (300-TK-
Variability testing was conducted on five primary samples 008) for cyanide destruction.
only, due to the amount of available mass for this ore zone, The DETOX circuit contains three agitated tanks, (DETOX
and the limited mass of the other ore zones (see Table 6). tank 1, 2 and 3 – 300-TK-008/009/010) and will be bunded
Table 4
Reverse circulation chip settling results.
Property SRC 730 SRC 711

Magnafloc 351 Magnafloc 345 Magnafloc 351 Magnafloc 345
Free settling velocity m/h 25.50 27.06 25.88 27.85
Hindered settling velocity m/h 1.14 0.78 0.72 0.60
Underflow % solids w/w 55.3% 51.6% 56.2% 44.9%

Karouni Gold Project from drill core to commissioning
Table 6
Variability leach results summary.
Sample Head grade Gold recovery (%) NaCN consumption Lime consumption
(Au g/t) Leach time (h) (kg/t) (kg/t)
0 2 4 8 24
P-V1 2.59 58.4 76.1 77.9 79.5 80.4 0.09 1.25
P-V2 11.12 42.6 78.8 87.3 92.4 92.8 0.12 1.15
P-V3 3.94 43.5 70.0 75.5 76.4 79.2 0.08 0.91
P-V4 7.80 79.3 88.4 92.2 95.6 97.4 0.10 0.83
P-V5 4.58 58.2 78.5 84.8 87.9 95.2 0.07 1.03
separately to the CIL tanks and operate in series. This will Air
provide three stages of DETOX to ensure that the cyanide is
Air is used throughout the process plant. This includes air
completely removed through the 7.5 hours total residence
for the workshop, sparged into the leach circuit to improve
time. As with the CIL circuit the DETOX tanks will be able
leaching kinetics, and instrument air. Air will also be used in
to have a tank bypassed as required for maintenance. The
the Detox circuit. This will be provided by blowers installed
DETOX tanks will all be dosed with sodium metabisulfite
in this area of the process plant. Air will be supplied from
(SMBS), while DETOX tank 1 will also receive a feed of
one of two air compressors (800-CP-001/800-CP-002). This air
copper sulfate (CuSO4) to maintain a minimum 50 ppm Cu2+
will be stored in the air receiver (800-RA-001) until required
as part of the detoxification process. Spent hydrochloric acid
throughout the process plant.
(HCl) and acid wash water from the elution circuit will also
be pumped into DETOX tank 3. The DETOX tanks will be
highly agitated and be air sparged to drive the reaction which
Diesel
destroys the cyanide. The slurry will gravitate between the Diesel will be used on-site as the primary source of fuel
tanks in this circuit. Controls are in place to alarm and even and heating. Diesel will be used for the power station, the
shut the plant down if cyanide is detected in tanks 2 and 3 smelting furnace, elution heater and the carbon regeneration
above a preset level. kiln, as well as fuel for site vehicles. Power supply is based on
packaged sets using diesel fuel. This will supply power to the
Tailings camp, process plant and facilities. Troy Resources will own
and operate the power station.
A conventional tailings dam will be constructed on-site. Due
to the high rainfall run-off water collected around the process
plant and TSF, or from mine dewatering activities, will be CONCLUSIONS
handled in stilling ponds prior to discharge into the river Small resource gold projects can be profitable and the risks
system. This will ensure the water released is treated and can be managed using suitable second-hand equipment. The
hence suitable for release. concept of minimal engineering, that is, project management
The tailings are not acid forming (benign sulfides) and this of a group of subconsultants to execute the project did result
is a positive for the project. in substantial savings in time and money. The use of a second-
hand ball mill (never used) saved time and money. The simple
Water management flow sheet has low technical risk and can be readily expanded
should the need arise. The DETOX circuit was critical given
Water will be fed to both the raw water dam (700-DM-001) and
past problems in Guyana with cyanide spills. Troy Resources
the water purification feed tank (700-TK-001) from the river or
has gained valuable experience in project development in a
bore field. The raw water dam will provide water to be used as
new international area and has established a presence which
both feed throughout certain areas of the process plant (elution
will allow the evaluation of other resource opportunities in
circuit, gravity fluidisation and reagents), and for top up water
the immediate area of the plant. The project impact will be
for the process water dam (700-DM-002). Mine dewatering
very positive for Troy and the local community. Exploration
water will be used preferentially for processing and the water
in the region is likely to find new resources within proximity
balance indicates excess water will need to go to stilling ponds,
to the process plant and the operations future may continue
tested and then disposed of into the environment.
beyond the present resources.
Reagents
The reagents and services area of the process plant will be used ACKNOWLEDGEMENTS
for both reagent preparation and storage prior to distribution The author would like to thank colleagues and persons
throughout the process plant. associated with the Karouni Gold project for the education
and life experience received during this exciting period of
The reagents will be mixed in a bunded area in covered
the development, birth, nurturing and growth of a new
tanks to prevent heavy rain damaging or causing losses to
gold project in Guyana, South America. Thanks are also
the reagents. Bunding will be placed around key areas to
due to the management of Troy Resources for permission to
also provide additional levels of separation between reactive
publish this paper.
reagents.
Cyanide, hydrochloric acid, sodium hydroxide, flocculant,
copper sulfate, sodium metabisulfite and antiscalant will be
mixed on-site for delivery and use within the plant.

Contents
Upgrades, modernisations, automation and

expansions … where will the expertise,
capability and skills come from in the future?
R Coleman1, J King2 and T Hunter3
ABSTRACT
The industry that supports operations and new projects is suffering a rapid decline
in the current severely curtailed capital investment environment. This is evident by a
decline in the number of skilled teams, who are now centred in fewer locations. Other
features are an aging and retiring ‘baby boomer’ generation, coupled with a low
uptake of new graduates and the uncertainty in predicting future capital expenditure
programs and requirements. Combine this with a reluctance of cash-poor operations
to modernise and adapt in an environment of low metal prices, and there is a real
uncertainty as to how service providers will meet industry needs as they evolve.
Trends amongst service providers will be put forward as well as developing options
and alternatives. These embrace a whole gamut of agile alternatives in the areas of
partnering, alliances, commercial and shared risk models and mutual innovations.
These will be discussed with some case studies and suggested approaches.
Particular emphasis will be upon brownfield expansions and upgrades which have
emerged recently as a preferred lower capital and focused way to achieve the best
practical outcome at lowest cost. There are several specific features of such brownfields
projects including the requirement for very close cooperation between the execution
team, operators, maintenance and suppliers and a maximum usage of site resources
and ingenuity. Often, required investment hurdle rates are high, downtime allowed
is minimal and site resources usage is maximised. This requires mutual trust, good
planning and high health, safety, environmental and community (HSEC) standards
and practice. In many ways, the practices and skills needed are the exact opposite of
the historical ‘big Greenfields’ project with which the industry has been geared for
more recently.
Another developing area has been with innovative commercial models and
practices. This is a direct result of the more challenging investment criteria being
adopted and the relative hunger of service providers. It has become more common
to look at these commercial-costing alternatives very early in the project cycle as an
essential project tool and not just as a possible option. These commercial solutions
cover such alternatives as:
•• leasing and rental
•• deferred payments
•• build–own–operate–transfer and build–own–operate
•• whole-of-mine preferred spares and maintenance terms
•• aggressive performance targets with real assurance penalties.
Once again this requires very close mutual dealings and trust, and sufficient time
allocation compared to a conventional approach.
INTRODUCTION
The broader Mining Equipment Technology and Services (METS) industry has
grown since the early 2000s in response to the demand from the industry and is a
significant employer and economic contributor in its own right. This has recently been
1. MAusIMM, Account Director, South East Asia
documented and recognised particularly by governments (Hunter, 2014; Austmine,
Pacific, Outotec, West End Qld 4101.
Email: rob.coleman@outotec.com 2013). The METS subsector dealing with metallurgical technology development,
process plant upgrades and new projects, had grown at a real pace and likely peaked
2. Head of Sales Minerals Processing, Outotec,
Frenchs Forest NSW 2086. in 2012. It is now in serious decline and losing capacity, capability and confidence.
Email: jason.king@outotec.com There are several reasons for the decline, which include:
3. FAusIMM(CP), Senior Consultant, Clayfield Qld •• the well-documented and recognised step change in metal prices since 2013
4011. Email: tomhunter134@gmail.com from previous buoyant levels (Fiscor, 2015)
107
•• the ‘hangover’ psychology from the project excesses, in •• practice and history
terms of bloated capital cost compared to expectations, •• options available in the market place
delayed deliveries and fierce project competition for
•• degree of technical innovation
scarce engineering and construction resources
•• competitive climate
•• the generational change occurring as the ‘Baby Boomers’
move on from active project involvement •• owner preferences.
•• the severe contraction in the skill level and number of The above list is not exhaustive and often combinations and
teams now available for projects (Federation University hybrids are adopted to fit. In the current climate of minimal
Australia, 2014). capital expenditure, fierce competition and an industry
So the challenge is how to provide the level of service demand for predictability and economy there has been a strong
demanded by our industry in a sustainable and competent trend towards lump sum and well-defined scopes rather
manner. This paper looks at the challenges, some potential than the more flexible reimbursable models. This approach,
solutions, illustrated by some recent project examples, however, does demand larger and more skilled owner teams
particularly in the ‘new’ environment since 2012. and may also involve greater preparation time and earlier
project expenditure on early contractor involvement (ECI).
BACKGROUND A key (and often overlooked) consideration for a project
typically going through the phases of studies, early
There is a wide range of options available to operations and
projects wanting to spend capital (Luxford, 2006; Hunter and involvement and approvals, engineering and execution, is that
Broome, 2008; Revy, 2008; Lane and Clements, 2012). Looking of trust and predictability between the parties. This is much
through them, one can identify the following six points: treasured and encapsulated in the salutation ‘no surprises’.
Constant multilevel open communication, sharing openly the
1. The classic ‘in-house’ owner’s team with several ongoing
inevitable project challenges, and accurate forecasting, are
projects, who know the owner requirements well and
intrinsic to achieving this aim.
have developed systems and techniques attuned to the
site. In summary, there are many project models and a large
number and variety of providers who are hungry and keen
2. The ‘hybrid owner’ model where a trusted outside
for work.
small group coordinates the project. This has many of
the advantages of the classic ‘in-house’ but can be more
flexible in handling a variable project portfolio. DRIVERS AND TRENDS
3. The ‘owner project team’ where a specific team is
recruited for the project. Typically, this works well Drivers in the new environment
for a substantial complete project cycle of some years, The industry itself and the associated METS area are acutely
for example prefeasibility study (PFS) to definitive aware of how the industry has changed since 2012. The
feasibility study (DFS) (Hunter et al, 2009), to execution previous extraordinary emphasis on capital expenditure (and
and commissioning. poor discipline in operating costs) in the drive to maximise
4. The traditional ‘engineering, procurement, construction production and capacity in the buoyant price environment has
and management’ (EPCM) approach where the owner’s been very rapidly replaced by a completely new discipline and
team engages a substantial study-execution team ethos. This is where capital expenditure is very constrained
on a reimbursable basis (Loots and Henchie, 2007). and mainly focused on essential HSEC drivers and often
The EPCM acts as the project’s ‘arms and legs’ under only where there can be a very rapid return on expenditure.
the direction of the small owner’s team. The owner is ‘Brownfields’ has become the main area of focus with the aim
ultimately responsible for the project. The ‘long-term of minimising operating costs with particular emphasis on
reimbursable alliance model’ is a variation of this, where ‘debottlenecking’ and ‘incremental efficiencies’. Many areas
the EPCM imbeds a long-term core team into the owner of previous outsourcing and use of contractors have now been
structure and uses that to generate project deliverables returned to owner involvement as this is perceived (often
and services. correctly) as adding value and promoting increased owner
5. The lump sum, engineering, procurement and awareness of operations and options. Particular drivers that
construction (EPC) or design and construct (D&C) can be identified include:
approach where the owner’s team carries out substantial •• a complete change in focus from production and tonnage
definition and invites lump sum prices for a well-defined maximisation to capital and operating cost discipline
scope-of-work with the intention of having minimal (Clarke, 2015)
variations upon execution. This involves potential risk
•• an acute bottom down driven capital expenditure
for both parties because of the tight contract definition,
discipline and budget
transfer of risk and the issue of variations (if and when
they occur). •• a fundamental review of operating practices and
expenditure on operating costs
6. The technology driven project where the owner’s desire
to involve a new, improved, or even revolutionary •• changed philosophies towards outsourcing and service
technology provider or partner in the project (Wasmund provider models
et al, 2011). The driver here is the technology provider •• changed perceptions by financiers and investors on
and the execution model is ‘adapted’ to meet the needs the sector’s prospects, relative attractiveness, future
of the researcher/technology owner or developer as the profitability and drivers; this has totally changed the
owner deems the greater execution risk is outweighed environment for the industry at all levels
by the perceived advantages of the technology.
•• the owner/operator view that the extraordinary
The chosen model will be governed by: demands up to 2012 for services had resulted in
•• company policy bloated costs (and charges) from the METS sector and

Upgrades, modernisations, automation and expansions … where will the expertise, capability and skills come from in the future?
poor performance in terms of productivity, relative providers. These arrangements tend to be more specific
international costs and project efficiency and early results driven compared to those of the early
•• competition for resources (particularly in engineering, 2000s which were more industry wide and often driven
project management and construction resourcing) was by the research or developer partner
a particular factor with coincident buoyant commodity •• a much higher degree of accountability in providing
pricing in energy and the ‘once in a generation’ services or solutions is being demanded
establishment of the huge regional liquefied natural •• the highly competitive new environment has
gas (LNG) projects in Western Australia, Queensland, concentrated the industry to fewer large providers and
Northern Territory and Papua New Guinea several new smaller entrants. The more junior sector
•• commercial discipline was viewed by many owner/ for instance, looks for early study engagement with
operator organisations as having been compromised commitment to their project and financing demanding
during the boom and needed to be ‘normalised’ a lump sum or EPC approach. This requires a nimble,
•• many services organisations had expanded laterally, client centric and innovative execution model well
usually in response to client demand, into extra services removed from the reimbursable EPCM model of the
outside their traditional core business. early 2000s.
This was therefore a new paradigm for many in the
industry. There were memories of years like 1987, 1992–1993 WANTS, NEEDS AND ISSUES
and 2001–2002 where downturns had occurred, but the speed, We have previously discussed the industry trends and
depth and completeness of the 2013–2016 downturn has been drivers, and will now focus on the wants, needs and issues of
viewed as almost unique probably due to rapid industry the models presented earlier.
growth in the decade to 2012.
Classic ‘in-house’ owners team
Recent trends Generally, a larger organisation with multiple sites and a
This paper has not tried to identify specific trends, which history/confidence in client knowledge/alignment and
can vary by company size, history and specific practices, projects delivery coupled with well-developed systems. This
geography and commodity, but have chosen those that in requires:
our view are the most germane and relevant to the Australian •• competence and alignment
industry in 2016. Obviously in the current environment, we
•• commercial/contract agreement
are faced with issues, such as high industry unemployment
and underemployment particularly at graduate level, and •• a predictable costing model justifiable to management
also in geosciences, engineering and construction, as well as •• ability to align with corporate HSEC imperatives
much reduced overall spend on services by the industry. •• comprehensive systems aligned to the owners
Relevant trends include: •• adherence to project milestone expectations/
•• a much reduced new project study and execution commitments.
capacity and capability
•• capability and project teams becoming concentrated into Hybrid owner model
fewer regional and global centres (for example, Perth, Here a specific outside team is recruited to act as the owner,
Brisbane, Toronto, Vancouver, Santiago and usually on a single project or program. What is required
Johannesburg), and many former ‘powerhouses’ having includes:
little real capability now (for example, Melbourne) •• a high degree of comfort and trust in the service
•• the increasing usage (particularly with demonstrated providers
and consistent performance) of ‘best value engineering •• a willingness to align and provide/source the areas that
centres’, such as those in Mumbai, Delhi, Dubai, the owner team are unable to provide
Shanghai, Kolkata and Manila. It does take time and real
•• a staged and flexible commercial arrangement with a
corporate commitment to incorporate this potential into
high degree of openness/transparency
reality and to achieve consistent performance
•• ability/willingness/appetite for varying execution
•• in-sourcing of many aspects for projects and execution,
models and partners (note that systems alignment can
and the re-emergence of some internal engineering
be an issue as can process coordination, so specialist
teams within organisations
help in these areas may be needed. Often metallurgical
•• a much more rigorous contractual, commercial and cost process expertise is provided by an outside expert who
environment may have historical owner links.)
•• a trend back to longer term alliancing often with
innovative and open commercial arrangements Owner project team
•• a very open dialogue between owners and potential or The owner project team may be used for a longer term or
incumbent service providers larger project of a cycle of some years. This is particularly so
•• opportunities for small and single-person consultants for a larger EPCM project where the owner team needs the
and contractors, particularly for studies, audits and authority and expertise and risk assumption inherent is such
upgrades arrangements. Special owner challenges exist for this case,
which is basically a combination or version of the two already
•• realisation by owners that significant technical and
considered, and includes:
productivity gains can only be achieved by long-
term alliances and joint developments with ‘original •• the use of systems and procedures that fit both the
equipment manufacturers’ (OEMs), research and parent owner and the contracted EPCM
technology developers and leading edge service •• retention and alignment of the team

•• decisions as to the degree of intrusion/checking into the •• for the contractor, the commitment to bid will only
EPCM and the degree of alignment and trust with the be made with a good certainty of project actually
EPCM (the ‘man matching’ syndrome) happening, a high probability of bidding success and
•• project manager alignment, retention and drivers can be a level of comfort as to the owner’s competence and
an issue. financial strength; often a contractor will strive to have
an ECI status where they are paid (partially or wholly)
There are many contractors who are very comfortable to
to invest the resources to give a bid
work on the owner side in this team environment and it fits
well in career development. One challenge is that each brings •• a lump sum bid by a substantial competent contractor is
a unique experience, preferences and skill set and that team very important to project participants such as banks and
alignment particularly on systems use can be a challenge. other lenders/financiers.
Some owners believe the risk is lowered once this approach
Traditional engineering, procurement, construction and management is adopted and that all risk is now with the contractor. This is
The EPCM approach is where the owner’s team essentially certainly not the case as a good owner’s team needs to check
uses an ‘agent’ to perform most of the tasks under the quality and progress, ensure what is being built is consistent
with owner expectations and the contractors obligations, and
direction of the owner team. This has worked well in the
to track and document inevitable variations.
past but perceived shortcomings were evident to many in
the period of 2002 to 2014 (Walker, 2015; Moore, 2015). These
issues involved: Technology-driven project
All the above approaches have assumed conventional,
•• a shortage of skilled project personnel for both
well-proven technology and process and would be greatly
EPCM’s and owners led to issues with team retention,
complicated by new, first-of-a-kind or largest-yet-built
incentivisation, alignment and performance
technology. The main driver in such projects is the special
•• intense competition for people driving up rates and needs of the process or technology provider and the degree
affecting productivity and quality to which the execution strategy and practice needs to be
•• excessive ‘project churn’ of people (turnover) adapted. These particular technology drivers may mean that
exacerbated these issues the OEM or technology provider (TP) may adopt any of the
•• the traditional clear delineation of owner and contractor above roles in addition to their speciality one and may have a
roles, functions and responsibilities became a real issue. great influence on the project model adopted and the project
participants. Areas of particular importance are:
It should be remembered that EPCM had evolved in the
industry to provide a solution to the need to efficiently exercise •• the relationship, confidence and trust between owners
projects in a timely manner with maximum owner control and OEM/TP
and accountability. When properly done by competent teams •• the level of development and maturity of the technology
in a ‘normal’ market it is still an efficient and well understood – there is a vast difference between a larger sized piece
project mechanism. of conventional equipment and a completely new
multiphased, integrated hybrid process only proven at
A recent trend has been to use the EPCM model and
demonstration scale; for example, mineral processing
expertise for longer term ‘alliancing’ where the client selects
with hydrometallurgical and pyrometallurgical
one or more alliance partners (usually from traditional EPCM
downstream steps (Mezei, Todd and Molnar, 2006).
practitioners) to work on a longer term embedded basis in the
owner team. The intention is to have a dedicated team that is •• clear definitions of roles and responsibilities need to
able to cover internal project generation, definition, approvals, be agreed and formalised with minimal number of
execution, commissioning and alignment. This has long been interfaces
the norm in the oil-gas-petrochemical area and combines •• the program needing to recognise the complexities
the advantages of a competitive contracting environment by having good ‘float’ and a realistic commissioning/
with dedicated owner teams who build-up client and project ramp-up schedule
familiarity and expertise, and project execution efficiency. •• multiparty involvement and complex reporting
imperatives
Lump sum engineering, procurement and construction •• each participant needing a technology ‘champion’
The lump sum, EPC or D&C approach involves slight who believes that the project/process complexities are
variations on a simple execution method namely that the outweighed by the advantages and potential.
project is defined by the owner who invites bids to do the
complete job. The issue for the owner is in defining the scope
of the project in sufficient detail to enable a contractor to cost
COMMERCIAL CONSIDERATIONS
and price their bid. Likewise, such an effort by the bidders is The typical execution models and study/practitioner issues
very expensive in terms of dollars, commitment of experts, have been examined but often of overriding importance will
the level of detail and clarifications needed, and in time taken. be the broader and detailed commercial matters. For the asset
It would not be unusual for a mid-size concentrator (if defined owner, these include:
to a 30 per cent engineering level) to take six to nine months to •• Risk profile and factors – a formal risk register needs to be
arrive at a bidding cost commitment of some millions. completed early and agreed internally by stakeholders. It
The issues here are: is often useful to do two; one from the owner viewpoint
and the other from that of potential contractors. This
•• the level and certainty of pre-work by the owner’s team helps in alignment later and identifying/quantifying
to reach the stage where it can be bid areas of potential dispute. A fundamental element of
•• the discipline required by the owner’s team in resisting risk mitigation on a project is that the particular risk
change, documenting and affirming variations and should always be allocated to the party best able (or
using the systems needed for project tracking willing) to manage that risk. If a party is uncomfortable

or unfamiliar with the particular risk it will be the subject Outotec itself adapts with its customers and for specific
of excessive provisioning and uncertainty. applications the preferred execution models, partners and
•• Is the potential contractor capable of completing the commercial model. There is no ideal single solution. We
task through to a satisfactory outcome? Factors such acknowledge that there are very competent competitors and
as: history, capability, capacity, team competence, alternatives, ensuring owners have good choice with their
contractor internal execution model, subcontractors, projects.
reference projects, reliance on variations for commercial The opinions expressed on these selected examples are
means should all be considered. This is often referred to those of the authors and are intended to generate ideas on
as ‘the smell test’. project improvement and how the optimal execution strategy
•• The importance of project staging, particularly with can be selected.
a large venture or one involving greenfields or new
technology features. This could involve multiple stages Alliances
from initial study/PFS/test work/pilot plant through ‘Alliances’ are very case- company- and site-specific and can
DFS, engineering, execution and commissioning. Is the cover the whole gamut of studies, engineering, approvals
contractor capable of these? and review, construction and execution, and commissioning,
•• Is the owner potentially committed for some years of upgrades, modifications, debottlenecking and operations/
working productively with that contractor? maintenance. There are many advantages for these
arrangements particularly for larger, multisite operations
•• Should an execution model be adopted which plans for such as Tronox, Alcoa, South 32, Rio Tinto and BHP. Owner
use of multiple contractors at different stages whilst management of the Alliance is best done with a specific
ensuring project knowledge is not lost? dedicated internal resource and an appropriate sponsorship
•• The execution model commercial/contract environment and hierarchy. The ability to Alliance with global miners is seen
will be fundamentally determined by whether it is as a key business imperative with many service providers who
reimbursable or lump sum. For a lump sum, very also emphasize the need to align with the customer’s systems,
careful and detailed definition of scope is required and procedures, standards, reporting systems, approval hierarchy
intrinsically the contractor will build in a risk premium and build-up ‘corporate and plant knowledge’ amongst the
and the cost of detailing variations. Conventionally this seconded employees. A key issue with the seconded team can
is estimated at about 15 per cent above that estimated be that they, over time, identify more strongly with the owner
with a reimbursable contract where the owner has a than with their parent service organisation.
much higher risk commitment but also greater flexibility Jacobs is a good example of the effective use of Alliances.
in changing scope. Jacobs have 30 years of experience in creating value within
•• The contract set-up needs to reflect the fundamentals the framework of alliance relationships which has led to
of how the job is being done and be consistent with the development of their alliance approach (Jacobs, 2013).
the practicalities of the parties’ performance. If it is too Their approach is based on the understanding that there is
one-sided, it will engender undesirable outcomes as the no ‘one size fits all’ formula to forming a successful alliance/
lesser party protects its interests eg excessive liability partnership. Each one has to be tailored to the objectives and
penalties will mean a behaviour focused on avoiding goals as well as the needs of the client. Jacobs have developed
those outcomes and this may be inconsistent with the a range of models of Alliances. They all have somewhat
best overall project result. different characteristics but all have the common goals of
There are many studies into project performance and continuous improvement and bringing best practices and
the factors behind underperformance or failure in the oil- lessons learned to their client’s business. More than 60 per cent
gas-petrochemicals area and more latterly in the minerals of their revenue is currently generated through Alliance-type
industry. The major conclusion for successful performance relationships. Therefore, forming and maintaining healthy
was the degree of front-end loading (FEL) on a project; the alliances that continually deliver value to their clients is key
more resources, effort, project team involvement and early to their success.
engagement of the contractor(s), the better. This would be
backed up by rigorous benchmarking and stage gate approval Owner project team
processes. Masan Resources (MR) in Vietnam has developed the Nui
Such systems have been adopted more widely in the Phao polymetallic project some 70 km NW of Hanoi and this
minerals industry over the last ten years, with some notable is a large modern project producing a large range of tungsten
successes. The system does, however, add earlier time and products, chemical grade fluorspar, bismuth metal and
costs to a project, but this is ultimately vastly cheaper than copper-gold as a concentrate (Masan Resources, 2016). The
making the mistakes later in execution. It does require a high process plant is specific and complex involving many process
degree of owner discipline and commitment, systems and steps and unique metallurgically (Morgan, 2016).
project team talent and input. Establishment of the project emphasized local capability
with western (Australia-India) engineering, along with
procurement and construction by the owner. For the ramp-
EXAMPLES up/debottlenecking/process optimisation phase, MR
Some indicative examples of recent projects will now be chose to use a model of a strong plant site metallurgical
provided. We acknowledge that there are many competing team, supported by a specialist engineering owner’s team
models, companies and service providers in our industry who in Australia. Emphasis was put on partnering with OEMs
energetically compete and this is very healthy. The project /service providers and maximising the use of in-country
developer faces the task of deciding how best to do the project construction capability. Projects are fast-tracked, fit-for-
and with whom. purpose and aggressively managed to align operations,
As a global supplier of equipment, industry solutions and plant improvement, OEMs and technology suppliers and
services, and a developer of new technology and techniques, the in-house team. This has enabled stepwise and regular

improvements in plant performance at appropriate cost and •• sufficient basic engineering on an open book basis to
tight capital control. A similar approach has worked very well derive an EPC cost for the project
recently at the Didipio project in Luzon, Philippines, operated •• fixing a mutually agreed EPC price
by OceanaGold (Walker, 2013).
•• EPC project execution.
Hybrid model This represents a very flexible contracting model aimed at
Carmen Copper Corporation (CCC) is a copper concentrator reducing the risks for both parties, and therefore reducing the
situated on Cebu Island in the Philippines. The concentrator risk premium associated with a traditional EPC contractual
was originally commissioned in 1955, and operated until 1994, model. This approach facilitates the client operating with a
when low copper prices forced the mine to cease production. small owner’s team, in line with an EPC strategy.
The mine was rehabilitated in 2007 and brought back into The initial design phase to produce a fixed EPC price is
production at a rate of 40 000 t/d, as outlined by Morgan et al based on a collaborative open book basis ensuring that Alkane
(2014). The decision was then made to increase the production has sufficient input into the plant design and the costs, before
from 40 000 t/d to 60 000 t/d. The decision was also made signing the EPC contract. This staged approach is designed
to utilise existing installed equipment as far as possible to to reduce the risks of both contractor and client, utilising
reduce costs. the prior client knowledge in the design, then agreeing on a
The client wanted a world-class project, with modern mutually acceptable EPC price.
process controls, utilising in-house capacity for construction,
and low cost detailed engineering. This resulted in a hybrid Build–own–operate and build–own–operate–transfer models
model being utilised for the project: Outotec has recently delivered a total system solution for a
•• metallurgical test work program designed and managed client in Australia to meet their underground mine backfill
by CCC and operational needs (Suvio et al, 2016). The contract
included an extensive backfill test work program, backfill
•• Outotec contracted to carry out the detailed design,
plant and underground piping EPC, system commissioning,
including a 3D model of the plant
complete mine backfill system operation and maintenance.
•• a local Philippine company contracted by CCC to do the This represents a classic build–own–operate (BOO) contract.
detailed engineering
This contract contains a higher level of backfill engineering
•• Outotec supplied and installed all the proprietary and expertise and provides the customer with unique
equipment expertise in state-of-the-art mine backfill engineering and
•• CCC procured all non-Outotec proprietary equipment plant operation. The major customer advantage is a lower
•• construction carried out and managed by CCC cost operation, operated entirely by Outotec and providing
the customer with lower risk to their mining operations
•• Outotec provided expert advice including construction
throughout the life-of-mine.
where requested
This type of contract represents the ‘Ultimate Process
•• CCC managed the commissioning supported by vendors
Guarantee’ as the contractor carries the full financial,
and expert advice from Outotec when required.
performance, operating and maintenance risk providing an
CCC selected the above contractual model to fast track over the fence service to the client.
the project compared to the standard EPC/EPCM model.
This enabled early works to be carried out facilitated by
utilising existing labour structures, circumventing the usual THE FUTURE
mobilisation phase of traditional projects. The project was Looking to the future, we need to make some assumptions
designed, constructed and commissioned in 20 months before venturing into this crystal-ball space. We have
from receipt of a letter of intent. This was an excellent result assumed a period of three to four years, with the industry
considering that the base starting document was a PFS, capital activity relatively restrained as it is at present. Whilst
compared to the standard practice of first doing a DFS as a this is hopefully a pessimistic view, we believe it is prudent
minimum. and consistent with many of the industry commentators and
the predictions made below:
Lump sum early contractor involvement •• more contractual, commercial disputes than hitherto
Alkane Resources Dubbo Zirconia Project (DZP) has one of •• the EPC model will continue to be more attractive than
the world’s largest in-ground resources of rare metals and the reimbursable EPCM method
rare earths (Alkane Resources, 2016). The mine is expected to
process 1 Mt/a over a period of 70 years or more. The DZP •• alliancing in all its forms will be a growing trend
will produce zirconium, hafnium, niobium, yttrium and rare •• dedicated owner teams for projects will be more widely
earth elements. This complex metallurgical plant flow sheet recognised as vital in terms of project success, necessity
fits in very neatly with the suite of process technology that and continuity
Outotec has developed over many years. •• there will be very few ‘new technology’ projects and
Outotec was engaged on an early contractor involvement plants as a result of owner conservatism and the much-
(ECI) basis for the project. The set-up is as follows: diminished recent capacity of in-house and industry
•• initial value engineering phase – reviewing the front-end collaborative research, development initiatives
engineering design (FEED) study carried out by others, •• process plants will become more complex and often
identifying gaps and opportunities for cost savings, and incorporate ‘hybrid processes’ eg mineral processing
process enhancements with hydrometallurgy
•• technical input into the pilot plant test work program •• innovation will be directed to ‘peripheral’ areas such
ensuring that the process is fully tested, and all market as water recycling/treatment, energy usage, advanced
and vendor sample needs addressed process control, remote operation and control, and

tailings management, as areas to give the greatest Hunter, T C, Borthwick, J, Douge, M and Yusi, M, 2009. Project
predictable advantages. feasibility studies – a necessary step or the best opportunity to
add value and certainty?, in Proceedings Project Evaluation 2009
Alternatively, if there is an appreciable recovery in metals Conference, pp 35–48 (The Australasian Institute of Mining and
prices and market demand, we would expect that many of the Metallurgy: Melbourne).
recent excesses and problems that emerged in 2008 to 2013 will
Jacobs, 2013. Partnering/Alliances white paper, Jacobs.
be repeated or even exaggerated. The capacity of the industry
for projects, from a skills and coordination viewpoint, will be Lane, G and Clements, B, 2012. Operations versus projects – how
much reduced due to recent events. One could foresee a need do people think and what are the implications?, in Proceedings
11th AusIMM Mill Operators’ Conference 2012, pp 11–15 (The
for strong self-discipline and clear thinking to avoid past issues.
Loots, P and Henchie, N, 2007. Mayer Brown article, worlds apart
ACKNOWLEDGEMENTS – EPC and EPCM contracts: risk issues and allocation [online].
The authors thank several Outotec and industry practitioners Available from: <http://fidic.org/sites/default/files/epcm_
for their input and comments, with special thanks to loots_2007.pdf>.
Alan Dennis. The support of Outotec’s management is Luxford, J, 2006. Project development and construction management,
acknowledged. As the authors collectively come from in Proceedings International Mine Management Conference, pp 97–105
backgrounds covering operations, research and development, (The Australasian Institute of Mining and Metallurgy: Melbourne).
project development, equipment and solutions supply, Masan Resources, 2016. Website homepage [online]. Available from:
services delivery and studies, we have tried to use that <http://www.masangroup.com/masanresources/en/projects/
experience to give a balanced and wide view. A keynote nui-phao/highlights> [Accessed: 10 April 2016].
opinion piece does involve judgement and personal opinions Mezei, A, Todd, I and Molnar, R, 2006. Can complex
and is intended to engender debate and discussion. As such, hydrometallurgical pilot plants effectively reduce project risks?
it is acknowledged that there will be a necessary degree of Part 1 [online], SGS technical paper 2006-2004. Available from:
interpretation and personal opinion. The authors are happy <http://www.sgs.com/en/mining/metallurgy-and-process-
to discuss this aspect and are open to further elaboration or design/pilot-plants>.
modification of their stated positions in this paper. Moore, P, 2015. EPCM and contracting – adapting to conditions,
International Mining, February, pp 78–86.
REFERENCES Morgan, S, Serdzeff, S, Malaloy-On, C and Pagdalian, L C, 2014.
Alkane Resources, 2016. Website homepage [online]. Available from: The benefits of technology partnerships during brownfield
<http://www.alkane.com.au/index.php/projects/current- upgrades, in Proceedings 12th AusIMM Mill Operators’ Conference,
projects/dubbo/project-overview> [Accessed: 14 April 2016]. pp 471–476 (The Australasian Institute of Mining and Metallurgy:
Melbourne).
Austmine, 2013. METS National Survey [online], in Australia’s
New Driver for Growth, funded by the Department of Industry, Morgan, S, 2016. Performance and safety optimisation in Vietnamese
Innovation, Climate Change, Science Research and Tertiary operations, in Proceedings Regional Technical Conference Mining
Education, p 2. Available from: <http://www.austmine.com. Vietnam 2016 (Mining Media International: Melbourne).
au/Portals/25/Content/Documents/Austmine%20Survey%20 Revy, T, 2008. Trends in project development: where are they
Highlights.pdf>. heading? [online]. Available from: <http://www.ibram.org.br/
Clarke, N, 2015. Improving productivity in processing operations, sites/1400/1457/00000190.pdf>.
paper presented at The International Mining and Resource Suvio, P, Palmer, J, del Omo, A and Kauppi, J, 2016. Holistic tailings
Conference (IMARC), Melbourne, November. management solutions [online]. Available from: <www.outotec.
Federation University, 2014. Skills and Innovation in the Resources com>.
and Mining Sectors. Walker, S, 2013. Didipio: a Philippines success story [online],
Fiscor, S, 2015. Copper at the crossroads [online], Engineering & Engineering & Mining Journal. Available from: <www.e-mj.com/
Mining Journal. Available from: <www.e-mj.com/features/5307- features/3502-didipio-a-phillipines-success-story> [Accessed: 10
copper-at-the-crossroads> [Accessed: 12 June 2015]. December 2013].
Hunter, T C, 2014. Concentrators – past, present and future trends Walker, S, 2015. The EPCM perspective [online], Engineering &
for operators and service providers, in Proceedings 12th AusIMM Mining Journal. Available from: <www.e-mj.com/features/5029-
Mill Operators Conference 2014, Townsville, p 3 (The Australasian the-epcm-perspective> [Accessed: 05 March 2015].
Institute of Mining and Metallurgy: Melbourne). Wasmund, B, Voermann, N, Haneman, B, Sarvinis, J and Sheehan, G,
Hunter, T C and Broome, A J, 2008. Current trends in project delivery 2011. Implementing new technologies in metallurgical processes:
to the minerals industry, paper presented at The AusIMM New building plants that work, in Proceedings 50th Conference of
Zealand Branch Annual Conference. Metallurgists, pp 1–19 (Canadian Institute of Mining, Metallurgy
and Petroleum: Montreal).

Contents
Is an 80th percentile design point logical?

D David1
ABSTRACT
Clearly, a plant designed only to treat average ore at the nameplate rate will fail to
achieve nameplate in any typical year. To insert the necessary capability to achieve
nameplate it is common process engineering design practice that a plant must be
able to treat an ore with 80th percentile value (hardness or competence values for
example) at the nameplate rate. In the author’s experience, apart from a couple of
notable cases, this has been applied reasonably well as a principle, with widespread
success. However, in a number of recent instances it became clear that the use of
the 80th percentile number would have resulted in significant under-design of the
plant. This paper makes the case that the 80th percentile, as a principle, can have
serious flaws and its use needs to be assessed on a case-by-case basis. The discussion
in this paper is the first step in developing a new design principle, and the associated
methodology for selection of a design value, that will ensure plants are designed to
achieve their nameplate capacities.
INTRODUCTION
Designing a process plant has many conventions that are often taken for granted. One
of these is that the 80th percentile value of a key measure, like the Bond ball mill work
index (BWI), will provide an unquestionable margin of design safety in the plant. The
normal procedure is to take a set of test results, for example 20 individual BWI values
from around the orebody, arrange them largest to smallest and take the 16th largest
value. This is placed in the design criteria as the 80th percentile, usually alongside the
average BWI value, and is subsequently used to design the ball mill. Many years later
that number can assume almost legendary status as the work index of the deposit or it
may have assumed infamous status as being the main cause of the failure of the plant
to achieve nameplate throughput or grind size.
The purpose of this paper is to explore circumstances where the 80th percentile
value will not provide a definitive design point. Guidelines for avoiding being
fooled by a false design value are provided, together with the basis for alternative
design methodologies.
PROBLEMATIC DATA SETS

The annual plan
If the 20 values of a particular measurement are provided from a 20-year annual mine
plan, then failure to design for variability for that measure is virtually assured. One
of the most common values that dangerously makes its way into design criteria from
mine plans is head grade.
A recent analysis of copper (Cu) head grade for a project determined the variability
of a number of related data sets ranging from the most inherently variable to the least.
The most variable data sets are those that look at the orebody in individual parcels
that may represent minutes or hours of plant feed. Two examples of such data (present
in virtually all projects) are the drill database and the block model. At the other end
of the variability scale is the annual mine plan. To arrive at the annual mine plan it
is normal to use the individual ore blocks contained in the block model in an orderly
and controlled fashion. The use of blocks (in an open pit example) is orderly because
it commences at the surface and must progress downward in some logical mining
sequence. The use of blocks is controlled because it is typical to attempt to obtain a
target head grade to the process plant when making selections from the blocks that
are immediately available to be ‘mined’. In many cases there is also a stockpiling
system in place (usually incorporated into the block sequencing procedures) between
to the mine and process plant to allow grade control to be achieved when the mine
1. FAusIMM(CP), Technical Director – Process, cannot directly provide it. Usually the measure being controlled is the head grade
Amec Foster Wheeler, Perth WA 6000. of the most valuable component, and this also is the measure for which believable
Email: dean.david@amecfw.com variability information is required for design.
115
D David
Almost immediately upon arrival at a copper mine for Interestingly, the core database 80th percentile would also
commissioning the author was greeted with the bunded underestimate the shift 80th percentile value, while both the
flotation area of the plant freshly filled with copper metallurgical test data set and the block data 80th percentile
concentrate, mostly from the cleaners and recleaners. The values would be acceptable estimates. However, all three
launders and pumps were continuing to overflow. On asking small-sample data sets overestimate the Upper Limit 95 value
what had caused the problem, the response was that the plant by at least 100 per cent compared to the shift data.
had been processing 8 per cent Cu in feed for the last day and When put in the context of real operational requirements for
they couldn’t do much about the problem immediately as the a process plant, it becomes clear that the even a weekly mine
stockpile was full of 8 per cent Cu ore! plan is not a valid source of process design information, let
The design criteria for this project were, in theory, prepared alone an annual plan.
on an even more conservative basis than the 80th percentile
because the ‘maximum’ grade had been used (100th The composited sample
percentile). This value was 1.6 per cent Cu and, obviously, To save money, a recent client had only conducted five sets
nowhere near what was being dealt with by the plant. of metallurgical tests at laboratory scale for a definitive
Although this value was in the design criteria, the maximum feasibility study (DFS) level design. Minimal background
case mass balance assumed a head grade of only 1.39 per cent information, apart from the data itself, was provided for
Cu, the 80th percentile value. Both the 100th percentile and a review. The five BWI values provided all lay between
the 80th percentile values were selected based on the data in 11.1 kWh/t and 11.8 kWh/t. The standard deviation (SD)
the annual mine plan. between the six tests was 0.3 kWh/t, approximately what
is accepted to be the inherent repeatability of the BWI test
An actual example from a design study of the variability itself. The standard deviation of the drop weight index (DWI)
levels for the various data sets described at the commencement results was only 0.4 kWh/m3, which is again similar to the
of this section is shown in Figure 1. The ‘upper and lower limit level of repeatability expected of that test.
95 per cent’ lines represent the range that contains 95 per cent
The 80th percentile of the five BWI values was only
of the data within that particular set. Note that this is not
2 per cent greater than the average value and the maximum
the same project that had the 8 per cent Cu in flotation feed
value was only 3.5 per cent above the average. For the DWI
problem. This ore has less variability and an average grade of result set, the 80th percentile value was only 10 per cent higher
only 0.51 per cent Cu. than average and the maximum value was not much higher
The 80th percentile value for the shift data set is 0.91 per cent at 13 per cent above average. There were two possibilities to
Cu and this would be a reasonable value to use for design explain these results, either the orebody was the first one in
purposes. The shift data set also predicts that the plant the author’s experience where the material was ‘all the same’
can expect to treat 1 per cent Cu (or greater) in feed for or, something very strange had happened in the constitution
2.5 per cent of the time (the ‘Upper Limit 95’ is actually the of the samples.
97.5th percentile value) which is 27 × 8 hour shifts per annum, In response to requesting more information, the core
or more than two per month. Clearly choosing to use the 80th intervals making up each of the samples were provided. The
percentile annual, monthly or weekly mine plan Cu grades for source of the problem was immediately obvious and it was
design purposes would dangerously underestimate the real composite preparation. Through compositing all semblance
Cu grade variability that the plant must be able to process. of variability had been eliminated from the test samples. Each
FIG 1 – Variability bands for grade data sets from a single orebody.

test sample was a composite of a minimum of 60 core intervals Provided the samples have been selected correctly, it is
from a minimum of four different drill holes. The stated certainly possible to derive a reasonable 80th percentile value
aim of this particular piece of work had been to define the for the fresh ore. Conversely, it is not even possible to estimate
properties of five different areas in the orebody. Obviously, it the degree of variability in key properties (let alone the 80th
was confirmed that the five areas can be considered virtually percentile values) that characterise the oxide cap using only
identical, on average, with respect to ball milling. However, two samples.
the test data set contains compositing that is equivalent to, If the oxide cap represents 100 per cent of plant feed for the
at least, the level of compositing inherent in an annual mine first six months, then understanding the properties of that
plan (and maybe even equivalent to a five-year mine plan material is of reasonably high importance. The plant will be
basis). As demonstrated in the last section, annual plan data is commissioned on that ore and it may play a significant role
unsuitable because it contains a level of variability far below when banker’s tests and warranty tests are being conducted.
what is needed for safe design. In this situation, a minimum of ten spatially distributed
Applying the relationship between annual variability and samples of oxide ore need to be tested before the variability
shift variability from Figure 1, it is possible to provide an can be estimated. There is usually one guarantee with oxide
estimate of what the true variability of BWI values might ore, it will be more variable than the fresh rock from which
have been, had multiple contiguous samples been tested it is derived.
individually. From Figure 1 the 80th percentile value from The transition ore presents similar problems, not least of
annual mine plan data is only 10 per cent above the average which is definition. An ore is called transition because it is
value. In contrast, the 80th percentile on a shift basis is part way between the original rock (in this case fresh ore)
74 per cent above the average value. Therefore, with the and the geological layer above it (in this case the oxide cap).
benefit of having all data sets available, a reasonable estimate Invariably, the transition ore will contain examples of the
of the design point, the shift 80th percentile, can be made from end members (fresh and oxide) together with everything in
the annual BWI 80th percentile as follows: between. The high degree of variability in typical transition
ore demands that a reasonable number of samples be tested,
80th BWI Shift = Average BWI + provided of course that the transition ore type represents
(80th BWI Annual – Average BWI Annual) × 7.4 a substantial plant feed component for a long enough time
period to be considered separately in the design process.
This equation has been applied to the five-sample case above
where the only data available has annual-plan-equivalent The more components in the ore the more important it is
variability and is unsuitable for design purposes. The outcome to have detailed mine plans guiding the design process. The
of this calculation is a more believable 80th percentile BWI most common misconception for the inexperienced is that
value of 13.2 kWh/t, 16 per cent above the average value of the proportions of each ore type in plant feed can be taken
11.4 kWh/t. Given the uncertainty in what the original data directly from the monthly or annual mine plan. The only
set actually represents, the real 80th percentile value on a way a realistic appreciation of the variability in ore type
proportions can be gained is to understand how such ore gets
production shift basis could be even higher.
from the mine to the plant.
This example clearly shows the danger of relying on
As an example, an orebody has two ore types with distinctly
composited samples for design purposes. The circumstances
different comminution properties and these ore types are
behind these composites are extreme, but it must be recognised
planned to be, on an annual basis, delivered to the plant in
that any compositing reduces the inherent variability that will
a 50:50 ratio. Simplistically the ore type properties can be
exist in the set of test results. Knowing how samples have
averaged and then used for design. However, discussions
been selected and prepared is essential to understanding the
with the geologists reveal that the first ore type is on one side
design implications of test data.
of the orebody and the other ore type is on the other side of
As a variation on the compositing theme, in many projects the open pit. Discussions with the miners reveal that they are
tests have been carried out on annual composites (for only intending to have one shovel in ore at any one time and
example Year 1 composite, Year 2 composite etc). Taking an that there is no intention to set up blending stockpiles, as it is
80th or 100th percentile value from the set of annual composite too expensive to double-handle all the ore. The 50:50 ratio is
results will, again, dangerously underestimate the variability not a controlled 50:50 blend and will be sequential processing
the process design needs to cope with. of one ore type, followed by the other. The resulting design
requirements are totally different to the requirements for
Multi-modal ore properties processing a controlled 50:50 blend.
Few orebodies consist of a single lithology or a single Where does the 80th percentile come into this discussion?
geological classification of material. In many instances (but The 80th percentile of the blend is obviously of little use. Each
not all) the comminution and separation properties of the ore type now needs to be understood individually in terms
different geological units can be distinctly different and need of variability, and the number of test samples required has
to be understood separately. If the geological differences also probably doubled. If the 80th percentile value is to be the
correspond to metallurgical differences, then the geological basis of design, then the properties of each ore type need to be
ore types are also valid geometallurgical ore types. In measured to the degree where a reliable 80th percentile value
addition, each geometallurgical ore type will display a range can be extracted from the results for each ore type.
of properties, and the proportions of each ore type in plant Before proceeding to design the plant to treat each ore type
feed will vary from shift to shift and year to year. separately, a constructive discussion is needed across all three
A classic example would be where the orebody has an disciplines to explore the implications of alternative mining
oxidised cap, a transition zone and fresh rock at depth. strategies. For example, introducing blending before the
The approach often seen by the author is for the fresh rock primary crusher or reducing the shovel size and having two
to be represented by 20 or more samples, the transition by (or more) ore faces supplying plant feed at all times. Beware
five samples and the oxide by two, neatly matching their of the argument that often comes up in such discussions that
proportion in the orebody. blending is happening in the coarse ore stockpile after primary

D David
crushing. Unless the coarse ore stockpile is a bed blending not suitable for design purposes. In any data set typically
arrangement where ore is stacked in layers and reclaimed available as a foundation for design, the average value will
across the layers, then it can be safely assumed that no blending be a much more reliable number than the 80th percentile or
of any consequence occurs in the coarse ore stockpile. the SD. In the example given for the effect of compositing,
an estimate was made of a believable 80th percentile value
based on an assumption about the variability differences
VARIABILITY EXPECTATIONS known to exist between data sets calculated on varying time
Steve Morrell (2011) published the distribution of variability scales of production. Notably the adjustment was made
levels that exists within the sag mill competence (SMC) test relative to this well-defined value, the average of the data
database. Typically, SMC test samples are from contiguous set. The method employing scaling of variability (or 80th
core intervals and are done in numbers large enough (per percentile) between data sets is one alternative approach to
orebody) to derive reasonable statistics. The variabilities of selecting the design point.
the drop weight index (DWI, in kWh/m3) results from the
The concept of the 80th percentile is one that provides some
650 orebodies represented in the database (at the time of his
comfort in design, but it is also one that is arbitrary. Why
writing) were distributed as shown in Figure 2. not use the 90th percentile, the 75th percentile or some other
For an orebody with an average DWI value of 5 kWh/m3 and value? In a recent design project some criticism was provided
a coefficient of variation (COV) value of 25 per cent, the 80th in a review that the 80th percentile had not been used and
percentile value must lie between 25 per cent and 50 per cent reliance was placed on what was considered a riskier
greater than the mean value (between one and two standard value, the 75th percentile. Regardless of the fact that the
deviations greater than the mean). An 80th percentile value 75th percentile was the agreed design point with the client,
of 7 kWh/m3 would not be unreasonable for this example. the results database was revisited and the design outcome
However, the COV could be any value in the range from recalculated using the 80th percentile value. In this instance
5 per cent to 60 per cent, so 80th percentile values of 10 kWh/ the difference was less than 0.5 per cent in the mill power,
m3 and 5.4 kWh/m3 are also possibilities for a 5 kWh/m3 ore. insignificant in the accuracy of the design. For this particular
All critical measures used for design will have a similar deposit the SD of the value in question was very low but the
(but usually less broadly spread) range of possible variability variability in the data set was considered valid for design.
values. Without having enough actual measurements from The data set consisted of about 50 results, all from individual
contiguous samples according to an Australian Mining
tests performed in the correct manner on appropriate samples,
Exploration Companies (AMEC) sample selection plan. The
it is not possible to achieve a reliable estimate of variability
insensitivity, in this instance, of the design outcome to the
for a particular property in an orebody. This statement also
selected percentile value was of concern, mainly because the
excludes the method applied in the compositing exercise
design point was not all that much greater than the data set
discussed previously. Although the final 80th percentile
average. This particular instance has led to a re-focusing of
estimate was superior to the estimate from the original data,
the basis of design towards using the mean value, rather than
it was also far from definitive.
any high percentile derivative (80th or 75th) of the data set.
Once the focus is on the mean value then the concept of
AN ALTERNATIVE APPROACH ‘confidence’ can be introduced to design process. It is possible
As has been demonstrated in the examples above, the 80th mathematically to derive the upper and lower confidence
percentile value can be totally misleading and dangerous limits for a given mean value of a data population by
as a design point if the data set from which it is derived is knowing the number of samples in the population, the SD of
FIG 2 – Distribution of orebody COV* values for competence measurement (after Morrell, 2011). *The coefficient of
variation (COV) is the standard deviation divided by the average value, expressed as a percentage.

the population and the degree of confidence that is required the average by 0.5 kWh/t to 9.6 kWh/t. The 30 extra tests
in the mean value. The higher is the required confidence level, have probably saved $1.5 M in capital costs by allowing a
the wider the range of possible mean values for any data set. smaller mill to be installed without compromising on design
A range of possible mean values exists because, if exactly the confidence. It is also clear that using the measured average of
same sample selection process was conducted on the same a data set always incurs higher risk than using the calculated
orebody, but different core intervals were chosen, the mean upper confidence level for the mean value.
result from testing this second sample set is almost certain to
Another good thing to check is if the upper confidence limit for
be different to the mean result from testing the first sample set.
the mean value calculates to be greater than the 80th percentile
Therefore, it is possible to know what the highest likely mean
of the data set. Consequently, by using the 80th percentile
value is (the upper confidence limit of the mean) and Microsoft
value the designer is inadvertently designing with a number
Excel provides a function to simplify this estimation. As the
that may actually turn out to be less than the upper possible
SD is an integral part of the calculation it is still necessary to
orebody average. This simple check shows definitively that too
re-estimate the SD to a higher value if compositing or annual
planning has smoothed the available data set. few samples have been tested for the inherent variability that
exists in the orebody.
Confidence can be thought of as the inverse of risk. You
have selected good test samples, generated a result data Having demonstrated a method to find a safe average value
set and you have averaged it. How do you know if your (the upper confidence limit) it is now possible to construct a
measured average value is a good estimate of the real average new design methodology based on estimation of data set SD
for the orebody? As an example, a set of ten results give a coupled with the statistically-derived upper confidence limit
mean BWI of 9.1 kWh/t and the individual test results range of the mean value. It will be argued (in a follow-up paper)
from 7 to 12 kWh/t. As a designer, you are happy to work that such an approach is more robust than the tried, and often
with a confidence level in the mean value of 90 per cent, as wanting, 80th percentile based method.
this means that the risk of your measured average being
materially wrong is 10 per cent. A one-in-ten risk of being CONCLUSION
wrong is unacceptable to most designers but the catastrophic
The 80th percentile value can be a useful design number,
design risk level is more acceptable at one in 20.
provided all the correct prerequisites are present in the
The 10 per cent risk associated with applying 90 per cent population that it is derived from. These are:
confidence limits is actually made up of two components.
1. The samples that the individual test results represent
The first 5 per cent risk component is that the measured
must consist of individual lengths of contiguous core,
average is outside of acceptable limits on the high side. As
or something close to this ideal.
a result, the real work index of the orebody is found to be
even lower than the lower confidence limit. The design risk 2. There must be some understanding of how the
is that a mill that is unnecessarily large and powerful will be variability of the available data set is related to the
selected by using the 9.1 kWh/t average value. The second variability that needs to be catered for in plant design.
5 per cent risk component is that the real average work index If a pathway to estimating shift-by-shift production
of the orebody is found to be much higher than the upper variability (for any variable) is available, then this can
confidence limit. In this case the risk is that the selected mill provide useful guidance and should be followed.
will be too small and consequently nameplate throughput at 3. There must be some relationship between the available
nameplate grind size will not be achieved. Mill under-sizing data and the time sequence of presentation of ore to
can be a major risk to project viability because the revenue the plant. For example, if the ore that the plant is to
stream is likely to be compromised. be commissioned on is not clearly understood, then
For the set of ten test results in the example above, the an adequate number of samples of commissioning ore
90 per cent confidence limits were calculated to be 10.1 and need to be tested before commencing definitive design.
8.1 kWh/t respectively. In accordance with the discussion
4. The relationship between the mining practices to be
above, there is a 5 per cent chance the real mean BWI for the
employed and the distributions of the various ores in
orebody is greater than 10.1 kWh/t and the 9.1 kWh/t average
the deposit need to be understood at an operational
value is too low and will compromise the design. There is
level, and not a smoothed monthly or annual ideal.
also a 5 per cent chance the real mean is less than 8.1 kWh/t
and a design based on 9.1 kWh/t will be conservative. As It is essential that the process design engineers have access
the catastrophic risk lies in under-sizing the mill the sensible to all raw test data, and that they understand the basis of
designer will take 10.1 kWh/t as the average work index, sample selection, they understand any compositing that
rather than 9.1 kWh/t. The risk of overdesign is increased has been performed on the samples and why it was done
significantly but the risk of under-design is acceptable. that way, and they have confidence in the testing that has
Now consider if 40 samples were tested and the individual been conducted. It is also essential to link the test results
results all lay within the same range, 7 to 12 kWh/t, and the to the ore types and to the time sequence in which the ore
average was again 9.1 kWh/t. The 90 per cent confidence limits is likely to be mined. Finally, the process engineers must
around this average value are much tighter and calculate to understand the operational reality of ore access and delivery
8.6 and 9.6 kWh/t. The increased number of samples tested to the plant, including any assumptions regarding blending.
has increased our confidence in the measured mean and A good start to gaining these necessary understandings
reduced the associated design risks. The sensible designer can is to ask the geologists for the drill database and to ask the
now use 9.6 kWh/t as a safe average value with exactly the mining engineers for a sequenced block list. The resulting
same risk level that existed when choosing 10.1 kWh/t based conversations are usually very enlightening.
on ten samples. If a valid 80th percentile design value is not obtainable from
In selecting a safe design point the ten sample set forces the available raw data, then a relatively simple pathway for
the designer to assume that the average could be as high as estimating a useful 80th percentile value (based on the relative
10.1 kWh/t. The 40 sample set allows the designer to lower variabilities within available data sets) has been demonstrated.

D David
To increase the confidence that can be placed in a design gooey ore and the crunchy blue rock, they are both the same
outcome, the basis for a new and robust design methodology geological classification. Acknowledgement is also given to
has been proposed that is linked to a worst-case estimate Steve Morrell, who has followed (and sometimes built) the
of the mean value (at a desired accuracy level) and with a pathway to understanding variability.
defined confidence level based on the client’s appetite for risk.
REFERENCE
ACKNOWLEDGEMENTS Morrell, S, 2011. Mapping orebody hardness variability for AG/
Anonymous acknowledgement is given to those projects SAG/crushing and HPGR circuits, in Proceedings International
supplying examples, good and bad, used in this paper. Special Autogenous Grinding, Semiautogenous Grinding and High Pressure
Grinding Roll Technology 2011 (eds: K Major, B C Flintoff, B
acknowledgement is given to the geologist who advised (in an
Klein and K McLeod) paper 154 (Canadian Institute of Mining,
operating pit) that there was no difference between the brown Metallurgy and Petroleum: Montreal).

Contents
Measuring and taking notice of orebody variability

– an essential ingredient for reliable plant design
D David1
ABSTRACT
The majority of semi-autogenous grinding (SAG) and ball mill design methodologies
are applied using a single design value for mill selection, such as the 80th percentile
ball mill work index or SAG mill comminution (SMC) value. Inherent in the selection
of this design value is a desire for the mill to accommodate unfavourable, but not
absolute worst-case, feed conditions. It has been observed in numerous ‘post-
mortems’ of unsuccessful designs and also in many optimisation studies, that the
selection of the design value has been poor. Statistical analysis of test work data
sets shows that inherent variability in the data can, by itself, lead to design errors.
Metallurgical testing is an attempt to measure the characteristics of millions of
tonnes of ore using only a few kilograms of sample. In this paper, some realistic but
synthetic test work data sets are used to explore old and new methods of defining
the design envelope and subsequently selecting a reliable point. The aim of the
paper is to provide practitioners with tools that reveal the adequacy (or otherwise)
of metallurgical test programs intended to support various levels of project design. It
also provides guidance as to how it is possible to statistically justify a comparatively
high design point in the absence of adequate test data.
INTRODUCTION
The words ‘it’s all the same’ are regularly heard in relation to orebody properties and
this phrase should always be an alarm bell to a mineral processing plant designer.
Nature does not provide orebodies that have consistent properties and this means
there will always be a range of values for any measure, be it for hardness or separation
characteristics.
Some of that variation in measurement comes from the repeatability of the
measurement technique itself and this should not be confused with ore property
variation. Test repeatability must be known and understood before using any test
results for design purposes.
The variation that will be discussed in this paper is the spread of test results
that naturally occurs when a set of samples, each sourced from a different three-
dimensional location in an orebody, is tested.
The degree of variability that is measured for the material will determine both the
design envelope and the safe design point for that measure.
It is easy to incorrectly measure variability through poor sample selection and
through compositing of samples inappropriately.
Correctly measuring variability leads to appropriate selection of design property
values and this, in turn, leads to a process plant design that can achieve its nameplate
performance requirements.
QUANTIFYING VARIABILITY
Ore variability measurement has two components. The magnitude of variability that
is measured and the belief you can place in that answer.
The magnitude of variability in a set of numbers is typically described as its standard
deviation (SD). The SD is almost always stated together with the average or mean value.
A typical example would be 10±1.0 where the average or mean is 10 and the SD is 1.
The belief you can place in that average value is called the Confidence Level. This is
mathematically determined using an equation that includes the number of test results
(n) and the SD value. As you conduct more tests you get more confident that you have
actually measured the mean value and also that the SD magnitude is correct.
1. FAusIMM (CP), Technical Director – Process,
Amec Foster Wheeler, Perth WA 6000. A typical statement of confidence is that there is 90 per cent confidence in the mean
Email: dean.david@amecfw.com to an accuracy of ±5 per cent. The statement ‘90 per cent confidence’ can also be
121
D David
written as ‘one time out of ten’. So, if the mean was measured TABLE 1
ten times (If the original calculations were based on a set of 20 Synthetic grindability data set.
correctly selected test samples then it is necessary to repeat
this by measuring another nine sets of 20 correctly selected Sample Result (kWh/t)
samples, 200 samples in all) we would expect nine of those
1 10.5
mean values to be within 5 per cent of this first mean value.
One of those readings would be expected to be greater than 2 9.5
5 per cent different from this first mean value. 3 8.8
Now if a 5 per cent error is enough to cause serious problems 4 11.0
in the design then you have a one in ten chance of getting
the design wrong. If a 5 per cent high result is acceptable but 5 12.2
a 5 per cent low result is not, then things improve because 6 10.2
confidence works both ways. The one in ten problem result 7 10.5
could be a high result or it could be a low one. There is a
one in 20 chance that the result will be high (greater than the 8 11.6
original mean +5 per cent) and a one in 20 chance it will be 9 8.4
low (less than the original mean – 5 per cent). Because only a
10 8.8
low result is a problem we now have a situation where there
is a one in 20 chance of the design failing. 11 7.9
Is this acceptable or do we need to be more accurate? Do we 12 10.6
need to reduce the risk further? Do we simply overdesign and 13 10.2
mitigate the risk?
14 9.8
The answers to all these questions depend upon the risk
profile of both the project and the client. For new technology, a 15 9.5
one in 20 chance of failure may be an acceptable risk level. For 16 9.9
the application of well-established technology, it is usually
17 10.0
expected that the design should have effectively no risk of
failure, say a confidence level of 99 per cent or one in 100. 18 10.0
For a major mining house, where the project is one of 19 11.1
many and access to expertise is relatively easy then perhaps 20 9.7
a confidence level of one in 100 is also required. Often such
a project will be competing with other projects in which the Mean or average 10.0 kWh/t
probability of success is similarly high. Standard deviation 1.05 kWh/t
For a junior miner with a single potential project and a high 90% confidence in average 0.39 kWh/t
degree of capital sensitivity and time pressure, perhaps a
one in 20 chance of failure is acceptable. It is up to those that
would supply the money to determine if they should ‘take
the chance’.
The unfortunate part about statistics is that uncertainty is not
removed by analysis, it is simply quantified. For example, a
one in 20 chance of failure means that the set of data you have
generated and analysed could represent that one in 20 problem,
but there is no way to identify this without more testing.
Variability example
To demonstrate the quantification and implications of
variability a constructed example has been prepared. A set
of 20 ‘results’ of Bond testing is listed in Table 1, but in fact,
these numbers have simply been fabricated by the author to
illustrate the points. FIG 1 – S-curve of Bond work index result set with
The data set is plotted using an S-plot in Figure 1 to show normal distribution for comparison.
its variability in a visual fashion. For comparison, a normal
distribution line having the same mean and SD is also plotted. samples that we know the mean of the orebody hardness to
within 4 per cent either side of the measured mean value of
The comparison shows that the data set can be considered
10 kWh/t. Making the simplistic assumption that SD can be
as closely following a normal distribution pattern. In other
applied across the full range of the data set, a first-pass design
words, the data is regularly distributed above and below the
mean according to known mathematical principles and that envelope can be prepared as is shown graphically in Figure 2.
the SD is statistically valid. Because only 20 samples have been tested out of countless
The 90 per cent confidence level in the average or mean millions of possibilities in an orebody, we do not know exactly
value is 0.39 kWh/t. This means that if another 20 samples what the average Bond work index (BWI) is, we only know
were selected and tested there is a 9 in 10 chance the mean will that it probably lies within the two red lines on the 50 per cent
be between 9.61 kWh/t and 10.39 kWh/t. There is a one in 20 rank horizontal line in the graph. To draw Figure 2, it has
chance the new mean will be greater than 10.39 kWh/t and been assumed that the confidence band remains unchanged
a one in 20 chance it will be less than 9.61 kWh/t. Literally, at 0.78 kWh/t (2 × 0.39 kWh/t) across the entire range of the
we have 90 per cent confidence after analysing these 20 measurements in the data set. This is convenient and looks

Measuring and taking notice of orebody variability – an essential ingredient for reliable plant design
FIG 2 – 90 per cent confidence envelope around Bond work FIG 4 – Design envelope after considering 90 per cent confidence in
index data set assuming consistent standard deviation. mean value and 95 per cent confidence limit in the standard deviation.
sensible on the graph, but it is not correct as the confidence Rank Confidence Limit (RankCL) at various ranking levels.
calculation has only provided information about the mean The RankCL value is half the span from the upper line to the
value, not the remainder of the distribution. There will also be lower line. At the mean the 90 per cent confidence interval
uncertainty in the value of the SD. As will be demonstrated, the and the RankCL values are almost identical (0.39 kWh/t
uncertainty in SD is more of a problem for traditional design 90 per cent CI versus 0.41 kWh/t RankCL).
methodologies than is the uncertainty in the mean value. In this data set the uncertainty in knowing the 80th
Upper and lower limits to the value of the SD can be calculated percentile value is 0.71 kWh/t which is 75 per cent higher than
from the data set using the method described by Sheskin (2007). the uncertainty at the mean. Uncertainty in knowing the 90th
From this BWI data set the calculated SD is 1.05 kWh/t but the percentile is more than double the uncertainty in the mean.
upper limit for SD (to 95 per cent confidence) is 1.55 kWh/t. From Figure 3 and Figure 4, the lower limit (with 95 per cent
The lower 95 per cent confidence limit to the SD is 0.80 kWh/t. confidence) for the 80th percentile design point is only just
This method provides us with a measure of uncertainty about above the measured mean at 10.3 kWh/t. The upper limit
the level of uncertainty we think we have measured. When is 11.7 kWh/t. The difference between these two values
the design envelope is now modified (selectively) using this represents a 14 per cent difference in power requirement
information Figure 3 is the result. ultimately selected for plant design.
The term ‘selectively’ is used because the envelope could The statistical analysis is indicating that these 20
have been drawn in a number of ways, but the calculation that measurements give a particular picture of the distribution
maximises the possible data spread in the upper half of the of hardness in the orebody. However, testing another 20
graph has been used. This analysis clearly demonstrates that samples from this same orebody and simply choosing the P80
the range of possible values for a typical design number, like value for design could have just as easily resulted in design
the 80th percentile value, are significantly less certain than the points anywhere from 10.3 kWh/t to 11.7 kWh/t without
range of possibilities for the mean value. An alternative drawing being statistically inconsistent with the data set that has been
of the envelope to emphasise the effect at the lower end of the analysed here.
data range would demonstrate a similar level of uncertainty
In a traditional analysis, the actual 80th percentile ranked
in the 20th percentile ranking value as has been demonstrated
data point is towards the lower limit of this statistical band
for the 80th percentile ranking. When the maximum level
at 10.6 kWh/t. The mathematically calculated 80th percentile
of uncertainty is estimated across the full distribution the
value (using the Percentile function in Microsoft Excel for
envelope can now be realistically drawn as per Figure 4.
example) is consistent with the normal distribution curve at
Figure 4 represents a much more realistic understanding of 10.9 kWh/t. The major point of this paper is to argue that
the uncertainty associated with choosing any percentile value these traditional estimates are inadequate for design using
in the distribution than is represented by Figure 2. This issue this data set because this orebody, sampled and tested in the
is further quantified by plotting what has been termed the traditional way could just as easily have provided a design
point as high as 11.7 kWh/t.
Alternative design point selection methods

One obvious alternative is to choose a percentile rank higher
than the 80th percentile that gets closer to the statistically
valid upper limit of 11.7 kWh/t. The 90th percentile point is
only 11.1 kWh/t and the 95th percentile point is 11.6 kWh/t.
Therefore, in this data set the second highest measured value
would become the design value. However, from Figure 5
it has also been shown that this 95th percentile point could
easily be 1 kWh/t higher.
If the normal distribution curve was used in mathematically
calculating a percentile (rather than simply choosing the
second highest measurement) it would be very safe to adopt
FIG 3 – Effect on the design envelope of applying the upper the 95th percentile point (11.74 kWh/t). If the Excel percentile
standard deviation (SD) limit to the upper mean value function was used, then the 90th percentile would also appear
and the lower SD limit to the lower mean value. to be suitable as it gives a value of 11.55 kWh/t.

D David
The ultra-conservative approach equation

Design point = UCL90 + UCL95SD [2]
where:
UCL95SD is the upper limit, at 95 per cent confidence,
of the SD of the data set
For the example data set this approach gives a design value
of 11.93 kWh/t.
This approach is recommended where the data does not
follow a normal distribution and is especially applicable when
insufficient tests (say less than 10) have been conducted to
establish the true mean and variability in the design property.
The outcomes of the various design point selection methods
FIG 5 – Calculated confidence in measured Bond work index are compared in Table 2.
value either side of the mean (50th percentile) value. This comparison suggests that the traditional design
method is likely to underestimate the power requirements of
The first option, choosing a measured point, is the least the circuit by at least 7 per cent. A plant failing by 7 per cent
favoured method. In this data set, choosing a particular to achieve its nameplate throughput rate may be enough to
measured point is relatively simple and presents few problems reduce the profitability of the project to zero, or even to a loss-
apart from the relatively low confidence level that has been making scenario.
identified when choosing points far from the mean. But this
data set is regular in its shape while many other real sets are Detailed equations
not. One or two outlying points at the top of the distribution The equations behind the two methods will be explained in
would invalidate any method where an upper measured a form suitable for use in Microsoft Excel. It is assumed the
point is simply chosen. 20 measured data values are in a single range of 20 cells that
The second method, calculating percentiles based on a has been assigned the name BWI. This means that the SD term
fitted standard distribution, also assumes regularity in the in Equation 1 is calculated as STDEV.S(BWI). The capitalised
data set. If the set follows the normal distribution curve function name STDEV.S is a new function in Office 365 version
closely, then using the basic normal distribution techniques of Microsoft Excel and has equivalents in older Excel versions.
are appropriate, although a value ranked higher than the UCL90 can be calculated as:
traditional 80th percentile is needed to be prudent in design. UCL90 = AVERAGE(BWI) + Conf90[BWImean] [3]
The third method (using the percentile function) seems where:
to lie somewhere between the first two and is subject to the
constraints of the normal distribution assumption. Conf90[BWImean] = CONFIDENCE.NORM(0.1,BWISD,20) [4]
The recommended method for choosing a safe and justifiable for 20 samples, and where:
design point is to apply some of the statistical calculations
BWISD = STDEV.S(BWI) [5]
used to analyse the problem above. The difference between
this approach and the traditional methods is that it recognises This allows Equation 1 to be rewritten as:
both the uncertainty in the mean value and the uncertainty in
Design Point = AVERAGE(BWI) + CONFIDENCE.
the degree of variability of the data.
NORM(0.1, STDEV.S(BWI),20) + STDEV.S(BWI) [6]
Recommended design point selection procedures The value 0.1 is derived from the choice of confidence level
Elsewhere, the author proposed alternative methods for of 90 per cent. The value 20 is the number of test samples.
selection of the design point based on the mean value Equation 6 is the recommended equation for selecting the
and variability in the data set (David, 2017). Equation 1 is design point.
the recommended method and is considered a prudent The calculation of UCL95SD for Equation 2 is more
conservative approach. Equation 2 represents an ultra- complicated, as described here.
conservative approach and is advisable when risk tolerance
is very low or the data set has high uncertainty. UCL95SD = BWISD*((n-1)/CHIINV((0.05/2), n-1))0.5 [7]
The conservative approach equation For 20 samples this equation simplifies to:
Design point = UCL90 mean + SD [1] UCL95SD = BWISD*((19)/CHIINV((0.025), 19))0.5 [8]

where:
TABLE 2
UCL90 is the upper limit, at 90 per cent confidence, of the Comparison of design point selections.
likely values for the mean value of the data set.
SD is the standard deviation of the data set Method Design value Relative magnitude
For the example data set this approach gives a design point (kWh/t) (%)
of 11.44 kWh/t. 80th ranked point 10.3 90
This approach is recommended where the data follows 80th percentile() calculation 10.6 93
the normal distribution to a reasonable degree and also Conservative recommendation 11.44 100
where sufficient tests have been carried out to establish the
variability in the SD. Ultra-conservative approach 11.93 104

FIG 6 – Hard and soft ore random data sets.
And as a result, Equation 2 can be written as such: based on the mean and SD of each set individually, are
shown for comparison.
Design Point = AVERAGE(BWI) + CONFIDENCE.
NORM(0.1, STDEV.S(BWI),20) + STDEV.S(BWI)*((19)/ To illustrate the resulting bimodality, all 40 values are
CHIINV((0.025), 19))0.5 [9] treated as a single set and shown as an S-curve in Figure 7.
An overall calculated normal distribution line has been
where: added that obviously does not represent the set well. A closer
0.025 is derived from the choice of SD confidence level (in fitting trendline has also been added using the polynomial
this example at 95 per cent) and 19 is the number of function in Excel.
samples minus one The distribution cannot be considered to follow the normal
pattern and presents a few problems for design thinking.
Problems with the data set To see the effect of the increase variability across the
For convenience and clarity, a fabricated data set that closely 40 samples, the upper and lower 90 per cent confidence values
complies with normal distribution parameters was used in the for the mean and the upper and lower 95 per cent confidence
preceding discussion. Many data sets generated in mineral values for the SD were used to construct the potential upper
processing test work do not comply with this convenient design envelope as per Figure 3. The resulting bimodal
structure and can lead to problems with design, ranging from envelope can be seen in Figure 8.
minor to catastrophic.
The 80th percentile could statistically be anywhere in the
range 13.9 to 16.3 kWh/t.
Multimodal data
When two or more separate populations of properties are The four calculation methods used to generate Table 2
embedded in the one set of data many simplistic analysis were applied to this new 40-point data set. The results are
assumptions are no longer valid. To illustrate multimodal compared in Table 3.
data the statistics (mean, SD and assumption of normal Again, the ‘conservative recommendation’ calculation
distribution) of the example data set were used to generate method is preferred to the traditional 80th percentile
20 new random results representing what is a soft ore. methods as the result is closer to the upper end of the
A second set of 20 hard-ore results were generated, also statistically valid range. In this instance, the degree of
somewhat randomly, by adding an amount varying between potential underestimation of design point using the
4 and 6 kWh/t to the first set. Normal distribution curves, conventional methodologies is about 5 per cent.
FIG 7 – Bimodal combined data set.

D David
FIG 8 – Multimodal upper design envelope.
Small data sets measurement or mathematically calculating the 80th

To illustrate the effect of using smaller and smaller data percentile, always underestimate compared to the maximum
sets, subsets of the 40 bimodal data points used in the above envelope value.
example were used as the basis for design calculations. A The conservative method either overestimates this value
random selection method was again used to avoid bias. or estimates a design point consistent with the value (it gave
The base case for comparison in this analysis is the outcome 16.0 kWh/t but only after 40 samples had been tested).
achieved using 40 data points. Example calculations have been The ultra-conservative method provided an extreme result
conducted selecting approximately 20 samples, 10 samples (35.2 kWh/t) when only three samples were tested. This result
and five samples out of the 40. Note that five samples are is a warning of the statistical inadequacy of testing only a few
generally considered a reasonable number of samples on samples and then trying to infer too much from the results.
which to base a scoping study. The ultra-conservative method provided reasonable design
Note that the subset numbers are only approximate. This is points when nine or more samples were tested.
because the random selection process as it has been applied It may seem unreasonable to adopt 21.6 kWh/t as a design
cannot (by definition), result in a fixed subset size. value after testing nine points and when the 80th percentile
Each time the calculation is updated the selections change value of the set is much lower at 15.1 kWh/t (Table 4).
as does the apparent shape of the S-curves. This is illustrated However, the statistical analysis is saying that after nine
by a selection of five graph updates shown in Figure 9. samples have been tested the variability at the top end of the
On all five graphs the original blue data set is constant. All distribution is extremely poorly defined. It is feasible (without
other data sets are random plausible outcomes when less than knowing anything about the next samples to be tested) that
the full 40 samples are tested. there is a significant amount of 20 kWh/t and harder material
As statistically expected, if three to seven samples are in the orebody that could be detected when more tests are
selected for testing (the ‘Approximately five samples’ cases carried out.
ranged from three to seven samples) the range of possibilities In line with the naming of the methods, the conservative
for the design point outcome is highly variable. Even choosing and ultra-conservative approaches give higher design points
the maximum test result may not get close to being a safe the fewer tests that have been conducted. This provides an
design value as Figure 9a illustrates. improved, some would say excessive, level of safety (compared
Design calculations were also conducted according to to traditional methods) if minimal testing is conducted. Most
the four methods previously described. The outcomes for importantly these high design points provide real incentive
the full and reduced sample number cases represented in for conducting more tests as the capital cost is most likely to
Figure 9e are shown in Table 4. reduce as confidence improves.
From Figure 8, the upper boundary of the design envelope A number of additional random calculations were captured
is known to be 16.3 kWh/t. However, this value can only showing the potential for problems with small sample sets
be known after 40 samples have been tested. Each subset In figure 10 and the associated Table 5 it can be seen that
result needs to be assessed as if the 40 samples have not been only conducting six tests leads to a potentially problematic
tested. The first two methods, selecting the 80th percentile hardness underestimation even using the conservative
TABLE 3 TABLE 4
Comparison of design point selections – bimodal data. Comparison of design point selections for reduced
sample numbers – graphical case (E).
Method Design value Relative magnitude
(kWh/t) (%) Samples tested 40 21 9 3
80th ranked point 15.3 95.1 80th point 15.3 15.8 14.8 15.8
80th percentile() calculation 15.3 95.2 Percentile (80) 15.3 16.0 15.1 14.9
Conservative recommendation 16.0 100.0 Conservative 16.0 17.0 17.8 19.0
Ultra-conservative approach 16.6 103.6 Ultra-conservative 18.8 19.4 21.6 35.2

A B
C D
FIG 9 – (A–E) Random subset expectations if bimodal data set was only partially sampled.
TABLE 5 method. The ultra-conservative method is the only one to give

Comparison of design point selections for reduced a safe answer. The traditional methods provide dangerously
sample numbers – graphical case (F). low design values. The results are a strong argument for
doing at least nine tests.
Samples tested 40 20 9 6 Another random cycle of the same calculation is shown in
80th point 15.3 14.8 13.6 11.1 Figure 11 and summarised in Table 6.
Percentile (80) 15.3 15.1 14.3 11.1 Again, the conservative method is preferred to the
traditional selection methods.
Conservative 16.0 16.8 16.4 14.8
Based on conducting this calculation a number of times it is
Ultra-conservative 18.8 20.0 20.1 19.1 recommended that if less than ten samples have been tested

D David
FIG 10 – Random subset expectations if bimodal data set was only partially sampled, case (F).
FIG 11 – Random subset expectations if bimodal data set was only partially sampled, case (G).
TABLE 6 TABLE 7
Comparison of design point selections for reduced Guide for use of conservative design methods.
sample numbers – graphical case (G).
Stage Scope PFS DFS
Samples tested 40 20 13 4 4 to 8 samples C UC N/A
80th point 15.3 14.8 16.0 13.6 9 to 16 samples C C UC
Percentile (80) 15.3 15.3 15.7 14.5 17+ samples C C C
Conservative 16.0 16.1 17.0 17.7 Notes: PFS = prefeasibility study; DFS = definitive feasibility study; UC = ultra-conservative;
Ultra-conservative 18.8 19.0 20.5 26.3 C = conservative; N/A = not recommended.
increase the sample numbers in the first column of Table 7 in

the ultra-conservative method is required. However, if the line with the increase in uncertainty in the data.
ultra-conservative method is arriving at design points twice
the magnitude suggested by traditional methods this should It is proposed that the 90 per cent confidence value, as a
be taken as a warning that inadequate sample numbers have percentage of the mean value, be used as a guide to sample
been selected for design purposes. numbers.
However, at a scoping level it is often not possible to test In the bimodal data example described above the
more samples, and the implications of slight underestimation confidence level in the mean can be expressed as a fixed
of power are not significant. This suggests a better approach value for the unchanging set of 40 samples but as ranges for
is to vary the design method according to the project stage the subset cases (as these change randomly) when the sheet
and available sample numbers. A guide for designers is is recalculated. The confidence levels (as a percentage of the
proposed in Table 7. hardness measurement) are compared in Table 8.
This analysis appears to suggest that 16 samples is sufficient Designers can use Table 8 as a reference to see if their data set
for a definitive design. Using the conservative method and is similar to this example or if it is much worse than the example.
with an orebody that appears to give test results of similar Simply calculate the 90 per cent confidence limit in the mean
variability to these examples the guidelines are useful. for the data set and compare it with the table. If, for example,
Where the orebody is less variable the methods will simply 15 samples have been tested for project XYZ and the 90 per cent
work better. Where the variability is much higher than these confidence limit is 15 per cent of the mean value, then the XYZ
randomly generated sets, then it is likely to be necessary to data set (and more importantly, the XYZ orebody) is much

TABLE 8 is shown to approach twice the uncertainty of knowing the

The 90 per cent confidence limit in the mean measured value as a percentage mean. Any measured data sets’ inherent variability can be
of the mean hardness measure – relationship to number of samples tested. evaluated using these same readily available statistical tools.
The mean value of any data set is its best-defined characteristic.
Samples Average confidence as Range The level of confidence inherent in values distant from the
percentage of mean mean (such as the 80th or 90th percentile) is much lower than
4 21.5 18–26 the confidence that can be placed in the mean.
Design difficulties arising from variability issues are
5 15.0 11–21
exacerbated when the ore is bimodal and multimodal.
6–8 15.2 12–20 Design difficulties are also exacerbated when small
9–12 11.4 9–13 numbers of samples have been tested. The number of samples
13–16 9.0 7–10 needed to be tested to provide design confidence is directly
proportional to the inherent ore property variability. In a sort
17–25 8.0 7–10 of catch-22, the number of samples that need to be tested to
40 5.6 develop a reliable design for a particular orebody can only be
determined once enough samples (typically seven or more)
more variable than the bimodal example. Fifteen samples of have been selected and tested.
XYZ ore have only given the statistical confidence level that six Two methods of selecting design points have been proposed
to eight samples of the fictitious bimodal ore provided. and compared with traditional and simple methods. The
A new version of Table 7 would result for XYZ ore in which conservative method employs the confidence limit calculated
the column one sample numbers would double. For orebody for the mean value coupled with the SD of the data set.
XYZ the 15 samples do not provide enough confidence to even An ultra-conservative method has also been proposed
use the ultra-conservative method for a definitive feasibility which considers both the confidence limit in the mean and
study (DFS) design. More XYZ samples must be tested before the confidence level the data set provides in the measured SD.
a DFS can be contemplated. Traditional design methods are prone to design point
The analysis shows that the confidence in the mean is underestimation at all sample sizes and are particularly poor
strongly related to sample numbers. However, the calculated with sample sets of six or less.
SD values were relatively independent of the number of The conservative method is reasonably good for estimating
samples tested. This is useful because it means that the SD is reliable design points with sample sets of eight or more in
a guide to overall variability, regardless of how much testing the examples analysed. With samples sets of four to eight the
has been performed. It follows that Sample XYZ would display conservative method is prone to some underestimation, but
a higher SD than the bimodal sample under all circumstances. less severe than using traditional design methods.
The ultra-conservative method is prone to conservative
CONCLUSIONS overdesign with sample sets of eight to fifteen. The method is
The analysis in this paper shows that the inherent variability unreliable for sample sets less than eight and may randomly
in a measured data set is not considered fully in a traditional return a conservative or optimistic result.
design point selection method, such as calculating an 80th The traditional and proposed methods can be usefully
percentile value. compared when selecting design points with real data sets.
It has also been shown that as the number of samples They can also be used to analyse under or overdesign situations
tested decreases the traditional method is relatively poor at in operating plants as this may provide a guide to avoiding
providing a design point that can be considered safe. mistakes in future projects. Simply performing the statistical
confidence calculations and plotting the design envelope is a
Two statistical measures not typically invoked in
useful guide to the adequacy of the number of samples tested
traditional design methods are the inherent uncertainty in
and can be used to justify conducting additional tests.
the calculated mean value and the uncertainty in the SD for
the measured sample set. Traditionally the mean and SD
values are taken as measured and the 80th percentile value REFERENCES
is considered a reliable indicator of where the plant needs to David, D, 2017. Is an 80th percentile design point logical?, in We are
be designed to operate. Metallurgists, Not Magicians (eds: D Pollard, G Dunlop and J Herzig),
In the synthetic examples analysed here the uncertainty in
Melbourne).
knowing the mean of a measured data set is quantified and
the uncertainty in knowing a value like the 80th percentile Sheskin, D J, 2007. Handbook of Parametric and Nonparametric Statistical
Procedures, fourth edition, pp 197–198 (Chapman and Hall/CRC).

Contents
Getting optimum value from ore characterisation

programs in design and geometallurgical
projects associated with comminution circuits
S Morrell1
ABSTRACT
It is extremely difficult if not impossible to ensure optimum process efficiency without
having a thorough knowledge of the orebody. For the comminution circuit this means
having detailed data on the breakage characteristics of the ore, in particular how it
varies spatially. When planning campaigns to determine this information, questions
such as ‘What breakage tests should be used?’, ‘How many samples are required?’
and ‘What equations should be used to forecast comminution circuit performance?’
are often asked. The following paper presents a series of statistical analyses to help
answer these questions. Perhaps not surprisingly there are no simple answers, as
it depends upon how the characterisation data are planning to be used, as well as
the inherent variability of the orebody. Guidance is provided to help metallurgists
make informed decisions, which should result in a characterisation programme that
is cost-effective and statistically sound, providing a clear picture of the comminution
behaviour of the orebody in question.
INTRODUCTION
Inevitably the post-boom era has changed the emphasis of many mining companies
away from new project development and expansion to optimisation of existing
operations. However, regardless of the project, high quality laboratory ore
characterisation will remain of the utmost importance. In the case of new project
development, the need to understand the breakage characteristics will remain
as important as it has always been. In optimisation projects, however, where it
will be required to get the very best from existing comminution circuits, accurate
geometallurgical modelling will take on an increasingly important role. However,
these requirements, in both new development and optimisation projects, are likely to
be at odds with budgets as these will be tight. However, this does not necessarily mean
that accuracy should suffer, only that the ore characterisation test work program must
be more focused and streamlined. The following paper discusses several factors and
guidelines that should be considered when faced with developing such programs.
NECESSARY ATTRIBUTES OF A TEST WORK PROGRAM

From a comminution perspective, an effective model, whether used for design or for
geometallurgical purposes should be able to accurately predict the throughput of the
grinding circuit from information concerning the breakage characteristics of ores that
are planned to be delivered to the processing plant. To do so there are at least five
important requirements:
1. sufficient and relevant samples have been identified, extracted and tested
2. appropriate ore characterisation tests have been chosen to describe the
comminution properties of the orebody
3. the model(s)/equations chosen to describe the comminution equipment in
the circuit respond realistically to the values obtained from the chosen ore
characterisation tests
4. all of the above are integrated into an overall description of the operational
response of the grinding circuit that also takes into account non-ore related
influences eg equipment size, speed, ball load etc
1. Managing Director, SMCC Pty Ltd, Chapel Hill 5. the final model/equations can convincingly demonstrate their accuracy through
Qld 4068. Email: steve@smccx.com validation using real plant data.
131
S Morrell
HOW MANY SAMPLES? values and hence unfortunately the author is not in a position
to determine whether this bi-modality can be traced back to
There is no easy answer to this question though it is true to
say ‘the more – the better’. If the deposit is highly variable broad geological descriptions of the nature and genesis of the
the required number of samples will be higher. Also, the orebodies. However, it is tempting to hypothesise that there is
end use will also drive the number of samples required. a very good physical reason why the distribution should have
Hence if samples are required for a prefeasibility study the such a bimodal shape.
number will be relatively low, whilst if samples are required The data in Figure 1 can provide the basis for some simple
for the development of a geometallurgical model that can calculations that can be used to guide the metallurgist’s choice
accurately forecast daily grinding circuit throughput, the of how many samples should be taken and analysed in this
number required will be at least an order of magnitude first step to investigate the orebody. Using classical statistics
higher. In all cases a staged approach to sample selection and and assuming the orebody has a variability of 20 per cent
laboratory test work is recommended to ensure that costs are (ie the lower of the modal values), then choosing a total of
kept to a minimum. Each stage should be designed to build 10–15 samples should provide a mean hardness value for the
on the knowledge gained from preceding ones, particularly deposit with a precision of approximately 10 per cent at the
concerning variability, both spatially within the pit as well as 90 per cent confidence level. However, if the variability is in
in terms of absolute hardness values. the 30 per cent class, to obtain the same level of precision the
The prefeasibility study level is often the best opportunity requisite number increases to about 30 and if the variability
to start accumulating useful information of the comminution is 40 per cent then sample requirement escalates to 45. The
properties of the orebody. At this stage, little or no minimum initial number of samples is therefore recommended
information is likely to exist on the comminution properties to be 10–15. Analysis of the data from these samples will
of the orebody and hence the metallurgist is faced with the provide an indication of the true variability and can then be
decision of how many samples should be treated for this used to estimate how many more samples (if any) are need to
very first investigation. A good starting point is to use the satisfy the accuracy for the prefeasibility stage. If, for example,
distribution shown in Figure 1. This comes from the semi- the indicated variability from the initial 10–15 samples is
autogenous grinding (SAG) mill comminution (SMC) Test® 40 per cent then an additional 30–35 samples will be required.
database which currently numbers over 40 000 separate test As the design stages progress through to a final bankable
results covering over 1500 different ore deposits. The figure study, more definition is required to enable more accurate
shows the coefficient of variation (standard deviation/mean forecasts of ore properties and hence throughput to be made
expressed as a percentage) of the measured drop weight during at least the first few critical years of production. This
index (DWi) values from each deposit (Figure 2 shows the requires further ore characterisation. However, the results
distribution of mean DWi values from each deposit). The from the initial stage of testing should provide a firm basis
DWi has been shown to be a very accurate measurement of on which to choose both the numbers and locations where the
the hardness from an autogenous grinding (AG)/SAG and samples should be taken.
high-pressure grinding rolls (HPGR) perspective and hence is It is pointed out that these guidelines assume that the
highly relevant in this context. The distribution of coefficients samples are representative of the whole orebody and should
of variation is bimodal, modal values being at 20 per cent and be suitable to provide mean hardness values that relate to
30 per cent, the average being 30 per cent. Unfortunately the the deposit in a global context only. For the development of
database does not contain information other than SMC Test® a geometallurgical model much more detailed information
FIG 1 – Distribution of drop weight index (DWi) variability values from the semi-autogenous grinding (SAG) mill
comminution (SMC) Test® database (based on 40 000 SMC Tests® covering over 1500 deposits).

Getting optimum value from ore characterisation programs in design and geometallurgical projects associated with comminution circuits
FIG 2 – Distribution of drop weight index (DWi) mean values from the semi-autogenous grinding (SAG) mill
comminution (SMC) Test® database (based on 40 000 SMC Tests® covering over 1500 deposits).
is required, eg such a model may need to predict the use of non-standard equipment. This problem recently arose
week-by-week or month-by-month throughput. In such when analysing the data shown in Figure 3. When considering
cases much more definition in terms of the hardness in crushing work index (CWi) data from machines conforming
specific parts of the deposit is required. In relation to this, to Bond’s original design a reasonably good correlation had
knowing how hardness varies with depth is particularly
been found between the parameter CWi*sg (specific gravity)
important as it indicates the extent to which throughput of
and the DWi (DWi units are in kWh/m3 so the use of CWi*sg
the comminution circuit will increase or decrease as time
progresses. Such knowledge is vital both for day-to-day and gives the same units). However, when data were added from
long-term planning and is critical to the continued financial a modified machine design a very different correlation was
success of the mine. As a result it is not uncommon for mines found. The differences between the CWi values from the two
to conduct ore hardness testing on hundreds of samples per types of machine are of the same order as those obtained from
annum throughout the life-of-mine. the Dunne and Angove program. Metallurgists involved in
ore characterisation test work need to be aware of such biases
WHICH TESTS? to ensure that the laboratory chosen to conduct the relevant
Clearly, the laboratory tests that are chosen need to be test are using standard equipment as well as understanding
compatible with the modelling approaches subsequently
used at the design stage or in the development of the TABLE 1
geometallurgical model; for example, if the AG/SAG mill Accuracy of Bond tests (after Dunne and Angove, 1997).
model in the JKSimMet software (by Julius Kruttschnitt
Mineral Research Centre (JKMRC), University of Queensland) Sample P G R Av
is to be used the laboratory test has to generate an A and b Crushing work index (CWi)
value as well as the specific gravity. In addition to this, the
Min (kWh/t) 12.4 8.1 12.1 10.9
metallurgist should also be aware of what are the inaccuracies
inherent in the laboratory tests that have been chosen. Max (kWh/t) 21.3 17.2 25.6 21.4
The study conducted by Dunne and Angove (1997) helps Coefficient variation (%) 27.3 43.5 40.9 37.2
to illustrate this point. Table 1 has been generated from the Rod mill work index (RWi)
data given in their study which looked at the variability in
Min (kWh/t) 15.9 16.9 15.8 16.2
results from sending the same samples to three different labs
and having Bond crushing, rod and ball mill work index Max (kWh/t) 20.9 18.2 17.9 19.0
tests carried out. It is clear from their results that the inherent Coefficient variation (%) 11.4 3.8 6.3 7.2
variability in the crushing work index test is very large,
values from some laboratories being half those of others. Rod Ball mill work index (BWi)
mill and ball mill tests fared much better, the results from the Min (kWh/t) 17.1 16.4 15.5 16.3
ball mill test showing a very good precision of 3.4 per cent Max (kWh/t) 19.3 17.6 16.2 17.7
on average. The reasons for the very high variability in the
crushing work index test results is quite likely due to the Coefficient variation (%) 4.6 3.6 2.2 3.4

S Morrell
FIG 3 – Correlation between drop weight index (DWi) and crushing work index (CWi*sg), where sg – specific gravity.
that even when using standard equipment, some tests are derived from the SMC Test® database. The magnitude of the
inherently less accurate than others. coefficient of variation is very low, indicating a very precise
In a study where the same sample was sent to eight different test. The result can be compared with the value of 5.7 per cent
quoted by Stark, Perkins and Napier Munn (2008) from doing
labs to ascertain the variability from SMC Tests®, the results
repeat full drop-weight tests using the same drop-weight
shown in Table 2 were obtained and reflect the variability of
machine at JKTech, University of Queensland.
the test itself plus variability associated with any differences
between labs in testing machine or operating standards. As
can be seen the coefficient of variation was only 3.9 per cent WHICH MODELS/EQUATIONS?
for estimates of the DWi and 3.8 per cent for the estimated A*b The choice of which models or equations to be used should
value. It should be noted that the SMC Test® estimates of A*b be driven by their demonstrable ability to predict the
were done without reference to associated drop-weight tests, performance of existing plants. Once again ‘the more – the
ie they were not ‘calibrated’ with actual drop-weight test better’ rule applies. The more (and varied) data that exists
data but used the raw SMC Test® results and global factors to prove the accuracy of the technique, the stronger is the
argument to adopt it.
TABLE 2 In design studies the most relevant data with which to
Accuracy of semi-autogenous grinding (SAG) mill comminution (SMC) Tests®. evaluate the suitability of a technique should be those from
existing comminution circuits whose performance has
Drop weight index (DWi) A*b been predicted then checked against high quality operating
data from the same circuit. Developers of design and
Lab 1 9.6 29.2
geometallurgical models will claim their techniques and
9.5 29.3 models to be suitable and accurate, but if a large volume of
Lab 2 8.9 31.6 appropriately varied data cannot be presented to validate the
claimed accuracy, the metallurgist should be extremely wary
8.7 32.2 of utilising such models and techniques.
Lab 3 9.7 28.8 The development of the SMC Test® and the use of the
9.5 29.5 parameters that it generates in predicting comminution circuit
equipment and circuit performance has been well publicised
9.0 31.2
in respected international technical journals (Morrell, 2004a,
8.8 32.0 2004b, 2008, 2009, 2010). Figures 4–6 demonstrate its accuracy
Lab 4 9.3 30.0 in each of these roles using a large database of operating
plants. More recently the Global Mining Standards and
Lab 5 9.4 29.9
Guidelines (GMSG) Group has adopted the so-called ‘Morrell
9.6 29.3 method’ as one of its guidelines for predicting comminution
Lab 6 9.1 30.9 circuit specific energy (GMSG Group, 2016).
8.6 32.5 The SMC Test® is unique in that from the one test a number
of parameters are generated which can be used for a variety of
Lab 7 9.6 29.3 equipment, thus saving money which would otherwise have
9.5 29.5 to spent on separate tests for each different type of equipment.
Such is the case with Bond’s suite of tests which need separate
Lab 8 9.1 30.8
tests for crushers, rod mills and ball mills.
9.1 30.9
For validation of geometallurgical models, the most
Mean 9.2 30.4 appropriate data are from production records over relatively
Standard deviation 0.36 1.2 long periods. Such an example is shown in Figure 7 and
demonstrates the accuracy of a model based on the use of
Coefficient variation 3.9 3.8 SMC Test® and Bond ball mill work index data.

Getting optimum value from ore characterisation programs in design and geometallurgical projects associated with comminution circuits
FIG 4 – Observed versus predicted tumbling mill circuit specific energy.

(Note: ab – ag/ball; ag –autogenous grinding; abc – ag/ball/crusher; sab – sag/ball; sag – semi-autogenous grinding;
sabc – sag/ball/crusher; ss ag – single-stage ag; cr-ball – crushing ball, hpgr – high pressure grinding.)
FIG 5 – Observed versus predicted crusher specific energy.
FIG 6 – Observed versus predicted high pressure grinding rolls (HPGR) specific energy.

S Morrell
FIG 7 – Example of the accuracy of a geometallurgical model (after Alruiz et al, 2009).
CONCLUSIONS Global Mining Standards and Guidelines (GMSG) Group,

2016. Morrell method for determining comminution
Given the tighter constraints on finances for ore circuit specific energy and assessing energy utilization
characterisation programs, metallurgists will need to be far efficiency of existing circuits [online]. Available from:
more selective in their choice of tests in future to ensure that <http://www.globalminingstandards.org/wp-content/
accuracy is not unnecessarily sacrificed. uploads/2016/08/20150821_Morrell_Method-GMSG-ICE-
v01-r01-.pdf> [Accessed: 29 May 2017].
The choice of appropriate test(s) should be made based
on measured precision as well as demonstrated accuracy Morrell, S, 2004a. Predicting the specific energy of autogenous and
of the techniques that subsequently use the test results to semi-autogenous mills from small diameter drill core samples
predict plant performance. This accuracy can only be truly [online], Minerals Engineering, 17(3):447–451. Available from:
<https://doi.org/10.1016/j.mineng.2003.10.019> [Accessed:
demonstrated from analysing large varied databases of
29 May 2017].
relevant plant performances. Regardless of claims by the
developers of various design and geometallurgical models Morrell, S, 2004b. An alternative energy–size relationship to that
proposed by Bond for the design and optimisation of grinding
and techniques, about their suitability and accuracy, if
circuits [online], International Journal of Mineral Processing,
data cannot be presented to validate this claimed accuracy, 74(1–4):133–141. Available from: <https://doi.org/10.1016/j.
metallurgists should be extremely wary of utilising such minpro.2003.10.002> [Accessed: 29 May 2017].
models and techniques.
Morrell, S, 2008. A method for predicting the specific energy
In terms of the required number of tests, a staged approach requirement of comminution circuits and assessing their energy
is recommended in which, as projects develop from the utilisation efficiency [online], Minerals Engineering, 21(3):224–233.
prefeasibility stage, knowledge of the variability of the Available from: <https://doi.org/10.1016/j.mineng.2007.10.001>
deposit is progressively built and used to drive the number of [Accessed: 29 May 2017].
tests required in subsequent stages. Morrell, S, 2009. Predicting the overall specific energy requirement
of crushing, high pressure grinding roll and tumbling mill
circuits [online], Minerals Engineering, 22(6):544–549. Available
REFERENCES from: <https://doi.org/10.1016/j.mineng.2009.01.005>
Alruiz, O M, Morrell, S, Suarzo, C J and Naranjo, A, 2009. A novel [Accessed: 29 May 2017].
approach to the geometallurgical modelling of the Collahuasi
grinding circuit, Minerals Engineering, 22(12):1060–1067. Morrell, S, 2010. Predicting the specific energy required for size reduction
of relatively coarse feeds in conventional crushers and high pressure
Dunne, R C and Angove, J E, 1997. A review of standard physical ore grinding rolls [online], Minerals Engineering, 23(2):151–153. Available
property determinations, in Proceedings World Gold ’97 Conference, from: <https://doi.org/10.1016/j.mineng.2009.10.003> [Accessed:
Singapore, September, p 139. 29 May 2017].
Stark, S, Perkins, T and Napier-Munn, T J, 2008. JK drop weight
parameters – a statistical analysis of their accuracy and precision
and the effect on SAG mill comminution circuit design, in
Proceedings MetPlant 2008, pp 147–156 (The Australasian Institute

Contents
Cost-effective concentrator design

G Lane1, P Dakin2 and D Elwin3
ABSTRACT
This paper discusses the factors that contribute to the cost-effective design of a
concentrator.
Concentrator design and layout outcomes are functions of the team (engineer’s
and owner’s) participating in each particular project. The benchmark for relatively
modest projects was set in the 1980s and 1990s during the ‘gold boom’ when
numerous cost-effective plants were designed and constructed on a lump sum basis
in a very competitive market. A number of factors contribute to lack of transference
of the lessons learnt in gold plant design to concentrator design including established
paradigms in the design and layout of concentrators, lack of experience in cost-
effective design, operator’s preferences for flow sheet and layout and simple lack of
appreciation of the impact of plant footprint on materials quantities and resultant
capital costs.
Experiences with recent copper concentrator projects (both small and large) are
used as case studies.
INTRODUCTION
Cost-effective concentrator design is not an isolated paradigm. It needs to interface
with the project infrastructure constraints, owner’s needs, vendor’s capabilities,
constructor’s logic and operator’s and maintenance team’s preferences. However,
cost-effective design has some ‘rules of thumb’, namely:
•• keep the execution strategy and plan simple, sift the ‘baggage’ from the facts
early, have a plan (and agreed scope) and stick to it
•• minimise the number of interfaces across all parties as every interface requires
‘management’
•• invest in good equipment as it saves you money
•• reduce plant footprint as capital and operating costs increase with increasing
plant footprint.
Capital costs will escalate if:
•• scope is poorly defined and the execution strategy meanders (scope and design
are not frozen)
•• simplicity is replaced with opportunism (hope)
•• pipe rack locations are used as the basis of plant layout
•• ‘expandability’ is a necessity.
SETTING THE SCENE

Nirvana for the project developer and plant designer is a ‘cost-effective plant’ that
meets expectations. There are many approaches to plant design that range from the
grandiose to the shoddy and mean. ‘Cost-effectiveness’ is a different paradigm. It
relies on sound judgement and a balanced assessment of what is required for the
circumstance. The ‘art’ of designing cost-effective carbon-in-leach (CIL) plants was
mastered in the late 1980s by Australian engineering companies who competed on a
lump sum turn key basis for plants in the 100 kt/a to 5 Mt/a throughput range. The
1. FAusIMM, Chief Technical Officer, Ausenco need for these plants arose through the development of CIL technology, the plethora
Minerals & Metals, South Brisbane Qld 4101.
Email: greg.lane@ausenco.com
of modest grade opportunities in Western Australia in particular, the high gold price
and the flexible nature of gold metal marketing (Close, 2002). These circumstances
2. Principal Designer, Ausenco Minerals &
Metals, South Brisbane Qld 4101. resulted in the need for innovation, short project schedules, low project capital costs
Email: phil.dakin@ausenco.com and a ‘money making machine’ approach. On occasions the commercial and design
3. Principal Designer, Ausenco Minerals ‘recipe’ came unstuck. Notable projects, such as Three Mile Hill, initially failed to
& Metals, South Brisbane Qld 4101. meet throughput targets due, in part, to a ‘one size fits all’ approach and insufficient
Email: derek.elwin@ausenco.com test work to define the design.
137
The lessons learnt from the ‘cost-effective’ design examples Key drivers for a cost-effective plant design are listed below:
of the Australian gold industry in the 1980s and 1990s are still •• Optimise the plant footprint with the aim of reducing
the basis of sound CIL plant design although the frequency of concrete, structural steel, piping and electrical/control
projects in Australia is now low and there is greater focus on cable and raceways. Push all areas close together,
Africa. The ‘art’ is practiced by some of the same individuals eg grinding close to CIL/flotation, desorption and
and mostly in small to medium sized engineering companies. reagents close to CIL/flotation, air /water services close
Interestingly, as the size of the engineering company increases to plant, big drives close to the switchroom.
the ‘art’ is diluted by other commercial imperatives. Larger
•• Keep elevation of the equipment to a workable
engineering businesses target Tier 1 mining clients. These
minimum. The elevation of run-of-mine bins, mills,
clients may express interest in cost-effective engineering but
cyclones, flotation cells and thickeners are key drivers
the demands of their business, the number of interfaces and
the physical size of the projects makes meeting the aspirations (Lane et al, 2005).
much more difficult. Tier 1 companies generally have long life •• Do not use dominating piperacks or large platform areas.
projects where the desire for durability, flexibility, life cycle •• Have common platforms, stairs, pipe and cable ladder
optimisation from the outset override the urgency and design supports.
practicalities often associated with smaller projects. •• Platework and lining should be kept lean, eg only put
‘Value engineering’ is often used to trim scope and cost wear liners in the chute areas exposed to wear and not
but the fundamentals and paradigm of the project are often all internals.
immovable or have a large coefficient of restitution (or
•• Design with a fit for purpose attitude as though it is your
resistance) due to complex standards, systems and approvals
money you are spending.
processes.
•• Do not accept second best.
Cost-effective design needs good engineers and designs
and motivation from the client. For their six large copper To progress design in a cost-effective manner the following
concentrators, Xstata has opted for a ‘standard concentrator’ guidelines need to be applied (Lane and Dickie, 2009):
approach as a method of reducing design costs. Another •• The orebody and its mineralogy, geometallurgy and
project in South America opted for a ‘low cost’ design. process responses need to be sufficiently understood to
However, the design failed to meet many of the ‘rules of allow process and market risks to be managed effectively.
thumb’ and the plant is struggling to meet throughput. •• Process flow sheets need to be ‘signed off’ in the first few
Another copper concentrator currently in design has had an weeks of design. Any changes result in design change
interesting history where the initial ‘high value centre’ design notices that cause rework and administrative churn
failed to deliver a cost-effective outcome. Subsequent external within the project. If the flow sheets cannot be ‘signed
review raised a number of pertinent issues and proposed an off’, detailed design is not ready to commence.
alternative design that went to the other extreme and was too
•• Duty/standby equipment needs should be defined in
‘low cost’ to meet process needs. The final outcome will be a
the process flow sheets.
balance between capital cost aspirations and sound design.
•• Process instrumentation and preliminary piping
diagrams need to commence early and be frozen from
KEY DRIVERS a scope perspective at the 40 per cent design complete
Typical concentrator layouts tend to be designed very stage.
conservatively because designers:
•• Survey and geotechnical data generally hold up progress
•• Either don’t know how ‘close to the wind you can sail’ when finalising the location and earthworks detail. Site
(fit for design) or aren’t sufficiently experienced to survey and geotechnical studies need to be completed in
understand the cost/risk/benefit relationships. This the study and front end engineering phases.
results in the additional elevation of equipment based
•• Client maintenance preferences and local crane
on false ‘standards’ or previous practices, eg concentrate
launder slopes, setting height of mills and flotation cells. availability impact on the decisions to use overhead
cranes, tower cranes, monorails or davit cranes for
•• Lack of innovation in design (as this requires experience various duties. These decisions need to be made early in
and effective risk management). This can include using the plant layout process to avoid rework in all disciplines.
the topography and gravity instead of pumps, not using
pipe racks to dominate layouts or building close-stacked •• Local weather or environmental issues may define the
vertical processes. Good examples are placing the need for a plant under roof, inside buildings or with
mixing tank on top of a storage tank or the motor control other protection. Clear definition of environmental
centre (MCC) under the mill feed platform. needs is required prior to commencement of detailed
layout design.
•• Tend to ‘re-invent the wheel’ instead of re-using,
improving or adapting previous proven design. Existing •• The cost of installing plant in buildings, particularly in
design libraries aren’t well known about or published. the typical South American style, is high. Clear definition
of client preferences is needed prior to commencement
•• Don’t have field installation, commissioning and site
of layout design.
as-built experience, particularly on projects they have
designed. Designers with this experience are better •• In-country materials of construction costs need to be
equipped to understand the basics of layout and understood in order to make cost-effective structural
operability and then implement these lessons in future decisions (eg concrete versus steel).
designs. •• Project expansion requirements and timing need to
•• Come from other industries, eg petrochemical or oil be clearly defined in the front-end engineering design
and gas industries. As a result of their prior history they (FEED) phase.
have little or no relevant commissioning or operational •• Concentrate transportation methods need to be defined
experience relevant to flexible but compact plant design. in the FEED phase (eg truck, rail or donkey).

•• Reagent delivery and on-site storage requirements need some locations where on-site fabrication costs are high
to be defined in the FEED phase based on plant access or people with the requisite skills are in short supply. A
limitations (seasonal weather and/or other social and logistics study is required at an early stage.
environmental factors). •• Steel sections standards differ between countries and
•• The water balance needs to be finalised by area as the need to be reconciled with fabricator’s norms.
design is developed with particular focus on the storage •• Project manager capability ‘to support the team
method (ponds versus tanks). to perform at maximum capability’ is a key driver
•• Environmental approvals need to be finalised and particularly in maintaining a high level of clarity from
permitting requirements (traffic, run-off, dust, noise, the client interface to the drawing floor.
fumes, and materials safety) need to be clearly defined •• Engineering managers need to be able ‘to lock down
in the FEED phase. the scope’, understand ‘fit for purpose design’, assign
From project management and execution perspectives the responsibility, support the leads and motivate the team.
following issues need to be considered:
•• The ‘options’ need to be considered and evaluated prior CASE STUDIES
to detailed engineering and project execution proper.
Value engineering assessments can occur during the Introduction
design process but these need to be limited to low level
Project names are not used when discussing most of the case
issues and not matters of scope or issues material to the
study examples herein. Photographs from other publications
schedule. Value engineering exercises to contain capital
and the public domain are used as examples to illustrate
cost after 30 per cent engineering completion mean
particular design features.
that the project was not set-up initially with the correct
capital and/or design expectations. Large and small concentrators present different challenges.
Small concentrators (less than say 10 Mt/a and single
•• Critical vendor equipment certified data needs to be
concentrate) are simpler to arrange as there is typically
expedited. Detailed design can continue without vendor
one crusher, stockpile, semi-autogenous grinding (SAG)
data if the team has sufficient experience to understand
mill, ball mill, flotation train and tailings thickener. Larger
the impact that vendor data can have on design. Vendor
concentrators with multiple SAG and ball mills, multiple
specifications by the engineering company may need to
be prescriptive to accelerate schedule. This compliments flotation trains and large service runs demand an additional
the use of good equipment as if this equipment is similar level of complexity for maintenance access, service equipment
to that used on other jobs the vendor information can be and service runs.
more easily expedited.
Small concentrators
•• Simplicity in approach is ‘king’. Packaging aspects of
the engineering for completion by ‘low cost engineering In many respects, small concentrators of less than 10 Mt/a
centres’ can be a recipe for disaster unless the packages and particularly less than 5 Mt/a are easier to layout in a cost-
are well defined and managed. Engineering needs to effective manner provided that all contributors are of a like
be progressed to between 40 per cent and 70 per cent mind.
complete prior to remote completion, and slightly less if The following examples indicate what to avoid in order to
key lead engineers migrate with the packages. optimise project value. Figure 1 illustrates a number of design
•• The fabrication strategy needs to be developed cognisant features that increase plant footprint.
of local (to the project) capability and capacity, low The space between the unit process operations are for
cost offshore alternatives and logistical issues such as mobile crane access and potential expansion (regrind mills).
the consolidation of equipment and fabricated items. Mobile crane access is most effective when there is no impact
Preassembled modules may provide opportunities for on plant arrangement. If pipe racks need to be installed or
FIG 1 – An example of a possible small concentrator layout.

extended to allow adequate access, the use of mobile cranes leads to a more complex decision-making process and a greater
may not be cost-effective and alternatives such as tower or tendency to conservatism. The approach to design can also
portal cranes should be considered. change to one that is driven by key engineers and designers
The arrangement of the unit processes in Figure 1 requires to one that is driven by a more over-arching approach based
extensive pumping of slurry between unit processes and the on ‘proven track record’ or prior designs.
installation of large pipe rack ways. These can be avoided by At 12 Mt/a capacity with a single train SAG and ball mills
thoughtful design in most circumstances. in the grinding circuit it is relatively easy to design a cost-
The derivation of the Figure 1 design is interesting in that the effective concentrator (Figure 2). These projects generally
original definitive feasibility study (DFS) design and layout require modest size teams and can be lead effectively by a
was a typical open air design on relatively small footprint that competent engineering manager using simple engineering
systems. As projects become larger, the team grows and
had small pipe racks between facilities and maintenance access
the infrastructure and technical issues increase, particularly
by local davits and monorails. A metallurgical review with
when equipment selection considers large capacities and/or
about 15 per cent of the engineering completed resulted in the
novel design features. However, it is still possible to design
inclusion of flash flotation and this changed the cyclone tower
concentrators with twin train grinding circuits with up to
design substantially. In addition, it was decided to leave room
25 Mt/a capacity with relatively simple and compact layouts
for a possible regrind mill and the milling facility and flotation
(Figure 3 for example) if the concept is set early, agreed by
circuit were separated to affect this change. Flash flotation
the owner and conveyed effectively to the engineering group.
cleaning was added when engineering was about 25 per cent
completed and installed in the location allowed for the future There are numerous examples of different approaches to
regrind mill. A maintenance study was conducted at about concentrator design throughout the world. Plants with large
40 per cent engineering complete and access for mobile cranes
increased the separation between the unit processes and pipe
rack lengths increased. In addition, overhead gantry cranes
were allowed over cyclones and the primary crusher requiring
significant structural strengthening due to the local seismic
conditions. Hence, this is a good example of a project where a
design approach was not frozen and maintained throughout
the engineering design and where a series of relatively minor
modifications led to a less than optimum outcome due to their
incremental impact on the layout.
Large concentrators
The design of large concentrators requires a large team of
designers, often multiple parallel lines of equipment and
critical consideration of operating and maintenance activities
due to the size of the wear items, the weight of replacement
equipment and the volume of consumables. In general, this FIG 2 – The 12 Mt/a concentrator (Lane et al, 2008).
FIG 3 – The simple layout of a 25 Mt/a twin train SABC (semi-autogenous grinding/ball/crushing) concentrator.

capacities, such as at <http://www.citicpacificmining.com>, out in this paper. However, the critical consideration is to
with multiple trains of the largest grinding mills have large give the owner(s) what they want in meeting targets, budgets
footprints. The grinding area layout is at the other end of the and project timing. To achieve this, it is the engineer’s role to
spectrum from the concepts promoted in this paper. The mills optimise the design within the owner’s constraints to achieve
are elevated and separated and as a consequence the bulk maximum value from the project.
materials quantities are high. This combined with location-
related costs for materials leads to high capital cost outcomes. ACKNOWLEDGEMENTS
There are reasons for the layout, particularly associated with
The authors would like to acknowledge Eddie McLean
the mill erection process where the mills were assembled
for reviewing the paper and providing suggestions for
overseas, transported to site and lifted into place using a
improvement. As well as all those engineers and designers
purpose designed system, but the proportional costs of civil
who have contributed to the authors experience over the
and structural works associated with this style of plant design
years, including those from Ausenco.
are significantly greater than those for smaller plants such as
in Figure 2.
The South American market benchmark for large REFERENCES
concentrators has been set by Bechtel, eg Los Pelambres and Close, S E, 2002. The Great Gold Renaissance, 282 p (Surbiton and
La Candelaria. However, the style of these plants leads to Associates: Melbourne).
high capital cost due to relatively large footprints compared Lane, G and Dickie, M, 2009. What is required for a low cost project? in
to Australian counterparts such as Cadia (Staples et al, 2008). Proceedings Project Evaluation 2009, pp 199–204 (The Australasian
One of the more recent concentrators constructed in South Institute of Mining and Metallurgy: Melbourne).
America is illustrated at <http://www.amec.com/andacollo> Lane, G, Green, S, Brindley, S and McLeod, D, 2005. Design and
and is said to be an ‘innovative, low cost design due to the engineering of flotation circuits in Australia, in Proceedings
low-grade of the copper deposit’. However, there are aspects Centenary of Flotation Symposium, pp 127–140 (The Australasian
of the comminution circuit layout and maintenance strategy Institute of Mining and Metallurgy: Melbourne).
that offer opportunities for further improvement in the context Lane, G, Staples, P, Dickie, M and Fleay, J, 2008. Engineering design
of cost-effectiveness. For example, the use of a Tower Crane of concentrators in Australia, Asia and Africa – what drives
may have allowed a significant reduction in structural steel in the capital cost, in Proceedings Procemin 2008, p 29 (Gecamin:
the cyclone tower by allowing overhead cranes to be removed Santiago).
while the grinding floor layout is ‘relatively spacious’. Staples, P, Lane, G and Messenger, P, 2008. Horses for courses –
tailoring front end design to project requirements, in Proceedings
40th Annual Meeting of the Canadian Mineral Processors, Ottawa,
CONCLUSION (Canadian Mineral Processors: Toronto).
There is no panacea solution but there are some key issues
to consider in the design and layout of any plant as pointed

Project management
and delivery
Contents
Fatal flaws in technical due diligences

A J H Newell1
ABSTRACT
The background and the basis for conducting a technical due diligence are discussed
as well as the source information, typical studies and potential limitations. Typical
short comings including ‘show-stoppers’ are presented for the metallurgical area,
with examples. Issues found in other areas such as geology, mining, infrastructure
and financial evaluations are briefly reviewed, highlighting the input and interaction
required by metallurgists with these disciplines. A checklist is provided as a guide
to maximise the success of a due diligence as well as a check on the quality of a
Feasibility Study.
INTRODUCTION
A due diligence is primarily required when funds are sought to develop a new
project or expand an existing operation or when there is a proposed merger or
acquisition. An investor, typically a bank and occasionally a trading house and/or
the merging or acquiring entity (usually a mining company or group of companies),
will select and appoint consultants to assist in the assessment of the project or
evaluation of the target asset.
There are two types of due diligence, namely legal/financial and technical, both
equally as important and both need to be found satisfactory before an acquisition or
merger can be progressed.
Historically, most technical due diligences conducted by RPM Global (RPM) have
been for financial institutions. Today, there are more potential operators (ie mining
companies) in the mix.
LEGAL/FINANCIAL DUE DILIGENCE

In a legal/financial due diligence, the investor conducts an investigation into both the
financial and legal status of the company, in an attempt to identify any current and
legacy issues, such as financial commitments, legal and ownership arrangements of
the company and the directors. It should be noted that the desirability for a merger
or acquisition can quickly evaporate with poor findings. This type of due diligence
should be conducted prior to undertaking a technical due diligence.
As an example, a client in the industrial minerals space planned a merger with a
Chinese producer. The technical due diligence went fairly well as one would expect
with an existing operation, except that the resources were not compliant with JORC
standards. However, the legal/financial due diligence (in this case completed in
parallel with the technical due diligence) discovered that not all of the shareholders
had been disclosed and moreover that there was money owed. Needless to say, the
merger did not go ahead.
TECHNICAL DUE DILIGENCE

A technical due diligence assesses the reasonableness of the technical and economic
aspects of a project in terms of practicality, viability and risks (RPM Global, 2011).
It examines the methodologies, supporting data and underlying assumptions used
for evaluating or managing a project based on supplied information, technical and
operating data from similar projects, site visits and interviews with key project or
operating personnel.
The main purpose of a technical due diligence is to identify ‘show-stoppers’
– aspects of the project that make it impractical or unviable – and thus protect the
investor from potentially losing money by investing the project. Often the technical
due diligence uncovers a raft of assumptions or unusual methodologies – ‘devils that
you don’t know’ – that seriously undermine the project. This does not necessarily
1. MAusIMM(CP), Executive Consultant, mean that the investor would not become involved in the project, particularly when
Processing, RPM Global, Brisbane Qld 4000. the underlying asset has significant potential (eg good orebody); it usually encourages
Email: anewell@rpmglobal.com a re-development of the project by the owners.
145
A J H Newell
A successful and timely technical due diligence relies A Feasibility Study, when properly developed, has
on experienced personnel with a background in project progressed through the Scoping Study and Pre-Feasibility
development as well as operations and an overlap with Study stages. When a study stage has been bypassed, it
other disciplines. For example, a metallurgist, besides being is quickly revealed in a Technical Due Diligence; there is a
competent in the processing technology associated with general lack of methodology in the study approach, only a
that commodity as well as typical metallurgical responses focus on selected areas, various options remain open and
and costs, would require some knowledge of geology (the unresolved, marketing studies are not as advanced as they
geometallurgical aspects, such as ore types, as applied in should be, insufficient test work to support the study level –
the resource model etc) as well as mining (mainly mine captured in a document that is incomplete, contains gaps and
scheduling, potential for and nature of dilution etc) to assist does not flow.
in a more accurate and timely assessment. It goes without saying that studies should be conducted by
Depending upon an individuals’ experience, a metallurgist competent and experienced personnel with a proven track
is often involved in assessing site infrastructure, except that record, either within the company or outside (eg consulting
power stations, tailings storage facilities and rail are the companies or engineering, procurement, construction
management – EPCMs). As an example, a recent due diligence
domain of infrastructure specialists. In addition, a metallurgist
of an industrial operation wishing to expand revealed a very
often reviews the marketing arrangements and the nature of
poor quality Feasibility Study, conducted by a fabricator. It
the smelter/buyer contracts.
was the show stopper of all show-stoppers, fatally flawed in
Technical due diligences are typically high level and many ways, as highlighted later in this paper. Technically,
conducted over the time span of three to four weeks, depending it lacked metallurgical detail, such as mass balances, design
upon the size of the project and the number of assets. For a criteria, flow sheets, process and instrumentation drawings
major acquisition, a high level technical due diligence may (P&IDs), design criteria etc and was based on limited test
run into eight weeks or so. Occasionally a detailed technical work employing samples from another deposit. Financially,
due diligence may be requested, which would take at least six there was no breakdown of the capital and operating costs,
months or longer. while the methodology and supporting data was poor.
Two types of assets are the target of a technical due The main areas examined in a technical due diligence
diligence. Most assets that are investigated in technical due parallel those presented in a Feasibility Study, namely:
diligences are projects supported by a Feasibility Study. •• Geology – is often the most important area and has
Consequently the main technical due diligence activities are significant potential for a fatal flaw; there needs to be
based around examining the reasonableness of the study, a high level of certainty that an orebody exists and
approach/methodologies, supporting data and assumptions supports economic extraction, allowing subsequent
and thus the likelihood of the project being technically and recovery of any project expenditures. Some significant
economically feasible. In addition, attention is paid to the flaws that can be encountered include:
nature of the project’s potential/upside as well as whether •• geological interpretation and the subsequent
risks and potential mitigations have been adequately conversion of drilling core data into a model
considered.
•• basis for classification of resources (inferred and
The other asset types considered in a technical due diligence indicated).
are an existing operation or operations, often becoming
•• Mining – has potential for fatal flaws, particularly
available for either merger or acquisition at either the top or
underground mines; major shortcomings include:
the bottom of the mining cycle. These types of technical due
diligences are in some ways easier to conduct, since the asset •• mine design and schedule – output not in terms of
has passed the Feasibility Study stage and a body of industrial ore types and is ‘lumpy’; significant variations in ore
grades, ore type blends and volumes
data and actual costs is available for review. Occasionally, an
operating asset under technical due diligence is planning an •• the optimum mining rate has not been determined
expansion. The expansion, which would be generally based •• inappropriate underground mining method
on an internal study, would hopefully follow the rigour and •• proposed levels of costing accuracy not achievable
accuracy of a Feasibility Study.
•• limited supporting geotechnical and hydrogeological
As a final comment, a generally less intense variation on a studies (particularly pit slope assumptions).
technical due diligence is an independent technical review
•• Metallurgy – often has a multitude of minor flaws
(ITR) or independent engineer’s report (IER), which are
and occasionally some fatal flaws; suitable flow sheet
mainly used for an initial public offering (IPO) or stock market and equipment with reasonable costs are the main
listing. ITR/IERs have a similar content and approach to that considerations; more details are presented in the
of a due diligence, however, are conducted at a higher level following section.
(ie less depth) and often take the form of a fatal flaw analysis.
•• Environmental – is a potential show stopper, sufficient
The nature of a Feasibility Study is discussed in some detail studies need to have been conducted and permitting
in Appendix 1 since in the assessment of a project, the quality must be in place.
of a Feasibility Study has a significant bearing on the outcome
•• Infrastructure – can become a significant show stopper,
of a technical due diligence. Some typical characteristics of
particularly for larger projects in remote locations and
non-Western feasibility studies are also presented.
typically for industrial minerals; inadequate water and
When technical due diligences are conducted on non- power supply studies; inadequate tailings disposal
Western Feasibility Studies, projects need to be evaluated planning; trade-off studies have not been conducted;
within the context of the local commercial, technical and local climatic, seasonal or seismic conditions have not
political environment and not necessarily compared with the been fully considered. In addition, the lack of or limited
standard Western approach which addresses the risks raised sterilisation drilling as well as geotechnical studies for
by Western investors and developers. the locations of the processing plant and tailings storage

facility can introduce unexpected costs and significantly executed, then the flow sheet may be fatally compromised and,
delay a project or limit production (eg Escondida). as a consequence, the process design and associated costs.
•• Social/indigenous – has potential to be a show stopper •• Test work sample issues:
(cf Garg Island); local support for the project needs to •• samples for separation test work not being representative
demonstrated. •• limited range of samples used to conduct
•• Marketing – occasionally a show stopper (eg product comminution testing
quality and or volumes not accepted by market). •• suitable test work sample locations for bench scale,
•• Financial evaluation – only as good as the quality and locked cycle and pilot plant test work.
reliability of the inputs. Note that any project valuation •• Limited mineralogical studies:
(eg net present value (NPV)) must be based on at least
•• particularly for more complex ores and subsequent
Indicated Reserves if a public document and for a
test work.
Feasibility Study financiers generally like to see at least
six years of reserves. •• Limited comminution test work:
•• Project execution – potential for flaws, occasionally •• none – common in Russian and Chinese work
fatal flaw, generally due to optimistic construction •• only Bond ball mill work index (BBMWi) and Bond
schedules (impact of seasons and climate has not been abrasion index (Ai)
fully considered), equipment delivery schedules and •• poor understanding of liberation requirements.
ramp-up times. •• No preconcentration test work (potential not identified).
An issue that can initially arise when conducting a technical •• Poor understanding of metallurgical losses.
due diligence is getting access to documentation and data in
•• Limited amount of separation test work:
a timely fashion. This information is typically provided via
an online data room through a series of access protocols. The •• insufficient amount of test work to convincingly support
quality of data can vary greatly and be in unusual places; for flow sheet interpretation and metallurgical response
example, in a recent technical due diligence of a large nickel •• bench scale flotation test work – no locked cycle test
operation, some metallurgical studies were in geological work (LCT)
reports – which is not a place that a metallurgist would •• no reproducibility studies, particularly when results
naturally search. are variable – test work and thus flow sheet decisions
Another strategy employed by target companies is the being made based on one result
provision of copious quantities of data, of variable relevance •• no establishment of feed grade-recovery relationships
and often out of date. Often the site visit produces the critical (important for financial model).
data sought, however due to bureaucratic protocols, there may •• Limited amount of dewatering test work:
be delays, some of which can be a strategy employed by the
•• insufficient body of data to convincingly size and
target company. Both of these approaches were experienced
select dewatering equipment
in a previously mentioned technical due diligence.
•• where concentrates are produced, no transportable
While site personnel tend to be knowledgeable, helpful and
moisture limit (TML) studies
communicative, this is not always the case. A relatively recent
technical due diligence site visit to a moderately large eastern •• where ‘dry stacking’ is selected for the tailings, a range
Australian copper operation revealed that a complete turnover of tailings samples reflecting the likely range in the
of technical staff had occurred within the previous nine months quantity of fines needs to be tested.
or so and that site personnel had limited detailed knowledge. •• Flow sheet:
Generic flaws, which often prove fatal, result from: •• not finalised
•• insufficient or unavailability of supporting data •• unresolved technical issues
(eg geological data, metallurgical samples etc) •• insufficient test work to support flow sheet selection
•• optimistic, incorrect or poorly based interpretations •• not suitable for all ore types/blends that would be
of data (eg geological data, future metal prices, metal presented to the plant
recoveries etc) •• insufficient flexibility
•• proposed levels of accuracy not achieved (eg capital and •• ‘novel’ flow sheets:
operating costs). •• not been proven on a demonstration scale (cf pilot
A processing technical due diligence checklist has been plant and bench scale tests).
provided in Appendix 2 highlighting the many elements that Other areas that are often flawed include:
need to be considered. For a project, it is also a guide to what
•• Metallurgical recoveries:
a successful Feasibility Study should be addressing.
•• limited understanding of sample or test work error
(reproducibility)
PROCESSING FLAWS
•• no establishment of feed grade-recovery relationships
The most commonly encountered processing flaws are related
to the flow sheet. The flow sheet attracts much attention •• no allowance for scale-up.
because it is the basis for the processing plant design through •• Plant design:
the mass and water balance, design criteria, equipment sizing •• design criteria does not allow for the full range of
and selection as well as the determination of consumable and processing requirements over the life-of-mine (LOM)
labour requirements and finally operating and capital costs. and may require a later expansion
A flow sheet is based on test work samples, mineralogy •• equipment selection and sizing unsatisfactory and not
and the test work program. If any of these tasks are poorly well supported

A J H Newell
•• limited or lack of modelling particularly for milling •• was mainly pyrite, assaying only 30 to 80 g/t Au when
circuits; eg transfer sizing (semi-autogenous grinding 5000 to 8000 g/t (0.5–0.8 per cent Au) is the threshold for
(SAB) / ball mill circuits). direct smelting
•• Operating costs: •• could not be easily upgraded
•• consumable estimates not based on first principles •• had poor gold recovery downstream (low gold
(eg test work or simulation results), however not recoveries found with intensive leaching).
always an issue, benchmarking may be acceptable
•• no current consumable unit cost quotations including Case study 3
power The project assessed was a multibase metal deposit (Cu-Pb-
•• proposed levels of costing accuracy not achieved Zn-Ag-Au) in Peru. The gravity concentration vendor had
•• no personnel list (employees and on-site contractors). taken over the test work program and not surprisingly, the
proposed flow sheet had eight stages of gravity separation,
some differential flotation and no tailings stream. Needless
‘SHOW STOPPER’ PROCESSING CASE STUDIES to say, this flow sheet was fatally flawed and completely
Three case studies are presented based on process development undermined the Feasibility Study.
issues concerning flow sheet design and equipment selection.
Each technical due diligences found a ‘show stopper’ that
was caused by the equipment vendor being involved in test
CONCLUSIONS
work and the flow sheet selection. Notably, the test work was Technical due diligences are often challenging experiences
designed around the equipment proposed to be supplied. As a that generally reveal flaws, sometimes fatal, that justify the
result of the due diligence findings, the interest of the investor expense of conducting such investigations. ‘Forearmed is
waned and in two cases, the projects were not developed. forewarned’ and identifying fatal flaws or ‘show-stoppers’
provides the investor or potential owner with the opportunity
The key issues found were:
to either walk away from the project or become involved
•• some misrepresentation about the capability of the by applying a new approach to the project development or
equipment operation.
•• some misinterpretation of the test work results While it is difficult to predict general fatal flaws or ‘show-
•• metallurgical ignorance by the project owners (no stoppers’ for specific projects, they typically arise from:
technical competency).
•• insufficient or lack of supporting data
Case study 1 •• optimistic, incorrect or poorly based interpretations of
data
The Chinese client (an investor without any technical
personnel) was proposing to invest in an iron ore project in •• poor costing methodologies
Australia (hematite property in Western Australia). Initial •• proposed levels of study accuracy not being achieved.
test work had been conducted with known equipment For processing, when assessing a project, the primary source
(Eriez wet high intensity magnetic separation; WHIMS) with of potentially fatal flaws is associated with the development
positive results; however, these units have high unit capital and interpretation of the flow sheet, which is in turn based on
costs, high power requirements and low unit capacities. The the nature of samples and the adequacy of test work.
equipment vendor proposed to change the equipment to
conventional drum wet magnetic separators with rare earth Process design including equipment selection and sizing
magnets, which would have significantly lower unit capital and the development of capital and operating costs represent
and operating costs. However, they would be unable to two other areas where potential flaws may be found.
generate the magnetic forces required to recover the hematite. It is important that technically competent people manage
The proposed equipment was not tested and the metallurgy the test work program and subsequent process design and
generated with the WHIMS test work was adopted for the study.
proposed equipment which indicated a very robust project In the case of existing operations, the availability of records
value (NPV) for the project. and the ability to inspect the operation makes conducting a
technical due diligence relatively straightforward. Potential
Case study 2 flaws may arise through current or potential bottlenecks, the
A major Western gold producer allowed geological staff quality of operational staff, the condition of the plant and the
to conduct a metallurgical study without metallurgical ability of the plant to handle future ore types.
staff input. An equipment vendor conducted the test work
(centrifugal concentrator). While the gravity concentration
stage was investigated, the remainder of the proposed flow
ACKNOWLEDGEMENTS
sheet, namely crushing, grinding, gravity concentration, direct The author wishes to acknowledge assistance and input
smelting of gravity concentrate and cyanidation of gravity from RPM colleagues namely Dick Addison and Don Larsen
tailings, was not tested. Furthermore, no mineralogical studies (Principal Metallurgical Engineers, Denver) as well as Bob
were undertaken on feed samples or intermediate products Denis (Executive Geological Consultant, Brisbane).
eg gravity concentrate. In a subsequent study, it was assumed
that it would be technically and economically possible to REFERENCES
direct smelt the gravity concentrate with 100 per cent gold Pincock (now RPM Global), 2011. Independent engineer/due
recovery. When the technical due diligence challenged this diligence reviews [online], Perspectives 111. Available from:
assumption, further test work was undertaken and it was <http://www.rpmglobal.com/wp-content/uploads/2015/08/
discovered that the gravity concentrate: Issue111-IndependentEngineer.pdf>.

Pincock (now RPM Global), 2015. Minimum study requirements update and project execution plan. Unlike a Feasibility Study, where
[online]. Available from: <http://rpmglobal.com/wp-content/ all of the project information is located in one document, this
uploads/2015/08/Issue128-UPDATE-Minimum-Engineering- body of information is separate to the FEL study process and
Study-Requirements.pdf>. generally not readily accessible.
For the assessment of projects, Feasibility Studies are the
APPENDIX 1 – FEASIBILITY STUDY main source of information while supplemented by site visits,
A Feasibility Study, whether claimed to be Bankable, other documentation as well as meetings with managers
Definitive or some other rarely met descriptor, is the and subject experts, the outcome of a technical due diligence
culmination of a series of studies, typically starting with depends heavily on the quality and accuracy of the associated
Conceptual and progressing through Scoping and Pre- Feasibility Study.
Feasibility. The intent of the study process is to decrease the Table 1 presents a summary of the study progress, showing
project risk with increasingly thorough technical studies that the decreasing level of risk and increasing level of detail
define an optimal project development with accurate costings and accuracy by decreasing the number of options through
and financial evaluation. trade-off studies, particularly in mining, processing and
Note that the major mining houses, which typically conduct infrastructure.
these studies in-house, use different terminologies for As in a technical due diligence, a Feasibility Study address
these study phases however employ the same processes in each major study discipline, that is geology, mining,
determining the optimum project solution. Some companies processing, infrastructure, social/indigenous, environmental
(eg Vale) have a preference for the front end engineering and financial analysis, except that in a technical due diligence,
design, which typically involves three front end loading (FEL) it tends to be undertaken by individual subject experts rather
stages (FEL 1, FEL 2 and FEL 3). This approach has a primary than a team.
focus on establishing the engineering design and subsequent The relationship between studies, with a metallurgical
capital and operating costs; it relies on a body of information flavour, and the project development cycle is presented in
(eg metallurgical test work) of Feasibility Study standard Figure 1.
that needs to be completed before the preliminary study A number of projects experience problems when a study
(FEL 1) commences. FEL 1 typically consists of a mass, water stage is bypassed, for example jumping from a Scoping Study
and energy balance while FEL 2 prepares the preliminary to Feasibility Study or from a Conceptual Study to a Pre-
equipment design, layout, schedule and cost estimates. FEL 3 Feasibility Study. This occurrence lies almost exclusively in the
is more akin to the final design phase (the EPCM stage in domain of junior mining companies and appears to occur due
Figure 1), where major equipment specifications are prepared, to a combination of inexperience, attempts to save money and
definitive estimates are prepared as well as electrical decrease the project development period. In fact it rarely does
equipment, line and instrument lists, preliminary 3D model and often substantially increases both costs and the chances of
FIG 1 – Relationship between project studies and project development cycle.

A J H Newell
TABLE 1
Project studies.
Project Stage Conceptual Study Scoping Study Pre-Feasibility Study Feasibility Study EPCM
Activity Examine concept Examine options Trade-off studies One option Construction
Resource Internal Internal/consultants Internal/consultants/EPCM contractor Internal/consultants/EPCM contractor EPCM contractor
Basis Comparable projects Database Fixed flow sheet 20% of the engineering 100% of the
Experience Probable flow sheet Layout drawings Engineering drawings engineering
Benchmarked costs Estimate based on data from similar Preliminary process and information Process and information drawings Engineering drawings
projects drawings Material take-offs Material take-offs
Some cost factoring Some estimates based on similar Firm quotations (most equipment, all Contracts
projects consumables and contract rates)
Budget quotations (key equipment and
consumables and contract rates)
Capex accuracy ±100 ±50 ±25 ±15 ±5
Contingency ±50 ±20 ±15 ±10 ±10
Opex accuracy ±50 ±40 ±20 to ±25 ±10 to ±15 ±5
Study/activity costs 12.5–25 k 125–250 k 625–1500 k 2.5–5 million 25–50 million
Duration 4 weeks 3–6 months 6–12 months 12–18 months 18–36 months
Geology Limited drilling More drilling Infill drilling Infill drilling Drilling
Geological understanding Develop resource model Indicated Resources Resource model Continue to upgrade
Ore types Inferred Resources Condemnation drilling resources
Inferred Resources Exploration potential Measured and Indicated Resources
Geotechnical and hydrological studies Geotechnical and hydrological studies Geotechnical and hydrological studies
Mining Desktop study Whittle Studies Whittle Studies Mine design Prestripping
Probable production rate Mining options Optimum production rate Schedules Developing audits
Probable mining method Mine scheduling studies Mine scheduling studies Mining fleet and drives/shafts
Equipment and labour Capex and opex Reserves Reserves Purchase equipment/
requirements Blast fragmentation studies Detailed capex and opex hire contractor
Typical capex and opex Capex and opex
Processing Desktop study Mineralogy Bench-scale testing Pilot plant (if required) Site preparation
Typical flow sheet and plant Limited bench-scale testing Variability testing Detailed comminution testing Source equipment/
Product volumes Ore characterisation Comminution testing Comminution and process modelling installing
Typical flow sheet Basic comminution data Dewatering testing Mass and water balance equipment/building
Mass and water balance Process options Flow sheet fixed Equipment selection and sizing and commissioning
Typical capex and opex Probable flow sheet Mass and water balance Power and water requirements plant
Mass and water balance Equipment selection and sizing Detailed capex and opex firm Training operators
Equipment selection and sizing Power and water requirements quotations: equipment, consumables
Power and water requirements Develop capex and opex: budget
Develop capex and opex: budget quotations for key equipment and
quotations for key equipment and consumables, labour
consumables, labour
Infrastructure Review Infrastructure requirements: sources of Infrastructure options Infrastructure design Install infrastructure
Outline power and water Indicative contracts for power and (tailings dam)
Roads, accommodation, logistics water (if appropriate)
Environmental Regulatory requirements Data collection and baseline studies Prepare and submit environmental Monitoring programs
(usually takes two years) impact statement
Social and Review Discussions with locals Develop strategy and initiate Continued negotiations
Indigenous Establish needs negotiations Signed agreements
Marketing Desktop study Product uses/potential for substitution Discussions with potential buyers Product samples for buyers Off-take agreements
Identify market and buyers Refine study Negotiations
Existing and future market size
Competitors: existing producers and
future projects
Future prices
Financial analyses Assumed values Exchange rates Exchange rates Detailed NPV analysis: various scenarios
High level analysis Discount rate Discount rate
NPV: used to compare options (trade- NPV: used to compare options (trade-
off studies) and sensitivity analysis off studies) and sensitivity analysis
Risk and mitigation Identify Preliminary Detailed Review
analysis
Based on Pincock (now RPM) (2007, 2009). EPCM – engineering procurement construction management; NPV – net present value.
project failure. Moreover, this approach assumes that the project for example, serves a very important role and represents a
is technically and economically viable, which undermines critical stage where the options are investigated, tested and
the purpose of the study process. The Pre-Feasibility Study, resolved (so-called trade-off studies), allowing a single project

development pathway to be interrogated in greater depth with •• No comminution (also with Russian TEOs) or dewatering
regards to costs during the Feasibility Study. test work conducted.
A problem often encountered with studies conducted by •• Equipment sizes are generally much smaller in size and
junior mining companies is that they have been conducted more numerous:
in a haphazard fashion, typically by different groups of •• spiral classifiers commonly used; unusual to see
consultants assigned with a particular section obviously hydrocyclones
working in isolation. This is typically driven by ‘the rush to
•• thickeners are rarely used.
market’ syndrome, where attracting finance overshadows
the need for a thorough understanding and study of the •• Marketable product grades are lower with higher
project. This approach is often characterised by the presence levels of penalty elements accepted (eg 18 per cent Cu,
of conflicting data indifferent parts of the study as well as 40 per cent Zn and 55 per cent Pb).
limited test work, general lack of supporting detail, optimistic •• Rarely issues with infrastructure – sources power and
capital and operating cost estimates and the use of current water are typically provided by the state.
and unrealistic metal prices for revenue calculations. •• Different methodology for calculating operating costs:
Employing adjectives such as Definitive (should be by •• the value of the mined material is included in the
definition) or Bankable (confirmed by a bank?) to a Feasibility mining operating cost, which is included as a cost in
Study does sound warning bells, suggesting a marketing the processing operating cost
rather than a study document.
•• depreciation is included as an operating cost which
Often significant differences in critical numbers are found in is normally considered as a separate cost in typical
various parts of the study and other supporting documents, Western financial analyses
making it difficult to form an opinion. In these cases, there
•• product transport costs for projects located in
would not appear to have been a study manager controlling
development regions are typically subsidised by the
and integrating the overall study process, resulting in little
buyer.
or no communication occurring between the disciplines.
Studies are an iterative process and require integration A Feasibility Study considers the following disciplines,
through review and communication by all disciplines. For a which parallel the technical due diligence assessment process:
processing engineer, knowledge of the geological (or block) •• Geology – ownership of tenures; regional and deposit
model, the mining schedule as well as what constitutes a descriptions; geological or block model; data acquisition
marketable product are critical elements in developing and (drilling methods, core recovery, logging, sampling,
designing a flow sheet that would successfully process LOM assaying, twinning etc); density determinations; quality
feed and thus maximise the project value. of the resource estimate (inferred and indicated – data
Note that there are so-called Western and non-Western type validation, modelling method, QA/QC, interpretation
Feasibility Studies; in most cases, non-Western Feasibility (‘geo-imaginitus’), grade/tonnage relationships etc),
Studies, with the general exception of the Russian Technio- exploration history and upside.
Economicheskiye Obosnovaniye (TEO), do not meet the •• Mining – open cut relatively straightforward,
standards of a Western Scoping Study let alone a Western type underground more complicated and depends
Pre-Feasibility Study. They tend to follow a formulaic approach upon the deposit style (cf thin veins and need for
as prescribed by guidelines held by the Design Institutes. backfilling), mining permits, mining design, waste
The main differences encountered with non-Western dumps and stockpiles, trade-off studies, mine schedule
Feasibility Studies are: and production plans, pit wall slope, rock stability
and strength (geotechnical studies), dewatering
•• Often a geological model may not exist (except Russian
requirements (hydrogeology), equipment, capital and
TEOs):
operating costs (methodology, budget quotations,
•• does not include ore types quantity take-offs, consumable and personnel estimates,
•• issues with quality assurance / quality control (QA/ accuracy, contingencies, equipment/services sources
QC) procedures as well as assaying methods. (currency), transport and insurance, dewatering and
•• Resources are not JORC compliant; while other reserves (proven and probable)).
methods of estimating resource are reasonable (for •• Metallurgy – location and nature of samples, test work
example, Russian), they cannot be simply ‘converted’ program, mineralogy, preconcentration, comminution
into a JORC classification without undertaking a full parameters, separation and dewatering; water
review, which is very difficult to do when: quality; technical management of waste streams
•• the drill core has been destroyed or lost (cf when the (eg detoxification); flow sheet; mass and water balances;
metal is classified by a country as strategic eg uranium design criteria; equipment selection, sizing and list;
in the USSR or tungsten and rare earths in China) comminution and preferably flow sheet modelling;
process description; drawings – general arrangement,
•• the actual coordinate system used to locate drill holes
process flow diagram (PFD), P&IDs and one-line
is not known.
electrical; consumable requirements (power, water,
•• Limited mine planning and lack of detailed mining labour, milling media etc); capital and operating costs
schedules. (methodology, budget quotations, quantity take-offs,
•• Source of samples not always revealed and often some factoring, accuracy, contingencies, mobile equipment,
uncertainty about representativity: equipment/services sources (currency), transport and
•• where samples cannot be obtained for test work insurance etc); product schedule.
(mainly occurs earlier study phases examining •• Environmental – suitable and sufficient baseline studies
underground deposits), metallurgical responses are have been conducted; appropriate regulatory approvals,
based either on local, provincial or country metallurgy permitting and licences eg details of environmental
for that deposit type. impact assessment or statement (EIA/EIS); stakeholder

A J H Newell
engagement, environmental management systems Site inspection

(EMS) and environmental/social management plans
(EMP) completed; reclamation and closure plan; Operation
operating costs; closure costs. •• how the plant is being operated – number of operators,
•• Infrastructure – requirements/sources/distribution/ operator activity, housekeeping etc
reticulation of power and water; if connected to grid •• condition of the plant – need for equipment replacement,
power, route map; logistics (roads, rail, airports, nature and magnitude of maintenance costs
ports etc – getting equipment, consumables and •• process bottlenecks
personnel to site; getting product to market); buildings •• review sample collection including final concentrates,
(offices, warehouses, maintenance/repair shops, assay facility, laboratory (routine and future ore testing).
accommodation); allowance for location, climate and Future plans – ore sources, grades, ore types, mineralogy etc.
seismicity (buildings); construction schedule. •• Will future ores have different characteristics or grades
•• Social/indigenous – needs to demonstrate a good that will affect the metallurgical performance in terms:
relationship with local people; support for the project, •• throughput
program of assistance and ongoing interaction. •• recoveries
•• Marketing – suitability of product and product •• product grades
specifications, size of market, future of market •• What procedures are in place to measure and record any
(competition, growth, substitution etc), product terms changes in ore types? For instance:
(penalties, credits etc), smelting and refining terms,
•• weekly testing on site
future product prices, product transport route map,
•• off-site testing
arrangements and costs etc.
•• frequency
•• Financial evaluation – typically based on discounted cash
•• How would the current flow sheet and equipment
flow methodology, with NPV used as the measure;
handle any changes in ore characteristics?
only as good as the quality and reliability of the inputs:
capital and operating costs (eg estimate basis, nature of •• Any potential bottlenecks?
the contingencies), future metal prices, exchange rates, •• increased hardness – comminution circuits
inflation rates, discount factor etc). Any NPV must be •• grade changes – separation and dewatering circuits
based on at least Indicated Reserves if a public document •• Flow sheet changes required?
and for a Feasibility Study bankers like to generally see •• potential costs
at least six years of reserves.
•• downtime for installation
•• Project execution – management plan and procedures •• Equipment changes or upgrades required?
(owner/EPCM structure and organisation, key
•• potential costs
personnel and qualifications, project responsibility list
•• downtime for installation.
(owner-managed and EPCM-managed projects), EPCM
contractor procedures (engineering, procurement, Future water and power supplies – quantities, availability,
construction and reporting etc)), implementation supply agreements.
schedule, list of probable contracts and facilities, likely-
qualified contractor listings, equipment delivery (long
Reconciliation between mine and plant
Tailings Storage Facility (TSF) capacity – need for another TSF.
lead items, logistical plans etc).
Future costs – basis for sustaining capital, any capital
In addition, bankers considering projects in non-Western
upgrades/expansions, supporting studies, production and
parts of the world also like to know that the project would revenue forecasts etc.
meet the Equator Principles (Equator Principles Association,
2011). These are series of guidelines adopted by financial Project (largely based on a Feasibility Study)
institutions for determining, assessing and managing
environmental and social risks. Nature of Feasibility Study
•• internal or external?
REFERENCES •• where they suitably competent to conduct or manage
Equator Principles Association, 2011. Equator Principles [online]. the Feasibility Study?
Available from: <http://www.equator-principles.com/>.
Pincock (now RPM), 2007. Minimum report contents for engineering
Ore types
studies, Perspectives, 70 (December). •• have they been identified?
•• reasonableness of classification (degree of oxidation,
Pincock (now RPM), 2009. Minimum engineering study requirements,
degree of alteration, mineral gangue ratios
Perspectives, 95 (March).
(eg chalcopyrite/pyrite ratio), mineral/mineral ratios,
impurity or gangue mineral (eg As level))
APPENDIX 2 – PROCESSING DUE DILIGENCE CHECKLIST •• included the geological model (and thus the mine
schedule)?
Existing operation Samples – nature of samples and composites: number, type
Historical and current records – production, operating costs, (drill core, air core etc), location, representative in nature,
recoveries, product grades etc. mining year etc.

Mineralogy – conducted for all ore types and composites; •• source for values used in calculations (eg ore, water,
identified economic and impurity types, associations, grain product and intermediate stream specific gravity
size and rock types; degree of liberation. (SGs), mass and metal stage recoveries).
•• sufficient flexibility to handle potential process
Test work program variations (‘average’ and ‘design’ cases).
•• undertaken by reputable and independent test
work provider (cf equipment vendor – under some Metal losses
circumstances)
•• metal losses for each ore type has been satisfactorily
•• conventional or ‘novel’ or using ‘novel’ equipment identified:
•• if ‘novel’, a pilot plant or pilot scale test and •• addressed the potential for improved recoveries or
demonstration plant required concentrate grades.
•• has addressed the obvious flow sheet options
•• sufficient test work has been conducted on an Metal recoveries and product grades
appropriate range and number of samples •• demonstrated as a function of ore types, mining
•• has preconcentration been examined composite types and head grades
•• has site water or synthetic water been used in test work •• full analysis of final concentrate (credit and penalty
•• ore variability test work been conducted elements and levels identified).
•• material handling – flow, rill angles, abrasiveness, slurry
viscosity etc
Design criteria
•• basis and source for selection of Design Criteria; is it
•• other characteristics: oxygen demand etc
reasonable and supported?
•• feed grade-recovery relationship been established for
•• transfer sizing for SAG – ball mill circuits
each ore type and mining composite
•• has the product been satisfactorily characterised in •• circulating loads
terms of: •• product and intermediate product per cent solids
•• full assay – credit and penalty elements •• media and reagent consumption rates
•• dewatering requirements and equipment requirements •• scale-up factors used to estimate residence time
•• if a concentrate, TML (eg conditioning, flotation, heap leaching etc).
•• other properties such as size range, friability etc. •• addresses needs of climatic and operating culture
requirements (eg cold climate, regulated holiday
•• have the tailings been satisfactorily characterised for
range of ore types and blends in terms of: periods etc).
•• mineralogical characteristics:
Equipment
•• potential for pollution and acid generation. •• basis for selection, number and sizing: reasonable
•• dewatering properties
•• equipment source
•• backfill and ‘dry stacking’ (where the size range,
•• equipment vendor: reputable, critical spares identified,
particularly the amount of fines, is important).
process and equipment warranties, availability for
•• environmental studies such as detoxification of process commissioning
water for release into the environment.
•• duty specifications: reasonable.
Flow sheet
Mine schedule
•• basis for and reasonableness of flow sheet selection
•• an optimum mining rate has been determined:
•• conventional or ‘novel’
•• basis: process capital and operating costs used
•• unnecessarily complex
•• limited by market.
•• appropriate use of surge capacity within the flow sheet
(eg stockpiles, bins or holding tanks with sufficient •• includes ore types and thus mining composites.
residence to decouple critical parts of the flow sheet)
•• potential bottlenecks.
Water
•• water quality:
Modelling •• impact upon separation and final product:
•• conducted in a satisfactory fashion for the comminution •• need to wash final product
circuit:
•• need to produce pure water.
•• basis for circuit and equipment sizing and selection?
•• fresh and process water circuits:
•• identified behaviour and impact upon design criteria
•• complexity
of different ore types and mine composites.
•• adequate storage.
•• where appropriate, for the separation circuit.
•• recovered water: assumptions, recovery from TSF,
Mass, water and energy balance allowances for evaporation losses etc
•• basis: optimum mining rate study, flow sheet, other data •• water sources available to sustainability supply
or sources: processing water requirements (including mine water).

A J H Newell
Power requirements •• cost elements:

•• basis (installed equipment, typical operating load, •• quantities reasonable: consumables, power, water,
simulations etc). labour schedule
•• unit prices reasonable: with quotations for key
Trade-off studies equipment; based on data from similar, current projects
•• have been applied in a satisfactory manner to select •• EPCM costs – reasonable for the location and
between processing, flow sheet or equipment options: complexity of the project
•• for example, ‘dry’ stacking, finer grind etc •• what has been factored: piping, installation,
•• comminution circuits. commissioning, first fill etc – reasonable and supported.
•• basis for determination of sustaining capital costs and is
Product schedule it reasonable?
•• basis: •• contingency: appropriate for commodity, project
•• appropriate mine schedule complexity and country location.
•• reasonable metal recoveries
Marketing
•• achievable product grades
•• has a marketing study been conducted?
•• varies appropriately with ore types or mine blends.
•• suitable standard?
Process description •• has it satisfactorily identified:
•• satisfactorily describes the process flow sheet. •• current market, uses and growth
•• potential buyers and actual market size for
Operating cost proposed product quality and volumes
•• methodology/basis: reasonableness (benchmark against •• competition:
similar projects in same operating environment) •• other operations (expansions) or projects
•• accuracy (typically ±10 per cent feasibility study (FS)): starting up
has it been achieved or demonstrated? •• potential substitution.
•• cost elements: •• product specifications and terms (credits, penalties,
•• what has been included? size ranges, moisture content, other properties)
•• quantities reasonable: consumable, labour schedule •• price history and future prices.
•• unit prices reasonable with quotations for key
consumables including power Risks
•• labour costs: salaries and on-cost reasonable •• has a risk analysis been conducted?
•• maintenance cost basis: factoring, reasonable. •• how does it assess the degree of risk?
•• does it identify and rank all potential risks?
Capital cost •• does it provide any recommended mitigating measures?
•• methodology/basis: reasonableness (benchmark against •• does it analyse residual risk (after application of the
similar projects in same operating environment) mitigation)?
•• accuracy (typically ±10 per cent FS): has it been achieved
or demonstrated?

Contents
Guidelines for mineral process

plant development studies
P R Whincup1
ABSTRACT
This paper presents guidelines for studies required for the development of mineral
processing facilities from initial feasibility studies through to commissioning. Mining
project schedule and cost overruns can often be attributed to inadequate metallurgical
test work, engineering and cost estimating leading up to commitment to the project.
In some cases this may result from lack of understanding of, and commitment by
the project proponent to, the requisite metallurgical and engineering studies during
the development stages. Guidelines for metallurgical test work, process development,
engineering and estimating requirements for each stage of precommitment studies
are described together with those for the engineering phase.
INTRODUCTION
The continuing rapid rise in metal prices has resulted in an unprecedented global
increase in the number of mineral project developments. There have however been
instances where companies in their rush to exploit resources, have overlooked or
cut short some of the necessary metallurgical and processing studies necessary to
ensure that a project is properly implemented and performs in line with expectations.
Consequences have included cost and schedule overruns and less than optimal plant
performance. This has led to disaffected shareholders, non-performing loans and
involuntary and disruptive changes at board and senior management level.
Requirements for the various levels of study leading to commitment of funding
for mineral project construction have been well documented (White, 2001; Noort
and Adams, 2006; Cusworth, 1993; Warren, 1991). This paper focuses on the mineral
processing aspects of these studies for which the outputs are:
•• throughput and recovery models as well as operating cost and capital cost
estimates for the project financial model
•• realisation cost information comprising transport cost, treatment and refining
charge (TC and RC) data including penalty element deductions and paid metal
recoveries
•• process plant operating costs to mine planners for pit shells/cut-off grade
determination and mining schedules, which are used in an iterative financial
modelling process to determine the project scale
•• flow sheet and design criteria for the process plant that provides for process
variability.
These outputs result from metallurgical test programs, engineering cost studies and
this paper provides guidelines for the study managers and project metallurgists at
each study level:
•• scoping
•• prefeasibility
•• feasibility
•• engineering.
Processing studies will usually interact closely with other studies contributing
to an assessment of project feasibility, which include mineral resource, mining,
infrastructure, environmental and marketing studies.
SCOPING STUDIES
A scoping study would typically commence following an exploration success to:
1. FAusIMM, Whincup and Associates,
Lower Templestowe Vic 3107.
•• define the range of process options
Email: whincup@optusnet.com.au •• establish the project scale
155
P R Whincup
•• provide first pass metallurgical recoveries and ore •• direct cost estimates for other commodities (for
processing costs for resource cut-off grade estimates example, steelwork, concrete and piping) factored
•• provide first pass cost estimates for a preliminary from the estimated mechanical equipment cost and/
evaluation of the prospect. or estimated installed electrical load; most mineral
process plant engineers will have in-house factors for
Expenditure on extensive sampling and metallurgical test
determining these, for example, percentages of the
work is usually not justified at this stage. It could be limited
mechanical equipment cost or $ kW-1 installed
to optical mineralogy followed by the minimum bench scale
test work necessary to establish indicative metallurgical •• indirect costs (for example, engineering, procurement
parameters and would be based on an assumed flow sheet. and project management) determined as percentages of
Examples include an agitated cyanide leach or a roughing/ the directs total
cleaning flotation test at one or two grind sizes with a typical •• from the ‘base case’ capital estimate, estimates covering
reagent regime. Limited comminution would be undertaken, the range of treatment rates would be made by scaling
which may include determination of approximate work ‘base case’ capital using, for example, the 6/10 rule:
indices using comparative methods. Capital 2 = Capital 1 × (Rate 2/Rate 1)0.6
Samples would typically be diamond drill hole quarter core •• review of step changes in capital. These could be additional
covering identified major mineralisation types and should costs arising from issues such as additional process lines,
be selected in consultation with study geologists. For most change in water supply or electricity sources.
projects a sample weight of between 5 and 10 kg for each Figure 1 shows an example of the capital estimating
mineralisation type should be sufficient. Scoping study test process for a large simple base metals mineral processing
work would typically cost US$30 000 to US$50 000 including plant covering the required range of processing rates for the
sample collection and freight. example described.
Results of test work would provide the basis on which In this example a contingency of 20 per cent was allowed
to develop process options. For each option a strengths, and represents the lower limit of the range of contingency
weaknesses, opportunities and threats (SWOT) analysis is allowances applicable to a properly conducted scoping study.
recommended as at a low level of cost certainty this may be At this and subsequent stages of the project an enthusiastic but
the only way to differentiate between options. From these inexperienced project proponent may be tempted to delete or
options a process route would be selected as the basis for the reduce the contingency. This is an early warning sign that a
scoping study. project could be heading for cost overruns.
It is important to focus on selecting a processing rate or Benchmarking at this level of study is valuable as a check but
project scale at this time. It is surprising how often detailed care needs to be taken that a comparative project and scope are
studies are attempted without serious efforts to establish the being examined. Capital estimates using this methodology are
project scale. It requires capital and operating estimates to considered to be accurate to no better than ±30 per cent.
be conducted over a range of treatment rates for the entire Preliminary processing cost estimates for each treatment
project including the mine, infrastructure and services. The rate would be produced from either current cost data from
‘base case’ mining and treatment rate may be determined by: similar operations or from first principles. The setting-up of
•• Observing the best project net present value (NPV) a processing cost model that reflects fixed and variable cost
return over the range examined, although sometimes components is recommended. Once established, the model
the numbers are too approximate and unless there is an can then be used over a range of processing rates and refined
obvious NPV difference or ‘step change’, this method as the project develops.
may be unreliable. Table 1 shows typical sources of scoping level processing
•• Using a rule of thumb by assuming a mine life (no less cost estimates.
than five years or greater than ten years) and applying For projects where a concentrate would be produced for
this to the expected mining inventory size. For example, transport to a downstream processing facility realisation costs
the base case treatment rate of a potential base metal must be taken into account at the scoping study stage as they
resource of 80 Mt could be 8 Mt/a based on a ten-year usually impact materially on project economics and resource
project life. In this case a range of preliminary capital and cut-off grade. Realisation costs include concentrate transport,
processing cost estimates at say 4, 6, 8, 10 and 12 Mt/a treatment and refining charges and can amount to 10–
would be conducted. 15 per cent of the in situ ore value. Indicative transport costs
During the preliminary evaluation one also needs to consider can be obtained from specialist road transport and rail freight
the likely price cycle of the commodity. In the above example operators. Treatment and refining charges are available from
the life of the project would be expected to see production commodities research groups.
through at least one base metals ‘low cycle’. It is recommended that at the scoping stage a risk and
Unless capital data are available from a recent ore processing opportunity register be established and reviewed during
plant of the type and capacity envisaged, preliminary capital each subsequent study stage.
estimates will usually require some engineering and vendor Time required for a scoping level processing study will
pricing, which would typically cost US$50 000 to US$100 000. be dependent on availability of data; however, for planning
Capital estimates would be based on: purposes a minimum of six months is recommended.
•• assumed flow sheet showing all major mechanical
equipment and ‘base case’ major process flows PREFEASIBILITY STUDIES
•• preliminary layout sketches The prefeasibility study (PFS) has three functions:
•• ‘base case’ mechanical equipment and electrical load lists 1. evaluate all process options by establishing preliminary
•• mechanical equipment pricing using recent pricing from financials for each
other projects or single vendor budget pricing 2. select one or two options for more detailed cost analysis

Guidelines for mineral process plant development studies
FIG 1 – Scoping level comparative capital estimates (US$M).
Table 1
Basis for preliminary ore processing cost estimates.
Expense element (notional, simple base metals plant) Basis

Operating and maintenance labour Conceptual manning schedule. Total employment costs from recent industry remuneration surveys or
similar operations. Employee related government charges can be sourced from government websites.
Grinding metal Annual grinding mill relines cost from other similar projects or single vendor pricing.
Annual crusher relines cost from other similar projects or single vendor pricing.
Typical grinding media consumption and current pricing.
Consumables Typical or preliminary test work consumptions and current pricing.
Maintenance materials and services 5 per cent of direct capital cost. Allowance for lubricants.
Technical services (eg assays, metallurgical consultants, audits) Allowances based on similar projects.
Services (eg freight, engineering, other consultants) Allowances based on similar projects.
Energy Preliminary electrical load list, diversified load or an allowance of 35 kWh t-1 of plant throughput. For grid
power use available gazetted prices. For diesel generated power use current or recent comparable build-
own-operate vendor pricing and the diesel price selected for the project.
Water Allowances for other energy sources. Project unit cost based on a consumption of 1 kL t-1 treated.
3. refine capital and operating cost estimates, metallurgical concentration or extraction testing. An additional 80–100 kg of
recoveries and concentrate quality ranges for project unbroken composite core sample material would be required
financial modelling. for comminution testing. Residual sample and selected test
These objectives would be met by metallurgical test work, products should be retained in storage until at least completion
and engineering and cost studies. of plant performance testing or abandonment of the project.
Test work would be aimed at providing sufficient data on It is strongly recommended that the detailed test program
which to: be developed well in advance of sample selection, in
•• undertake comparative evaluations of process options consultation with the selected laboratory and with one or
•• establish the key preliminary design criteria on which more specialist metallurgical consultants to reduce the risk
to base the engineering work needed to upgrade capital of significant additional sample material and test work being
and processing cost estimates to prefeasibility level. required at detailed feasibility or design stage to resolve
Test work will typically also produce samples for tailing flow sheet uncertainties.
storage, environmental and marketing studies. The scope of comminution test work on a composite sample
The source of sample material for prefeasibility test or, depending on variability of the lithology, a number of
work would be as for the scoping study test work, drill individual samples of the major lithology types, should be
core. However, a minimum total sample weight for each sufficient to establish the comminution circuit. Test work
mineralisation type of ~50 kg would be required for bench scale would usually include:

P R Whincup
•• unconfined compressive strength (UCS) Prefeasibility engineering would typically cover delivery of
•• Bond crushing work index (CWI) an engineering and cost study covering:
•• Bond rod mill work index (RWI) •• key design criteria
•• Bond ball mill work index (BWI) •• preliminary flow sheets and piping and instrumentation
diagrams (PIDs)
•• abrasion index (AI).
SAG Mill Comminution Tests (SMC Tests®) may not be •• preliminary mass balance, including a water balance
required at this stage if there are other strong indicators that the •• site selection and layout drawings
mineralisation would or would not be suitable for SAG milling: •• a limited number of preliminary general arrangement
•• mineralisation is not from a very competent uniform (GA) drawings, plans and sections taking into account
zone or a fully oxidised clayey zone safety, operability and maintainability; it is not
•• UCS >180 MPa unusual to commence the development of 2D and,
in some instances, 3D computer-aided design (CAD)
•• BWI >20 kWh t-1
models at this stage to provide GA and plan layouts
•• RWI is not significantly higher than the BWI and both with sufficient detail, enabling preliminary materials
are not significantly >15 kWh t-1. takeoffs (MTOs) for cost estimating
Morrell (2009) has provided guidelines for the number of •• preliminary mechanical and electrical equipment lists
comminution samples required using classical statistical
analysis of comminution parameters starting with a minimum •• preliminary electrical load list
of ten samples, respectively representative of each production •• preliminary commodity pricing
year, if possible, as well as guidance on selection of tests, test •• capital cost estimate
equipment and modelling techniques. It is advisable to have
•• processing cost estimate
preliminary comminution parameters of potential ore types
benchmarked for SAG milling amenability using a specialist •• preliminary schedule including a capital disbursement
comminution consultant. schedule
Initial prefeasibility leaching, flotation, gravity and other •• study report.
beneficiation test work would focus on elimination of process Engineering design should take into account known
options. For example, combinations of flotation, gravity and environmental and regulatory constraints.
cyanide leach tests on a copper–gold ore would be aimed
Capital estimates would be typically based on:
at resolving questions such as whether to include a gravity
circuit for gold removal, merits of intensive cyanide leach on a •• mechanical and electrical equipment pricing using a
gravity gold concentrate or cyanidation of an auriferous pyrite single vendor quotation
flotation concentrate. Comparative capital and processing •• structural steelwork, plate work, concrete, major
costs may be required to identify preferred process options. piping and architectural MTOs and single vendor
Bench-scale batch tests will usually suffice but the number written quotation
and complexity of tests required will be specific to the •• factored costs for other commodities shown in Figure 1,
mineralogy of the prospect. including architectural – use of factors for prefeasibility
Following determination of a preferred process route some direct capital estimates assumes that the process plant
optimisation test work should be undertaken, particularly is typical of those from which the engineer has derived
to determine the grind/recovery relationship for major the factors; however, if the process plant is known, for
mineralisation types, and in the case of a concentrate, the example, to have an unusual amount of pipe work or
grind/recovery/concentrate grade relationship. speciality pipe or plate work, MTOs and pricing should
A single locked cycle test and single test on reground be used for that commodity
middling for each major mineralisation type would usually •• owner’s preproduction capital from preliminary
be the limit of prefeasibility test work. quantities and current rates, and should contain an
Reagent optimisation would not normally be done at this allowance for spare parts based on a percentage of
stage unless reagent selection has potential to materially mechanical equipment capital
impact project viability. •• estimated feasibility level metallurgical test work and
The study metallurgist should consider engaging an engineering costs
independent third party to review the metallurgy and •• an assessment of working capital
processing aspects of PFS and subsequent studies leading up
to commitment to the project. •• other indirect costs as percentages of directs
An allowance of at least US$70 000 to US$100 000 is •• a preliminary engineering and construction schedule.
recommended for prefeasibility level test work for simple It is recommended that during prefeasibility engineering a
metallurgical processes. Complex processes such as those for work breakdown structure (WBS) be developed for the entire
refractory gold mineralisation treatment would need to be project (including mining, infrastructure and indirect costs)
estimated on a case by case basis but a cost within a range of and the estimating package set-up.
US$200 000 to US$500 000 would not be unexpected. Capital estimates produced for a PFS should have an overall
Engineering at prefeasibility level would usually be accuracy in the range from ±20 to ±25 per cent.
undertaken by an engineering consultancy experienced in The processing cost model developed at the scoping study
the design type of the mineral processing facility anticipated. stage would be updated and refined, and include:
For most ores any one of over 20 internationally recognised
engineers would be appropriate. Selection of the engineer •• a preliminary ore processing and production schedule
would be based on considerations of cost, relevant experience, •• a revised manning schedule and current industry rates
quality, availability of people and location. applicable to the location

•• estimates associated with on-site accommodation and Morrell (personal communication, 2008) has advised that
rotational travel while large diameter diamond core PQ (85 mm diameter)
•• consumable costs determined using rates from test work size samples may be taken, use of smaller diameter core,
and current vendor pricing for example, NQ (50 mm diameter) as comminution sample
material is satisfactory. Generally few contemporary test
•• maintenance materials as a percentage of the direct
capital cost procedures, in particular the drop weight test that forms
the basis for the SMC Test®, make any practical use of the
•• allowances for services information from larger rocks.
•• electrical energy costs based on electrical load list and
Sample weight for comminution testing would typically be
written vendor pricing
700–1000 kg for each lithological domain. Ideally the domain
•• other energy from estimated consumption derived should be defined in terms of comminution properties, which
from preliminary equipment vendor data, engineering may not necessarily coincide with the mineralogical domains.
and current pricing, taking into account freight and
The additional sample weight required for other design
storage for items such as diesel fuel oil and liquefied
level test work and variability testing is likely to be an
petroleum gas (LPG).
additional 200–500 kg per geological/mineralogical domain
The processing cost model should be extended to cover if these domains are not the same as those identified by
preproduction capitalised processing costs and set-up on a comminution properties.
quarter by quarter basis for at least four years from project
commitment and annually thereafter. The required time to Mineralogical investigations should be conducted on
commission the process plant and ramp-up to capacity needs samples or specimens from each geological/mineralogical
to be considered at this stage to assist in assessing working domain before finalisation of the test program and include:
capital requirements. •• mineralogical examination including multiple optical
Realisation costs should be updated based on: evaluations
•• product quality determined from metallurgical test work •• mineral liberation analyser (MLA) or quantitative
evaluation of minerals by scanning electron microscopy
•• preliminary transport studies including vendor budget
(QEMSCAN) bulk modal analysis.
pricing
The following two-stage approach to comminution testing
•• current or predicted industry treatment and refining
costs, penalties, deductions and price participation is also suggested by Morrell (2009).
arrangements; for smaller companies, use of a mineral In the first, limited stage, sufficient samples are tested
commodity marketing consultant is suggested. to carry out a statistical analysis, following which a more
At this point the financial and technical aspects of the extensive program is undertaken, based on results of the first.
project are reviewed and further test work, options evaluation The wider the spread of results from the first stage, the more
and value engineering may be required before committing samples would be needed for the second. The first stage would
to feasibility level studies. Corporate self-discipline may be typically involve four to five samples from each domain.
required so as not to rush into a detailed feasibility study with Comminution test requirements for samples from each
significant technical issues unresolved. domain are as for the comminution testing recommended for
A time of 8–12 months for the PFS could be assumed for PFS together with:
study planning purposes. •• Bond AI and UCS if not included in previous test
programs.
Detailed feasibility studies •• JK Mineral Research Centre (JKMRC) drop weight
At this stage of the project there is a reasonable expectation tests or the recently developed JK rotary breakage tests
that the project will proceed and metallurgical test work and (JKRBT).
process plant engineering would be undertaken on this basis.
•• SMC Test®.
Study results may form the basis for a project funding request.
•• If geotechnical core is being point load tested then
During the process plant feasibility study all design level
consideration could be given to having point load testing
metallurgical test work should be completed together with
done on comminution test samples to provide a link
30 per cent of the engineering.
between the two databases. Point load tests correlate
Dependent on equipment lead times, it may be prudent to quite well with the SMC Test® results and hence the
complete sufficient test work and engineering to allow ordering geotechnical data can provide a good indication of SAG
of long delivery equipment (eg grinding mills) prior to project
mill competency variability.
approval. On more than one occasion urgent additional
comminution test work and grinding mill specification work The crushed products from the drop weight and SMC Test®
have been required after project commitment to allow the can be reused for the Bond mill work index work if sample
mills to be ordered to meet the committed project schedule. quantity is a problem.
Design level metallurgical test work should be commenced Pilot scale comminution testing is not generally required
early in the study and should be scoped in consultation as comminution consultant databases are now sufficiently
with the proposed laboratories, a recognised comminution large to preclude the need for pilot scale testing; unless
consultant and, if applicable, a metallurgical consultant rarer circuits are being designed (eg single stage autogenous
specialising in the subject metallurgy and processing grinding milling or high pressure grinding rolls).
techniques (for example, flotation). Other feasibility bench scale test work is process specific
Early in the feasibility study dedicated metallurgical but, for example, for a large copper–gold orebody for which
samples should be taken. Sample locations should be selected treatment by flotation to produce a saleable concentrate is
in consultation with resource geologists and a consulting proposed, the test work should include for each ore type as
mineralogist. Samples should include dilution waste rock. a minimum:

P R Whincup
•• roughing and cleaning batch tests to establish baseline •• providing any necessary processing cost input data (for
flotation conditions example, manning schedule).
•• bench scale locked cycle tests to establish optimum grind Engineering deliverables will include:
sizes, flotation conditions and reagent regime leading to •• detailed design criteria
definition of a standard test flow sheet.
•• detailed flow sheets
Variability testing for the recovery and throughput using
•• mass balances for both design and operating departures
the standard test flow sheets should be undertaken:
•• life-of-mine ore treatment and production schedule by
•• by production year in which composites representing
ore type
production periods are evaluated
•• PIDs
•• characterisation of the deposit by testing a variety of
samples representing the spatial distribution of each •• detailed site layout drawings showing site roads,
mineralogical and lithological zone within the deposit. hardstand, plant service buildings and services
(consideration may need to be given at this stage of
Pilot scale beneficiation testing needs to be considered. For
the project to provision for future expansion of the ore
simple mineralogy, and established unit processes, pilot scale
processing facilities)
test work is usually not justified. Indicators of the need for
pilot scale testing include unusually complex mineralogy •• GA and plan/section drawings taking into account
and use of new or uncommon technology. Between a clear safety, constructability, operability and maintainability.
case for not including pilot scale testing and clear necessity The applicable CAD 2D or 3D model would be
lies a range of situations for which consideration would be considerably refined and optimised from that
given to time, cost and risk to arrive at a decision. As a general commenced at prefeasibility level
principle pilot scale testing hydrometallurgical processes •• mechanical and electrical equipment lists
must be considered as issues such as penalty element build- •• electrical load list
up and side reactions leading to scaling may not be apparent
•• data sheets and specifications for any critical long delivery
during bench scale testing.
equipment (eg grinding mills and large transformers)
Test work samples should be made available to equipment
•• MTOs and written quotation pricing for all commodities
vendors to enable equipment specification and pricing. These
will include, for example, settling test work for thickener •• estimated construction hours and construction labour
sizing and viscosity testing for pump selection. Samples of rates
test work residues should be retained for testing by the tailing •• detailed capital cost estimate in the WBS format to an
storage facility engineer. accuracy of no less than ±15 per cent
The cost of feasibility level test work will vary but the •• processing cost estimate to an overall accuracy of no less
following may be taken as a general guide for a large orebody than ±15 per cent
with three domains: •• engineering, construction and commissioning plans
Mineralogy US$100 000 and schedules including a quarter by quarter capital
Comminution US$250 000 disbursement schedule
Bench scale testing US$300 000 to US$400 000 •• plant and unit process performance guarantees
Vendor test work US$50 000 •• study report.
Pilot scale test work US$250 000 to >US$1 000 000 The design criteria and mass balance should provide for a
certain amount of variability however to accommodate short-
These allowances exclude the cost of sample collection and
term variability may be unjustified from a capital perspective
freight.
and this variability would be taken up in operations by
Feasibility process plant engineering should be awarded stockpile management.
to an appropriately qualified and experienced process plant
The process plant production and processing cost models
engineer in a process where tenders are evaluated on:
should be refined and updated to include:
•• ability to meet the scope and deliverables
•• ore processing and production schedule by ore type on a
•• experience in the type of facilities proposed quarter by quarter basis for at least two years following
•• price commissioning and semiannually thereafter
•• acceptability to proposed financiers (usually decided at •• metallurgical parameters determined from test work
the prequalification stage) (recoveries and product quality) for each ore type
•• quality of the proposed study team •• realistic ramp-up factors (recovery, plant availability
•• availability and timing and product quality)
•• location. •• labour costs from feasibility study manning schedule
and rates agreed with operations management
The project metallurgist plays a role in the engineering and
cost study by: •• consumables usage determined from metallurgical test
work results and process engineering (metal wear, power
•• timely provision of results of metallurgical test work
and reagent consumptions) and written vendor pricing
•• providing input to the plant design operating and
•• maintenance materials as a percentage of the feasibility
maintenance philosophies
capital
•• making process related decisions
•• energy costs from feasibility engineering, electrical load
•• participating in hazard and operability (HAZOP) studies list and written vendor pricing
•• initiating value engineering, if required •• services costs (eg laboratory, freight and consultants)
•• signing off key process documents from written vendor pricing.

During the feasibility study revision of the resource model and will review critical documents such as process design
may be required to take account of updated metallurgical criteria, mass balance and flow sheets and recommend these
parameters and processing costs. for sign-off for construction. Some additional metallurgical
Realisation costs should be updated based on: test work may be required for which sample material from
the preceding feasibility study test work would be used.
•• product quality determined from metallurgical test work
During the engineering phase facilities required for process
•• detailed transport studies and vendor written pricing
monitoring and control will be defined and there may be the
•• negotiated offtake agreements with product purchasers temptation to include all the control systems that might be
(for example, smelters). required. On the other hand, there will sometimes be pressure
The updated capital, processing and realisation cost by others in the owner’s team to remove control systems,
estimates will be included in the project financial model. The including sampling systems as the capital cost increases, often
project metallurgist should be proactive in reviewing the without undertaking any value engineering. One approach to
total model to ensure its completeness and accuracy from a resolve the potential conflict is to include all facilities that are
processing perspective and to reduce the risk of items being normally or typically included for the type of process plus any
omitted or double counted, particularly at the process plant/ for which there is demonstrated short-term economic value.
mine and process plant/infrastructure interfaces. In cases where uncertainty exists, allowances should be made
The feasibility study should include a project risk analysis in the estimate outside of the normal project contingencies
of risks associated with the delivery and operation of the for inclusion of these items post-commissioning and where
process plant. economic benefit can be demonstrated.
For planning purposes a minimum of 12 months should At a point where ~60 per cent of engineering has been
be allowed for completion of processing facilities feasibility completed, the estimate should be to an accuracy of no less than
studies. ±10 per cent, and becomes the control budget for construction.
ENGINEERING AND CONSTRUCTION REFERENCES

Cusworth, N, 1993. Predevelopment expenditure, in Cost Estimation
Engineering and construction is usually managed by an
Handbook for the Australian Mining Industry (eds: M Noakes and
engineering and construction company under, for example T Lanz) pp 252–259 (The Australasian Institute of Mining and
an Engineering, Procurement and Construction Management Metallurgy: Melbourne).
(EPCM) contract. This may include commissioning in
Morrell, S, 2009. Generating optimum value from ore characterisation
conjunction with the owner’s operations team. programs in design and geometallurgical projects associated
The quality of the preceding feasibility study notwithstanding with comminution circuits, in Proceedings Tenth Mill Operators’
the success of a project is very dependent on selection and Conference, pp 167–170 (The Australasian Institute of Mining and
management of, in this example, the EPCM contractor. As Metallurgy: Melbourne).
with the feasibility engineer, the EPCM contractor must be Noort, D J and Adams, C, 2006. Effective mining project management
selected on the ability to meet key performance criteria. A systems, in Proceedings International Mine Management Conference,
dedicated owner’s team with experience in project delivery pp 87–96 (The Australasian Institute of Mining and Metallurgy:
(as opposed to operations experience) would be formed to Melbourne).
support and manage the EPCM contractor. Warren, M J, 1991. Pre-feasibility and feasibility studies: a case for
For the study metallurgist the engineering phase in the improvements, in Proceedings Mining Industry Optimisation
Conference, pp 1–11 (The Australasian Institute of Mining and
development of a project commences the transition from
providing input to the final design and engineering of the
process plant to preparation for commissioning and operations. White, M E, 2001. Feasibility studies-scope and accuracy, in Mineral
Resource and Ore Reserve Estimation – The AusIMM Guide to Good
During detailed engineering the study metallurgist will Practice (ed: A C Edwards), pp 412–434 (The Australasian Institute
assist with equipment selection, layout and process control of Mining and Metallurgy: Melbourne).

Contents
Project delivery
G Lane1 and E Skinner2
ABSTRACT
Project delivery strategies are determined by project context, business case requirements,
scale of project, market conditions, owner’s capability and preference, contractor’s
experience, resource and skill availability, project location and type of project or facility.
Successful project management requires the setting of coherent, effective and
realistic business objectives and success criteria. The management team needs to lead,
prioritise, resource, align, plan, track and communicate the processes that enable the
realisation of these business objectives and success criteria.
For any given project, there is a unique balance between selection of a simple project
delivery strategy that minimises interfaces and a complex strategy that necessarily
engages the use of expertise and experience from numerous parties. The right balance
for a project between these ‘book-end cases’ is a function of how the project owner or
managing contractor defines its objectives and chooses to manage risk, cost, schedule
and project quality, in the context of the conditions affecting the project.
Selection of an effective project delivery strategy requires a detailed appreciation of
the range of external and internal environmental factors which shape and influence
the project. These factors include the project constraints imposed by the business,
market or operational environments and the selection of an effective mix of primary
(head contract), secondary (mix of vertical and horizontal packages) and tertiary
(type of individual contract) strategies. The selection of the primary, secondary and
tertiary strategies also shapes the project, the size and responsibilities of the owner’s
team and the framework for delivery of outcomes.
How does an owner or prime contractor decide? What is the optimum for a given
project? Those with operations management backgrounds will have different views
to those with project and construction management backgrounds.
Is the simplest practical approach the best approach? Each and every interface
needs to be managed.
The owner will want a very experienced team that is engaged directly with the
owner. The team will have a track record of successfully performing projects using
contractors skilled in their respective disciplines and selected using competitive
tendering on brilliantly designed scopes of work and good contract terms, with a fully
defined project scope and appropriate schedule, for which all significant risks have
been addressed and no scope changes are outstanding. Budget and schedule will be
expected to have been developed with a full understanding of the scope, situation,
risk management requirements and so on.
The problem is that these conditions rarely exist. The challenge in selection of a
project delivery strategy is to resolve how best to address the inadequacies between
the ideal case and the present reality. The mechanism for addressing the inadequacies
defines the likely outcome of the project.
INTRODUCTION
Project delivery strategies are often discussed in the context of the advantages and
disadvantages of the common primary strategies:
•• reimbursable engineering
•• procurement
•• engineering, procurement, construction and management (EPCM) versus lump
sum engineering
•• engineering, procurement, and construction (EPC) (for example Gabrielson,
2007; Hundertmake et al, 2008; Loots and Henchie, 2007).
1. FAusIMM, Chief Technical Officer, Ausenco These advantages and disadvantages usually revolve around the key criteria for
project delivery:
2. Vice President, Project Delivery, Ausenco. •• cost
Email: ed.skinner@ausenco.com •• schedule
163
•• quality. Decisions around the factors affecting the business case

Notionally, EPC provides cost security, but a poorly defined have a significant bearing on the investment decision, the
EPC scope can lead to additional claims. EPC contracts project delivery strategy, project timing requirements and
generally impose a quantum of risk on the contractor which subsequent operations.
may invite a significant cost premium. Poorly managed For any given project, there is a unique balance between
EPC contracts may see either the owner bearing more costs, selection of a simple project delivery strategy that minimises
or the contractor seeking to cut costs at risk of schedule or interfaces and a complex strategy that necessarily engages the
quality outcomes for the owner. EPCM has the advantage of use of expertise and experience from numerous parties. The
flexibility, transparency and quality control with the potential right balance for a project between these ‘book-end cases’ is
disadvantages of cost and schedule growth where project a function of how the project owner or managing contractor
definition and/or controls are inadequate. defines its objectives and chooses to manage risk, cost,
Beneath this primary strategy level there are a multitude schedule and project quality, in the context of the conditions
of options that determine the shape and outcomes of the affecting the project.
project. Selection of a complex packaging and delivery Post the business case planning phase, the commonly
strategy in the interests of say, cost economy or community applied options available to the project are summarised
engagement leads to an increased number of interfaces. below in the context of primary, secondary and tertiary
Increased interfaces lead to complexity for the owner and level strategies that address the head contract, management
contractors requiring higher levels of management and method and delivery approach, respectively.
control. This often provides fertile ground for interface
related claims from contractors. Primary delivery strategies
A small project using proven technology and applying At the primary level the delivery options are:
a simple project delivery strategy can be managed with a •• Owner self-perform – the project owner’s team directly
small team and simplified systems. Increased scale of project, manages and controls contractors or employees that
the introduction of one or more novel technologies or more complete the works.
complex interfaces often demand corresponding modifications
to team size and competency, and the complexity of systems •• Augmentation strategies – the owner’s team is supported
required to deliver the project. Interface management often by a specialist project delivery team to directly manage
becomes a key determinant of team size and project systems and control contractors or employees that complete the
for larger, more complex projects. Team capability needs to be works. The supporting mechanism may be one of the
matched by trust and accountability by the team to deliver a following:
large project successfully. •• Program management – adopts a holistic approach to
This paper discusses the project delivery options that are project delivery, with capabilities extending across
available to a project developer/owner. The various levels the complete project life cycle. Hence, program
of delivery strategy are summarised to provide context for management is used in the context where a contractor
how factors that are internal and external to the project drive takes on a broader role than just a project to deliver
strategy selection. multiple interrelated outcomes (Table 1). Program
management is sometimes also used to address a
Finally, some case study examples of strategy selection
series of concurrent projects which together form
processes are provided to illustrate both success and where
a program of works (eg complex sustaining capital
improvement is possible.
works projects). By contrast, ‘project management’
usually refers to a single project which can be delivered
PROJECT DELIVERY OPTIONS independently of any concurrent activities by an
Each project is unique, as are the factors affecting the owner and which for purposes of this paper, does not
business case for the project, the market conditions predicted extend across the project life cycle.
to prevail during the production life cycle and the owner’s •• Alliance arrangements – can result from the owner
circumstances. In forming a business case for a project, incorporating contractor’s personnel into the project
owners need to take a long-term view with careful and management team. True alliances result in ‘salt and
holistic consideration of all the factors affecting a financial pepper’ teams where project outcomes are shared on
investment decision. These considerations may include the a pain/gain basis. Pseudo-alliances can result from
following interrelated factors: ‘body-hire’ style models where the contractors risk
•• size and richness of the resource exposure is negligible.
•• costs of extraction of the resource •• Outsourcing strategies – the most common primary
•• return on investment (ROI) strategies:
•• market for the product •• EPC (fixed price) or lump sum turnkey (LSTK)
approaches are used to transfer the bulk of the project
•• production scale of the project
design and construction risk (and potential reward) to
•• costs of establishment of production facilities the contractor. To be successful, the realised rewards
•• sources of labour skills for construction and operation need to be significant for both parties and the scope
•• sources of construction materials and operating of delivery needs to be well-defined. LSTK projects
consumables may involve extended performance warranties over
and above typical EPC delivery where the contractor
•• external factors (government, community, competitors,
accepts responsibility for performance of the entire
stakeholders)
project as a functional unit. EPC delivery may include
•• time to market nearly all project scope of work, or subelements of the
•• long-term positioning of the owner’s operating cost project which can be adequately segregated and tested
profile in context of the market. for performance in isolation from the interfacing

Project delivery
Table 1
Comparison of projects with programs.
Project management Program management

Fixed duration Undetermined duration, or determined across a suite of deliverables which may have imprecise delivery timeframes
Defined objectives Defined/negotiated objectives accommodating a broader delivery framework
Task focused Goal focused
Focused on specific task/project Focused on multiple projects or across multiple phases in a single project life-cycle
Manager as overseer Leader as creative thinker in addition to manager as overseer role
Single deliverable Multiple interrelated outcomes
elements of the project. In such cases, contractors are Project framing considerations will determine logical
reluctant to accept risk of performance outside their selection of compensation methods. Common choices include
control. This classically arises where the feed products schedule of rates, cost plus and lump sum basis. The pros and
from one subelement of the production chain are cons of these approaches are provided in Table 2.
required in sufficient quality, quantity and feed rate to As seen in Table 2, increasing time risk for lump sum
performance test another system. compensation arrangements relates to time investment by
•• EPCM (fee for service) is the most common head the owner to achieve sufficient definition to support a lump
contract model used in the minerals business. All sum contract. Given that the lump sum contractor is accepting
services expended to meet the scope of services cost and time risk, lump sum compensation arrangements
defined in the head contract are reimbursable and represent an incentive for contractors to complete works at
performance warranties are typically diminished high levels of efficiency.
when compared with EPC contracts due to increased
Owners and EPCM contractors can choose to ‘self-perform’
owner input to engineering design and construction.
project works. This typically occurs:
EPCM delivery is generally for a scope that includes
the majority of design and construction works within •• when the owner or EPCM contractor has access to
a project. Some elements may be retained by the equipment and can perform certain works economically
owner, typically those related to mining and external (eg use of a mine fleet for bulk earthworks)
authorities. •• for minor works which may not justify costs of
equipment mobilisation and demobilisation
Secondary delivery strategies •• where contractors have their own integrated or related
Secondary level strategies relate to the method of managing workforce
the scope of services. The packaging strategy is tied to the
•• where contractors are unwilling or unable to accept the
primary delivery strategies and may be constrained to
risk associated with location, scope or supply.
respond to external factors (eg financing, equity partner
requirements, permitting, community requirements, time to Mixtures of the above and hybrid approaches are also
market) or internal factors (eg owner organisation, capacity, common. Table 3 illustrates the contracting approach used
level of project definition). Secondary level strategies can by a major copper-gold project. This model included a head
relate to front end engineering design (FEED), organisation contract with an EPCM contractor.
of early site works, or permitting. Activities may be tied to
the packaging approach and selection between vertically Tertiary strategies
integrated packages, horizontal discipline based packages Tertiary level strategies relate to the mix of methods of
and/or ‘EPCM self-perform’ delivery. delivery. These include:
Vertical packages are typically for blocks of work that •• design only
are supplied as complete systems, where the contractor •• supply (and fabricate) only
is required to perform all works from the ground up to
point of handover of a functioning asset. Examples include •• design and supply
accommodation villages, primary crushing stations, power •• design, supply and construct (D&C)
stations, technology packages or other similar elements •• construct only
of project that can be effectively designed, procured and
•• hybrid types.
constructed as a packaged scope of work. Vertical packaging
is viable where the contractor can economically perform an An example of the mix of delivery methods for a large
entire package of works and operate relatively independently copper concentrator within an EPCM contract and using
from other interfacing works. various horizontal, vertical and EPCM self-perform delivery
models is provided in Table 4.
Horizontal packages represent a practical solution where
quantities are large and/or scope overlaps between packages The selection of the primary, secondary and tertiary
provide a viable case for deploying contractors geared for strategies shape the project, the size and responsibilities of the
performance of works on a bulk scale. Examples include owner’s team and the framework for delivery of outcomes.
site survey, bulk earthworks and drainage, roads, concrete, At the tertiary level, the owner may choose (for a range of
steelwork, site erected tanks forming part of a broader reasons), to combine supply and fabrication or fabrication
system, electrical bulk scopes or combination works such and installation scope, free-issue equipment and/or materials
as steelwork, mechanical equipment, platework and piping by the owner or the head contractor, and a myriad of other
(SMP) or bulk electrical and instrumentation packages (E&I). procurement, fabrication and contracting options.

Table 2
Pros and cons of contract types.
Contract type Cost plus Schedule of rates Lump sum

Positives Can begin very quickly Certainty on unit rates (for example, $/t steel, Certainty of price
$/m3 concrete etc)
Can commence quickly while design is being
completed
Negatives • no certainty of price • owner takes the risk on quantities Needs a higher level of engineering and
• no incentive by Contractor • must define what each rate includes (for example, scope definition prior to contract bid
to minimise cost or time for concrete)
• detail excavation
• formwork
• rebar
• concrete
• hold down bolts etc
Increasing cost risk
Increasing schedule risk
Table 3 •• market conditions (revenue side considerations)

Example of contracting strategy – used for a major copper-gold project. •• financing requirements
•• context and location
Area Type
•• external environmental variables with potential to affect
Water storage Vertical, D&C Fixed price project delivery
Concentrate and return water lines Horizontal Fixed price •• market conditions (availability of owner/contractor
Tailings dam Direct hire/sub-contract resources/competition for these resources, offshore versus
Site preparation Horizontal Fixed price on-shore supply, stick build versus modular construction)
•• technical factors
132/33 kV substation Vertical, D&C Fixed price
•• operational needs
HV distribution Vertical, D&C Fixed price
•• project scope
Administration building Vertical, D&C Fixed price
•• level of definition around basic data (process material
Store and workshop Vertical, D&C Fixed price quality/morphology, services, geotechnical conditions,
Field erected tanks Horizontal Fixed price permitting etc).
Primary crusher facility Vertical, D&C Fixed price All the above can be quantified and valued as inputs into
the assembly of the delivery strategy. Additionally, several
Conveyors and pebble crusher facility Vertical, D&C Fixed price
known but unquantifiable factors may be significant in
Concentrator area concrete Horizontal Fixed price strategy selection. ‘Unknown unknowns’ are generally
Concentrator structural steel Horizontal Fixed price considered as part of contingency assessment for the project
and do not form part of a packaging strategy. The strategic
Mills installation Horizontal Fixed price
element in conceptualisation relates to the ability of the team
Mechanical – piping installation Horizontal Fixed price to identify the issues (and opportunities) that are impacting
Concentrator electrical and control Horizontal Fixed price and/or may impact on the project, to diagnose the causes, to
develop solution frameworks and to map these against the
Tailings and water pipelines Horizontal Fixed price
available options.
Concentrate dewatering facility (Blayney) Vertical, D&C Fixed price On some occasions project owners/managers/directors
Process water ponds Horizontal Fixed price come to a project with experiences (‘baggage’) that
Site access road Self-perform predetermine the strategy for the project. Some prefer the self-
perform approach, some EPC, others favour alliances, and
still others choose hybrid approaches that have worked in the
SHAPING FACTORS THAT INFLUENCE THE STRATEGY past. Whilst experience from past projects is highly valuable, a
The selection of the best delivery strategy depends on the healthy project climate is one in which decisions are based on
significant factors affecting the project and the strategic facts and robust debate occurs around appropriate strategies
elements related to the assembly of the selected solutions. to address the facts.
Early attention to shaping factors drives project stability
(Merrow, 2011). The early sharing of knowledge regarding
Strategy complexity
The size of the combined owner’s and EPCM teams will be
shaping factors and structure facilitates coherent alignment
driven by the complexity of the strategy and the accompanying
among the project delivery team and stakeholders.
management systems and protocols.
Significant factors include: Simple contracting strategies need a high level of clarity,
•• project business case accountability and trust to be successful. The advantages

Project delivery
Table 4
An example of a mix of contract types for a large copper concentrator.
Activity Facility
Primary O/L Stockpile Grinding Pebble Flotation Thickening Moly plant Concrete HV power Tailings Water
crusher conveyer and reclaim crusher handling supply piping supply
Site preparation C1
Engineering C6
Concrete C7 C8
Steelwork – C9 C10
fabrication
Steelwork – C11
installation
Site tanks C12
Platework – C13 C14
fabrication C4
C2 C3 C5
Platework –
installation
Mechanical
installation C15 C16
Piping – installation C16A
Piping – spool fab C17
Electrical and C18
control
Major equipment Free issue to contractors
and disadvantages of simple and complex strategies are a project containing relatively novel technology on a lump
summarised in Table 5. sum basis and completed engineering in several engineering
Larger (owner) companies attempt to manage risk through hubs. This mix of simple head contract and complex
complex systems and, on some occasions, complex management engineering delivery, combined with the novel technology
structures. If not well managed, these risk management caused the project substantial hardship in construction and
strategies can hinder project success due to the inherent lack early operation.
of clarity and accountability, combined with complex decision-
making and approval processes. The decision process becomes Owner capacity and structure
defensive rather than balanced and accountable. Unless an owner’s business is structured for planning and
delivery of successive projects as occurs in the construction
An example of simple contracting strategies working
industry and elements of the minerals and resources sectors,
successfully is in the West Australian goldfields where the
several considerations arise for an owner embarking on a
construction of small gold plants has predominantly been
development project:
based on simple LSTK or EPC contracts to experienced small
to mid-tier engineering companies. There were relatively •• What is the execution of the works program ‘core
few failures in this model in a very competitive market place business’?
(Close, 2002). •• What are the requirements of investors/partners?
One example where an EPC contract failed to deliver to •• What is the legacy requirement for the business after the
expectations, in the same geographical area, was the Murrin work is complete (eg completed asset or ongoing project
Murrin Nickel Project (Taylor, 2000) where the contractor bid delivery capability)?
Table 5
Summary of advantages and disadvantages of simple and complex strategies.
Packaging Advantages Disadvantages

approach
Simple Owner’s team can be smaller where the level of confidence in If the project is complex, higher levels of risk are assumed by the contractor and higher levels of
contractor delivery is high. cost are implicit in contractor’s price.
Low levels of complexity in interfaces lead to leaner control A contractor who fails to perform their obligations represents a risk to the owner, which may
structures. be difficult to address.
Very successful with experienced contractors. Not suitable for novel technologies.
Complex Opportunity for owner to more significantly affect cost and risk Higher numbers of interfaces invite commercial complexity and higher levels of complexity in
outcomes to potential advantages. control interfaces with consequent increase in project management systems and resources.
Opportunity for owner to share work where a contractor may be Owner needs to invest more in up front definition work to mitigate commercial exposure to
under-performing. claims arising from poorly defined scope interfaces.

•• How well can the owner’s team perform the works? •• adequacy of resources, utilities and services required for
•• What is time criticality? the project
•• What is the level of outsourced support required? •• governance and security around achievement of the
project deliverables in accordance with the business
•• Management of delivery risk.
case, cost and time parameters
Secondary considerations include:
•• tranched investment structures which are tied to
•• What organisation structure is required to manage the achievement of certain milestones (approvals, permits,
project delivery and adequately deal with risks the progress), capacity of contractor/supplier market to
owner envisages could arise? deliver equipment and services in context of the project
•• What technical requirements and controls should be program
applied? •• level of certainty around repayment (strength/depth
•• What are the key implementation risks and how should of market, probable fluctuations in revenue, defensive
they be managed? measures against unforeseen events)
•• The impact of external factors on the project and how •• controls around surety of contractor performance, cash
best to manage them. flow and milestone payments
•• What processes, procedures, tools and systems are •• insurance conditions imposed on the project delivery
required? •• interface/external environmental risk management
•• Management of quality. •• cost implications arising from key risks occurring
Gearing an organisation to deal with project delivery (sovereign risks, schedule slip, commodity movements,
consistently represents a substantial commitment on the labour/resource costs.
part of the owner. The sustainment of this organisational Figures 1 and 2 illustrate the linkages between project
capability between projects represents a carrying cost which financing and the contracting strategy. Figure 1 compares
must be traded off against alternative strategies. Common simple and complex financing arrangements and Figure 2
considerations in organisational planning are listed in Table 6. provides a simplified view of the payment interactions
Two examples of owners who have successfully delivered between financing and contracted services and goods. As
self-managed projects are First Quantum and PanAust. Both the financing structure becomes more complex, the owner’s
ability to prescribe the contracting strategy reduces and the
have outsourced engineering and self-managed construction.
financier’s influences increase.
The success of this methodology is linked to the ability of the
owner’s project team to maintain a project focus and resist Export credit finance may represent an attractive source of
operational mindsets (Lane and Clements, 2012). funding for projects incorporating international contractor
content. Export credit finance arrangements are usually linked
Financing structures to governmental incentives from the contractor’s country of
origin and commonly are tied to incorporation of goods and
Capital requirements are significant for large projects. Where
services in the export contract from the contractor’s country
an owner requires recourse to external funding for project
of origin. Terms are generally attractive for projects in that
delivery, conditions may be imposed by equity partners or
the financing terms are generally low interest, long-term and
lenders on how the project is shaped. Typically these will be
the opportunity to leverage imported goods and services may
related to:
represent an attractive alternative to locally sourced goods
•• capacity of owner to succeed in its objectives and services. A disadvantage of this form of finance is the
•• confidence in business case and factors critical to the time it takes to secure arrangements with the relevant foreign
business case governments via the selected contractor.
•• off-take sales contracts secured
Insurance
•• proof of concept where novel solutions or technologies
Risk appetite varies between owners and is a function of the
are proposed (eg prototyping, ore processing trials)
owner’s capacity to absorb costs in respect of unplanned
•• permitting and required approvals in place outcomes affecting a project. The addition of equity partners
•• community and governmental support for the project and lenders in a project financing strategy imposes additional
Table 6
Common considerations in organisational planning.
Considerations Owner Self Perform Integrated Team PMC EPC EPCM Construct Only
Level of owner control Full Shared Shared/devolved High on owner controlled facilities/services,
lower on execution of works
Time to build/develop Slow Slow to moderate Fast Mgmt team required to overview contractor
team (Owner/Integrated/PMC)
Procedures Develop in house Build/adapt from contractor Adapt from PMC Owner/ Owner/ Owner/
standards Standards contractor contractor contractor
Credibility for external Low unless proven track Depends on calibre of contractor Higher Higher Higher Depends on level of project
investors record and team org structure definition and package structure
Cost of team Lower Higher Higher Blend of owner's mgmt team and contractors mgmt team indirects
Holding cost of team Higher Medium Lower (flexible pool) Contractor risk Contractor risk Contractor risk

Project delivery
FIG 1 – Simple and complex financing instruments.
FIG 2 – Sample payment release process incorporating banker’s engineer.
complexity around the conditions required to secure financing A typical project insurance scheme as used on the Goro (New
for a project. Caledonia) and Telfer (WA) projects is shown in Figure 3.
An owner will need to balance the amount of cost it bears
in context of an insurable event, against how much risk a
Stakeholders and local content
contractor and a third-party insurer would bear. On large The owner is often developing and operating an asset which
projects it is common to engage a broker for advice around a risk will have a macro effect on local communities for a significant
period of time. Large projects can substantially distort factors
management framework as package conceptualisation occurs.
affecting employment opportunities, compete with or disrupt
This assures that contracts are structured with a considered
local business and community, create demands on community
position around leveraging contractor standard insurance
infrastructure which did not exist prior to the project, introduce
programs as an integral part of the overall risk management nuisance factors related to emissions/effluents, noise and dust,
program the owner will assume. Early engagement of the competition for local resources, create disruption of cultures
underwriter market is often cost beneficial as costs of the overall and/or religious traditions, create fly-in fly-out (FIFO) cultures
risk management program and their effect on the project can be where they may not have previously existed.
tempered from the pretender period. Conversely, the owner is also faced with needing to assure
Owners typically need to think horizontally in terms of itself of adequate access to utilities, services, transport
phases of work (supplier works insurance, marine transit, infrastructure and skilled/unskilled resources required to
construction, commissioning and operations phase) and sustainably operate the production assets.
vertically in terms of the classes of cover required to form an A sensitivity analysis of the needs and wants of stakeholders
adequate ‘wrap’ across the project. and affected communities is essential to sustainable project

FIG 3 – Project insurance scheme.
construction and subsequent operations activities. Early a process of assessment as to the true progression of key
conditioning of the project to engage the affected groups in an activities and will assure that resourcing and expenditure is
empathetic and constructive manner and an ongoing positive matched to activities most necessary to progress the project
engagement with these groups is essential to good project and keep within the cost and schedule parameters. On the
governance and leads to reduced possibility of disruption and first stage of the Goro nickel project, permitting requirements
related costs. were substantially underestimated and engineering required
For the Telfer gold project in WA, care was taken to genuinely to support permit applications was intensive. Construction
engage local communities early through a combination of activity at the site commenced early and the unexpected
Indigenous land user access agreements, positive engagement delays in engineering delivery resulted in cost blowouts
associated with construction activity until the project was
strategies through a local community engagement plan
suspended and the costly process of demobilising contractors
against which contractor performance was measured and
and re-assessment of the ‘go forward’ plan was completed.
active engagement plans affording local communities early
access to information concerning opportunities to participate
in the project. For the Goro project in New Caledonia, a range DELIVERY STRATEGY SELECTION
of Indigenous community groups and a small number of How does an owner or prime contractor decide which project
monopolistic businesses heavily influenced project delivery/ delivery strategy is most appropriate? What is the optimum
contracting strategy. Additionally, regular engagement for a given project?
sessions with the local community and government were Personal experience and personality plays a significant
required in addition to periodic reporting on local contractor role. For example, those with operations management
engagement to the territorial government. backgrounds will have different views from those with
project and construction management backgrounds (Lane
Permitting and Clements, 2012). Financiers want certainty of outcome
Permitting is integral to the selection of the contracting (cost and schedule).
strategy. Permit applications in Australia typically involve Major mining companies tend to be protocol driven and risk
federal, state, local government and often land-owner averse. Their businesses are often sufficiently large that they
tiers, with various classes of construction phase permits have a macro effect on the regions or even countries in which
required. Large projects in Australia commonly require they operate. They are therefore burdened with complex
several thousand permit applications and the processes can regulatory arrangements and protocols around their systems
be lengthy. Often traditional-owner groups will also impose of business, financial arrangements, community engagement,
certain conditions on the project and the negotiations can be project delivery and operations which may not apply for a
lengthy and cost intensive. Whilst environmental and access junior mining company.
controls vary in jurisdictions outside Australia, the permitting Junior mining companies are often capital constrained and
process is often of similar complexity. more accepting of risk. By contrast with the major miners
It is important to consider the timing of activities for project who often have established finance facilities and large asset
delivery and to periodically take stock of true project progress bases against which to leverage credit, junior miners are more
(toll-gating reviews). A measured and gated approach forces likely to need to take a flexible approach to project financing

Project delivery
and the concessions such arrangements may impose. Equity •• the capacity of each contractor, based on shop visits and
partnerships are more common with the junior mining houses discussion with existing clients (some local contractors
and often a mix of credit and debt facilities are required. oversell their capacity)
The challenge in selection of a project delivery strategy is to •• track record for similar works
resolve how best to address the inadequacies between the ideal •• capacity to innovate/add value for the owner
case and the present reality. The mechanism for addressing the •• financial strength, including recent commercial history
inadequacies defines the likely outcome of the project.
•• other issues that may impact on cost and schedule such
It is therefore beneficial to keep an open mind in the early as location, rates etc.
stages of project definition with regard to delivery strategy
and solutions. As project definition improves, options for Contracting strategy
delivery become constrained to the point (usually when a There are four key components to developing a viable
financial investment decision is made) where the project contracting strategy for a project:
execution plan can be finalised.
1. understand contractor capability
Good projects are characterised by solid basic data, high 2. determine the capacity and regulatory constraints to
levels of definition around the factors affecting project delivery, mobilising suitable contractor resources (which may
a robust design with a well-defined production system and include resources from other countries)
aligned and appropriately motivated delivery resources among
3. deeply understand the project
the owner management and contractor teams. Projects which
are founded on a solid business case, a well-articulated delivery 4. develop a work breakdown structure based on the above.
plan and a skilled and experienced team which is encouraged
to work within the project plan and regularly challenge the Local contracting capability
project to deliver on and improve the factors underlying the One of the deliverables from the feasibility study for a large
business case are likely to be more successful. project should be a contractor capability report that clearly
defines the contracting capabilities of local fabrication and
As an example, consider a 5 to 8 Mt/a concentrator. An construction contractors with respect to the following:
established mining house may have sufficient income/
•• type of contractor by discipline, eg concrete, steel
assets to substantially self-fund the project, perhaps with
fabrication, plate work, electrical etc
minor assistance from banks or forward sales agreements.
As the banks would have ample surety, the owner is free to •• the capacity of each contractor, based on shop visits
determine the project delivery strategy which would typically and discussion with existing clients (local contractors
oversell their capacity)
be EPCM where the owner can maintain substantial control
over the project with a modest project team, supported by the •• the ability of the contractor to meet QA/QC
corporate and operations team as needed. requirements, based on shop visits
A junior explorer would need to interest other parties •• financial strength, including recent commercial history.
through equity arrangements, but maintaining control would Transport and logistics should be included in the contractor
require substantial financiers’ support. This support would framework.
be provided at increased risk and necessitate that the owner The above data is required to assess whether/how local
passes as much risk as possible on to supporting contractors. contractors can support a large project and to define the
This scenario leads to LSTK or EPC contracting arrangements requirement for external contractors (higher cost). Note that
involving performance warranties that mirror, in part, the the use of external contractors as opposed to local contractors
financiers’ needs (Lane et al, 2007). As well as derisking the needs to be justified and the basis of the justification explained.
project for the financier, the owner’s team is smaller and the
contracting environment is simpler (if project definition is External contractors
sufficient). This approach can be extended to single sourcing External contractors may be sourced from neighbouring
equipment and contractors where existing relationships countries or overseas. For example, for projects in South-East
add value to the project (shorten schedule and increased Asia, heavy steel fabrication may be sourced out of China,
confidence in delivery). Thailand, Vietnam etc.
EPCM and LSTK approaches considerations for a 5 Mt/a Some external sourced contractors may have local offices, for
project are summarised in Table 7. example electrical contractors, and may source their workforce
from a combination of national and international sources.
Once the local and external contracting opportunities have
CONTRACTOR SELECTION been identified, the project manager will be able to make an
Definition of project parameters is typically undertaken assessment of the most effective contracting strategy for the
during a feasibility study. The project activities are defined on project.
an area and activity basis to allow a contracting breakdown
structure to be prepared. Thus, the key role of the project
CONCLUSIONS
director/manager is to ensure that all scope items are
Successful project management requires the setting of
identified and key issues, critical paths and required actions
coherent, effective and realistic business objectives and success
are clearly defined.
criteria. The management team needs to lead, prioritise,
The contracting capabilities of local and other pertinent resource, align, plan, track and communicate the processes
fabrication and construction contractors need to be defined that enable the realisation of these business objectives and
with respect to the following: success criteria.
•• type of contractor by discipline, eg concrete, steel Success requires a viable business case coupled with clearly
fabrication, platework, electrical etc articulated project delivery strategies. The project delivery

Table 7
Engineering, Procurement, Construction and Management (EPCM) versus Lump Sum, Turn Key (LSTK) (self-perform) for a 5 Mt/a concentrator.
Parameter EPCM LSTK (self-perform)

Owner’s team size Up to 10 Up to 5
Owner’s team capability Sufficient to report to owner’s corporate management and Sufficient to deal with variations in contract and monitor
provide technical input/review adherence to scope
Contract complexity Higher, primary, secondary and tertiary contracts and Low, only one contract if head contractor has construction
associated interfaces delivery capability
Suitability for simple technology OK OK
Suitability for novel processes More suitable Less suitable
Cost structure Reimbursable Fixed price with variations
Contractor’s fee structure Up to 3% direct costs plus multiplier on cost of Hidden, but typically cost of employment plus 12% to 25%
employment of direct costs
Process/performance risk Process risk borne by owner Additional fees for process risk
strategies are the product of adequately funded project Lane, G S, Davis, M, McLean, E J and Fleay, J, 2007. Performance testing
definition processes with respect to basic data, external – when, what and how?, in Proceedings Project Evaluation Conference
environmental variables, process requirements, resourcing, (The Australasian Institute of Mining and Metallurgy: Melbourne).
community requirements, capital and operating funds. Loots, P and Henchie, N, 2007. Worlds Apart: EPC and EPCM Contracts:
Risk issues and Allocation (Mayer Brown: London).
For any given project, there is a unique balance between
selection of a simple project delivery strategy that minimises Merrow, E W, 2011. Industrial Mega Projects, Concepts, Strategies and
interfaces and a complex strategy that necessarily engages Practices for Success (Wiley: USA).
the use of expertise and experience from numerous parties. Taylor, A, 2000. The outlook for the PAL process [online], in Proceedings
The balance between these ‘book-end cases’ is a function of World Nickel Congress, Melbourne. Available from: <http://www.
how the financier, project owner or managing contractor altamet.com.au/wp-content/uploads/2012/12/The-Outlook-for-
defines its objectives and chooses to manage risk, cost, the-PAL-Process.pdf> [Accessed: 1 August 2017].
schedule and project quality, in the context of the conditions
affecting the project. GLOSSARY OF TERMS
An experienced project delivery team with a proven Alliance – where the owner and contractor work as a common
track record on similar projects is essential for successful team to achieve performance outcomes agreed using open
delivery of large projects. Projects with the best levels of and transparent decision-making processes and shared goals.
definition and coherence of vision among the team will still A good application of alliances arises where project definition
generate a wide range of unforeseeable challenges through is poor, or where the capacity of the parties to segregate and
the execution process. Teams able to rapidly and effectively allocate risk is poor. In such cases, alliances seek to draw
identify and mitigate adverse conditions or risks are essential upon the best resources of the owner and contractor teams
to minimisation/avoidance of consequent delays and costs. to achieve outcomes required for achievement of the project.
In general, simple contracting strategies are adequate D&C – A variation on the EPC formula where
to deliver smaller projects of known technology and low conceptualisation of the scope may be provided under a
complexity. This approach should be extended to single separate contract and the contractor’s responsibilities are
sourcing equipment and contractors where existing limited to design finalisation and supply/construction and
relationships add value to the project (shorten schedule and usually, commissioning duties.
increased confidence in delivery).
E&I – Electrical and Instrumentation – a typical aggregation
Complex delivery strategies should be carefully articulated of construction services provided by contractors.
and managed. They require higher levels of interface
EPC – Engineering, Procurement and Construction – these
definition and control and, hence, more management,
contracts place responsibility for delivery of the scope and
higher cost and result in greater ‘churn’. However, they do
deliverables in the hands of the contractor with minimal
offer the ability to include specialist (owner’s, contractor’s
owner interface.
and consultant’s) expertise and capabilities that can add
substantial value to the overall project. EPCM – Engineering, Procurement, Construction,
Management – delivery of scope and deliverables is placed
with the contractor however owner involvement can range
REFERENCES from minimal to significant levels. Where owner involvement
Close, S E, 2002. The Great Gold Renaissance (Surbiton and Assoc: Melb). is significant, it can be challenging for owners to hold the EPCM
Gabrielson, A, 2007. Current trends in project delivery, CIM contractor to account for adverse outcomes with the project.
Bulletin, 2(7). FEED – Front End Engineering Design – typically follows
Hundertmake, T, Valle Silva, A O and Shulman, J A, 2008. Managing a feasibility study as the lead in phase to project engineering.
capital projects for competitive advantage, McKinsey Quarterly,
LSTK – Lump Sum, Turn Key – such projects are generally
June.
designed around a contractor delivering an entire production
Lane, G S and Clements, B, 2012. Operations versus projects – how do system which must meet performance requirements across
people think and what are the implications? in Proceedings 11th multiple subelements.
AusIMM Mill Operators’ Conference (The Australasian Institute of
Mining and Metallurgy: Melbourne). SMP – Structural, Mechanical, Piping – a typical aggregation
of construction services provided by contractors.

Contents
Mineral project management – a perspective

from four decades in the industry
J S Dunlop1
ABSTRACT
At its most basic level, minerals project management may be broken down into a series of
logical chronological phases, described in this paper as: exploration, discovery; scoping
and prefeasibility; feasibility and approval; project financing and commissioning;
operation; expansion and closure. Each phase has its own peculiarities and challenges,
any one of which can potentially upset the forward progress of a minerals project.
Reference is made to each of these project phases based on the author’s many years of
firsthand experience, to identify issues within the project management spectrum where
additional caution may be warranted. Left unaddressed, these issues can cause undue
delays in the project timeline, ultimately delaying approval at significant project cost.
Finally, reference is made to the Mine Manager’s Handbook (MMH), published by
the AusIMM. This is an excellent project management resource and is considered
to be very relevant for minerals treatment plant managers, whether the context be
constructing, operating, expanding or shutting down.
INTRODUCTION
We all recognise that minerals projects don’t just happen. They usually span years
from discovery to stable production, during which time many obstacles arise,
each having to be overcome whether they be technical, economic, environmental,
regulatory or various external factors. It is not surprising, therefore, that many
projects do not survive this process. Given that mineral deposits are becoming more
elusive and either deeper or harder to find, logic suggests that worthwhile projects
should not be shelved simply because they could not overcome the manageable
post-discovery hurdles.
Care therefore needs to be taken to identify the possible project ‘stoppers’ and manage
them effectively before they materialise. The concept of project phases is developed in
this paper with appropriate reference to previous publications on this subject.
The identified chronological phases of a mineral project are frequently referred
to in published literature. For example, Noort and Adams (2006) refer to project
phases in relation to the feasibility study process. Van der Merwe (2008) presented
a very similar view. In this paper, a somewhat broader view is taken, commencing
with a mineral deposit discovery and ending in mine closure. Each project phase is
summarised in the list below. They are expanded on by some personal observations
under each section which follows.
Where relevant, reference will also be made to previous MetPlant keynote addresses
on this subject, with the aim of providing continuity with some of the excellent
previous papers which touch on some of the themes discussed in this paper.
The project development phases discussed here are as follows:
•• exploration
•• discovery
•• scoping, prefeasibility and definitive feasibility
•• project approval
•• project financing and commissioning
•• operations
•• expansion
•• project closure.
Additional observations will be made in relation to:
1. FAusIMM(CP), Chairman, Alliance Resources
Limited, El Arish Qld 4855.
•• capital markets
Email: jsdunlop@bigpond.com •• government understanding of the minerals industry.
173
J S Dunlop
EXPLORATION specialists in this area should be added to the management

team as soon as the feasibility process seems inevitable.
Reference here to mineral exploration is made in the general
context of exploration activities carried out prior to the making
of an economic mineral discovery. More specifically, reference SCOPING, PREFEASIBILITY AND DEFINITIVE FEASIBILITY STUDIES
here is to the exploration budget and how to manage it. The feasibility stages of project development are well
From a project management perspective, the questions often understood and will be reiterated upon here. Noort and
asked are, ‘How much should we spend and over what period?’ Adams (2006) provide a very appropriate refresher which
This issue often arises at the annual budget approval process. includes orders of accuracy for each stage of the study
A theoretical approach to these questions was proposed by advancement towards final or ‘definitive’ status. Table 1
Binon (1981) who described how to arrive at a budget figure depicts the development study process, in terms of option
after establishing an agreed profit level and time frame. identification and estimation accuracy.
In actual practice however, the theoretical or optimum Aspects of the 2012 edition of the Joint Ore Reserves
budget may not be available due to a shortage of funds, Committee (JORC) Mineral Resource and Ore Reserve
especially for small capital explorers in a depressed share reporting code (JORC, 2012) are relevant here, and readers are
market. In such a situation, the above questions remain directed to clauses 37–40 of the new code.
unanswered. This author has found it useful to start with the The use of discounted cash flow (DCF) methods to arrive
same objective (deposit size, type etc) and then to assign an at a net present value (NPV) of a project during the scoping
acceptable maximum exploration cost, such as unit cost/oz and prefeasibility stages is common practice in the minerals
of precious metal or unit cost/t of economic mineral resource. industry. If such estimates of project value are used in these
In this way, the project manager may determine the point stages, it is likely that they will be based primarily on Indicated
where acquisition may be a cheaper alternative to grass roots and Inferred Resources (as was the case in the Paddington
exploration. For example, if the exploration budget implies a example above). Such estimates should be for internal use
gold discovery cost exceeding $50/oz (in the ground), it might only and not for external or public reporting.
be cheaper to acquire already discovered mineral resources. The new code requires a prefeasibility study to have been
Unit discovery cost benchmarks are also recommended completed before an Ore Reserve may be reported (clauses
as a management tool when testing the effectiveness of an 38 and 39). It follows from this that a DCF analysis could be
exploration campaign. reported along with such an Ore Reserve announcement. Still
being debated, however, are the conditions wherein Inferred
Exploration funding can also be managed by introducing a
Resources should be included in the production schedule
joint venture (JV) partner to assume exploration costs by way
underpinning the DCF model. In a previous MetPlant
of a ‘farm-in’. This subject will be addressed later in this paper.
keynote, Card (2011) provided extensive reference to the
AusIMM’s emerging economic modelling guidelines.
DISCOVERY It is the author’s view that Inferred Resources only be used in
A mineral deposit discovery is the natural (though seldom DCF-based project evaluations under the following conditions:
achieved) end point of a mineral exploration program. When
•• scoping studies – not at all
a discovery is made however, additional management issues
can arise. For example, is the deposit immediately economic or •• prefeasibility studies – in conjunction with (and following)
are there remaining issues (such as metallurgical) unknowns? Measured and Indicated Resources, but only to assess
Is the market fully informed (as the discovery is clearly ‘upside cases’ and not be included in the ‘base-case’
material)? What are the next steps? (Some stock exchanges, for example Hong Kong and
Singapore have strict rules concerning inclusion of
Frequently, the market loses sight of the lead time to bring Inferred Resources in DCF models. Hong Kong forbids
a project to production, with the consequence that it loses their use entirely)
interest, further impeding the availability of capital required
for mineral resource delineation, let alone feasibility. •• definitive feasibility studies (DFS) – as for prefeasibility
studies.
A useful perspective on this can be gained from a paper by
Duncan (1985), who described the development plans for the In the case of mineral project valuations under the VALMIN
Code (VALMIN, 2005), Inferred Resources should similarly
Olympic Dam discovery at Roxby Downs, some ten years
only be included in ‘upside cases’ and not be included in the
after the initial discovery hole, (RD1) was drilled in 1975.
‘base-case’ evaluation (The code was updated in 2015, and
Another worthwhile example would be the history of the clarified this point, tending to support this interpretation.).
Paddington gold mine in Western Australia, still working at the
Another potential issue arising from the 2012 JORC Code
time or writing. The early history is described by Nice (1986):
is centred on clause 51. This clause refers to the inclusion
the discovery was made in 1981, the decision to develop in 1984,
of ‘in ground’ mineral value in public reports. Such in situ
based on Measured and Indicated Resources of 5.6 Mt which
or in ground financial valuations must not be reported
later were drilled out to reveal an orebody of 8.4 Mt @ 3.2 g/t
by companies in relation to Exploration Results, Mineral
(based on a gold price of $310). The plant was commissioned in
1985 at a throughput rate of 850 000 t/a. Being a less complex
Table 1
and smaller project to Olympic Dam, the project lead time from
discovery was approximately five years, 18 months of which Project development phases (after Noort and Adams, 2006).
was taken up by approval and construction.
Study phase Scoping study Prefeasibility Definitive FS
This latter point is significant in the light of today’s far
Iterations Multiple 2–5 1 or 2
slower project approval/construction timelines. It has been
this author’s experience that stakeholder expectations (not just Accuracy ±30–50% ±20–25% ±10%
the public and the local community, but also the regulators, Contingency ~ 25% ~15% ~1%
permit issuers and engineers) need to be actively monitored
and addressed both widely and early. For this reason, Options Full range 2–5 1–2

Mineral project management – a perspective from four decades in the industry
Resources or deposit size. Whilst this is well overdue, a Unfortunately, some benchmarks may not be known, such
potential problem arises in relation to compliant use of as the required liquidity ratios, hedging levels and debt
the VALMIN Code for mineral deposit evaluation. In this service requirements of the lender. Ultimately, these gaps
instance, ‘in ground’ mineral value is still used as one of the may materially reduce the debt level that the project can
rules-of-thumb, described by Lawrence (1994). This author achieve. In a tight capital market, this factor alone may be a
argues that in situ value still be retained as a valid rule-of- potential project stopper.
thumb in mineral asset valuation work, but not in the context
of mineral resource reporting under the JORC Code. Feasibility studies and joint ventures
Readers will be familiar with the very common joint venture
PROJECT APPROVAL ‘earn in’ condition, referred to in joint venture agreements
Whether a project advances to the ‘go-ahead’ stage is largely as the ‘completion of a bankable feasibility study’. Despite
dependent on three factors: what this author considers to be a professionally sound
understanding in the minerals industry as to what a BFS
1. the inherent techno-economic strength of the DFS,
requires, there has been much JV litigation resulting from
primarily through its financial argument
farm-in partners ‘passing off’ feasibility studies which were
2. the completeness of the DFS, often referred to as its claimed to be, but later shown not to be bankable.
‘bankability’
It is therefore recommended that where JV agreements in
3. the speed of its approval by the project owners and the future require this sort of clause, the agreement should
regulators. set out clearly the minimum standards for bankability status
These will now be discussed in the following subsections. to be achieved, and preferably with reference to an agreed
independent third party to arbitrate, should there be no
Feasibility studies agreement.
The content of the DFS is not usually the cause of a project Readers will be familiar with the term ‘fast-tracking’ where
delay. The most likely causal factor is the quality of the feasibility study preparation, engineering and approval are
content. No attempt will be made here to discuss the optimal integrated to reduce project commencement delays. Often, this
or recommended DFS format. A detailed outline of this may situation results from a deadline of some sort to be met, such
be found in the MMH (AusIMM, 2012). as tenure of the project tenements or logistic or climatic issues.
Feasibility studies can be weakened by two critical factors: An excellent account of fast-tracking was prepared by Hinz
excessive contingency and failure to recognise and satisfy and Aseervatham (1999), citing Rio Tinto’s experience with
the critical ‘bankability’ issues. The first of these two factors its Hail Creek, Baranji and Gokwe projects. The compression
has two components: of tasks which results from fast-tracking increases the overall
1. excessive conservatism in ore grade, productivity and risk of a successful project risk and should, in the author’s
costs view, be avoided if possible.
2. layer upon layer of uncontrolled contingency. Project approval

The second of these factors comes into play when the DFS Project approval in this context has four significant stakeholders:
is completed without establishing achievement of critical
bankability issues (sometimes referred to as ‘fatal flaws’). 1. project owners who are responsible for the equity funding
Usually these include: 2. banks (or others) who are to provide the project debt
•• an inadequate ‘resource tail’ for the project 3. regulators (government or government agencies)
•• unproven metallurgy 4. the general public whose support for the project must
•• half-committed market arrangements with customers be obtained.
•• a failure to recognise the minimum set of environmental It is worth remembering that the project owners cannot
standards approve a technically and financially sound project if their
company is not capable of raising the equity proportion of the
•• an unclear position as to the lender’s minimum lending total project funding. Often the equity to debt ratio is only
terms. known as negotiations for the project debt are advanced, and
For an account of project estimation conservatism leading to in recent times, debt ratios as low as 20 to 25 per cent have
a complete project re-think, readers are referred to Needham been offered as a maximum funding level.
(1985), who described in detail why the original Kidston gold
Investment banks, which traditionally have offered project
project was uneconomic until a leaner approach was taken to
finance, are becoming less popular since the Global Financial
every facet of the project. It was subsequently successfully built
Crisis (GFC) as they have become progressively more risk
and commissioned and operated successfully for many years.
averse and some would say, more aggressive towards
It is suggested that detailed analysis of key project minerals ventures. Term sheets and loan agreements, in the
development and operating assumptions aimed at discerning author’s view, have of late been characterised by unattractive
the true base-case parameters, free of all contingency. Allocation terms such as: high fees; intrusive security demands; tough
of the project contingency should then be added only when the liquidity coverage; challenging project ratios and recourse
absolutely bare base-case has been attained and validated. that places the borrowing entity’s directors at significant risk.
Metallurgists have a key role to play here, as described by Compounding this funding dilemma is the emerging issue of
Whincup (2008), McCarthy (2002, 2011) and other previous excessive project approval lead times. In one recent example,
MetPlant keynote speakers. approval for a NSW based gold project (Alkane’s Tomingley
Finally, the author recommends that, to avoid bankability gold project; despite this delay, the mine was constructed and
gaps, a list of all key bankability benchmarks (potential commissioned, achieving target head grade and production
fatal flaws) are established for each section of the feasibility guidance in its first two years of operation) took more than 18
study, and systematically addressed as the study advances. months for its DFS and environmental plan to achieve final

J S Dunlop
approval for site earthworks. By this time, the capital cost of •• detailed loan agreement
the project had increased by about $25 M and the gold price •• drawdown of the loan funds
was in retreat. The subsequent ramifications were significant
•• post-drawdown monitoring of repayments.
as the reduced cash available for debt service (CADS) reduced
the available debt level, increased the required hedge cover Project finance has lost much of its appeal in recent years for
whilst reducing the overall project rate of return (IRR). a range of reasons, which the author considers include:
A notable description of the conventional regulatory process •• high fees
appeared in 1997 (Chamberlain, Johnston and Joyce, 1997), •• high levels of security
where the concept of stakeholder groups was identified in the
•• onerous debt service arrangements
context of the approval of the Cadia mine project in NSW.
Readers interested in this subject in more detail are referred •• conservatism
also to chapter 4 of the MMH (AusIMM, 2012). •• inflexibility to the client’s time frame
Approvals can also be overturned by regulators, thus •• operational needs.
increasing uncertainty and project risk. This was illustrated
in 2010 when the Northern Territory government reversed Other forms of finance
its approvals and support for the Angela Pamela uranium Other forms of finance discussed here include:
project (Australian Uranium Association, 2010), citing concerns
expressed by local residents. At the time of writing, the project •• convertible notes and bonds
remains stalled, further disadvantaged by a low uranium price. •• export credit funding
Finally, the importance of the support of the general •• customer terms.
public cannot be underestimated. The example above serves These are all alternative funding options which have arisen
to illustrate how pressure can be brought to bear on the to offer alternatives to conventional project debt. Convertible
regulators by a range of project opponents. It is therefore notes allow major shareholders (with available cash) to retain
considered essential to monitor, establish and maintain an option to convert loaned funds to shares in the project
public support for new projects if potential delays are to be company or alternatively to redeem the loaned funds (with
avoided. The need for allocating high priority for this activity interest, referred to as a coupon rate).
is currently clearly reflected in the current opposition to coal
Export credit funding arises where the project’s product
seam gas (CSG) exploration on pastoral land (no literature
reference is suggested here, as any online CSG search will customer is able to make use of funds provided by its host
produce a plethora of argument both for and against). government (usually at a lower interest rate compared to
conventional debt). For example, the Ichthys liquid natural
gas (LNG) project in Darwin has an estimated US$20 B
PROJECT FINANCING, CONSTRUCTION AND COMMISSIONING capital cost. Of this, US$11.2 B is to be provided by Japanese
Project financing, construction and commissioning are given Bank for International Cooperation (JBIC), Export Finance
mention here because they each play a more important role for Australian Companies (EFIC), Nexi and five other export
than perhaps was the case in the past. The comments below credit authority (ECA) agencies in the form of direct loans or
attempt to answer the question, ‘Why is this so?’ guarantees. This actually resulted in the conventional debt
being scaled back (Schuler, 2013).
Project finance mix Finally, customer terms might include advance payment
The term project financing mix refers to the various component for mineral output (sometimes referred to as a ‘prepay’),
parts of a total project financing package. Most typically, such a capital contribution to the project itself, tied to supply
a package consists of equity and debt, though other forms of arrangements, technology sharing or the purchase of an
finance are emerging as conventional debt has begun to lose equity share in the project itself.
its appeal. Murray (1995) provides more background on this
theme and expands on frequently used banking terms.
Construction and commissioning
Equity finance It is generally accepted that over the last decade, capital
and operating costs have escalated at rates which exceed
The common forms of equity are:
the traditionally enjoyed operating margins, based on the
•• free cash within the project company prevailing metal prices. This, together with the project
•• cash raised by sale of existing assets funding challenges referred to above has caused a sharper
•• cash inflow from a farm-out focus to be drawn on project construction and commissioning
costs. An excellent account of these activities was presented by
•• cash raised by rights issue or placement (dilutive)
Luxford (2006), from which this author’s principal learnings
•• cash raised by share purchase plan (SPP) (not dilutive). are summarised in the following sections.
Equity raising is vulnerable to the state of the share market
and the project company’s share price. In short, when the Construction (project implementation)
capital markets dry up, it is very difficult to raise the level of Firstly, there is the engineering, procurement and construction
equity demanded by the lenders of debt. This is one reason management (EPCM) decision. Do you outsource it, share it
why funding is more complex than in the past, and why high or execute the roles in-house? Outsourced or shared EPCM is
levels of project debt are becoming less attractive. the most frequent choice, usually where the owner assumes
EPCM responsibility for the mine plan and mining fleet
Bank finance acquisition and the processing plant and infrastructure is
Conventional project debt arrangements consist of: handled by a specialist engineering firm.
•• project funding term sheet Next comes the selection of the owner’s management team.
•• credit approval Here the essential points, as suggested by Luxford, are:

•• knowledge and experience in what is to be designed (AusIMM, 2012). Of the various headings dealt with in that
and built text, the author sets out below his observations of what are
•• demonstrated track record in similar projects perhaps areas of potentially major project impact.
•• honesty and integrity
Regulatory considerations
•• intellectual and practical ability
For many years there has been confusion caused by the lack
•• ability to work well in a team. of congruence between the various metalliferous and coal
It is considered essential to recognise that an owner’s mining regulations across the Australian states. This began
project team, separate to any ongoing operations or company to change from the 2002 Conference of Chief Inspectors
administration, be appointed. The project manager need not of Mines and has now reached the point where in April
be, and more often than not is not, the eventual operations 2009 the Workplace Relations Minister’s Council (WRMC)
manager of the project. This arises from the fact that the project endorsed the creation of Safe Work Australia (SWA), a new
construction manager has specialised skills in managing the independent body charged with progressing the concept of
construction processes (such as contract management and harmonised work health and safety laws.
liaison with the EPCM engineers), which are themselves Following on from that, the National Mine Safety Framework
different to the operations skill set. (NMSF) steering group is working with SWA to develop mine
The need for a project execution plan (PEP) is obvious specific regulations with national outreach and application. In
but often treated in an ad hoc manner. Luxford provides a effect, each state was to have ‘mirror legislation’ in place by
comprehensive outline of what such a plan entails. In the 1 January 2012, though only the NT and the ACT actually met
author’s view a detailed PEP is essential, with clearly allocated that deadline (as recently as 2016, complete harmonisation
responsibilities for each set of tasks. has yet to be achieved).
Engineering management responsibilities are also well The end result will be that all states will adopt ‘core
outlined by Luxford, who makes the point that, in addition, drafting instructions’ but retain some freedom in ‘non-core
the project management team must, in a timely manner: areas’. The resulting legislative approach has been termed
•• provide user requirement specifications harmonised legislation.
•• examine and approve concepts and designs The three areas of key harmonisation may be summarised
•• review and approve equipment selections as follows:
•• approve standards. 1. roles, functions and powers under the model legislation
(such as managers and mines inspectors) for mining
operations
Commissioning
Project commissioning is often misunderstood by operating 2. management plans and records under the model (such
management. There are several stages involved in this as hazard analyses)
process. Luxford describes this process below: 3. duties (such as duty of care) and other requirements.
Construction completion involves completing installation These three points are expanded upon in the MMH
and erection work. The phase is complete when all (AusIMM, 2012) at chapter 7, and familiarisation with them is
components of the completed system have successfully recommended. The harmonisation process will be completed
passed inspection and testing to verify they comply with the before too long and it is recommended that minerals
contract. The only equipment operation at this point is to professionals keep apprised of developments.
check motor rotation.
No-load commissioning involves testing and verifying that Operations management
all equipment is ready to run, followed by operation of all Perhaps the most controversial aspects of operations
equipment under zero-load conditions. At the completion of management in the author’s experience has surrounded issues
this phase, the plant has achieved mechanical completion and to do with owner versus contract mining and joint ventures.
is ready for commissioning. The plant is ready for handover Some observations on these are offered below.
and is accepted by the owner. At this point, ownership is
usually transferred to the owner and the defects liability Owner mining versus contract operations
period commences. Wet commissioning then follows with Detailed accounts of the main issues dictating a decision on
the introduction of loads to the plant. In the case of a process which operating methodology to adopt may be found in
plant this would involve the loading of process media and Dunlop (2002) in the case of open pits, and Luxford (2005) in
introduction of ore followed by: the case of underground works. The arguments for and against
•• adjustments and minor modifications each operating strategy are clearly set out in these papers, along
•• process guarantee performance testing with the operational backgrounds dictating each situation.
•• completion of as-built documentation In recent times, commodity prices have been falling whilst
operating costs, capital costs and foreign exchange rates
•• final punch list completion.
have all been moving adversely, thus threatening operating
Finally, it is recommended that the construction handover margins and making the economic viability of minerals
phase include a checking that all as aspects of the PEP projects all the more challenging (In 2017, we are again seeing
have been completed and that the EPCM engineer’s list of effects of this in the so called ‘end of the mining boom’).
deliverables has been completed and that those deliverables As a direct result, most operations in this country are now
are in place. back on a cost cutting footing. This has caused owners once
again to look at the contractor’s margin and demand that the
OPERATIONS advantages of contracting out still justify the additional cost.
An excellent account of mine (and mineral processing Whilst this is a natural consequence of the economic times,
plant) operations management may be found in the MMH the author is still of the view that the reasons for and against

J S Dunlop
contracting services out are still not sufficiently recognised Substandard or otherwise deficient reporting has had
or understood. It is therefore recommended that decisions to the result that many mines have no adequately detailed,
contract out or owner-operate are taken following a detailed permanent record of their month-by-month operations, going
cost and benefit analysis following a careful review of the back in time. It is simply a matter of sound management that
literature on this subject. adequate operation reports be prepared and appropriately
backed up and stored.
Joint ventures Readers are urged to adopt, as a minimum standard the
The next issue of potential major project impact centres on pro forma operations report format set out in Appendix 3 of
JV. For an excellent general overview of JVs in the minerals the MMH.
industry, readers are referred to Reynolds (1983). In Reynolds’
view, the principal reason for the creation of JVs is the need Offshore projects
to share project risk. He then describes the (presumably most Minerals professionals of all disciplines will readily recognise
significant) risks as: the importance of identifying, monitoring, fostering,
•• project size in relation to the market maintaining and always improving stakeholder relationships
on their project. These issues are discussed in some detail in
•• financial magnitude of the project
chapter four of the MMH.
•• whether new technology is involved
Offshore projects introduce another dimension to the picture,
•• lack of certainty. where cultural or other norms may not be as we, the expatriate
In this author’s experience, a large proportion of minerals skills providers, may well expect. It goes without saying that
industry litigation has arisen out of JV arrangements, simply great care and sensitivity is required in this situation and
because these unincorporated vehicles offer the greatest careful choice of staffing for foreign work is advisable.
potential for divergent participant interests. A worthwhile Another issue that often arises is that of the project
example is that of the original Ok Tedi JV where the three establishment coming ahead of basic regional infrastructure
major participants, BHPB, Amoco Minerals and the PNG such as power and water supplies, communications and
Government had differing project development priorities. associated basic services. In the author’s experience, it is
Namely these were capital cost minimisation; maximised essential that the project address this imbalance in some
NPV and early production, respectively. meaningful and ongoing way, accepting that, in principle,
It is probably a truism to suggest that farm-outs are only the prosperity of the regional stakeholder must advance in
born of necessity – that is, when all other options have been line with the project itself, if stakeholder harmony and trust
investigated and shown to be impracticable. The party farming- is to be maintained.
out has to cede some control (often total management) in ‘Sustainable development’ was studied by the Mines and
order to see the project proceed, exacerbated by the farming-in Minerals Sustainability Development Project (MMSD), which
party assuming autonomous control and paying scant regard reported its findings (International Institute for Environment
to the new ‘minor participant’. Unless, therefore, there is good and Development, 2002). There are useful observations in that
will on both sides and strictly ethical behaviour, disputation report which are still considered relevant more than ten years
is highly likely. Should such disputation proceed to litigation, after its publication.
the possibility is real that the project may be delayed for years.
The areas of disputation considered by this author to be
Ethics
most common include: A simple definition of a professional (courtesy of the Oxford
dictionary) is:
•• failure to prepare and adequate JV agreement
...a person engaged or qualified in a profession: professionals
•• failure to provide for equitable voting rights on the JV such as lawyers and surveyors; a person engaged in a specified
operating committee (with regard to key decisions such activity, especially a sport, as a main paid occupation rather
as the life-of-mine plan and annual budgets) than as a pastime; a person competent or skilled in a particular
•• wilful abuse of JV terms as defined in the agreement activity: she was a real professional on stage.
•• lack of good faith by the manager, possibly coupled with Perhaps a more relevant definition for the minerals industry
a failure to exercise a fiduciary duty to the other party might read something along these lines: ‘a person of specialised
(after writing this paper, the author became aware of a training or skills who, acting always with ethics, applies those
landmark Supreme Court judgement in South Australia skills for the betterment of society and the industry, before
where it was held that a manager had no fiduciary duty him or herself’. This definition is manifest in the first bullet
to a minor JV participant – details upon request). point of the AusIMM Code of Ethics (AusIMM, 2013).
At the outset, therefore, it is strongly recommended that a In this latter definition, it is implied that a professional is a
detailed and painstaking review be applied to all JV agreement person with a degree; one qualified to become a Member of
drafts, and that expert advice be obtained before they are the AusIMM or an equivalent professional institution, though
executed. It is also suggested that the disputes clauses in the there is no reason why tradesmen, para-professionals and
agreement be broadly worded and provide for specialist, other service providers should not act ‘professionally’.
binding arbitration as opposed to litigation. It is often assumed in general trade and commerce that
the rules of ethics do not apply – or at the very least have
Operations reporting low priority – whatever those rules may be. Consequently,
Brief reference is made here to operations reporting, which, circumstances can arise where a professional is confronted
in the view of the author, have lacked consistency over many with disingenuous behaviour by others, and thus be tempted
years. There are many reasons for this, but at the very least, to adopt the same tactics.
there has never been a widely adopted template or other The Commonwealth Trade Practices Act (TPA) and its more
reporting convention to guide site operations personnel. recent replacement, The Competition and Consumer Act 2010

(CCA, 2010), are silent on ethics but make the following little to say about sustainability in general in the locations
requirement very clear: in trade and commerce it is expressly where mining has ceased. For example, it is difficult to find
prohibited to intentionally or unintentionally to engage in any accounts of communities which have not only benefitted
conduct which might mislead or deceive another party. This from the mining operations themselves, but have also
requirement covers a key part of professional ethics (not acting benefitted thereafter, by virtue of sustainable small business
in a misleading or deceptive way), but does not cover the entire ventures and the like which have endured post-closure
spectrum of ethical practices a professional must follow. (The Bougainville Copper and Ok Tedi projects, (the author
As a general rule, the best defence is what is termed the worked at both) when looked at today, have achieved little for
‘due diligence defence’ which, in plain language holds that the local communities in which the projects were established,
you cannot be guilty of misleading or deceptive conduct if despite the best intentions).
you can show that every effort was made not to do so. For Readers interested in this aspect of the mining cycle are
example, peer review of reports or opinions, risk audits and referred to McGuire (2003) who describes the mine closure
other forms of due diligence will be taken into account in this process at Rio Tinto’s Kelian gold mine in East Kalimantan.
context. Mere disclaimers, absent the above actions, will not In that paper, the concept of a ‘Mine Closure Committee’ is
achieve the same level of protection. introduced, along with the concept of involvement by the local
The obvious dilemma which emerges for professionals (and broader, regional) stakeholders in the committee’s work.
leading public companies is embodied in the question: is Today, a key part of annual audit requirements is a review
his or her primary obligation to the (profit of) company of mine closure plans and financial provision for those plans.
shareholders or to the (betterment of) the public? As long as It is evident that much greater thought and detailed planning
company directors and executives continue to disregard this and resourcing needs to be applied to this phase of the
dual obligation, corporate ethics will continue to lag behind mining cycle in the future. The sustainable mining principles
the required standards of profession ethics. dictate that when mines close, the land and the communities
left where the mine once occupied ought to be better off for
PROJECT EXPANSIONS the mine having been there, not just while the mine was in
operation.
Project expansions are becoming more and more
commonplace, probably for two major reasons: the high
capital cost of start-ups making the ‘start small’ approach THE CAPITAL MARKETS
attractive; and the simple economics of increased capacity The changing pattern of investment in world mining is
reducing unit production costs. Consequently, expansions of constant and the effects of those changes have a profound and
operating plants are becoming a feature of a minerals project’s continuing effect here in our Australian minerals industry.
normal operating life. These observations were made by Aldous (1993) whose paper is
In the past, however, the obviously disruptive practice of still recommended for those wishing to understand the concept
expanding an existing plant whilst continuing operation at of mineral capital allocation in the world’s capital markets.
the old rate until ‘cut-ins’ were complete, was rare. This is no If we were to consider Aldous’ view in 1993 that capital
longer the case and it is suggested that modern day mining comes from where it is most available and goes to where it
managers need to be equipped with the skills to recognise the is most welcome, and compare that view with the situation
key requirements of an ‘internal’ expansion and to resource today, it is clear that these fundamentals have not changed.
the work appropriately. Suppose we were to rank Australia’s ‘capital attractiveness’
In practice, all of the guidance set out under the construction according to Aldous, we might well be disturbed by the list of
and commissioning headings (earlier in this paper) apply negatives, which currently include:
equally in this context, though operations managers often •• lengthy project approval timelines
fail to recognise the fact sufficiently. There is a tendency for
•• relatively high project capital and operating costs
operational and construction responsibilities to overlap unless
the operations manager takes organisational steps to separate •• inflationary trends which outpace metal price growth
the two activities, as recommended by Luxford (2006). •• relatively high environmental and other compliance costs
These construction projects also require special care regarding •• relatively low rates of return compared to the risk profile
fitness-for-work and inductions for casual contractors •• a small capital market compared to the rest of the world
associated with the construction work. Contractors who come
•• an uncompetitive foreign currency exchange rate.
onto the designated construction-site (as distinct from the mine
site) need to be inducted and cleared for work even though The combination of these factors, coupled with financial
their site residence time may be as short as one day. Time must instability in most of Europe and a sluggish US economy, has
therefore be allowed for medical screening and drug clearance, caused the minerals equity and debt markets to dry up almost
so as to ensure contractors do not commence work whilst their completely for ‘small cap’ and mid-tier companies on the
prestart screening results are awaited. Australian Stock Exchange (ASX) and the Toronto Venture
Exchange (TSXV). If this situation is allowed to continue,
Finally, it is essential that any additional site regulations
significant damage will be done to what some would refer to
that apply pursuant to construction-related (non-mining)
as our country’s cornerstone industry.
regulations are clearly understood by all concerned.
Whilst the capital markets will recover, following the endless
cycles we all recognise, the industry needs to find ways to
PROJECT CLOSURE continue in the meantime. The solutions, it is suggested, lie in
Project closure (and, perhaps suspension of operations) is an the following strategies:
integral part of the overall mining cycle. Despite this, a search
•• operating cost reduction
of the published literature on ‘mine closure’ reveals most
papers which focus on the environmental compliance issue •• continuous improvement to lift productivity per
post-operations. Whilst such compliance sits well as part of employee
the MMSD principles referred to earlier, the literature has •• leading in the application of advanced technology

J S Dunlop
•• faster project approval mechanisms liaison with our industry at all levels and commit to a much
•• favourable government policies for projects of ‘national higher priority level being accorded to the industry’s needs.
importance’
•• favourable taxation incentives for small cap explorers CONCLUSIONS
•• further development of Free Trade Agreements (FTAs) The personal observations of critical issues presented in this
with our closer trading nations, to further unlock paper span the many phases of mineral project development
additional export credit. from exploration, through development, operation, expansion
and finally closure. In summary form, the critical issue
identified are as follows:
GOVERNMENT UNDERSTANDING OF THE INDUSTRY
•• exploration – the market continues to lose site of the real
Following on from the aforementioned observations, there is
lead time to develop a project following its discovery
likely to be little disagreement with the proposition that the
current mining boom in Australia has stalled, to a greater or •• discovery – stakeholders need to be identified and
lesser degree, due to the lack of capital attractiveness, to which engaged with as soon as possible
has been referred. •• scoping, prefeasibility and definitive feasibility –
The Minerals Council of Australia (MCA) addressed this in potential fatal flaws need to be identified and resolved
a paper by Ergas and Owen (2012), which concluded: early in the process
The easy gains from Australia’s early 21st century mining •• project approval – current processes are far too slow and
boom are over, though large and enduring benefits are still completely uncompetitive when seen in an international
there to be secured from further resource-intensive growth context
in emerging Asia. •• project financing and commissioning – conventional
Rebooting the mining boom calls for renewed policy focus on debt funding is losing its attraction, in favour of
securing the next generation of mining project investment alternative financing means
and delivering on potential export volume growth out •• operations – issues of concern relate to JV and corporate
to 2025 (Author’s note: Very pertinent still in 2016. ethics
Perhaps the ‘transition from the mining boom’ will not •• expansion – operational and construction responsibilities
be to something else but rather back to the next mining to overlap unless the operations manager takes
boom in the endless minerals cycle). organisational steps to separate the two activities
That focus should be on tackling the ‘unfinished business’ •• project closure – mine closure plans need to be detailed
across Australia’s export supply-chain, from exploration and involve a mine closure group of all relevant
and initial development through to final shipment, where shareholders.
cost control, timeliness, flexibility and adaptability present
Additional observations have been made in relation to:
critical challenges. It is only by stripping out inefficiencies
across our export supply-chain that Australia can reap the •• capital markets – emphasizing the Australian minerals
rewards on offer from the next phase of the boom. industry’s current capital unattractiveness, and
corrective measures
As Marius Kloppers (then CEO of BHP Billiton) has
observed, companies themselves need to do the heavy •• government understanding of the minerals industry –
lifting to make our mining projects more cost competitive. some thoughts on how the government might better
But they can only do so if the policy environment provides view and assist the industry.
the framework and tools for this to occur. At the moment,
it doesn’t. ACKNOWLEDGEMENTS
To ‘re-boot the boom’, then, governments at all levels must The author wishes to thank the organising committee of
improve their understanding of the industry – particularly MetPlant 2013 for the opportunity to present this keynote
now when it contributes to approximately 50 per cent of paper. He also gratefully acknowledges the many authors
national exports and is increasing. At present, its level cited in the paper, whose contributions to the topics discussed
understanding, in the author’s view, is superficial at best and, in the paper provide an expanding pathway for those wishing
at worst, lacking in both breadth and depth when viewed in to concentrate further on any of the issues touched upon.
the context of its national economic significance.
To illustrate the proposition, the industry can be viewed in REFERENCES
a range of ways: As far as practicable, all references cited in this paper may be
•• commodity (base metals, precious metals, bulk found on <www.onemine.org> and downloaded for free by
commodities, industrials and so on) members of the AusIMM, SME or SAIMM.
•• industry grouping (metalliferous, coal, petroleum, Aldous, R T H, 1993. The changing pattern of world mining investment
industrial) and the implications for Australia and her mining industry, in
Proceedings Centenary Conference, p 267 (The Australasian Institute
•• activity (exploration, production or downstream of Mining and Metallurgy: Melbourne).
processing)
AusIMM, 2013. Code of Ethics [online]. Available from: <http://
•• size (small caps, mid-tiers producers and multinationals). www.ausimm.com.au/content/docs/code_of_ethics_stand_
Each of these groupings require a profound understanding alone.pdf> [Accessed: July 20, 2017].
to formulate policy settings which will not only be of benefit AusIMM, 2012. Mine Managers’ Handbook, Appendix 3 (The
to those concerned, but will also address the structural Australasian Institute of Mining and Metallurgy: Melbourne).
efficiency areas referred to by Ergas and Owen. It is the view Australian Uranium Association, 2010. NT decision on Angela
of this author that those responsible for governing and setting Pamela causes uncertainty [online], press release, 28 September.
minerals policies must improve their understanding of, and Available from: <www.aua.org.au> [Accessed: 8 May 2013].

Binon, L C, 1981. Determining the exploration budget, in Proceedings Luxford, J, 2006. Project development and construction management,
SME-AIME Fall Meeting and Exhibit, Denver, November, Preprint in Proceedings International Mine Management Conference, p 97 (The
No 81–311 (Society for Mining, Metallurgy and Exploration; Australasian Institute of Mining and Metallurgy: Melbourne).
American Institute of Mining, Metallurgical, and Petroleum
International Institute for Environment and Development, 2002.
Engineers: Englewood).
Breaking New Ground: Mining, Minerals and Sustainable Development
Card, P, 2011. Guidelines for economic evaluation of mineral projects, [online], 450 p. Available from: <http://pubs.iied.org/9084IIED.
in Proceedings MetPlant 2011, August, 12 p. html> [Accessed: May 2013].
CCA, 2010. The Commonwealth Consumer Act [online]. Available McGuire, G M, 2003. Managing mine closure risks in developing
from: <http://www.comlaw.gov.au/Details/C2007C00619> communities – a case study, Kelian Equatorial Mining, Indonesia,
[Accessed: May 2013]. in Proceedings Mining Risk Management Conference, p 385 (The
Chamberlain, C, Johnston, G and Joyce, C, 1997. Regulatory
framework for Mine Project Permitting, in Proceedings Mindev 97, Murray, J, 1995. Project finance in the minerals industry, in Proceedings
November, pp 193–198 (The Australasian Institute of Mining and Young Professionals Conference, p 59 (The Australasian Institute of
Metallurgy: Melbourne). Mining and Metallurgy: Melbourne).
Duncan, I J, 1985. The competitive challenge in marketing Olympic Needham, R, 1985. The Kidston Gold Mine: a development
Dam copper, uranium and gold, in Proceedings The AusIMM challenge of a low grade deposit, in Proceedings AusIMM Annual
Annual Conference, p 6. Conference, p 121 (The Australasian Institute of Mining and
Dunlop, J S, 2002. Contract versus owner mining – an update on
Australasian open pit mining practice, in Proceedings Iron Ore Nice, R W, 1986. The Paddington Gold Mine Project – a case study,
2002, Perth, p 205 (The Australasian Institute of Mining and in Proceedings Second Project Development Symposium, Sydney, p 299
Metallurgy: Melbourne). (The Australasian Institute of Mining and Metallurgy: Melbourne).
Ergas, H and Owen, J, 2012. Rebooting the boom: unfinished business Noort, D J and Adams, C, 2006. Effective mining project management
on the supply side. Available from: <http://www.minerals. systems, in Proceedings International Mine Management Conference,
org.au/file_upload/files/publications/mca_rebooting_the_ p 87 (The Australasian Institute of Mining and Metallurgy:
boom_FINAL.pdf> (Minerals Council of Australia: Canberra) Melbourne).
[Accessed: May 2013].
Reynolds, G C, 1983. Unincorporated joint ventures in project
Hinz, R and Aseervatham, R, 1999. Feasibility studies, EPCM and the developments in the mining industry, in Proceedings Project
environment – case studies in integrated mine development, in Development Symposium, Sydney, p 349 (The Australasian Institute
Proceedings PACRIM 99, Bali, October, p 123 (The Australasian of Mining and Metallurgy: Melbourne).
Schuler, S, 2013. Financing projects in hard times: export credit agencies
JORC, 2012. Australasian Code for Reporting of Exploration are the smart alternative [online]. Available from: <http://
Results, Mineral Resources and Ore Reserves (The JORC Code) www.corrs.com.au/thinking/insights/financing-projects-in-
[online]. Available from: <http://www.jorc.org> (The Joint Ore hard-times-export-credit-agencies-are-the-smart-alternative>
Reserves Committee of The Australasian Institute of Mining and [Accessed: May 2013].
Metallurgy, Australian Institute of Geoscientists and Minerals
VALMIN, 2015. Australasian Code for Public Reporting of Technical
Council of Australia).
Assessments and Valuations of Mineral Assets (The VALMIN
McCarthy, P, 2002. Setting plant capacity, in Proceedings MetPlant Code) [online]. Available from: <http://www.valmin.org/docs/
2002, Perth, pp 21–30 (The Australasian Institute of Mining and VALMIN_Code_2015_final.pdf> (The VALMIN Committee,
Metallurgy: Melbourne). a joint committee of the Australasian Institute of Mining and
Metallurgy and the Australian Institute of Geoscientists).
McCarthy, P, 2011. Predicting variations in mill feed, in Proceedings
MetPlant 2011, pp 19–25 (The Australasian Institute of Mining Van Der Merwe, M J, Vegter, N M and Momberg, N 2008. Value
and Metallurgy: Melbourne). driven resource allocation and focus in the early stages of project
development, in Proceedings Lead and Zinc 2008, p 255 (The South
Lawrence, M J, 1994. An overview of valuation methods for
African Institute of Mining and Metallurgy: Johannesburg).
exploration properties, in Proceedings Mineral Valuation
Methodologies, p 205 (The Australasian Institute of Mining and Whincup, P R, 2008. The role of the metallurgist in project
Metallurgy: Melbourne, in conjunction with The Mineral Industry development, in Proceedings MetPlant 2008, p 25 (The Australasian
Consultants Association: Sydney). Institute of Mining and Metallurgy: Melbourne).
Luxford, J, 2005. Future directions in Australian contract mining, in
Proceedings Underground Operators, p 3 (The Australasian Institute

Contents
Keeping projects on the rails

J Canterford1
ABSTRACT
Although we have access to a wide range of sophisticated process design and
improvement tools in combination with a significant range of case studies, the mineral
resource industry still manages to be plagued by the negative publicity surrounding
operating plant failures because they were neither smart nor safe, together with
proposed projects that fail at the first or second hurdle. The public and the finance
industry are generally totally uninformed about projects that ramp-up to name-plate
capacity on-time and on-budget and even more so when ahead of schedule and
under-budget. As the saying goes, good news does not ‘cut the mustard’.
The following is a summary of the author’s sometimes jaundiced view about why
metallurgical plants do not live up to their technical and commercial expectations.
In no way should the summary be construed as a negative reflection on the skills of
the metallurgical profession. Rather, it is to be taken as a positive acknowledgement
of those skills and is intended to highlight some of the major considerations that
need to be assessed as a potential project transforms into a sustainable technical and
commercial reality.
While a considerable portion of the summary is directed at developing
hydrometallurgical flow sheets, the general principles are equally important to all
aspects of extractive metallurgy.
For convenience, the summary covers the following five basic discussion points –
project, people, process, patents and politicians.
INTRODUCTION
On any given day, every successful metallurgical plant can be characterised as being
a positive compromise between a series of potentially conflicting criteria, the more
significant including:
•• the mineralogical complexity of the feedstock
•• market requirements
•• owner/developer expectations
•• technical complexity
•• ongoing operating cost scenarios
•• sustainable environmental footprint.
Although we can talk about generic flow sheets, no two flow sheets are identical
when dissected in detail. While we certainly can and should learn from past and current
experience, what we must accept is that the balance between the conflicting criteria will
almost certainly alter during the life of any individual project. In turn, this means that
the ‘best’ flow sheet should have a reasonable level of flexibility to cater for a realistic
range of inputs and outputs, both metallurgical/environmental as well as corporate.
Some of the anticipated variations may be self-induced while others will be a result of
external forces over which the project owner and/or developer has no control.
If the conflicting criteria cannot be balanced and that balance maintained, then
the outcome will almost certainly be closure or at least the requirement for major
modifications to plant layout, operating procedures, changes to volume and
specifications of outputs etc. The capital and operating cost implications of such
modifications can be mind blowing. Without going into any detail, experiences at
Ravensthorpe and Goro are stark reminders of what can go wrong.
As a person with no practical metallurgical plant design experience but who is
sometimes asked to comment on what, how and why a flow sheet appears to have
failed or at least throw enough hand grenades at the evaluation stage of flow sheet
development, the simplistic answer provided is itself multifaceted covering:
1. FAusIMM(CP), Process Technologies Australia
Pty Ltd, Deloraine Tas 7304.
•• the chemical/mineralogical integrity of all test work
Email: pr37666@bigpond.net.au •• realistic technical risk analysis
183
J Canterford
•• treatment of process and financial models with a fairly be described as conventional in that they yield standard
high degree of scepticism products such as gold bullion via heap leaching of an oxide
•• understanding what the market will buy and why and gold ore or cathode copper via smelting/electro-refining
how these requirements can be met. of high-grade chalcopyrite concentrates can be relatively
straightforward in that product specifications are well known
It will come as no surprise that this author is very strongly of
and there is an active, open market.
the view that throwing hand grenades and proper evaluation
of the metallurgical plant design should be initiated at the The situation with projects where each end-user sets their
scoping stage and should be a continuing exercise well into own specific product specifications is quite different. For
the operating phase and until name-plate capacity has been example, virtually every end-user of dead burned magnesia
achieved or exceeded. will have quite specific requirements for crystallite size and
shape in addition to other chemical and physical properties.
There are no universal product specifications. While magnesite
SUCCESSFUL PLANT DESIGN DRIVERS calcination may seem to be a relatively straightforward
First and foremost, it is essential to match output volumes and metallurgical process, it is the physical structure of the
qualities with a realistic assessment of market requirements. magnesium carbonate raw material as well as the calcination
As noted below, some 15 years ago magnesium metal was conditions that determine the properties of the magnesium
the metal of the future. Numerous proposals were being oxide product. It follows that the owner/developer of a
bandied about. In Australia alone, the combined capacities magnesia production facility must be fully aware of detailed
of the proposed magnesium metal projects in Tasmania, market requirements and understand that it may not be
Queensland, Victoria, South Australia and Western Australia practical to service a broad spectrum of customers. Thus there
between 1995 and 2005 was in excess of the 300 per cent of has to be an ongoing assessment of potential markets as it is
then total world consumption and more than 200 per cent not a good option to tie production to a very limited number
of the projected demand in 2015. Individually each of the of customers.
proponents considered their proposed projects to be superior Practical metallurgical projects are generally based upon
to all others and that only they would attain commercial status. consumables that are widely available although some, such
Of course reality was quite different and not one of those as sodium cyanide, will require a high level of process control
proposed Australian projects has seen the light of day. As the and mitigation strategies to be put in place to overcome any
author of this paper participated in some of these proposed plant failures however caused.
projects he must confess to being guilty by association.
Consistent with his novice status as a metallurgical plant People
‘expert’, he would like to put forward what he has termed As practicing metallurgists, it is important to maintain a
the ‘Five P’ set of criteria that should be seen as one of several high level of realistic technical input into senior management
sets of criteria that need to be considered when developing considerations and, where appropriate, curb the misplaced
and maintaining a state-of-the-art metallurgical plant. This conviction by the technically illiterate that reliance on
set of criteria is made up of Project, People, Process, Patents simplistic in-ground evaluations and avoidance of locked
and Politicians. Clearly there is a strong interplay between cycle continuous test work at the pilot and demonstration
each of these components and each may individually or in level are just two of several ways of achieving a project that is
combination lead to initiation of an outstanding technical and not technically and commercially sustainable.
commercial success but conversely to an absolute failure.
It is not unusual for initial capital and operating costs to
From the author’s perspective it is important to expect the be considerably higher than originally anticipated so that
unexpected and avoid any major deviation from the KISS cost-cutting procedures are instigated, particularly by the
(‘keep it simple’) principle. The title of this presentation corporate finance management team. This may involve,
indicates acquisition of information by osmosis and as such for example, reduction in pipe and cable runs, reducing
the following comments cover some of the more relevant overhead access, reduction of automation and monitoring
observations made over the past 40 years. instrumentation etc. While there may be a saving on structural
steel, concrete, process control facilities etc, the result will
Project ultimately lead to increased maintenance costs, more complex
Geometallurgy is now properly recognised as a critical occupational health and safety issues, less reliable adherence
evaluation step in establishing a technically and commercially to operation within the designated bandwidth and potentially
viable process flow sheet. Unless a detailed evaluation of the extended shutdown periods. These technical issues must be
chemical and mineralogical complexity of the resource under properly flagged. The project’s technical champion has a very
consideration is executed then failure in some form is highly significant role to play.
likely to follow. One of the real challenges is to ensure that test ‘Bean counters’ need to be controlled to avoid equipment
work samples should be properly characterised and that the selection based on simple cost and availability terms, rather
selected samples are representative of the expected resource than on a genuine ‘fit for purpose’ basis. There is no point
overall aspects of the entire project life. This means that there in gold plating the processing plant, but it should not be
must be close cooperation and coordination between the mine deficient in practical operational and safety terms. Similarly,
planners (including the geologists), the client’s project team there is no value in over-promoting the economic benefits that
and the process flow sheet developers. might accrue.
While the distribution of the marketable metal(s) and
product(s) has a major influence on mining and processing Process
options, in reality it is how, where and in what form the It is possible to develop a process flow sheet that is chemically
non-value components are present and how to reject them sustainable, at least at the theoretical and initial pilot/
efficiently that will be the key to success or failure. demonstration scale, but which never reaches commercial
Geometallurgy is also a major factor that assists in the status because it is ‘impossible’ to engineer and operate
determination of realistic product outputs. Projects that can it and/or economic reality sets in. There are quite a few

Keeping projects on the rails
proposed processing flow sheets that are best described as the The cover gas technology developed by AMC and CSIRO
‘tail wagging the dog’. This is particularly the case where the was and remains technically astute – it is unfortunate that it
flow sheet incorporates recovery of every possible product in has not been commercialised due to the total collapse of the
the purest possible form. For example, many of the proposed AMC project.
nickel laterite flow sheets incorporated recovery of metallic As noted previously, it is important to expect the
magnesium, given that the magnesium content of the pregnant unexpected, especially when considering some of the more
leach liquor is many times greater than that of nickel. Only a complex hydrometallurgical flow sheets. For example, during
little bit of evaluation from the other side of the fence clearly carbon dioxide leaching of several caustic calcined magnesia
indicates that this approach is not sustainable. If the project feedstocks it was found that the resultant liquors had quite a
does not make commercial sense based on the nickel cash high soluble ferric concentration (several g/L) even through
flow with a nominal cobalt credit then the presumed financial the bulk pH was greater than 9.5. Under the operating
benefits from additional by-products will just be an illusion. conditions it is possible to form soluble iron (III) carbonato
As a general but not totally universal comment, the ‘best’ complexes even though all the available thermodynamic
process flow sheets are developed and commercialised by data suggested that this was ‘impossible’. Development
project owners/developers. This is particularly the case with of a suitable technique for selective iron removal from the
mineralogically complex resources since exploitation of the pregnant liquor proved to be quite challenging.
resource will not proceed without the successful development During the 1970s and 1980s the majority of hydromet flow
of the required flow sheet. sheet development centred on the initial leach step, with
less emphasis on the downstream purification and recovery
This comment should not be taken as an unqualified
steps. For example, chloride hydrometallurgy of base metal
criticism of research and development (R&D) undertaken
sulfide ores was seen as the panacea of all environmental ills
by universities, research organisations and independent
associated with sulfur dioxide abatement with conventional
companies. They certainly generate some great concepts.
pyrometallurgical operations, although it is appropriate to
However, such groups rarely have the financial and technical
note that commercial reality is yet to be achieved. While
expertise and abilities to undertake and achieve commercial
detailed knowledge of the leach step was generated, many
status for their concepts and are reliant on selling their know-
of the challenges with chloride hydrometallurgy relating
how. One must always question their ability to provide the to separation and recovery of elemental sulfur, recovery of
necessary support when their know-how is implemented precious metals from leach residues, deportment of nasties
by others. In other words, the end-user should always very such as arsenic, regeneration and recycle of the leachant, sulfate
seriously question the real value (technical and commercial) control etc, were subjected to far less rigourous evaluation.
of the licence to use purchased technology. Fortunately this situation of unbalanced unit step development
As noted above, magnesium metal has been vigourously for most hydrometallurgical flow sheets is now being corrected.
promoted as the metal of the future. Australia was at the Two of the driving forces for a balanced assessment of all
forefront of developing what was being claimed as novel hydromet unit steps are the need for and influence of a range
state-of-the-art processing regimes. Three major areas of of internal recycle stages and the maintenance of a workable
sustained R&D centred on: process water balance. In some locations and for some flow
1. purification and dehydration of concentrated sheets there may be an excess of process water that needs to
magnesium chloride liquors be discharged into the local environment in an acceptable,
2. development of alternative electrowinning electrolytes benign manner. In other situations, there will be a potential
deficiency of fresh suitable process water at reasonable cost,
3. improved cover gas technology. so recovery and recycle of process water is a requirement that
Despite spending more than $50 M on their demonstration adds to both capital and operating costs. Desalination plants
facility, amongst other things the Australian Magnesium have their own problems including power supply, plant
Corporation (AMC) purification and dehydration technology duplication for continuation of supply and waste disposal.
failed to deliver the anticipated outcomes. The chemical/ For hydrometallurgical flow sheets in particular, all
thermal decomposition of the organic dehydration agents definitive test work from feed preparation, which may
was soon found to be quite deleterious. involve grinding and flotation, right through to final product
Electrowinning magnesium from fused salt electrolytes is recovery, must be carried out using process water that will
highly energy intensive as well as challenging in engineering be available at site. The physical and chemical properties of
and operational terms. Molten magnesium metal is not the the process water will have a significant effect on parameters
simplest material to handle. Energy consumption during such density, viscosity, redox potential, oxygen solubility,
electrowinning can be reduced by modification to the ionic strength, leach and precipitation kinetics etc, let alone
composition of the fused salt electrolyte. One such option is materials of construction considerations and actual equipment
to add neodymium chloride. Unfortunately one side effect design. It also follows that care must be taken in preparing
is that rather than pure magnesium being produced at the process models since many of the thermodynamic inputs for
cathode, the end product is a magnesium-neodymium alloy. most models are clearly deficient.
Such alloys have several useful physical properties, but it was Infrastructure requirements and power generation
soon worked out that the ‘loss’ of neodymium to the alloy (particularly if ‘peak’ oil status is fact not fiction) will impact
product was such that the volume of make-up neodymium on all remote sites as a cost burden as governments continue
chloride would soon exceed that currently available. In other to bail out of these areas.
words, the proposed magnesium metal production facility
would be dependent upon the establishment of an ongoing Patents
rare earth production facility with a significant neodymium While it is not unreasonable to protect intellectual property
output. The technical and commercial constraints so imposed (IP) by means of filing and executing patent applications,
basically canned the concept of the use of neodymium chloride from a practical perspective the situation is tending to get
additives to the magnesium electrowinning cell house. ‘out-of-hand’. For example, since 2000 more than 90 patent

J Canterford
applications on sulfuric acid leaching of nickel laterites have a high level of derision – relates to the ‘neutralisation of an
been filed via the World Intellectual Property Organisation acidic stream using sized limestone’. It was originally allowed
(WIPO). The claimed flow sheets cover heap, vat, atmospheric by the United States Patent and Trademark Office (USPTO)
and high-pressure leach circuits, sometimes in combination on the basis of the word sized. Fortunately, industry finally
with downstream unit steps such as iron precipitation, got the USPTO to accept a dose of technical reality and the
intermediate product recovery, solvent extraction and application was forced to lapse.
electrowinning. Sometimes it is quite difficult to discern any
realistic technical differences between two or more claimed Politicians
flow sheets. Part of the problem relates to the definition of The mining and metallurgical industries face several
patent laws and concepts such as novelty and inventiveness. challenges as they are forced to cope with ever increasing
It must also be remembered that a patent is basically an idea volumes of more and more diverse bureaucracy. It is
– it is not necessary to prove chemically and/or physically unfortunate that the majority of our politicians and their
that the concept actually works. In fact in some instances a advisers are technically illiterate and are easily persuaded
little technical nous indicates that the concept cannot work. by pseudo-scientific commentary. This can and does lead to
Another feature of patent law that is sometimes difficult constrained operating parameters because of the imposition
to fathom in technical terms is that known processes and of illogical rules and regulations as a means of placating vocal
concepts can be incorporated into primary and subordinate opponents of our industries. Perhaps the most bizarre was the
claims. For example, many of the nickel laterite sulfuric acid push to ban the use of stainless steel because nickel has some
leach applications incorporate current commercial practice rather ill-defined carcinogenic properties.
such as nickel-cobalt separation by solvent extraction using
Our industries must and do accept the imposition of logical
Cyanex 272. From a practical perspective, it seems difficult to
environmental constraints but it is a pity that the overall
understand how any claims relating to the use of this reagent
record of compliance is not recognised by our politicians
in recent applications are sustainable.
and the public. Accidents do happen, but it would be safe to
One apparent reason for the proliferation of patent assume that our record is somewhat more positive compared
applications relates to the fact that patent challenges/ with those of many other primary and secondary industries.
litigation is extremely costly in terms of time, money and user
The challenge for us is to educate our politicians.
licence fees, and conditions are often unacceptably onerous.
Thus ‘new’ flow sheets are devised to overcome claims of
existing novelty/inventiveness even though such flow sheets CONCLUSIONS
may not be technically and/or commercially optimal. To continue to learn by osmosis it is constructive to
It is now common practice to initiate a ‘freedom to operate’ periodically assess why and how ‘wayward’ metallurgical
review of potential patent infringement as part of the overall plants come into being. This paper indicates that this occurs
risk analysis procedures during completion of feasibility because of scientific-engineering incompetence, project
studies. owner/developer avarice, insufficient attention to future
markets and product requirements, imposition of impossible
There are numerous examples of what are best described
operating constraints, right through to obvious bad luck, but
as ‘nonsense’ patents. Probably the most technically
it is more likely to be a combination of all such factors.
‘challenging’ – and that description is certainly offered with

Contents
Operations versus projects – how do people

think and what are the implications?
G Lane1 and B Clements2
ABSTRACT
The operations and project environments are very different, but overlap in many
areas. The development of mineral processing projects is goal oriented, schedule
focused and a contractual environment. The operating environment for a mineral
processing plant is oriented around longer-term commitments, focused on delivery
against budget expectations, and an environment heavily influenced by long-term
personal relationships, community relationships and people issues.
Further contrast is evident in the way expenditure is managed. Projects commit
large amounts of money quickly to meet schedule-based objectives. Operations
manage expenditure to minimise operating costs and are able to, or are forced to delay
decisions on large expenditures. However, these two environments need to come
together in many ways, ranging from greenfields plant handover to a new operations
and maintenance crew, to day-to-day interaction and management of brownfields
projects, often in close proximity.
This paper considers the different perspectives people bring to projects and early
operations, including the advantages and disadvantages of ‘operations’ input to
a project and models for managing value adding during project development,
implementation and debottlenecking.
INTRODUCTION
This paper is about the way people need to respond to the management of operations
and projects. The mining environment in 2012 was characterised by a shortage of
engineers and managers with project experience. This situation has existed for at least
three of the preceding five years with a short hiatus driven by the Global Financial
Crisis (GFC). The shortage caused a significant movement of personnel from mining
operations into projects, on either the owner’s or the contractor’s side of the fence.
The consequences of this are numerous but this paper focuses on one area key to
the delivery of a successful project, the behavioural characteristics of the project
management and leadership team.
Experience in mineral processing plant operations brings substantial value to
a project in terms of quality of outcome; however, the mindset associated with
operations is not the same as that required for a project. The development of mineral
processing projects is goal oriented, schedule focused and a contractual environment.
Because projects are not just driven by the quality of the outcome, schedule and
cost are the key measures of success up to, and including, commissioning. Quality
is heavily influenced by the design, procurement and contracting strategies. The
simplicity and effectiveness of the management systems, alignment of, and trust
between, the participants and speed of (correct) decision-making, become the key
behavioural requirements.
PROJECT MANAGEMENT 101

A successful project is one that meets or out-performs the project criteria agreed by
the contracting parties in terms of cost, schedule and quality (performance). The
project criteria are often based on the outcomes of a feasibility study, adapted to meet
financing and other external drivers.
Successful project outcomes are based on the following:
1. FAusIMM, Chief Technical Officer, Ausenco •• a well-gated feasibility study (Biery, Hollands and Young, 2009)
Minerals & Metals, South Brisbane Qld 4101. •• an experienced project team
•• sound project planning and effective communication
2. Former President/General Manager,
Sociedad Minera Cerro Verde, Freeport
•• simple interfaces (contracting and execution)
McMoRan Copper and Gold, Arequipa, Peru. •• fast and effective macro-scale decisions.
187
The different types of execution strategy require very the technical scope and simplifying and clarifying the interfaces
different behavioural needs (Gabrielson, 2007; Hundertmark, between the service providers.
Siliva and Shulamn, 2008). Engineering, procurement,
construction, management (EPCM) style projects require KEY REQUIREMENTS OF A GOOD OPERATIONS MANAGER
ongoing collaboration and interfacing with the project owner
An operations manager’s performance is measured against
and the associated operational team. This contract style
a budget that is usually based on previous performance
is characterised by ongoing scope change and continuous
(benchmarked or historical). Behaviour is heavily long-term
change management. EPC or lump sum turnkey (LSTK)
people oriented with a clear focus on incremental operating
contracts require much more rigid scope management and
and maintenance costs. This results in a thinking process,
change is managed on a contractual basis (rather than a ‘trend
which is heavily expenditure-driven and uses tools, such as
register’ used for reimbursable EPCM projects). Hence, the
systems applications and products (SAP) in data processing,
EPC approach gives cost surety, but is less flexible than the
to focus on cost management.
EPCM approach, and quality outcomes can be sacrificed in a
poorly managed EPC environment. Incremental improvement is a critical part of operations
management. Reducing the time taken for effective
Operations management is about managing continual
maintenance increases plant availability and revenue. The
change on a relatively micro scale. Trends are compared
balance between maintenance and availability is often the
with annual budgets in a personalised environment that is
key focus of the incremental improvement process and the
subject to continual change. Whilst the annual budget sets
management process is more ‘microscale’.
a time frame for performance (and reward), the operating
environment requires management on a long-term horizon In this process, the speed of decision-making is not critical.
to optimise project net present value (NPV) and other issues Making a quick decision may result in an adverse outcome
such as community relations. that is easily compared with expected budget outcomes. In
this environment, it is more important to make the right or
In a simplistic sense, there is a continuum of behavioural
optimised decision, than the quick decision.
environments that stretch from the microscale, continuous
improvement environment of operations, through various Because of the long-term focus on availability and ease of
forms for collaborative and reimbursable project management, maintenance, operations personnel are typically much more
to the heavily contractual and short-term focus of EPC/LSTK focused on installations designed for long-term performance
delivery. For the purposes of simplicity, we have characterised than on project deadlines and delivery to scope and budget.
this continuum using the three aforementioned examples:
1. operations management KEY REQUIREMENTS OF A GOOD PROJECT MANAGER
2. EPCM reimbursable project management Due to the nature of their roles, project managers are more
3. EPC or LSTK project management. transient and less risk adverse. They are used to spending
large amounts of money quickly in order expedite the job or
The EPCM reimbursable approach also includes the maintain the schedule. Schedule often takes precedence over
owner’s self-perform approach where an owner’s team takes cost, in part due to the high cost of delayed production.
on the role of EPCM management rather than contracting to
a separate party. A good project manager is milestone driven and requires
a management tool that is heavily focused on schedule and
work breakdown structure (WBS). Hence, SAP systems are
PROJECT ENVIRONMENTS not appropriate and are replaced by project delivery systems
In addition to the three management scenarios listed in the and other suitable workflow management tools that are cost
previous section there are other factors that influence the and schedule focused.
type of behaviour required for a successful project outcome. Time is a key differentiator between operations thinking
These include: and project management thinking. During a design phase,
•• greenfields and brownfields project environments – a typical medium sized copper concentrator project will
brownfields projects typically require greater operations burn $2 M to $4 M per month. This excludes the cost of lost
input production due to delayed delivery. Hence, delays in project
•• small or large owner’s team environments – larger decision-making and poor decisions that result in rework
owner’s teams typically include more operations and need to be avoided.
maintenance experience/focus Micromanagement is to be avoided and it is important that
•• conventional or novel technology environments – decision-making authorities are set at the lowest practical and
novel technologies require an additional level of risk effective level. This requires an experienced team capable of
management through test work, scale-up, engineering managing issues at the lowest level. Without a high quality
design and start-up and are accompanied by increased team, project managers find more and more of their time
complexity in project relationship management. taken up by low-level issues, to the extent that the focus on
the broader issues may be lost. This is a major challenge in the
A well-defined (and gated) greenfields project utilising
current project environment as the large number of projects
conventional technology with a small owner’s team is much
and lack of experienced personnel mean there is little chance
more likely to be a successful project than a brownfields project
of truly having an experienced team on your project.
with a large owner’s team using novel technology, simply due
to the complexity of the latter case and the number of interfaces If people from an ‘operations environment’ fill these project
that require managing (personnel and technical). Clear management positions without adjusting their behaviour, the
project definition becomes increasingly more important as the result is extended schedule and increased cost.
complexity of the project increases. Clear project definition is Unlike operations managers, project managers are not
not a function of increased management, increased complexity interested in process or maintenance optimisation once the
of procedures or layering of approvals. Clear project definition scope of work and services is agreed. Generally, a successful
is brought about by reducing the uncertainties associated with project has a high-level of definition during its design phases

Operations versus projects – how do people think and what are the implications?
and manages project execution to a defined scope. Variations •• during project review to check for any fatal flaws that
cost time (and money). One small design change, such as a may creep in during detailed design
change in valve size may require modification of 50 engineering •• as part of the operations readiness planning where
documents. With associated approvals and sign-offs this will ‘mod-squad’ projects can be identified early and plans
cost much more than the new valve. Hence, whilst ‘fit for put in place to address serious defects.
purpose’ is a paramount criterion, changes that refine this
concept are not critical in a project manager’s paradigm.
WHAT CAN GO WRONG?
The contractual and cost environment for a project manager
Input from operations is crucial to ensure rigorous project
is demanding. Performance against contract timelines is
design and without it projects can encounter trouble over issues
critical and if an existing strategy needs to be changed,
that could easily have been avoided. For example, poor chute
alternative suppliers are not easily sourced. Similar strategies designs can create plugging segregation, or high-wear areas.
as those of an operations manager may be engaged, for More critical are designs that have not considered maintenance
example using two suppliers for one commodity, but the access to equipment. In some cases, poor access to maintenance
complexities of transferring structural steel fabrication or equipment can create safety hazards where maintenance teams
erection scope from one supplier to another are far greater have to handle heavy materials in areas with little room to
than, for example, sourcing an alternative grinding media manoeuvre. Finally, it is important to review the flow sheet
supplier (in most cases). and layout for ‘fatal flaws’, ie design flaws that will prevent the
The implications of the above for an EPC contract are plant from performing as desired. The best time to catch these
greatly magnified. Profit in a competitive bid environment is in the prefeasibility phase or at the latest during the feasibility
is driven by completion ahead of schedule, cutting corners study. Safety issues and lack of access are issues where it can
and maximising the margin in variations to the contract. be necessary to make design modifications post-20 per cent
This environment magnifies the differences between the engineering complete, even if it impacts the schedule. Hence,
operations paradigm and the project management paradigm. it is important to include operations and maintenance thinking
prior to freezing the design.
OPERATIONS INVOLVEMENT – WHEN AND HOW MUCH? For greenfields projects it is usually more difficult to
Key questions from a project owner’s perspective are: finalise operator input because the operations team is often
not assembled until engineering has begun, or sometimes
•• When is the best time to freeze design? during construction. In these cases the success of the design
•• What defines a change outside the defined scope? is dependent on the quality of the design team. For the best
At some point during project execution the ambitions results, it makes sense to have people on the design team that
of operations and the project collide. From the operations have been involved in the post-commissioning, operations
perspective, operating knowledge needs to be embedded in and maintenance processes, but who understand the project
the project so that the plant can be effectively maintained implications. Otherwise designs can be very logical, but not
and operated. However, from the project perspective, early consider common operations/maintenance issues.
involvement of the operations team will lead to increased Above all, clarity regarding cause and effect is a key
capital cost and extensions to schedule. requirement to manage change during the reliability,
From both owner’s and EPCM contractor’s perspectives, accessibility, maintainability, build-ability and operability
one of the biggest issues is when to get the operations (RAMBO) process. ‘Wish list’ changes need to be differentiated
representatives involved in the project. If the operations from fatal flaws in the design. Strong and consensus
and maintenance personnel are brought on too late in the leadership from the owner’s and Contractor’s teams is
execution process their input is not able to be included in necessary to ensure that design change is minimised once
the design. If they are brought on too early and the input is critical commitments are made. Designers must understand
not critically assessed against project needs, the project cost the impact of their design on plant operation and operations
personnel must understand the cost and schedule (cash flow)
generally increases. The best approach is to have plenty of
impact of design change.
operations and maintenance input during the feasibility stages
of the project. This is much more easily done on a brownfields Another common problem, especially in greenfields projects,
project or where the owner has existing operations, because the is that operational maintenance management systems are
operations and maintenance groups are well established and implemented too late in the process or are incomplete. This
are aware of the types of issues created by existing designs. can cause many problems during the commissioning phase
because critical spare parts or warehouse stocks may not be
In general, any change creates rework and can delay the
available. This problem can be magnified if the feasibility
project. The design usually needs to be frozen in parallel with
study does not consider or underestimates the cost and time
the receipt of vendor data, at about 20–25 per cent engineering
involved for operating spares. Cutting operational spares is
complete. After this point even minor changes result in
an easy way to reduce capital spending, but is a false economy
numerous changes to documents. This is often an issue at
if operating time is lower during the commissioning/early
the time when the client’s operating team joins a project as operating phase of the project.
their preferences and paradigms are often not met by an
existing design. The resultant tension needs to be carefully There is nearly always a tension on the project between what
managed. The first stage of managing this process is mutual the operational team would like to have, what the project
understanding of the needs of each party. This was one of the can afford and what the designer would like to supply. The
degree of tension is a function of the contract type. Fixed
drivers for this paper.
price contracts necessitate a highly contractual (and clear)
Operations team involvement is most effective: decision-making and change control process. This makes
•• during the project feasibility study when the operations owner input more difficult once contracts are let and requires
wish list can be assessed based on a cost/benefit analysis a more detailed and extended design period with associated
without impacting on project schedule project schedule risk. Reimbursable contracts tend to be less

rigorously managed and even though change control is in there was substantial cost escalation associated with the plant
place, issues can arise due to poor decision-making processes design in the transition from option study to execution that
and rework and wasted time within the project. Projects with was related to the growth in bulk quantities in the concentrator
large owner’s teams tend to have greater issues with decision- design to meet the expectations of the contractor’s and
making than projects with small owner’s teams. Hence, owner’s teams as they evolved.
major mining houses have more issues controlling their Construction and commissioning of the brownfields
project budgets and schedules than junior mining houses project adversely impacted on the operating environment
(irrespective of the size of the project). and production from the existing plant. Although this was
anticipated, the extent of the disruption caused by the addition
CASE STUDIES of new front end processes was underestimated. This was
These case studies are based on the authors’ recent experiences particularly true of the amount of relearning required by the
with major projects. They represent a contractors’ perspective operators as the plant changed over a period of 12 months.
and an owners’ perspective of typical issues that arise during Lessons learnt included:
project execution, mostly due to the way the various parties •• keep the project execution plan simple and minimise the
react to project issues. contractor’s interfaces
•• clearly identify and continually communicate the vision
Contractor’s perspective for the project
From an EPCM contractor’s perspective there are three broad •• critically review and assess the project direction and
categories of project owner: make courageous decisions (as a team) and thus avoid
1. major mining house the need to ‘cost-cut’ to maintain budget
2. mid-tier mining house •• unless tightly controlled by the project and owner’s
3. junior mining house (and explorers). project directors, brownfields projects will maintain the
paradigm set by the plant operators
The extent of owner input, review and approval process
differs across this spectrum with major mining houses being •• brownfields projects impact adversely and materially on
the most demanding and complex in terms of owner’s team existing plant operations and are harder to commission
interface and impact on the control of project schedule and and ramp-up than a separate stand-alone (pseudo-
cost. The following case study pertains to an experience with greenfields project).
a project with an owner in the range of mid-tier to major
mining house. Owner’s perspective
Cost-effective design is the aspiration of most project owners. As previously stated, project schedule is arguably the most
Several years ago when considering project development critical factor in determining the success of a project. This
options for a major new mine, discussions between the owner argument is supported by the economic drivers of almost
and the EPCM contractor occurred around the nature of every project. A large-scale concentrator project has a high
the plant design for a new processing plant. The owner had capital cost and the best economic return of the project is
several operating plants and one of these was in the vicinity achieved by completing the project on time and ramping it
of the new project. The agreed concept at executive level up smoothly to design productivity, including both design
was to design and build a ‘cost-effective’ plant and critically performance of the equipment as well as design asset
assess operations and maintenance input to the design prior efficiencies. So it is important, once the project scope is locked
to inclusion of ‘nice to haves’ in the design. down, for the operating and maintenance team to focus on
ensuring a smooth start-up.
Several option studies were completed that focused on
the process flow sheet and whether to upgrade a nearby One of the most valuable activities for the operations/
concentrator to treat the ore from the new mine or build a maintenance team to engage in is what is commonly called
new stand-alone plant. The upgrade option (brownfield) ‘operational readiness’. This involves ensuring that all of the
allowed the use of existing capital equipment and associated necessary resources are ready for the handover of the plant.
infrastructure. The stand-alone plant option (greenfield) This involves hiring and training of the team, assembling the
allowed some of the existing paradigms to be tested, so both required operational spares, and developing the maintenance
had pros and cons. system such that surprises can be minimised. It is essential
during the commissioning phase that routine maintenance
As the work progressed from option studies to prefeasibility activities are handled effectively. The best outcome is when
to feasibility to implementation, the objective of establishing a these activities can be planned into the scheduled outages
new paradigm became more and more diluted as new people required by vendor run-ins. What needs to be avoided is
came onto the project. New people in the EPCM contractor’s having a critical piece of equipment breakdown and not to
team brought their design experiences and preferences that have the required materials on hand to complete the repairs
tended to be conservative rather than aggressive. As the efficiently. Since cash flow is so critical during this period, lots
owner’s team grew, influence from the operations side of of time and money can be wasted in having to rush materials
the owner’s business increased and the pre-existing design in, including priority freight etc. In the worst-case, when
paradigm re-established itself within the project. starting a project in a remote location, parts can take weeks
A complex execution strategy further complicated the or longer to arrive on-site. The effort of finding work-arounds
execution phase and made project communications and for these issues is draining, and the lost opportunity can be
interfaces more complex. financially disastrous to the project.
The final outcome was a plant design that was robust and In this case study, the operators’ team was involved early
flexible rather than ‘lean and mean’. Given the cost profile on with the feasibility design. As a result, once the scope
of the project where the process plant costs are small when was locked down, they were able to focus on the operational
compared with the mine infrastructure, this outcome is readiness phase of the project. The start-up team was hired
reasonable in the context of the overall project. However, and they used a local technical school to train a large number

Operations versus projects – how do people think and what are the implications?
of workers to be used in both operating and maintenance •• experienced personnel who can focus on the operations
activities. They also ensured a sufficient number of experienced and maintenance needs and not be distracted by the
workers were on-board (hired from other operations) to operate construction phase of the project
and maintain the equipment effectively. Numerous operating •• a dedicated budget.
and maintenance manuals were used to gain awareness of the
•• Project debottlenecking is an important part of project
specific equipment used in the plant. One of the challenges in
development and can’t be avoided by late design
this particular start-up was that the plant was designed with
changes during the initial project execution.
equipment that was novel to almost everyone on the team,
so there were many unknowns creating distractions from the •• Plants containing new technology are more complex to
establishment of routine maintenance activity. design, engineer, construct and commission due to the
learning curve.
Another challenge encountered in this case was that despite
it being a brownfields project, the projected plant was a
concentrator being constructed on a site with an existing CONCLUSIONS
leaching plant. This minimised the amount of site knowledge In a recent communication, a senior project executive from
available to enhance the design. In addition, the projected an owner’s team commented: ‘lessons are never learnt’ and
plant was much larger than the existing operation, so existing without an ‘A-team’, project teams will observe ‘that stuff
resources were ill-prepared for the size and scope of the project. will never happen to me’ and ignore wisdom gained from
Whatever the causes, the maintenance systems for the experience.
new plant were not completely addressed during the This paper states ‘the obvious’ in many ways, but ‘the
precommissioning phase. This made the routine maintenance obvious’ only becomes obvious through experience. The
functions much more difficult than usual, because the pressure on the current project development environment
maintenance team were not always able to find the necessary means that experience can be in short supply. The behaviours
materials to repair the equipment. This resulted in many of, and interactions between, key project personnel drive the
instances of having to use suboptimum materials to repair outcomes of projects.
equipment or to rush materials to site, including air-freight in Project managers need clarity and a clear path to timely
some instances, in order to keep the equipment operating. Since completion.
the correct materials were not always available, and because
of the lack of experience with the high-wear environment Operations managers need a plant that is easily operated
in a concentrator, ‘breakdown maintenance’ (rather than and maintained, supported by a robust operational readiness
preventative maintenance) became more common, reducing program.
operating time of the plant. The tension between the two sets of needs can add
Over time, the systems were implemented and the team tremendous value to a well-managed project when all those
was able to install high-level maintenance systems including contributing to a project understand and respect the diverse
reliability centred maintenance (RCM) techniques, predictive requirements of a successful project.
maintenance etc. As a result the plant is currently achieving
excellent asset efficiency numbers. However, the same results ACKNOWLEDGEMENTS
could have been achieved earlier if the maintenance systems The authors would like to acknowledge Ausenco and
had been in place prior to the always-hectic commissioning Freeport McMoRan for permission to publish this paper and
period. the various people who contributed insights from within and
In the end, a very successful debottlenecking project was outside these and other organisations.
implemented. Many of the experiences and lessons from
commissioning and early operation were used in this project. REFERENCES
Some of those were probably thought of during the later Biery, F, Hollands, A and Young, R, 2009. Minerals and metals project
engineering phases and construction. But it is unlikely that performance and improvement opportunities, in Proceedings
many of these plant modifications could have been imagined Project Evaluation 2009, pp 21–26 (The Australasian Institute of
prior to start-up. So, trying to implement these ideas during Mining and Metallurgy: Melbourne).
the engineering phase could have improved some areas Gabrielson, A, 2007. Eye on business – Current trends in project
marginally, but would have been costly, and a debottlenecking delivery, CIM Magazine/Bulletin, 2(7).
phase would still have been necessary.
Hundertmark, T, Siliva, A and Shulamn, J, 2008. Managing capital
Lessons learnt: projects for competitive advantage, McKinsey Quarterly, June.
•• Operational readiness requires:

Contents
Performance testing – when, what and how?

G Lane1, M Davis2, E McLean3 and J Fleay4
ABSTRACT
Performance testing is often used by project owners and financiers to provide
protection from, or assessment of, technical aspects of project design, procurement,
construction and commissioning. The performance tests can range from comprehensive
performance warranties relating to process performance and production in the case
of lump sum turn key (LSTK) projects to less onerous demonstration tests for cost
reimbursable contracts.
This paper will discuss performance testing from both the contractor’s and owner’s
perspectives and focuses on:
•• when performance tests are required
•• what type of performance test/warranty is optimum for each type of project
•• how project performance should be evaluated.
The requirement for performance testing can be driven by the project financiers,
particularly in the case of smaller resource development companies. In such cases,
differentiation between owner’s risk and contractor’s risk can become contentious. At
the other end of the spectrum, larger mining houses may choose to heavily influence
the engineering and design process with the contractor supplying services on a
reimbursable basis. In that case, the performance testing process is often limited to
demonstration tests that validate performance for the owner’s corporate governance
requirements.
This paper was prepared due to the lack of published benchmarking of performance
testing and the resultant differences in expectations between a contractor and a project
owner that often surface late in the contract negotiation process. The paper provides
a checklist of requirements for both the owner and the contractor that may be used
to provide ‘common purpose’ early in the contract negotiation process.
INTRODUCTION
Process plant performance testing is often used by project owners and financiers
to provide protection from, or assessment of, technical aspects of project design,
procurement, construction and commissioning.
This paper was prepared in response to the lack of benchmarking of performance
testing and the resultant differences in expectations between technology vendors,
contractors, project owners and financiers that often surface late in the project
contract negotiation process.
The discussion pertains mainly to minerals processing projects, for example gold
plants, base metals concentrators and hydrometallurgical plants, and minerals sands
plants.
Performance testing obligations vary with the type of contact. For LSTK work,
performance warranties and resulting performance test requirements are used
to define the minimum performance required for a plant or project soon after
commissioning. For engineering, procurement, construction and management
(EPCM) work, performance testing is typically conducted within 12 months of the
1. FAusIMM, Chief Technical Officer, Ausenco completion of plant commissioning.
Minerals & Metals, South Brisbane Qld 4101. Achieving minimum performance is particularly important for LSTK contracts
where a contractor is generally given more freedom inflow sheet definition and
2. FAusIMM, Managing Director, Sivad Resources equipment selection, and uses this freedom to optimise between plant performance
Pty Ltd, Darlington WA 6070.
and plant capital cost. For reimbursable contracts, where the owner’s input into plant
Email: michael.davis@sivadresources.com.au
design can be extensive, the role of performance testing is more complex, and the
3. FAusIMM, Manager Minerals Consulting,
testing process can be linked to contractual bonuses and/or be used to protect the
Ausenco Minerals & Metals, South Brisbane Qld
4101. Email: eddie.mclean@ausenco.com
contractor from ad hoc claims by the owner.
4. FAusIMM, Manager Metallurgy, Minnovo Pty Performance tests are expensive to undertake and should only ever be conducted if
Ltd, Leederville WA 6007. aspects of the plant or project, during the pretrial period post-ore commissioning, are
Email: john.fleay@minnovo.com.au underperforming with respect to the warranted criteria. Poor performance can be due
193
G Lane et al
to obligations not being met by the project owner as well as Over a 25 year period from the early 1980s to the current
poor performance by the contractor. A clearer understanding time, the use and application of performance warranties has
of the issues may result in additional and more cost-effective departed significantly with respect to the purpose and intent
rectification work being completed prior to a reassessment of from those initially used by the industry.
whether actual performance testing is required. Of course, performance testing has been used for many
This paper discusses performance testing from the years to ensure unit process performance where technology
technology vendor’s, contractor’s and owner’s perspectives. packages have been provided by vendors to projects, no
The testing processes associated with plant preoperational matter what the size and style of the project.
testing and commissioning are not specifically addressed,
other than where they interface with performance testing AIMS OF PERFORMANCE TESTING
post-commissioning.
A performance warranty, and the associated testing process,
is a method of risk management and quality control for
HISTORICAL PERSPECTIVE IN AUSTRALIA project development.
From a historical perspective, performance warranties played The requirement for performance testing can be driven by:
a notable role in the early development of some major mining
•• the project financiers, particularly in the case of smaller
companies and the, now ‘mid-tier’, Australian engineering
resource development companies
companies. Close (2002), in reporting the recent history of the
development of the Australian gold mining industry, noted •• larger mining houses where the testing process is often
that the developing Australian engineering companies in the limited to demonstration tests that validate performance
1980s: for the owner’s corporate governance requirements
… provided a completion guarantee that the plant would •• engineering contractors where a vertical package LSTK
function to the specifications set out in the contract subcontracting approach is used
that … this significantly reduced the risk for project owners •• engineering contractors where equipment performance
and made it easier for them to procure funding … criteria, such as mill power draw, are critical components
of a vendor supply package.
… significantly reduced the need for technical expertise in
the newly emerging gold mining companies ... and that ... It is not uncommon for junior resource companies and their
financiers to attempt to manage process risk by using the
… the large firms were simply unable to compete with the
LSTK contracting approach and passing on as much of the
keen pricing and aggressive smaller local suppliers for the
process risk to the contractor as possible. This approach is only
design and construction of smaller plants. Neither were the
successful if the project and process design criteria have been
large traditional design and construction engineering firms
such as Fluor, Davy McKee, Bechtel and Dravo willing to adequately defined in preceding feasibility study assessments
provide the completion and performance guarantees and and the ore reserve and mine plan are well defined.
the financing options that the smaller companies such as
Minproc offered their clients. TYPES OF TESTS
The early 1980s saw the emergence of the LSTK contract and The types of performance tests range from comprehensive
accompanying performance warranty in preference to the performance warranties relating to process performance
traditional EPCM contract. These were generally preferred and production in the case of LSTK projects, to less onerous
and commonly used by emerging, junior and new mid-tier demonstration tests of major equipment for reimbursable
companies. Typically the project was their first, or among contracts.
the first of their ‘company makers’. Nearly all were for gold The consequences of failing a performance test vary and are
projects. typically defined in the head contract or term sheet. Failure
In the 1990s, the proportion of LSTK projects decreased by the contractor can require full rectification in the case of
and there was a return to EPCM type contracts favoured LSTK works or re-performance of services in the case of a
by the newly established companies and major mining reimbursable EPCM contract. The consequence of test failure
houses. These companies had larger technical resource bases, by equipment vendors is typically limited to the maximum
either in-house or at-call from their portfolio of operations. liability under the liquated damages clause of the contract.
Projects were typically carried out by personnel with project The styles and types of tests are numerous. However, to
development experience (from the LSTK days). Many of the simplify matters for discussion, the performance tests can be
aspects of performance warranties were retained in the EPCM categorised into:
work. This period saw extension of performance warranties to
•• financier’s tests – typically between a financing body
base metal projects.
and the project owner and linked to the conditions of
Toward the end of the 1990s, several nickel laterite projects the loan (for example interest rate) to the project owner
were developed with varying contract types and associated
•• LSTK performance tests – typically between the project
warranties. Despite relatively onerous warranties and claims
owner and the LSTK contractor
on some of the engineers, the project owners and shareholders
were not protected from financial distress following poor initial •• EPCM performance tests – typically between the project
performance of some of these operations, whatever the cause. owner and the EPCM contractor for reimbursable
contracts
The trend in the first part of the 2000s has seen a continued
dominance in EPCM type jobs with increased technical •• vendor tests – required from major equipment vendors
participation and project direction by owners as well as to demonstrate that the supplied equipment performs in
a tightening of expectations and expansion of scope for accordance with the contract (for example power draw
performance warranties by owners and financiers. This for grinding mills).
period has seen further extension of performance warranties Each of these is discussed in further detail in the following
to the mineral sands industry. sections.

FINANCIER’S TESTS •• where the owner, typically a junior company, does not
have adequate technical expertise and experience to
Financier’s or banker’s performance tests can be the most
manage a reimbursable contract
extensive form of performance tests, as they can cover an
extensive range of project activities. These tests are typically •• where a financing institution insists on a LSTK contract
linked to the financing conditions for a project with a pass for the above reasons
resulting in reduction in project risk and a lower interest rate •• where a separable portion of a larger contract can be
for project financing to the project owner. The following list is identified and a specialist contractor can supply a
indicative of the type and extent of tests that may be required: packaged plant at a lower cost to the project.
•• mine reserve test – validation of the mine reserve estimate In all the above cases, the bidding process will force the
based on early mine development and production contractor to cut costs by simplifying the process flow sheet,
•• ore to waste ratio test – based on early mine performance simplifying the engineering and construction services and by
comparing the basis of the financial model with mine procuring low-cost equipment, where possible.
practice The aim of the performance warranty is to ensure that the
•• ore delivery test – based on early mine performance outcome of the LSTK contract is a plant that delivers the
comparing the predicted delivery rates with mine practice required performance. The performance may be measured by
items such as:
•• operating cost tests – based on early project performance
comparing the basis of the financial model with actual •• plant throughput
project performance •• grind size (P )
80
•• environmental/social compliance – based on project •• for a gold plant:
performance comparing the environmental/social •• leach efficiency
management plan with actual project performance
•• adsorption efficiency
•• management and staffing – comparing actual management
•• elution circuit efficiency
and staffing levels with planned performance; both under
and over manning may require explanation •• WAD cyanide reduction circuit efficiency.
•• contract compliance and close-out – evaluation of •• for a concentrator:
whether financial obligations have been met and •• recovery of metal(s) to concentrate
contracts closed in a satisfactory manner •• concentrate grade
•• financial management tests – assessment of project •• filtration circuit performance.
financial management processes
•• process or project specific performance criteria.
•• product quantity and quality tests – assessment of
Thus, the LSTK performance warranty provides a measurable
plant production and product quality compared with
quality control step at contract completion. In the 1980s, it was
forecasts, typically with a focus on measures of plant
relatively common for the performance tests to be completed
throughput and metal recovery and quality even though the plant was meeting performance targets.
•• insurances in place – assessment of whether appropriate More recently, best practice is to demonstrate performance
insurances are in place targets have been achieved using plant operating records and
•• marketing tests – assessment of whether product waive the performance tests, thus saving on the testing costs
marketing systems are in place and capable of marketing and upset to overall project operation.
product as planned The costs associated with LSTK performance tests in the first
•• power supply test – assessment of whether power instance are typically met by the project owner. Repeat test
supply provides adequate starting and operating costs are typically met by the party at fault in the preceding test.
capacity and reliability With LSTK performance warranties the consequence of
•• water supply test – assessment of whether water supply failure by the contractor typically requires the contractor to
provides adequate operating capacity and reliability complete all necessary rectification works at no cost to the
owner.
•• tailings disposal test – assessment of tailings storage
performance against plan with a focus on settled
densities and impacts on any indicated sustaining ENGINEERING, PROCUREMENT, CONSTRUCTION
capital requirements above plan. AND MANAGEMENT PERFORMANCE TESTS
There are occasional attempts to ‘back to back’ as many EPCM performance warranties have a much reduced
as possible of the financier’s tests requirements with LSTK consequence for the contractor as failure to achieve required
contractor performance. However, in practice, many of the performance typically requires that the contractor provide only
above tests are related to aspects of the project that are not the EPCM services for rectification at no cost to the owner.
within the contractor’s scope, for example how well the Reimbursable EPCM contracts typically have substantially
project was assessed in the feasibility stages, and how well more input from the owner into the flow sheet, process and
the plant is operated post commissioning. As a result, ‘back plant design and engineering. Plant equipment is typically
to back’ or tripartite agreements offer little advantage other purchased by the owner and all drawings are reviewed and
than to bring the interested parties together (often at a stage signed-off by the owner.
that is too late for beneficial effect).
As a consequence, poor plant performance that results
The cost of financier’s performance testing is met by the from a reimbursable EPCM contract is the outcome of shared
project owner. responsibility, with the owner and contractor contributing to
the outcome.
LUMP SUM TURN KEY PERFORMANCE TESTS There is a recent trend toward EPCM project owners
LSTK or vertical contracts are used in the following cases: asking for more detailed and onerous warranties, while still

G Lane et al
having substantial input to design and equipment selection. •• additional definitions

This can lead to conflict during the design phase where the •• warrantor’s obligations and warranties
engineer may raise issues with client design or equipment
•• warrantee’s obligations and warranties
preferences and seek written instructions from the client that
effectively waive the contractor’s warranty obligations. This •• procedures for testing (sampling, characterisation, data
process needs to be thought through by both parties prior to capture and data analysis)
commencement of the project. •• performance assessment and pass criteria
The bulk of the risk associated with poor performance of •• test reporting and warranty claims
this type of contract is with the project owner in the form •• limits of liability
of lower than expected cash flow due to reduced throughput
or lower than expected metal recovery. Performance •• arbitration.
warranties are typically aimed at ensuring that materials Aspects such as payment of costs, access to data and access
handling systems (conveyors, pumps and pipelines) and to site are best addressed under the obligations section of the
services (water and power systems) have adequate capacity. document.
Thus, the performance tests are restricted to throughput and A few ‘rules of thumb’ apply:
capacity tests rather than metallurgical performance tests
•• keep the document simple
based on metal recovery or product quality.
•• don’t repeat information that can be referenced to the
EPCM performance warranties are often associated with a
main contract
‘fee at risk’. In many cases the project owner may do better
to expend the cost of the fee on quality control and technical •• don’t refer to other documents, such as design criteria
review during the EPCM phase rather than entering into unless they are attached as schedules to the contract
performance warranties with a contractor. •• clearly state the obligations of the warrantor and
warrantee with respect to services and work required
VENDOR TESTS prior to and during the performance tests
Vendor warranty tests include all tests specified in contracts •• clearly quantify the warranties required of both the
between the owner or contractor and equipment vendors warrantor and warrantee for each relevant component
for major equipment items, such as mills and filtration of the plant and whether these warranties are to be met
equipment. The contractor typically supervises the conduct independently or collectively; each warranty should be
and assessment of these tests and provides a written report to associated with a clear definition of the data required to
the owner on the outcome of each test. demonstrate compliance
Typical vendor warranties are listed below: •• clearly state the period over which each performance
•• primary crusher – throughput test shall be run
•• semi-autogenous grinding (SAG) mill – power draw •• detailed procedures are not necessarily as long as a
method of agreement and arbitration; agree immediately
•• ball mill – power draw on detailed procedures prior to the commencement of
•• regrind mills – power draw the performance test
•• flotation cells – solid suspension and air dispersion •• clearly define the pass/fail criteria for each performance
•• thickeners – overflow clarity test, including the timeline and process for test
•• filters – throughput and moisture content of filter cake completion, reporting of warranty claims, arbitration
and rectification activities
•• blower performance
•• refer to limits of liability as stated in the contract.
•• vendor package performance – for example lime slaking
and flocculant system.
Some of the listed vendor warranties require that test work
PERFORMANCE EVALUATION
is conducted on ‘representative samples’ to enable the vendor The desired outcome for performance warranties is that all
to size the equipment and measure commercial risk. tests are waived by the warrantee on the basis that project
operating records, during the pretrial period post-ore
Vendor warranties enable the process risk to be shared
commissioning, demonstrate warranted performance by the
but do not necessarily obviate the owner’s risk. If the
warrantor (Lane and Messenger, 2005). This avoids the cost
representative nature of the sample tested is poor, lack of
and imposition of performance testing for all parties involved.
technical precedents and poor risk management can result in
The failure of a performance test is of little benefit to either
relatively minor cost to a vendor and large ongoing losses to
party unless the relationship between the parties has broken
the project owner until the process is upgraded or rectified.
down to an extent that obvious areas of underperformance
cannot be negotiated (in the absence of legal teams) to a
STRUCTURE OF PERFORMANCE WARRANTIES mutually successful conclusion. The effort associated with
AND CONTRACT ISSUES undertaking the performance test is better directed at the
The structure of the performance warranty can be made engineering solution.
reasonably consistent across all contract types and Performance testing places significant obligations and
circumstances. The performance warranty document is warrantee requirements on the project owner as well as
typically a schedule to the contract documentation. As such, the contractor. Churchill and Lane (1997) noted that for an
the principal definitions and references are in the contract EPCM contract:
rather than the performance warranty schedule. There a (Successful) Performance testing is reliant on the owner
number of key components to the performance warranty supplying suitable material for testing. Failure to do so
schedule, typically consisting of: within a reasonable period will result in default completion
•• introduction of performance testing.

Of course, the perspective given to performance data and prior to the contractor demobilising from site. This model
project outcomes differs between parties and as a consequence, is suited to owners who are new to the mining business and
performance tests are, on occasion, necessary. To minimise benefit from the handover of a near-fully operational plant.
the likelihood of disagreement over whether the plant has Alternatively, both parties may be best served by divorcing
reached performance targets during the pretrial period, the initial handover of plant operations and completion of
following process is recommended: performance tests. In this case, demonstration of compliance
•• prior to plant commissioning, jointly agree the data is based on plant operating data. The best model is usually
that is to be used for assessing performance compliance project specific and dependent on location, the availability
and agree to collate the data on a periodic basis of technical support, owner’s experience and the project
contracting risk/reward model.
•• periodically, jointly review and evaluate plant
performance data to determine the nature of any cause Reimbursable EPCM contract performance tests may be
of poor performance completed after the plant has been operating for an extended
period, and typically within the first 12 months of operation.
•• as warranted levels of performance are met, each party
As the objective is to demonstrate compliance with the tests’
should inform the other of the compliance
objectives based on operating data, it suits all parties to work
•• any outstanding warranties should be discussed and toward test compliance without the cost and potential upset
plans to rectify performance developed. to normal operation of actually running performance tests.
The above process differs from the more common outcome The durations of reimbursable EPCM contract performance
for LSTK, EPCM and vendor contracts where the warrantor tests are typically shorter (24 hours) than LSTK contract
demobilises from site and lays low waiting for the liability performance tests with the focus on throughput testing rather
period to lapse if there is an indication of poor project than operational/availability issues. As discussed earlier, this
performance. This approach involves less cost (for both change of focus is due to the much greater owner input into
parties) in the short-term but can lead to acrimony and a more plant design and equipment selection.
costly outcome in the longer term.
A CHECKLIST FOR FINANCIERS, OWNERS AND CONTRACTORS
TIMING AND DURATION OF PERFORMANCE TESTS A checklist of requirements for each of the various types of
The timing and duration of plant performance tests varies with performance tests is provided in Table 1 over the page.
the test type and parameters. Vendor equipment testing may
be conducted over short durations, for example grinding mill REFERENCES
power draw tests, while financier’s or banker’s performance
Churchill, S and Lane, G, 1997. Effective commissioning, in Proceedings
tests are often over extended periods (months) to allow Mindev 97 Conference (ed: E Barnes), p 247 (The Australasian
assessment of operational and financial performance. Institute of Mining and Metallurgy: Melbourne).
LSTK performance testing is often carried out soon after Close, S E, 2002. The Great Gold Renaissance, The Untold Story of
commissioning and prior to final handover of the facility to the Modern Australian Gold Boom 1982–2002, p 141 (Surbiton
the owner. The test durations are sufficient to demonstrate a Associates Pty Ltd).
suitable level of operability and availability (usually seven to Lane, G and Messenger, P, 2005. Commissioning, in Advances in Gold
14 days, sometimes longer). LSTK contracts in the late 1980s Ore Processing (ed: M Adams), Volume 15, p 176 (Elsevier).
and early 1990s required completion of performance tests

G Lane et al
TABLE 1
A checklist of requirements for each of the various types of performance tests.
Performance test type Banker Lump sum turn key Reimbursable Vendor
Parties Banks and owner Owner and contractor Owner and contractor Owner or contractor and
vendor
Participant B O O C O C O/C V
Document structure
Interface with head contract or term sheet    
Description of purpose    
Definitions and terminology    
Warrantor’s obligations    
Warrantee’s obligations    
Warrantor’s warranties    
Warrantee’s warranties    
Procedures
• sampling and test duration
Can be agreed at a later date and subject to arbitration.
• characterisation test work
Arbitrating party to be named in Performance Testing document.
• data capture
• data analysis
Performance assessment and pass criteria    
Test reporting and warranty claims    
Limits of liability †
   
Plant performance warranty types
Throughput    
Grind size   ? (mills)
Plant availability  
Process metallurgy
• metal recovery  
• concentrate grade (concentrators)  
• solution losses (gold plants)  
• concentrate moisture (concentrators)  
• equipment performance    
• rheology   
• head grade   
• ore characteristics  
• elution efficiency  
• cyanide destruction
Power draw or equipment specific performance 
Non-process engineering related tests
Mine reserve test 
Ore delivery test 
Mine performance tests 
Operating cost tests 
Environmental compliance 
Management and staffing 
Contract compliance and close-out 
Product quantity and quality tests 
Financial management tests 
Insurances in place 
Marketing tests 
Power supply test 
Water supply test 
Tailings disposal test 
B = banker, O = owner, C = contractor, V = vendor. † – Typically covered in head contract.

Unit design and
development
Contents
Process development and throughput forecasting

at the Phu Kham copper-gold operation, Laos PDR
D Bennett1, I Crnkovic2, P Walker3, A Hoyle4, A Tordoir5,
D La Rosa6, W Valery7 and K Duffy8
The Phu Kham deposit in Laos People’s Democratic Republic (PDR) represents a
highly variable and complex copper-gold porphyry system, and the original 12 Mt/a
concentrator commissioned in April 2008 lagged a long way behind most others in
terms of recovery of copper and gold into concentrate.
A focus during the original Phu Kham concentrator design stage was to counter
the expected metallurgical challenges by including first class instrumentation,
on-stream analysis and sampling systems to be able to provide process measurement
and collection of data. Once in production, gathering key variability data from daily
laboratory test work and monthly mineralogical composites helped to unlock the
secrets of the ore, and laid the foundations for the subsequent capital investments
in extra capacity and increased recovery. Process control opportunities have been
advanced in parallel to manage the ore variability and to further enhance throughput
and copper-gold concentrate production. The involvement of people with the right
skills and experience throughout all stages of the Phu Kham operations development
has always been considered as critical to success.
The results of all the process development and enhancements from commissioning
until June 2016 is presented in Figure 1, with throughput increasing from 12 Mt/a to
19 Mt/a, copper recovery increasing from approximately 60 per cent to 80 per cent,
and gold recovery increasing from approximately 35 per cent to over 50 per cent.
ABSTRACT
The Phu Kham deposit represents a copper-gold porphyry system, with mineralisation
present in skarn, stockwork and disseminated styles. Significant folding and alteration
events have created a complex heterogeneous geotechnical and mineralogy horizon,
which affect plant throughput and metallurgical performance. Weathering and water
table contact have created a leached zone, overlying transition zones with supergene
1. MAusIMM, Principal Metallurgist, PanAust
chalcocite-dominant secondary copper mineralisation and clay-rich gangue.
Limited, Fortitude Valley Qld 4006.
Email: duncan.bennett@panaust.com.au Primary ore copper mineralisation is mainly chalcopyrite with minor bornite. The
major challenges to the copper-gold flotation process are a wide size distribution of
2. MAusIMM, Manager Process Technical
Services, Ok Tedi Mining Limited, Tabubil, chalcopyrite mineralisation and poor primary grind liberation, a high pyrite content
Western Province, Papua New Guinea. in skarn ore requiring aggressive pyrite depression conditions, clay-rich gangue and
Email: ivan.crnkovic@oktedi.com non-sulfide copper mineralisation in weathered zones, and a significant association
3. MAusIMM, Director, Minmet Services of gold with pyrite.
Pty Ltd, Wilson Beach Qld 4800. The Phu Kham concentrator has been developed as a conventional semi-autogenous
Email: peter29walker@yahoo.com.au
grinding (SAG) and ball milling circuit followed by rougher flotation, regrinding and
4. MAusIMM(CP), Manager Minerals and cleaner flotation to produce a copper concentrate containing payable gold and silver
Metals, Ausenco, San Isidro, Lima 27, Perú.
Email: andrew.hoyle@ausenco.com
values. The original 12 Mt/a copper-gold concentrator flow sheet design offered a
capital efficient compromise between high copper recovery bulk sulfide flotation
5. Lead Drill and Blast, Group Mining Technical
and Sustainability, AngloAmerican, London with large cleaning capacity, and lower recovery copper selective rougher flotation to
SW1Y 5AN, UK. ensure concentrate specification of 24 per cent copper grade could be achieved.
Email: alan.tordoir@angloamerican.com Phu Kham commenced production of copper-gold concentrate in April 2008.
6. Principal Mining Engineer, CRC ORE, Flotation copper and gold recovery from commissioning was consistently poor, due
Queensland Centre for Advanced Technologies to the high levels of pyrite and problematic secondary and oxide copper species and
(QCAT), Pullenvale Qld 4069.
non-sulfide gangue. Incremental improvements in copper recovery were achieved by
Email: david.la.rosa@crcore.org.au
2011 through conversion of conditioning cells to flotation cells in both roughing and
7. FAusIMM, Global Director – Consulting and
first cleaning and increased cleaner capacity through the installation of a Jameson
Technology, Hatch – Mining and Minerals
Processing, Brisbane Qld 4000. Cell. In 2012 the Phu Kham Upgrade Project (PKU) was commissioned to increase
Email: walter.valery@hatch.com mill throughput, along with additional rougher and cleaner capacity to increase
8. MAusIMM, Process Consultant / Mining and residence times and maintain copper production at the nominal 16 Mt/a throughput.
Mineral Processing, Hatch, Brisbane Qld 4000. The marginal reduction in primary grind size due to the additional ball milling power
Email: kristy.duffy@hatch.com also provided a small recovery improvement.
201
D Bennett et al
FIG 1 – Phu Kham mill throughput and copper and gold recovery from commissioning until June 2016.
The Increased Recovery Project (IRP) commissioned in April concentrator. The project is owned and operated by Phu Bia
2013 targeted the major causes of copper and gold loss from Mining Limited. PanAust Ltd based in Brisbane, Australia
the circuit. Based on fundamental and detailed test work and holds a 90 per cent interest in PBM through its wholly
mineralogical analysis of concentrator streams, an optimised owned subsidiary Pan Mekong Exploration Pty Ltd, with the
design was developed which has increased copper and gold remaining ten per cent held by the Government of Laos PDR.
recoveries by over five per cent and ten per cent respectively. The Phu Kham copper-gold deposit is located in the
The IRP positioned the concentrator to process high pyrite Xaisoumboun province as shown in Figure 2, approximately
ores that were previously considered untreatable, and opened 120 km north of the Lao capital Vientiane. Access to the mine
further opportunities for maximising flotation recovery by is approximately four hours by road from Vientiane.
debottlenecking plant regrind and cleaning capacity.
The Phu Kham 12 Mt/a concentrator was commissioned in
To evaluate how to maintain design throughput rates with 2008 for a capital cost of approximately US$150 M, placing it
increasing ore hardness over the life-of-mine (LOM), Phu in the lowest quartile for capital intensity for copper mineral
Bia Mining Limited conducted a throughput forecasting and processing projects.
optimisation project for Phu Kham with the assistance of
The installed plant was a compromise between a high
Metso Process Technology and Innovation (PTI). The project
recovery but high capital intensity design, and a lower
involved a review of the blasting, crushing and grinding
recovery but technically lower risk and low capital intensity
processes and development of a throughput prediction model
design. The selective rougher flotation design was driven by
based on geometallurgical modelling for long-term planning.
the complex and variable mineralogy and high pyrite content,
The scope also included identifying opportunities for
with over 90 per cent of pyrite required to be rejected in order
increasing throughput and improving overall comminution
to produce a final concentrate of over 23 per cent copper.
circuit performance when treating the most competent ore
With increasing depth of the pit since the commencement of
types. A shorter term objective was to identify if and when
operations, the weathering profile of the feed has changed
secondary crushing or other process changes will be required
such that the ore became primary dominant in 2010, with
to maintain the target throughput over the LOM.
chalcopyrite the main copper sulfide mineral. The complex
This paper examines and discusses process and concentrator folding and alteration of the ore zones has meant continued
flow sheet development, including projects implemented mining of supergene and oxidised areas within the pit, with
since commissioning in 2008 to improve throughput and the copper mineralogy remaining diverse and varying from
copper and gold recovery with decreasing copper grade and native, oxide, secondary and primary copper species within
increasing pyrite content of ore feed, and increasing hardness short time periods.
of primary ore.
The development of the Phu Kham flow sheet was driven by
the poor recovery in comparison to other low-grade copper-gold
Introduction ores, and a need to counter decreasing ore grades from 2013.
The Phu Kham operation consists of a copper-gold mine using Major projects implemented up until 2011 included increasing
conventional shovel mining and truck haulage to a flotation rougher capacity by 25 per cent and increasing first cleaner

Process development and throughput forecasting at the Phu Kham copper-gold operation, Laos PDR
FIG 2 – Location of the Phu Kham copper-gold mine.
capacity by 16 per cent, and the installation of a Jameson Cell in 23 per cent copper concentrate by regrinding of rougher
a cleaner scalper duty. In 2012 the operation was upgraded to concentrate to 20 µm and additional cleaning flotation
a nominal throughput of 16 Mt/a with installation of a second capacity. Following extensive test work at bench, pilot, and
13 MW ball mill, a further 33 per cent increase in rougher full scale which proved the concept and led to investment
capacity, 40 per cent increase in second cleaner capacity, and approval, in 2013 the operation increased total recovery of
33 per cent increase in third cleaner capacity. both copper and gold by approximately six per cent into final
In 2009 a project to achieve step-change in copper and concentrate through the Increased Recovery Project (IRP). A
gold recovery from Phu Kham was initiated. A process second filter was installed during the project to dewater the
development study was completed in 2011, which showed additional concentrate produced.
that it was technically feasible to generate a low-grade copper As mining extends deeper into the deposit, the operation
and gold mineral concentrate by bulk sulfide flotation of will experience an increased proportion of highly competent
concentrator tailings suitable for leaching for recovery of ores which will have the potential to limit plant throughput,
copper and gold into high-grade products. During the study, through the semi-autogenous grinding (SAG) mill. Phu
opportunities for increasing copper and gold recovery in the Bia Mining commenced a throughput forecasting and
existing concentrator using standard processing methods optimisation project in 2012 to evaluate how to maintain
became apparent, and detailed mineralogical work and design throughput over the LOM. Insufficient comminution
metallurgical test work was undertaken to determine the data in the mine block model created a lack of confidence in
causes of copper and gold loss to tailings. The mineralogical the ability to predict mill throughput, particularly in the later
work revealed that up to 60 per cent of copper sulfide mineral years of the mine life.
lost was in coarse non-sulfide gangue composites, and over
50 per cent of gold loss was in gold-pyrite composites. Geology and mineralogy
The work presented an opportunity to recover these The Phu Kham geology is highly variable due to weathering,
composites by less selective rougher flotation, before upgrade alteration, faulting and folding. The deposit consists of
and additional recovery for both copper and gold into a complex heterogeneous mineralogy horizons of copper-gold

D Bennett et al
stockwork and skarn mineralisation as shown in Figure 3. banded to massive pyrite skarns and veinlets containing
Weathering and water table contact have created a soft pyrite, chalcopyrite and bornite in garnet, magnetite
leached zone, overlying transition zones with supergene and hematite-chlorite skarns. Pyrite skarns are common
chalcocite-dominant secondary copper mineralisation throughout the mineralised system.
and clay-rich gangue. The rock mass strength and degree Stockwork mineralisation is present as fine fractures in
of weathering vary considerably across the deposit with quartz veins. The fractures host pyrite, chalcopyrite and
extremely competent (hard) rock found in the deeper levels. bornite sulfide minerals. Minor chalcopyrite mineralisation
Such variability causes a large range of plant throughput and is also present in quartz-carbonate veins. Disseminated
metallurgical performances. mineralisation consists of scattered grains of bornite and
Mineralisation is present in iron-rich skarns, silica-rich chalcopyrite in sericite altered host rock.
stockwork and altered disseminated styles. Chalcopyrite Gold occurs as small grains associated with pyrite and
and bornite are the dominant primary copper minerals in copper sulfides throughout the mineralised system.
skarn, stockwork and disseminated mineralisation. Gangue
mineralogy is mainly quartz, mica and pyrite, with significant
kaolinite clay and talc-related magnesium silicate content CIRCUIT DESCRIPTION
within the weathered zones. The original 12 Mt/a concentrator design and commissioning
A gold-enriched oxide zone on the Phu Kham orebody was in 2008 has been described in detail by Crnkovic et al (2009).
the resource for the heap leach gold mine which was built and The description of the 2016 concentrator following the PKU
operated by Phu Bia Mining during the 2005 to 2010 period. and IRP is provided in this section and the simplified flow
Below the oxide zone, there is a zone of supergene weathering, sheet is presented in Figure 4.
with copper leached from the oxide zone re-precipitated The crushing plant consists of a primary 55 in × 77 in
in contact with pyrite grains as particles and coatings of gyratory crusher, with single truck dump point above a pocket
chalcocite and covellite, with minor enargite and tennantite designed to hold 200 t capacity equivalent to two 777D haul
copper arsenic sulfides. Significant copper enrichment in the trucks. Crusher discharge drops to a crushed ore bin of 200 t
oxide and supergene zones is also present as oxide and native capacity. The crushed ore bin is emptied by a variable speed
copper species. apron feeder onto a crushed ore transfer conveyor belt (CV-
Skarns are present as replacement of carbonate minerals, 001). The CV-001 conveyor transfers the ore to an 890 m long
with disseminated grains of chalcopyrite and bornite in overland conveyor CV-002, which moves ore to the coarse ore
FIG 3 – Phu Kham geological zones.

FIG 4 – 2016 Phu Kham 18 Mt/a concentrator simplified flow diagram.
stockpile with a live capacity of approximately 24 000 t. There rougher cells in the No 1 rougher bank, with the conditioning
is additional dead capacity for storage of up to 300 000 t of ore. tank converted to a ninth flotation cell in 2009. In early 2011, a
Ore is reclaimed from the crushed ore stockpile by two tenth 200 m3 rougher cell was installed and commissioned in
variable speed apron feeders onto a SAG mill feed conveyor. the No 1 rougher bank.
SAG mill grinding media is added to the ore feed conveyor Cyclone overflow from the No 2 milling circuit reports to
via a spillage return hopper. Primary grinding is achieved in a 70 m3 agitated tank with two discharge pumps, one which
a dual pinion 13 MW variable speed slip energy recovery/ reports to the mixing box before the No 1 rougher bank, and
hyper-synchronous drive 34 ft × 20 ft SAG mill in closed circuit another which reports to the No 2 rougher bank of five 200 m3
with scats return conveying including a high-lift conveyor to tank cells via a multiple stage feed sampler.
overcome topography constraints. Dithiophosphate collector is used to recover copper sulfide
SAG mill discharge is classified using an integral mill minerals in flotation while maintaining selectivity against
trommel, with minus 12 mm product reporting to a 1.85 MW pyrite.
cyclone feed pump. In the original circuit, cyclone feed was Rougher banks tailings pass through static dual fin pipe
classified in a single cluster of 18 × 650 mm diameter cyclones, samplers before reporting to final tailings mixing box where
with cyclone underflow reporting to the No 1 dual pinion it is combined with cleaner scavenger tailings. The mixed
13 MW drive 40 ft × 24 ft ball mill in closed circuit. In the tailings discharge to a metallurgical multiple stage sampler
upgraded plant, a bleed stream of approximately 50 per cent and a final tailings sump. The final tailings sump discharges
of the SAG mill feed mass is diverted via a transfer pump slurry by gravity through two 750 mm diameter tailings lines,
to a 1.85 MW cyclone feed pump reporting to a cluster of which transport tailings approximately 1.5 km to a cross-
17 × 650 mm diameter cyclones, with cyclone underflow valley subaqueous tailings storage facility (TSF).
reporting to the No 2 dual pinion 13 MW drive 40 ft × 24 ft ball Rougher concentrate from both rougher banks reports to
mill in closed circuit. Quicklime slurry is added to the SAG a common cyclone feed pump hopper before classification
mill and ball mills for flotation pH control to depress pyrite. in a cluster of 12 × 400 mm diameter cyclones, with cyclone
Cyclone overflow from the No 1 milling circuit reports overflow reporting to a Jameson Cell feed hopper. Cyclone
to a mixing box, where it joins a bleed of cyclone overflow underflow reports to two open circuit M10000 IsaMill™
from the No 2 circuit to balance the volumetric feed split to regrind mills operating in parallel. IsaMill™ discharges
the No 1 rougher bank. The mixed cyclone overflow passes report to the Jameson Cell feed hopper. The 24 downcomer,
through a multiple stage feed sampler before the No 1 bank of 6500 mm diameter Jameson Cell operates in a cleaner feed
ten 200 m3 tank cell roughers. The original plant had a single scalping duty, with concentrate passing through a pipe
200 m3 rougher feed conditioning tank before eight 200 m3 sampler for process control, before reporting to the final

D Bennett et al
concentrate thickener feed sampler, and tailings reporting to Concentrator raw water is harvested from the Nam Mo River
the conventional cleaning circuit. before being pumped to a crusher process water tank and mill
The conventional cleaning circuit consists of three stages, header tank. The raw water is mainly used for cooling, pump
with the first stage in open circuit. The first stage of cleaning/ glands, flotation froth wash showers and fire water. Process
cleaner scavenging consists of two parallel banks of seven water is recovered from the TSF supernatant, and transferred
70 m3 tank cells for a total of 14 cells. Cleaner scavenger tailings to a process water tank via two transfer stations.
pass through static dual fin samplers before reporting to the
final tailings mixing box. First cleaner and cleaner scavenger process improvements and FLOW SHEET DEVELOPMENT
concentrates report to the second cleaner, which consists of
The Phu Kham 12 Mt/a concentrator was designed and built
seven 20 m3 cells. Second cleaner concentrate advances to the
to treat a high pyrite copper-gold skarn ore with significant
third cleaner of four 20 m3 cells, while second cleaner tailings
clay content, as described by Meka and Lane (2010). The plant
return to the first cleaner feed pump hopper. Third cleaner
was commissioned in April 2008 and had ramped up to meet
concentrate passes through a pipe sampler, before reporting
nameplate capacity and operating time design by November
to the final concentrate thickener feed sampler. Third cleaner
2008. A simplified flow sheet for the original 12 Mt/a
tailings return to the second cleaners.
concentrator is shown in Figure 5.
Final concentrate (combined Jameson Cell and third cleaner
A photograph of the 12 Mt/a concentrator in June 2008 is
concentrates) is sampled in a multiple stage metallurgical
sampler, before gravitating to a 15 m diameter high-rate shown in Figure 6.
thickener. Thickener supernatant flows to a thickener overflow
process water tank. Thickener underflow at a nominal density 2009–2010
of 65–70 per cent solids is pumped to a mechanically agitated
filter feed tank of approximately 24 hours surge capacity, with Flotation cell conversions
excess production bled to a second mechanically agitated filter In September 2009, the existing rougher conditioner and
feed tank of approximately 8 hours capacity. The thickened cleaner conditioner tanks were retrofitted with flotation
concentrate slurry in the 24 hour capacity tank is dewatered mechanisms, thereby increasing the roughing capacity from
using a 64-plate horizontal filter, with filter discharging into a eight to nine 200 m3 cells, and increasing the cleaner capacity
covered storage shed of nominal 8000 t capacity. The thickened from six to seven 70 m3 cells. The benefits arising from these
concentrate slurry in the eight hour capacity tank is dewatered changes amounted to increased copper recovery in the rougher
using a 40 plate horizontal filter, with the filter discharging into flotation circuit by 3.5 per cent, and increased copper recovery
a separate covered storage shed of nominal 5000 t capacity. in the cleaner circuit by 2.5 per cent. The increased residence
Concentrate is loaded into 25 t containers for transport by truck time in each of the circuits resulted in higher recovery of slow
to the Sriracha port in Thailand or ports in Vietnam. floating secondary copper minerals, and composite particles
FIG 5 – 2008 Phu Kham 12 Mt/a concentrator simplified flow diagram.

rougher residence time by 2.4 minutes at maximum design

throughput of 1750 t/h. In late January 2011, the tenth 200 m3
rougher tank cell was installed at the head of the rougher
circuit. At the time that the tenth rougher cell was installed,
the flotation feed rate was increased by about four per cent,
which effectively reduced the overall rougher residence time
increase from an expected 11 per cent to seven per cent. The
overall surveyed copper recovery improvement was between
0.4 per cent and 0.6 per cent, depending on throughput rate,
which met the project criteria.
FIG 6 – Southern view of the Phu Kham concentrator in June 2008. Cleaner circuit debottlenecking
One of the major limitations in the original plant design had
particularly while transitional ore types were dominant at
been a lack of cleaning capacity, particularly with respect to
this time. The improved recovery performance was validated
the second cleaner bank of four 20 m3 cells. The limitation
from using the database of daily rougher tail and cleaner was that of carrying capacity, rather than residence time. The
scavenger tail re-flotation tests. cleaner circuit performance would generally deteriorate when
the cleaner feed copper metal units exceeded 9.1 t/h copper,
FloatForceTM mechanisms limiting copper metal production to a sustainable maximum
An investigation into using Outotec FloatForce™ rotor-stator of approximately 8.5 t/h.
mechanisms in the rougher flotation circuit commenced To further investigate the cleaning circuit capacity, a cleaner
in 2009, starting with the first rougher cell 1B. The design circuit optimisation study was completed in February 2010.
of the new mechanism was to deliver improved recovery From plant data, a mineral based floatability component
from increased mixing efficiency, by not allowing any air in model was developed which allowed different cleaner circuit
the central mixing area of the impellor thereby improving configurations to be simulated. The option which gave the
mixing efficiency without affecting slurry pumping. The optimum copper grade and recovery result was to install
installation of the FloatForce™ mechanism was simple, and additional cleaning capacity ahead of the existing cleaner
was commissioned without any issues. The conclusions, circuit, so effectively cleaner feed scalping. Different flotation
drawn from extensive survey data around rougher cell 1B cell technologies were considered for this application, with the
(before and after installation), demonstrated an improvement Xstrata Jameson Cell meeting design criteria. The simulations
in rougher cell 1B copper recovery of 0.2 per cent. On the indicated that a 0.6 per cent improvement in cleaner recovery
basis of the survey data, the total predicted rougher copper could be achieved. The Jameson Cell was chosen because
recovery increase for the nine rougher cells with FloatForce™ of low installed cost, confidence in simulated performance
mechanisms was 0.3 per cent. The installation of the remaining results, low performance risk, moderate installation risk and
eight mechanisms was completed in June 2011. low production continuity risk during installation.
The circuit simulations including the Jameson Cell
2011 as a cleaner feed scalper indicated substantial recovery
improvement over the existing circuit at circuit feed rates
Tenth rougher cell installation greater than 150 t/h, due to elimination of the carrying
The ongoing rougher tail re-flotation tests continued to capacity limitation as shown in Figure 7.
highlight that there was a further copper recovery benefit On this basis, it was decided to proceed with this design, the
of approximately 0.6 per cent to be gained with the Jameson Cell in a cleaner feed scalping simulation recovering
addition of another 200 m3 tank cell, which would increase approximately 60 per cent of the copper present in the
FIG 7 – Effect of cleaner circuit feed tonnage on copper recovery for Jameson Cell cleaner feed scalping simulations.

D Bennett et al
cleaner circuit feed across a cleaner circuit feed rate range of of the upgrade designs for Phu Kham were the limitation
100 t/h to 300 t/h. The Jameson Cell concentrate grade from of available space for additional equipment, as the original
the simulation was 27 per cent copper, against a target of 12 Mt/a plant design had not specifically made allowance for
25 per cent copper. The simulations showed that substantial any expandability.
unloading of the remainder of the cleaner circuit would occur. The initial phase of the PKU included a plant debottlenecking
The result of this was that the third cleaner concentrate grade study, which consisted of analysis of the actual plant
was low at less than 22 per cent copper, however the net effect performance and capacity data from 2008 to March 2010
was to produce an overall circuit final concentrate grade of against the original plant process design criteria. The purpose
24 per cent copper. of the bottleneck study was to determine aspects of the original
Although there was a small cleaner recovery improvement plant that either were, or would become bottlenecks with the
shown from the simulations performed using a cleaner 25 per cent increase in mill throughput. The key findings
scalper, the real benefit was in maintaining cleaner recovery from the plant bottleneck study are shown in Figure 8, which
when the cleaner feed rate is greater than 150 t/h. shows that rougher copper recovery was 16 per cent below
The cleaner scalper cell was commissioned in March 2011. design, cleaner copper recovery was seven per cent below
Commissioning was carried out over a period of one week, and design, and mill throughput was three per cent below design
no significant problems were encountered. The performance at 12 Mt/a. The mill throughput variance was a function of
evaluation of the Jameson Cell was remarkably consistent rougher copper recovery and cleaning circuit capacity rather
with the expected performance from the equipment vendor, than limitations in the grinding circuit.
the simulation data and from Phu Kham Metallurgical The crushing and concentrate dewatering plant capacities
Laboratory flotation tests simulating performance of the were also considered during the PKU bottleneck study. The
Jameson Cell prior to commissioning. From surveys carried bottleneck study indicated that additional crushing capacity
out in February 2012, with the Jameson Cell online and would be required with capability for handling wet and
offline, the benefit of having the Jameson Cell in circuit was sticky ore which is a common feature of the transition zones
determined to be 0.8 per cent increase in copper recovery. of the orebody. A mineral sizer in parallel to the existing
In terms of overall cleaner circuit debottlenecking, the crushing plant, with product reporting directly to the coarse
objectives were achieved. The cleaner circuit with cleaner ore stockpile was included in the PKU designs. Although
feed scalping capacity is 10.1 t/h copper metal at 24 per cent the concentrate thickener and filter performance had not
concentrate grade for a total 16 per cent copper metal indicated that future concentrate production rate would
production increase. exceed capacity, limited data was available to confirm the
capacity against upgraded plant design criteria. Test work
2012 was conducted to determine settling rates and filtration rates
for concentrate during the PKU to obtain the required data.
Phu Kham Upgrade project The basis of design for the grinding circuit upgrade has
The Phu Kham Upgrade (PKU) project commenced in March been described by Hadaway and Bennett (2011). Two options
2010 with a study to develop designs to ensure that copper for increasing grinding circuit throughput after the SAG
in concentrate production was maintained over 60 kt/a mill to a nominal design of 16 Mt/a and primary grind at
after 2013 when plant copper feed grade was expected to 80 per cent passing 106 µm or 75 µm were reviewed. The first
decrease. In order to maintain copper metal production, plant option was based on the original 2008 plant upgrade design
nominal design throughput increased from 12 to 16 Mt/a incorporating an additional 6.5 MW single pinion ball mill,
(1500 to 2000 t/h) and maximum instantaneous design and the second for another 13 MW ball mill.
throughput increased to 2250 t/h. The plant upgrade concept Data from JKTech grinding circuit modelling in 2009 was
was not original, and had been studied in 2008 as part of a extrapolated using the Phu Kham mine schedule to determine
copper production expansion project concept. Key aspects throughput estimates at 106 µm and 75 µm for the two mill
FIG 8 – Phu Kham 12 Mt/a concentrator actual performance variances against design 2008–2010.

options. Pebble crushing was not considered in the evaluation. The dominant cause of the 16 per cent rougher copper
The 6.5 MW mill was able to increase throughput at a 106 µm recovery shortfall shown in Figure 8 was a combination of
primary grind to above the 16 Mt/a nominal design, however lower than design rougher residence time due to five per cent
was unable to meet the 16 Mt/a target at a significantly finer lower rougher feed density, and a cleaner circuit capacity
primary grind. The 13 MW mill was able to achieve above constraint which limited the rougher mass recovery. The
18 Mt/a for the 106 µm primary grind, and could achieve dominance of transition ores with significant slow floating
above nominal design throughput at a 75 µm primary grind. secondary copper mineral content milled during the March
2009 to February 2010 period and the under-representation
The effect of primary grind on flotation recovery was
of these ore types in the feasibility study test work provide
reviewed based on feasibility study work from bench scale
explanation for some of the copper recovery shortfall in
batch tests in 2005. The study work indicated that the major
cleaning stages against design. The debottlenecking study
primary ore sources, in particular stockwork primary, were was developed for the flotation circuit to determine increased
relatively insensitive to primary grind size. Plant operations capacity requirement at the PKU design 16 Mt/a throughput.
mineralogy data from 2008 to 2011 monthly composites
Rougher flotation feed density design for the 12 Mt/a plant
indicated that minor sensitivity existed, with increases of over
was 35 per cent solids. Actual operation rougher feed density
five per cent in copper sulfide liberation with a primary grind
averaged 30 per cent solids due to the higher slurry viscosity
size decrease from 80 per cent passing 106 µm to 75 µm.
from kaolinite clay content not quantified during the feasibility
An economic analysis was conducted based on differences study. An extra 200 m3 rougher cell was required to achieve
in capital and operating costs for the two options at 16 Mt/a the same residence time as at 35 per cent solids, which was
throughput and 106 µm and 75 µm primary grind. The increase achieved by conversion of the rougher conditioning tank to a
in operating cost for finer grinding versus revenue benefits in cell in 2009. The reduced residence time from operating at the
copper recovery showed that above $2.50/lb copper price, the lower rougher density at design tonnage throughput resulted
finer primary grind increased gross margin. Capital cost per in a three per cent decrease in copper recovery, based upon
installed megawatt was 26 per cent less for the 13 MW mill plant residence time–recovery data from July 2009 to February
option, and the capital payback period for the 13 MW option 2010. The PKU design therefore allowed for reduced rougher
was significantly shorter. feed pulp density, and a residence time calculation confirmed
that a 33 per cent increase in rougher capacity was required
A risk assessment was conducted for the 6.5 MW option
for the 25 per cent increase in mill throughput at 16 Mt/a,
and the 13 MW option. The risk of the 13 MW option was
which would also provide an additional one per cent copper
considerably lower than for the 6.5 MW option, mainly due
recovery. A total of five 200 m3 rougher cells in addition to
to the operating flexibility for periods of low-grade ore and
the existing ten cells were included in the design, for a total
ore types with higher sensitivity of recovery to primary
of 15 cells.
grind. The 6.5 MW option was not able to take advantage
of economies of scale gained by increased throughput, or The cleaning circuit was not expected to require significant
expansion as a result of the 16 Mt/a upgrade, as the lower grade
the estimated one per cent increase in copper recovery at
mill feed would result in equivalent concentrate production
the finer grinds, and would not be able to reach the nominal
to the 12 Mt/a design throughput rate. The PKU design for
16 Mt/a throughput at 106 µm primary grind after 2014. The
the cleaner flotation circuit also included the Jameson Cell
throughput at 106 µm and 75 µm primary grinds for the two
cleaner scalper although this had not been installed at this
mill options is shown in Figure 9.
time. However, cleaner circuit mass balance simulation data
Based on the results of the risk assessment, the including the Jameson Cell indicated that 40 per cent increase
recommendation for installation of an additional 13 MW ball in the existing second cleaner residence time and lip length
was accepted. Procurement of a second dual pinion 13 MW was required at 16 Mt/a. To gain this increase in second
drive 40 ft × 24 ft ball mill commenced in November 2010. cleaner capacity, the existing three 20 m3 third cleaner cells
FIG 9 – Throughput at 106 µm and 75 µm primary grind for additional 6.5 MW and 13 MW ball mill options.

D Bennett et al
were combined with the four 20 m3 second cleaner cells to 2400 t/h and target utilisation of 75 per cent of total time. With
create a second cleaner bank of seven 20 m3 cells, and four increases in haul fleet numbers for the PKU, improvements
new 20 m3 third cleaner cells were added for the PKU. The in run-of-mine stockpile inventory, and a standby loader
simulation data for the cleaners also demonstrated that available when there were delays in truck presentation to the
upgrade of the first cleaner capacity was not warranted, as crusher, the MSPR demonstrated that increasing crushing
the fine low-grade middlings recovered in the final cells were capacity was not required.
not able to be upgraded to near final concentrate specification.
The design specifications for the 64 plate and frame filter
Prior to commencement of the detailed design phase of the
were for a filtration rate of 225 kg/m2/h, with an annual
PKU study, maximum sustainable production rate (MSPR)
analyses were undertaken for the crushing and concentrate design production rate of 311 000 t of concentrate. Actual filter
dewatering plants to determine whether capacity expansion plant operating data was analysed to check filter performance
was required for these areas of plant based on 16 Mt/a against the design capacity. MSPR for the filter was
production schedules. The MSPR was defined as the best determined to be 18 per cent above the life-of-mine maximum
consecutive five days of performance, normalised using concentrate production schedule, leading to deferral of capital
plant-specific industry standards for annual availability to expenditure for the filtration plant. The main reasons for the
allow for major scheduled maintenance. The MSPR for the higher than design performance were; optimisation of filter
crushing plant also included seasonal variation due to the cycle settings following an improvement program including
tropical environment and the wet season impacts on crusher operations, maintenance, and vendor support input and
productivity. change to filter cloth media type.
The primary conclusions from the performance review of
Following the review of the PKU design, engineering and
the Phu Kham crushing plant were that it had demonstrated
procurement services commenced for the 16 Mt/a PKU
the target PKU production rate of 16 Mt/a over the June 2010
project in January 2011, with commissioning commencing in
period, and approximately one-third of total crushing plant
downtime had been caused by events up and downstream the third quarter of 2012. Ramp-up to nameplate capacity was
of the crushing plant while the plant was available to crush. achieved within two months, and over 17 Mt/a annualised
The low crusher utilisation of 64 per cent as a result of these throughput rate was achieved in the fourth quarter of 2012.
operating standby periods was equivalent to over 2.4 Mt/a A simplified flow diagram for the PKU plant is shown in
of additional crushing capacity at the target throughput of Figure 10, with new equipment highlighted in mauve.
FIG 10 – 16 Mt/a PKU simplified flow diagram.

2013 flotation option consistently achieved over 22 per cent copper

final concentrate grade, however ultimate copper recoveries
Increased Recovery Project to final concentrate were lower than the bulk flotation option.
The Phu Kham feasibility studies between 2004 and 2006 The design of the original 12 Mt/a Phu Kham concentrator
identified two options for flotation processing of Phu Kham was a compromise between the two process options, with
ore. The first option involved bulk flotation of the rougher partially selective roughing being applied to minimise
feed targeting a 25 per cent mass recovery into rougher pyrite gangue recovery into cleaner flotation feed. Sodium
concentrate using non-selective amyl xanthate sulfide mineral cyanide addition to the cleaners was included in the original
collector. The rougher concentrate was then reground to design, however was never used with concentrate grade
over 22 per cent copper consistently achieved following
80 per cent passing 38 µm and subjected to cleaner flotation
commissioning. This partially selective flotation process had
at pH 12 for pyrite depression. This process produced
significant capital cost advantages over bulk flotation at a time
high copper recovery results, however there was difficulty
when the long-term copper price forecast was less than $2.00/
achieving final concentrate grade of greater than 22 per cent
lb, due to the lower rougher concentrate regrind and cleaning
copper across all ore types, particularly transition chalcocite-
capacity required, and provided the best cost-benefit process
covellite secondary copper mineral dominant ores. A rougher alternative while reducing risk of being unable to achieve
feed photomicrograph is shown in Figure 11 with chalcocite- concentrate specifications using bulk rougher flotation.
covellite intergrowth with pyrite and rimming of pyrite.
Minimal work was performed during the feasibility
There was also indication of copper activation of pyrite from
studies to test the sensitivity of final copper grade and
soluble copper species in weathered and transition ores.
recovery on rougher concentrate regrind product particle
The second process option involved selective flotation in size. Grind size analysis was limited to two mineralogical
roughing at pH 11–12 using a copper sulfide selective collector. examinations which concluded that reasonable copper and
The rougher concentrates were again reground to 80 per cent gold recoveries to rougher concentrate would result from
passing 38 µm and lime to pH 12 and sodium cyanide was a primary grind of 80 per cent passing 106 µm, and that a
added to the cleaning stages to depress pyrite. The selective rougher concentrate regrind to less than 45 µm was required
to achieve an acceptable final concentrate grade. Mineral and
liberation analysis showed that associations between copper
sulfide minerals and pyrite did not indicate complex or fine
intergrowths that would adversely impact on the metallurgy.
The Phu Kham IRP commenced in 2009 as part of a
concept study to develop a process to increase copper and
gold recovery from Phu Kham ore. Since commencement
of operations in 2008, copper recovery had increased with
increasing proportion of chalcopyrite-dominant primary ores
replacing the chalcocite-covellite secondary copper mineral
dominant high clay and talc transition ores. The increasing
plant throughput and poor primary liberation with increasing
pyrite content caused copper and gold recovery to ‘flat-line’
as shown in Figure 12.
Bulk sulfide flotation bench tests in 2009 using isopropyl
xanthate collector on Phu Kham rougher tailings indicated
FIG 11 – Phu Kham transition ore rougher feed photomicrograph. that up to ten per cent additional copper recovery and
Cp – chalcopyrite, Ch/Cv – chalcocite/covellite, 70 per cent additional gold recovery could be achieved into a
Py – pyrite, Gn – non-sulfide gangue. scavenger concentrate of approximately 0.8 per cent copper.
FIG 12 – Phu Kham concentrator copper and gold recovery by year 2008 to 2011.

D Bennett et al
The initial test program was designed to determine whether a TABLE 1

low-grade copper-gold concentrate suitable for downstream Plant tailings flotation test results summary.
hydrometallurgical processing to saleable products could be
recovered from the concentrator tailings. Copper Gold
The preliminary test program showed that a bulk sulfide Stream Grade Distribution Grade Distribution
concentrate from plant tailings flotation could be upgraded to % Cu % Au ppm %
over ten per cent copper concentrate grade, depending upon
copper sulfide mineral liberation, using a roughing, regrind Plant tailings (feed) 0.17 100.0 0.15 100.0
and two stage cleaning process similar to the Phu Kham Concentrate 2.81 69.6 1.87 54.1
concentrator process. Figure 13 shows the copper grade- Final tailings 0.05 30.4 0.07 45.9
recovery relationships with varying rougher concentrate
regrind power input of 10, 20 and 40 kWh/t.
Gold recovery by flotation at Phu Kham had been poor
The test results presented in Figure 13 clearly demonstrated since commissioning, averaging approximately 40 per cent
that finer regrinding of rougher concentrates from plant to final copper concentrate product. Prediction of gold
tailings flotation would improve both copper grade and recovery had also been demonstrated to be inaccurate during
recovery. Further flotation tests on plant final tailings samples plant operation, due to a lack of understanding of the key
using roughing at pH 9 with amyl xanthate, followed by mineralogical characteristics of gold occurrence and the
regrinding of concentrate at 10 kWh/t power input, and two variability of gold occurrence across different Phu Kham
stages of cleaning consistently produced a low-grade flotation
mineral assemblages. As part of the recovery improvement
concentrate of approximately three per cent copper and 2 g/t
study, the mineralogical reasons for gold loss into Phu
gold. Average copper recovery was 69.6 per cent and average
Kham tailings were examined to determine any potential
gold recovery was 54.1 per cent from 24 flotation tests as
opportunities to increase gold recovery.
shown in Table 1.
Diagnostic leach tests were conducted on Phu Kham plant
Since 2008, monthly plant composites had been submitted for
tailings samples. Results of the diagnostic leaching are
quantitative mineralogical analysis, and this data provided the
presented in Table 3.
critical information used for recovery improvement process
development. The mineralogy data demonstrated that since The diagnostic leaching results in Table 3 demonstrated that
primary chalcopyrite ores had become the dominant source of over 60 per cent of the gold in tailings was available for cyanide
plant feed, the major cause of loss of copper in plant tailings leaching, either as free gold or partially liberated gold. This was
had changed from slow floating fine liberated copper minerals supported by the Albion gold leach test work on acid Albion
(Crnkovic et al, 2009) to chalcopyrite locked in poor quality copper leach residues, which demonstrated that extraction of
coarse binary particles with non-sulfide gangue. The copper gold to solution did not significantly increase with increasing
loss in plant tailings from November 2011, representing a sulfide sulfur oxidation, as shown in Figure 16.
typical month, is shown in Figure 14. Laser ablation testing was conducted on the rougher tailings
A digital photomicrograph of flotation tailings is shown in to determine the proportion of gold locked with pyrite and
Figure 15. The wide size range of the chalcopyrite particles in other sulfide mineral species. The laser ablation test results
non-sulfide gangue is evident. were combined with the results from the diagnostic leaching
Copper sulfide mineral grain size data for rougher tailings to provide a total gold association for the rougher tailings as
is presented in Table 2. The data showed that under selective presented in Table 4.
rougher flotation conditions, recovery of coarse low quality Diagnostic leaching provided a measure of the unlocked
binary copper sulfide and gangue composite particles was (cyanide soluble) and locked gold deportment, but the total
poor due to the fine copper sulfide grain size. unlocked gold could not be split between fully liberated gold
FIG 13 – Phu Kham tailings flotation rougher concentrate regrind power grade – recovery relationships.

FIG 14 – Copper sulfide loss in Phu Kham flotation tailings by size and mineral association.
TABLE 3
Results of diagnostic gold leaching of Phu Kham flotation tailings.
Plant tailings % gold cyanide % gold locked % gold locked

stream soluble in sulfide in non-sulfide
minerals gangue
Cleaner tailings 72 21 7
Rougher tailings 57 42 1
Final tailings 64 35 1
sample from the IRP laboratory test work was also conducted
to produce a gold-rich concentrate by gravity concentration
suitable for ADIS and photomicrograph analysis.
The ADIS work on the final concentrate showed an
FIG 15 – Phu Kham flotation tailings photomicrograph. average recovered gold particle size of 7.9 µm. The class and
mass distribution summary and area as a percentage of the
TABLE 2 observed gold particles is presented in Table 5.
Rougher tailings copper sulfide grain size by size fraction. Binary particles of gold locked with pyrite in concentrate
were of particular interest. Only 8 per cent of the observed
Size fraction Copper sulfide grain size (µm) particles were gold-pyrite binary composite particles,
>106 µm 19.6 however 71 per cent of the total mass of observed gold was
in these particles. The average surface area of the gold in the
<106 >75 µm 13.0 gold-pyrite binary particles was 87 per cent of the total particle
<75 >53 µm 11.4 surface area. An interpretation of this data indicated that for
<53 >38 µm 9.1 a gold-pyrite particle to float into concentrate, the gold:pyrite
surface area ratio must be sufficiently large to overcome the
<38 µm >C2 9.1 depression of the attached pyrite particle under high pH
<C2 >C4 9.8 flotation cleaning conditions. The liberated gold recovered
was typically fine, with an average particle size of 7 µm.
<C4 >C5 8.1
Photomicrographs of the gold showing some typical
<C5 3.7 particles in concentrate are presented in Figure 17.
The ADIS work found that the average gold particle size
particles, and partially liberated (exposed in composites) was 19 µm in rougher tailings, approximately 13 times
gold particles. Therefore, the diagnostic leach measure of larger than the average gold particle in final concentrate.
57 per cent cyanide soluble gold in Table 4 could not provide No binary particles were observed, only liberated gold and
definitive information for the cause of gold loss to tailings. gold-chalcopyrite-pyrite-gangue multiphase particles as
Final Phu Kham copper concentrate monthly plant summarised by mass distribution in Table 6.
composites were analysed in 2012 by automated digital image The liberated gold particles in tailings had a mean size of
scanning (ADIS) at G&T Metallurgical Services to determine 35 µm, with the data indicating that liberated gold particles
the characteristics of gold and gold composite particles above 20 µm in size are unlikely to be recovered to final
recovered in flotation. Laboratory work on rougher tailings flotation concentrate, with a 13 µm particle the largest

D Bennett et al
FIG 16 – Copper and gold extraction in Albion leaching as a function of sulfide sulfur oxidation.
TABLE 4 TABLE 5
Phu Kham rougher tailings gold deportment. Phu Kham final concentrate gold distribution data.
Gold mineral association % of total gold Liberated Gold locked in binary particles with:
Data Multiphase
Cyanide soluble 57 gold Cp Ch/Cv Te Py Goe Gn
Pyrite locked 25 Association class 38% 30% 8% 5% 8% 3% 3% 5%
Other sulfide locked 17 Gold mass by association class 11% 16% <1% 1% 71% <1% 1% <1%
Non-sulfide gangue locked 1 Particle area % gold 100% 73% 52% 51% 87% 89% 98% 9%
Total 100 Cp – chalcopyrite; Ch – chalcocite; Cv – covellite; Te – tennantite; Py – pyrite; Goe – goethite; Gn – gangue; MP – multiphase.
FIG 17 – Phu Kham August 2011 final concentrate gold photomicrographs. NB: Au – gold, Py – pyrite.
TABLE 6
Flotation rougher tailing gold mode of occurrence, size and mass data.
Gold mode of occurrence Average particle diameter (µm) Average % of particle mass
Au Cp Py Gn Au Cp Py Gn
Gold liberated 35 - - - 100 - - -
Gold adhesion multiphase 6 17 151 15 <1 <1 99 <1
Gold locked multiphase 8 28 104 27 3 2 80 15
Gold adhesion/locked multiphase 17 14 125 61 2 1 94 3

observed in the concentrate. Cyanide leaching would have of three per cent copper and 2 g/t gold, providing an overall
extracted 69 per cent of the gold particles observed. 15 per cent copper recovery and 25 per cent gold recovery
The significant characteristic of the multiphase particles increase for Phu Kham operations.
observed in tailings was that greater than 95 per cent of the The process concept utilised bulk sulfide flotation with
total mass of the particles were gangue mass. The gold in the potassium amyl xanthate collector (PAX), with the bulk
multiphase particles had an average diameter of 9 µm, while concentrate reground to 18 µm before cleaning flotation to
the pyrite gangue had an average diameter of 121 µm. produce a three per cent copper concentrate. The concentrate
Photomicrographs of the gold showing some typical would then be leached in an Albion leach process with copper
particles in rougher tailings are presented in Figure 18. cathode produced from the pregnant Albion leach solution
Based on the concentrate and rougher tailings ADIS data, a via a solvent extraction-electrowinning process. The Albion
summary of the estimated recoveries of the main Phu Kham leach residue would then be neutralised prior to carbon-in-
ore gold association classes is presented in Table 7. leach processing to produce gold dore.
The downstream extraction of copper and gold from a The Albion test work program included a flotation pilot
low-grade pyrite-rich concentrate produced by bulk tailings plant at Phu Kham in order to produce a concentrate from the
flotation was developed to a prefeasibility level during plant tailings for pilot plant testing at the HRL/Core Resources
2010 and 2011 using Albion Process™ atmospheric leaching Albion pilot facility in Brisbane. The pilot facility consisted
technology. The Albion test work showed that copper of an M20 pilot IsaMill and a 1.5 t/h three stage flotation
extraction of over 95 per cent using acid Albion leaching, and plant. During the early stages of the pilot testing in October
gold extraction of over 85 per cent using a standard carbon- 2011, an updated capital estimate for the tailings retreatment
in-leach process was achievable from a flotation concentrate flotation and Albion hydrometallurgical process plants was
developed, with the updated estimate significantly higher
TABLE 7 than previous estimates. The change in the project cost base,
Phu Kham indicative gold recovery by mineral association class. coupled with the higher project risks for a tailings retreatment
plant incorporating hydrometallurgical processes, led to a
Gold association class review of alternative recovery improvement strategies.
Parameter Liberated Au-copper Au-pyrite Au-gangue The prefeasibility tailings flotation study results had
Au sulfide indicated that there was potential for increasing copper
and gold recovery by mineral processing methods alone,
Gold recovery 70% 100% 10% 0% with reduced technical risk and capital intensity compared
FIG 18 – Phu Kham rougher tailings gold photomicrographs. NB: Au – gold; Cp – chalcopyrite; Py – pyrite; He – hematite; Ma – magnetite; Gn – gangue.

D Bennett et al
to hydrometallurgical processing. The results from the conducted to confirm the recovery estimates and improve the
detailed mineralogical examination of Phu Kham tailings certainty of the economic case.
during the Albion prefeasibility study were used to improve
understanding of the mechanisms of copper and gold loss Project selection
and develop a simplified process for improving recovery. The Phu Kham Albion flotation pilot plant work program
The first alternative was a ‘mainstream inert grinding’ was changed to test the ‘less selective flotation’ process,
process which classified the rougher tailings from the plant which successfully demonstrated that significantly increasing
to recover the plus 53 µm fraction containing the low quality the copper and gold recovery from roughing was achievable
copper sulfide and gangue composites. This coarse fraction utilising PAX with over 90 per cent copper recovery achieved
of tailings would then be reground to 80 per cent passing at 30 per cent mass recovery. The rougher concentrate grade
53 µm in order to further liberate the chalcopyrite prior to was significantly reduced due to the bulk sulfide flotation
an additional scavenging flotation stage, with scavenger regime employed during the test. The pilot scale tests proved
concentrate to report to regrinding with rougher concentrate. the concept that increasing mass recovery would lead to
Although the process was highly predictable (as it was based increased copper recovery, whilst highlighting the pitfalls of
on particle size rather than mineralogy), the capital and recovering too much liberated pyrite which would require
power costs for the coarse fraction grinding were very high extremely aggressive pyrite depression conditions in cleaning
due to the large tonnage of liberated and coarse silicate and to produce a saleable grade concentrate.
pyrite gangue in the rougher tailings required to be ground
In developing the design for a bulk sulfide rougher,
for no value.
regrind and selective cleaning flotation process, three key
The second alternative was a ‘less selective’ sulfide rougher parameters were required to be identified; rougher mass
flotation process followed by increased regrind capacity recovery to achieve maximum copper recovery across all ore
in order to liberate the copper sulfide from coarse and low types, rougher concentrate regrind product size, and cleaner
quality composites, and increased cleaner circuit capacity
capacity.
in order to treat the additional and lower grade rougher
concentrate mass. Absolute recovery improvement estimates Due to the high variability in plant rougher copper and
based on the test work for the ‘mainstream grinding’ and ‘less gold recovery between the low and high pyrite content
selective flotation’ processes were similar at approximately ores, the rougher mass recovery likewise varied depending
5–10 per cent for copper and gold. The advantage of the on feed mineralogy. In order to develop the design rougher
‘less selective flotation’ process was that it was not a tailings mass recovery target, a review of the flotation recovery data
retreatment process and project value could not be eroded from daily rougher flotation testing of plant rougher feed
with improvements in existing plant performance. Whilst over a three-month period in 2011 was used to allow for the
it was acknowledged that the mainstream inert grinding variability in mineralogy of plant feed. Extensive rougher
process could also be performed on the plant feed, the ‘less flotation rate test data that had been used to develop early
selective flotation’ process had significantly lower capital and Phu Kham recovery models were statistically analysed in
operating costs providing higher return on investment, so order to determine the mass recovery and copper recovery
was selected as the priority for further study. curves for differing ore types, from weathered and altered
The disadvantage of the ‘less selective flotation’ process high pyrite (high S/Cu) material to chalcopyrite dominant
was that it was based on ore feed mineralogy, and therefore and low pyrite (low S/Cu) primary material.
generated a higher risk of mass balance errors during the The statistical analysis showed that rougher copper recovery
process design and uncertainty of achieving target copper could be increased to 85 per cent by targeting a maximum
and gold recovery. The ‘less selective flotation’ process 20–25 per cent mass recovery from rougher flotation across all
development risks required that further test work be ore types, as presented in Figure 19.
FIG 19 – Rougher copper recovery and mass recovery for major ore types showing original mass recovery design and IRP mass recovery design.

The results show that the original design for rougher mass observed as shown in Figure 13. Samples of plant feed were
recovery of 11 per cent of rougher feed was suitable for the tested at bench scale using roughing, rougher concentrate
chalcopyrite-dominant primary ore types, but gave lower regrind at increasing power input, and three stages of cleaning
copper recovery from the high pyrite and non-sulfide copper to produce the copper and gold grade-recovery response
species in chalcocite-rich transition ores. The results indicated curves in Figure 21. The increase in rougher concentrate
potential for an additional six per cent copper recovery regrind power input shows the benefit of the increased
in roughing by increasing mass recovery to a maximum liberation for both copper and gold grade and recovery into
25 per cent of rougher feed for all ore types. flotation concentrates.
The results for rougher flotation gold recovery from the same With the variable mineralogy of Phu Kham ore, the
test work are shown in Figure 20. The gold recovery increase monthly composite mineralogy and laboratory scale test
in roughing with the higher mass recovery is more consistent work was used to determine the rougher concentrate regrind
across primary, high pyrite and transition ore types than for size required to achieve maximum copper sulfide liberation
copper, which reflects the association of gold with copper in cleaner feed, with a target of 80 per cent copper sulfide
sulfide minerals and pyrite. The mineralogical analysis of liberation considered to be required to maximise recovery
gold deportment and bench scale test work results concluded and maintain final concentrate specification. Scan data on
that with the design rougher mass recovery increase and finer two samples representing different Phu Kham ore types in
regrind, gold recovery to final concentrate could be expected Figure 22 showed that maximum copper sulfide liberation was
to rise by at least six per cent. typically achieved at 20 µm particle size, although 80 per cent
Rougher concentrate regrind size optimisation work liberation was not necessarily achieved for all ore types.
commenced in 2011 to support the IRP process design. The Power input to achieve 20 µm regrind product size was
benefits for copper recovery of finer grinding of concentrates calculated from laboratory regrind signature plot data,
from bulk flotation of Phu Kham tailings had previously been and daily surveys of the existing M10000 IsaMill to be 18–
FIG 20 – Rougher gold recovery and mass recovery for major ore types showing original mass recovery design and IRP mass recovery design.
FIG 21 – Copper and gold flotation grade-recovery responses as a function of rougher concentrate regrind power input.

D Bennett et al
FIG 22 – Copper sulfide mineral liberation with particle size.
25 kWh/t, with an additional 3 MW M10000 IsaMill included The design included 100 per cent increase in first cleaner
in the design to provide a total of 6 MW power input at the capacity to allow for the increased rougher mass recovery, and
maximum rougher concentrate mass recovery of 25 per cent. lower slurry density to provide improved dilution cleaning to
minimise fine gangue particle entrainment.
First cleaner design To validate the extensive test work and mineralogy data
The increase in rougher concentrate mass recovery to results, four full-scale process plant ‘less selective flotation’
25 per cent of rougher feed, and the reduction in cleaner trials were conducted at Phu Kham between January 2012 and
feed particle size to 20 µm required a corresponding increase April 2012. The trial method used was to decrease SAG mill
in first cleaner flotation capacity. The 12 Mt/a plant had throughput by 50 per cent to avoid overloading the cleaning
25 minutes total first cleaner residence time for a 38 µm circuit, reduce rougher cell residence time to the equivalent
cleaner feed size, and test work was designed to determine post-PKU 16 Mt/a time of 30 minutes by reducing levels
whether this residence time needed to increase at the 20 µm and air in three of the ten cells, and using increased collector
cleaner feed size due to potentially reduced kinetics of and frother addition to increase rougher mass recovery to
the finer particles. Bench scale flotation cleaning tests and 25 per cent. The IsaMill power draw was increased to 2.8 MW
cleaner circuit model simulations indicated that flotation to target a 20 µm regrind product size. Immediately prior to
kinetics remained similar for copper sulfide minerals due the trial periods, baseline plant surveys were undertaken to
to improved liberation, and that no increase in residence obtain comparison data for the same ore type. The key results
time was required for first cleaning until regrind product of the four plant trials are summarised in Table 8.
size was below 10 µm. Figure 23 shows the plant cleaner The first plant trial period was over four hours to ensure
circuit chalcopyrite recovery by particle class and size. critical test criteria could be achieved with rougher mass
The recovery of liberated chalcopyrite remained high with recovery target of 25 per cent. Although this was a preliminary
decreasing particle size, which confirmed that the increased trial, survey results for the IRP process were positive compared
chalcopyrite liberation from gangue at a 20 µm regrind size to baseline survey results, with eight per cent overall copper
would improve copper recovery. recovery and 19 per cent overall gold recovery achieved
FIG 23 – Phu Kham cleaner circuit chalcopyrite recovery by particle class and size.

TABLE 8
Phu Kham ‘less selective flotation’ plant trial results.
Trial 1 – 15 January 2012 Trial 2 – 27 January 2012

Preliminary trials
Baseline IRP Difference Baseline IRP Difference
Rougher recovery (%) 83 92 +9 76 88 +12
Overall copper recovery (%) 71 79 +8 68 79 +11
Overall gold recovery (%) 48 66 +19 45 56 +11
Concentrate grade (%) 22.1 24.1 +2.0 24.2 24.8 +0.6
Trial 3 – 21 March 2012 Trial 4 – 24 April 2012
Variability trials
Baseline IRP Difference Baseline IRP Difference
Rougher recovery (%) 87 93 +6 81 87 +6
Overall copper recovery (%) 81 87 +6 75 80 +5
Overall gold recovery (%) 40 53 +13 36 56 +20
Concentrate grade (%) 25.7 23.8 -2.0 25.5 21.7 -3.8
into a two per cent higher copper grade final concentrate. March 2012, following PanAust Ltd board approval for the
The second plant trial was conducted over a full 12-hour IRP, the M10000 IsaMill was ordered as a turnkey package
shift period, with improvement in both overall copper and including all the peripheral equipment to the mill along with
gold recovery of 11 per cent at a 0.6 per cent copper grade instrumentation. The package also included all steelwork but
improvement in final concentrate. specifically excluded mill foundations.
Two short four-hour variability trails were conducted in The project schedule was developed to include the
March and April 2012 on primary ore with a good flotation completion of a front-end engineering phase (FEED) level
response and high pyrite ore with a poor flotation response. design and cost estimate by May 2012 using an external
Copper recovery improvement was consistent at over engineering consultant. The FEED phase fixed the design
five per cent for both ores, and gold recovery improvement criteria, mass and water balances and tagged equipment lists
was 13 per cent and 20 per cent into final concentrate. The which were then be used in the detailed engineering and
short duration of the trials did not allow time for optimisation procurement (EP) phase.
of cleaner circuit performance. 100.0
During the EP phase, the engineering consultant was
The trials were not used to confirm design mass recoveries responsible
95.0 for engineering design and equipment selection up
or regrind size. Higher final copper concentrate grades to the recommendation for purchase decision, with approval
were achieved despite the higher rougher mass recovery, 90.0
and procurement performed by PanAust staff in Brisbane and
although the minor difference against baseline values was Laos. To avoid issues identified during the PKU with delays
Chalcopyrite Recovery (%)
not considered significant during the short trial. The ability 85.0
in receipt of vendor data, the engineering consultant was
to conduct a plant trial and successfully demonstrate a responsible for liaison with vendors once the purchase order
80.0
consistent metallurgical benefit at laboratory, pilot, and full was raised to ensure the engineering consultant maintained
plant scale prior to implementation underpinned the high schedule
75.0 control.
degree of confidence in the ‘less selective flotation’ concept.
The detailed mineralogy-based understanding of the causes The basic project schedule for the IRP is presented in
70.0
of copper and gold loss to tailings developed over many years Figure 24.
was fundamental in gaining this confidence. 65.0
The economic evaluation at 6 per cent increase in copper PROJECT DESIGN
60.0
and gold recovery after the initial plant trials provided
a compelling investment case, resulting in PanAust Ltd Feasibility
55.0
board approval in February 2012 for the development of the In order to develop a business case for the project, initial
re-named IRP at Phu Kham to allow early procurement of process
50.0 design was developed by PanAust during the
long-lead equipment. The plant trial results also provided 1 stage along with factored capital and operating
prefeasibility 10
significant confidence in the IRP basis of design. Particle Size (µm)
cost estimates. The cost estimate was considered to be at a
higher level of accuracy than a normal prefeasibility stage due
Liberated Chalcopyrite Chalcopy
Compression of the IRP schedule to the recent completion of the engineering for the PKU and
Chalcopyrite‐Gangue Binary Multipha
During the early process design and project scoping, the project construction of the Ban Houayxai CIL plant by PanAust Ltd.
Fig 23 – Phu Kham cleaner circuit chalcopyrite recovery by particle class and size
case for the IRP was determined to be compelling enough for
the project development to be fast tracked with commitment 2012 2013
Project Stage
for long-lead capital items approved before completion of Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
the feasibility study. In addition to achieving the increased Feasibility Study/FEED
plant recovery earlier, reducing the project schedule had the Long Lead Procurement
added benefit that the PanAust construction team would be Engineering and Procurement
able to commence work on the IRP immediately following Construction
completion of the PKU in August 2012. Commissioning and Ramp-up
The longest lead item for the project was the M10000 IsaMill Fig 24 – Increased Recovery Project schedule
with a delivery period of approximately 11 months. In early FIG 24 – Increased Recovery Project schedule.
23
D Bennett et al
In addition to process design criteria, the equipment lists on concentrate produced from the variability trials showed
and brownfield plant layouts and integration plans were that the existing 15 m diameter concentrate thickener would
developed to a detailed level by PanAust so that the FEED be able to handle the concentrate produced from the IRP,
stage was mainly focused on engineering detail and minor however settling efficiency could be improved over current
equipment sizing rather than process design. Plant survey operation by reducing the amount of entrained air in the feed.
work was performed during this stage to improve the Site investigation determined that air was being entrained
mass balance which was used to confirm pump and piping through the trash screen collection box, so the decision was
specifications. The initial flow diagram showing the additional made to add a de-aeration box prior to the thickener to
equipment planned to be installed as part of the IRP in red is improve thickener performance of the thickener.
shown in Figure 25.
Challenges in brownfields plant upgrades were highlighted
during the FEED phase with the marriage of new equipment
Front-end engineering
with existing equipment which cannot be relocated. Specific
GR Engineering Services (GRES) was selected as the FEED
design principles were included to minimise capital and
engineer for the IRP. During the FEED phase a number of
operating cost based on the learnings from the original plant
details in the plant design were resolved which resulted in
design and PKU. The design was developed with a focus on
minor changes to the process flow sheet. The most significant
the use of gravity flow wherever possible and gravity lines
change was the change to a second parallel bank instead of
were used in most tie-in locations despite the additional
additional series cells to the first cleaners. This was changed
mainly due to limited dart capacity in the first cleaner cells engineering design required. This design philosophy
as well as the cost associated with upgrading the first cleaner increased the complexity of the design effort, however
feed pumps. The change resulted in a lower project capital cost increased operability was achieved.
as capital was moved from debottlenecking the existing first The existing plant concentrate production capacity was
cleaner cells to installing a parallel bank of cells. The process analysed, with the existing filter able to produce concentrate
flow sheet was also simplified due to the change. The change below the transport moisture limit at a rate of approximately
also reduced the associated tie-in time for construction as the 330 000 t/a. This was determined to be marginal at the increased
new equipment could be added in discrete modules, and did concentrate production rates as shown in Figure 26. This was
not impact on the existing plant which was a key learning a key risk for the project as the filtration rate was expected
from the PKU for brownfields plant expansions. to decrease due to the finer concentrate size distribution. To
In parallel with the FEED phase, additional metallurgical eliminate this risk a second filter was included in the design.
test work was performed in order to assess the effect of the Although the second filter had fewer plates than the original
expected finer concentrate size on the settling rate of the filter, the same plate size was maintained to allow common
existing concentrate thickener. The settling tests performed spares with the existing filter.
FIG 25 – Initial process flow diagram for the Increased Recovery Project.

FIG 26 – Filter performance against increased recovery project concentrate production targets.
A design review was conducted during the FEED to look •• 40 plate Ishigaki filter
at installing a second filter in the existing shed without •• concentrate storage shed.
disruption of the existing filter operations. The review
determined that disruption would be significant, so an
additional smaller storage shed was included in the design to Increased Recovery Project COMMISSIONING AND RAMP-UP
allow for additional storage capacity during the wet season. Plant tie-ins were performed progressively as equipment was
The design of the new storage shed and filtration area was delivered to site and in line with the project schedules which
leveraged from the existing engineering designs to reduce were integrated with the planned shutdown schedules. The
costs for installation. Some minor modifications were made majority of construction and tie-in activities were completed
during the FEED phase to shorten the length of the shed in by March 2013.
order to reduce storage capacity in line with the additional
concentrate production. IsaMill commissioning
An additional item of risk which was identified during Commissioning of the new IsaMill concluded with a 72 hour
the FEED was the second cleaner flotation cell tailings darts acceptance test which commenced on the 26 March 2013 with
capacity not having sufficient margin above maximum both the new and old mills run in parallel for the first time.
flows predicted for the IRP. Due to the lack of mitigation Operational conditions during the period were identical for
measures available, contingency was allowed for in the each of the mills. During the period, the new IsaMill was run
estimate to install inter-cell bypasses should the flow rate over a range of power draws, with the ultimate design power
significantly exceed darts capacity. It was anticipated that of 2500 kW being achieved. During the test, no external signs
the second cleaner would become the circuit bottleneck in the of any problems were detected, however upon completion of
IRP circuit, and if liberation was not achieved in the regrind the test period, the mill was inspected and damage to the shell
mills, a recirculating load of middlings particles would build- liner (Figure 28), feed end liner and three discs was identified.
up between the second cleaner and third cleaner banks. The root cause of the damage to the mill liners was
This would result in recovery losses or below specification overheating of the mill slurry due to the high recycle of
concentrate grade. IsaMill discharge to the feed. The high recycle was caused
The FEED phase was completed on schedule with the capital by insufficient flow from the regrind cyclones which were
cost reduced by $10 M due to additional savings identified exacerbated by a biased flow from the underflow splitter box
during the design, and a reduction in the project contingency to the existing IsaMill. The new regrind cyclones had also
due to better project definition. The reduction in capital been modified by operations due to the need to run them prior
was achieved despite an increase in the scope of the project to IRP commissioning and did not have the design spigots
in order to minimise risk. As the project was brownfields installed prior to start-up. The high temperature in the IsaMill
integration, additional capital was allowed in the estimate for would have normally been detected by the process control
resolution of legacy pumping and spillage issues which had system however the resistance temperature device (RTDs),
affected the Phu Kham concentrator since the original design which were installed to protect the mill in such instances of
and the PKU. Following completion of the FEED phase the low feed, were found to be providing an inaccurate output.
final equipment list was confirmed along with the changes The RTDs were replaced along with the shell liner, and the
to the process flow diagram (shown in red) in Figure 27 to discs and feed end liner were reused.
optimise both cost and design: The mill acceptance test was re-run with additional wear
•• twelve 400 mm diameter regrind cyclone pack identified in the regrind mill due to media compression. The
media compression had been caused by a high slurry feed
•• M10000 IsaMill
density which was attributed to an incorrect and high solids
•• seven Outotec TC70 tank cells specific gravity input into the nuclear density gauge.
•• thickener de-aeration box In addition, the output value of the density gauge had been
•• concentrate filter feed tank mistakenly assumed to be a calculation of per cent solids.

D Bennett et al
FIG 27 – Simplified Phu Kham process flow diagram including Increased Recovery Project expansion.
extensive. Existing equipment, particularly slurry pumps,

were utilised within the new circuit.
During the ramp-up phase a number of issues with slurry
pumping were encountered. These were mainly confined
to original pumps which had been upgraded or had duty
changed as part of the IRP or earlier improvement projects.
The IRP construction and precommissioning phases had
not generally considered the condition or status of existing
FIG 28 – Damage to the shell liner during the first 72 hour acceptance test. equipment and as it had been previously operating
satisfactorily, and no program of inspection was initiated
This meant that the density in the mill was approximately prior to IRP commissioning. This resulted in delays in
eight per cent higher than the actual density target, with reaching design capacity.
the subsequent high viscosity in the mill causing the media As an example, the rougher concentrate pump system
compression. In order to ensure that this didn’t recur, the had the common discharge pipeline upgraded to match
target pulp density was decreased to 45 per cent solids the higher IRP design flow rates, with the existing duty
and the viscosity was monitored at hourly intervals with a and standby pumps and motors assessed as suitable for the
Marsh Funnel. higher design duty. During the ramp-up phase the pumps
Once the control measures were put into place the decision were not able to reach design flow rate which resulted in both
was made on the 18 April to run the mill for a 24-hour test. the duty and standby pumps being run to meet design flow
At the completion of this test, the mill was inspected and no rate. Investigation of the pump variable speed drive systems
further wear detected on the liners. The final acceptance test discovered that the original overload and frequency limit
was then performed for 48 hours at power draws of up to settings had been retained for the original motors, although
2600 kW. On the 21 April, the mill was inspected for the final the motors had been upgraded prior to the IRP. These settings
time. No wear had occurred over the final trial period and were limiting the pump speed and power draw to a level well
the mill was handed over by Xstrata Technology personnel to below maximum operating capability. Once the limit was
PanAust operations. removed the pumps were found to be easily able to meet
design flow rate.
Plant ramp-up Ramp-up was also hindered by power supply continuity to
Even with the modular design of the IRP which aided site, as there was often insufficient power able to be supplied
constructability, the integration with the existing circuit was from the hydroelectric schemes which provide all power to

the Phu Bia Mining operations. This was due to low water With the improvements in control, equipment performance
levels in the reservoir at the end of the dry season coinciding and operating knowledge for the new circuit, the ramp-
with the IRP ramp-up and extended planned maintenance up of the IRP was mainly completed by June 2013. Copper
periods on generating units. Operation response to power rougher recovery increased from an average of 78 per cent to
limitation was to load-shed power across the plant with 83 per cent within the first month of operation, in line with
power draw being lowered in the ball mills and IsaMills to expectations. The increase in gold rougher recovery was very
maintain throughput. This increased primary grind size and significant from an average of 55 per cent to 78 per cent as
regrind size lowering copper mineral liberation and reducing shown in Figure 29. Gold recovery at Phu Kham had always
plant copper recovery. The problem continued until June with been highly variable, which led to a conservative estimate
the onset of the wet season replenishing reservoir levels. of gold recovery increase used for the IRP investment case,
however the observed increase in gold recovery exceeded
the results achieved during the plant trials and even the most
INcreased Recovery Project PERFORMANCE optimistic estimates. The reasons for the improvement were
likely the improved rougher recovery of lower quality copper
Rougher recovery sulfide and pyrite composites containing gold and improved
Despite the described issues during ramp-up, rougher mass fine gold recovery.
recovery was quickly increased to 20 per cent of feed mass
Once operational control was improved and selectivity
following completion of commissioning.
against liberated pyrite and non-sulfide gangue was improved
At an early stage of ramp-up the operating strategy was to in the rougher circuit, the increased copper recovery was
effectively ‘fill up’ the cleaner circuit by recovering as much confirmed though a consistent reduction in the quality of the
mass as possible from the rougher banks. Cleaner circuit copper sulfide-gangue binary particles in the rougher tailings.
performance and overall copper recovery were generally Following the implementation of the IRP, the copper sulfide
compromised by this strategy because excessive gangue was content of these particles decreased from approximately
recovered in roughing requiring rejection in cleaning, and 15 per cent to seven per cent copper sulfide indicating that
the regrind product size to achieve liberation targets was not the IRP target low quality copper sulfide-gangue composite
being met. In addition, the feed ore quality was generally particles were being recovered (Figure 30).
challenging during ramp-up due to the requirement to mine A photomicrograph of typical rougher tailings is presented
ore from the upper areas of the pit which had very high pyrite in Figure 31, indicating the lean quality of copper sulfide-
and clay content, and a high proportion of easily oxidised gangue composites following the IRP.
secondary copper sulfide species.
During this period, operational strategies were redefined Overall recovery
and training undertaken with site personnel to ensure that The ramp-up in overall recovery was achieved in June 2013
copper recovery was maximised from the rougher banks and with 76.4 per cent average weekly recovery being achieved.
the cleaning circuit was not overloaded with liberated gangue. This was 3.8 per cent above the baseline PKU recovery of
Back-to-basics improvements were undertaken on operating 72.6 per cent. In addition during this time the variability of
parameters; in particular work was completed with Outotec the overall recovery decreased with the standard deviation
on optimising the rougher cell level and air addition profiles of weekly recovery reducing from six per cent to 4.6 per cent
which were then set to avoid excessive water and gangue of the mean. As presented in Figure 32, the number of weeks
recovery and improve low quality coarse composite recovery. with recovery below 70 per cent has decreased over time as
Process control improvements were also implemented the operational strategies implemented during the IRP have
during the ramp-up period, including commissioning of been further reinforced.
automated rougher mass recovery control through use of As for copper, initial increases in the rougher gold recovery
froth velocity measurements and level and air controls. did not correspond to a significant increase in overall gold
FIG 29 – Rougher mass, gold and copper recovery following the Phu Kham Upgrade and Increased Recovery Projects.

D Bennett et al
FIG 30 – Quality of the copper sulfide gangue binary composites in the rougher tailings following the Phu Kham Upgrade and Increased Recovery Projects.
recovery. Gold losses in the cleaner tailings were investigated

during this period in order to determine the mechanisms of
gold loss from the cleaner circuit.
Analysis concluded that although 25 per cent of the identified
gold containing particles lost from the cleaner tailings were
liberated gold, the majority of these particles were less
than 10 µm in size. Almost 50 per cent of the particles were
liberated or high quality composites with pyrite or copper
sulfide minerals. The other 50 per cent of gold particles lost
were in extremely low quality binary composites with pyrite
(Figure 33). These composites were required to be rejected to
maintain acceptable final copper concentrate grade. As gold
particles even in binary composites were fine (Figure 34),
FIG 31 – Photomicrograph of the typical rougher tailings gold recovery to the final concentrate relies on liberation
following Phu Kham Upgrade and Increased Recovery Projects. from pyrite due to the high lime pH environment for pyrite
Cp – chalcopyrite; Py – pyrite; Gn – non-sulfide gangue. depression, which also reduces fine gold flotation kinetics.
FIG 32 – Overall copper recovery mean and standard deviation following implementation of the Increased Recovery Projects.

FIG 33 – Quality of the gold particles lost from the cleaner tailings.
FIG 34 – Examples of poor quality composites and fine liberated gold losses in the cleaner tailings.
With improvement in plant performance, overall gold increased from over 16 during the PKU ramp-up period to 21
recovery increased, with weekly gold recovery averaging during the IRP ramp-up. In addition, head grades declined
over 50 per cent by September 2013 (Figure 35). No decrease in over the period which allowed the throughput to be increased
the variability of the gold recovery was achieved because the in order to maintain the copper production rates. Pivotal in
gold recovery is mainly affected by the variable mineralogy, this increase in mill throughput was the implementation of
liberation and gold feed grades to the plant. new plant control strategies (Baas, Bennett and Walker, 2014)
High gold recovery was maintained throughout the year along with the additional rougher concentrate, regrind and
following the IRP even though the plant experienced a cleaner capacities provided by the IRP.
significant change in the gold head grade from 0.3 g/t to The S/Cu ratio has been used at Phu Kham in recovery
0.2 g/t from September 2013 to September 2014 (Figure 36). modelling as a proxy for pyrite in the ore, which is the
Once head grades returned to normal levels in late 2013, overall dominant driver of metallurgical performance. Increased
gold recoveries of 60 per cent were being achieved which was S/Cu ratios result in decreased copper recovery, due to the
almost a 20 per cent increase over the pre-IRP recoveries. requirement to be highly selective against pyrite. On a S/
Cu ratio basis, the increased copper recovery achieved by
Recovery against increased recovery project investment case the IRP varies from seven per cent at a S/Cu ratio of over
The initial analysis of the plant data shows that the average 30 to four per cent at a S/Cu ratio of less than 20 (Figure 38).
increase in copper recovery was less than the six per cent The increase achieved by the project was determined to have
which was originally predicted for the project and the increase reached the six per cent goal when the ore quality schedule
in gold recovery was above the original six per cent target. was compared on a like for like basis.
This was despite other design targets (eg mass pull rate and
rougher recovery) within the plant being achieved. The source FUTURE OPPORTUNITIES
of the copper recovery difference was due to a number of other Since the IRP has been implemented, a number of further
operational and ore quality factors which were impacting on opportunities for recovery improvement have been pursued.
production during the period as discussed below. Major additional improvement in reducing the variation
It can be seen in Figure 37 that during the period of the in overall recovery has since been achieved through
ramp-up, the quantity of pyrite in the ore as measured by the implementation of a mass recovery control system which
sulfur to copper content fraction (known as the S/Cu ratio) utilises froth cameras on the rougher banks and first cleaner

D Bennett et al
FIG 35 – Overall gold recovery following commissioning of the Increased Recovery Projects.
FIG 36 – CUSUM chart for gold showing increase gold recovery despite declining head grades.
FIG 37 – CUSUM chart for copper recovery showing changes to operation and ore quality indicators.

FIG 38 – Copper recovery against S/Cu ratio in plant feed.
banks. This system has allowed maintenance of mass and Phu Kham Upgrade Project to process a nominal throughput
copper recovery based on concentrate grade targets and metal of 16 Mt/a with installation of a second 13 MW ball mill, a
input. The integration of second cleaner and third cleaner 33 per cent increase in rougher flotation capacity, a 40 per cent
control to maximise copper recovery and trim concentrate increase in second cleaner capacity, and 33 per cent increase
grade with the Jameson Cell concentrate grade to achieve a final in third cleaner capacity.
concentrate grade target has been implemented, with a particle The Phu Kham IRP was completed under budget and
size analyser used to control IsaMill regrind power draw. schedule in 2013, resulting in increased copper recovery
Finally, work has progressed to improve cleaner circuit pulp of five per cent and increased gold recovery of at least
chemistry following regrind. This body of work was identified ten per cent over the first year of operation at ore throughput
by site metallurgists during Jameson Cell optimisation test rates exceeding maximum design of 2250 t/h by up to
work where it was found that the copper sulfides in regrind ten per cent. The increased recovery was achieved despite
cyclone overflow exhibit significantly faster flotation kinetics decreasing ore quality over the period. In addition, the extra
than the IsaMill regrind discharge. This has been attributed to cleaner capacity allowed throughput to be increased in order
oxygen depletion via reaction of pyrite in the closed milling to increase copper feed tonnes to the plant despite the lower
system. Automated analysis of the pulp chemistry has been head grades.
implemented and a trial using peroxide (dosed to the regrind
The plant optimisation and development programs and
discharge to modify the pulp redox potential in the cleaner
major plant designs at Phu Kham have been supported
feed) is currently being advanced.
by extensive mineralogy, mineral association and mineral
liberation data from plant monthly composites and bench
CONCLUSIONS scale test products. The collection and analysis of this data
The installed 2008 Phu Kham 12 Mt/a concentrator was has revealed reasons for copper and gold losses and mineral
a compromise between a high recovery but high capital deportment, and significant opportunities for increasing
intensity design, and a lower recovery but technically low risk recovery of copper and gold have been identified.
and low capital intensity design suitable for the prevailing
low copper price market. The chosen selective rougher Throughput Forecasting and Optimisation
flotation design was driven by the complex and highly
variable mineralogy particularly in the transition ore zones, Project Background and Scope
and high pyrite content. Over 90 per cent of pyrite is required Following the completion of the PKU and IRP circuit
to be rejected in order to produce a final concentrate of over expansions in 2013, the processing plant capacity was
23 per cent copper. increased to a design maximum capacity of 2250 t/h,
The optimisation of the Phu Kham flow sheet was been driven equivalent to 18 Mt/a. As mining extends deeper into
by low copper and gold recoveries and the requirement to the deposit, the operation will experience an increased
maximise copper production by increasing plant throughput. proportion of highly competent ores which have the potential
Flotation circuit improvements and debottlenecking to limit plant throughput, in particular through the 13 MW
projects including additional roughing capacity and cleaner SAG mill. Insufficient comminution data in the mine block
feed scalping had increased copper metal production by model created a lack of confidence in the ability to predict
16 per cent since 2009, while maintaining copper recovery at mill throughput, particularly in the later years of the mine life.
the original maximum design throughput of 14 Mt/a. In 2012 Phu Bia Mining commenced a throughput forecasting and
the operation was upgraded and debottlenecked during the optimisation project in 2012 to evaluate how to maintain the

D Bennett et al
target throughput over the LOM. Metso PTI was engaged can be measured with laboratory tests such as Point Load
to conduct a full process integration and optimisation (PIO) Index (PLI), Drop Weight or SMC tests (DWi), and Bond ball
project. The main objectives were to: mill work index (BWi). The unconfined compressive strength
•• review the current blasting, crushing and grinding (UCS) is a common measure of strength and can be estimated
processes, identify opportunities for increasing from PLI values to reduce laboratory testing requirements.
throughput, and improve overall comminution circuit In general terms, the rock structure affects the coarse end of
performance when treating very competent ore types the ROM fragmentation, while the strength (hardness) affects
the generation of fines. Improved plant throughput can be
•• develop a throughput prediction model based on
achieved by manipulating ROM fragmentation through
geometallurgical modelling for long-term planning
optimisation of blasting to reduce top size and increase fines,
and optimisation
especially for SAG mills.
•• identify if and when secondary crushing or other process
At Phu Kham, ore domains were defined based on the RQD
changes will be required over the LOM to maximise
(for structure) and PLI (for strength) values in the geotechnical
plant throughput.
block model. This resulted in a matrix of nine ore domains as
shown in Figure 39. Ore within a domain will produce similar
Methodology overview ROM fragmentation for a given blast design.
Metso’s PIO methodology involves the development of An increase in harder ores is expected in the future at Phu
integrated operating and control strategies from the mine Kham as the pit deepens, and has the potential to reduce
to the plant that maximise throughput, minimise the overall throughput. Consequently, the focus of the project was on
energy consumption, cost per ton and maximise profitability. increasing throughput and improving overall comminution
This requires an understanding of the physical properties and circuit performance when treating the hardest ore types.
composition of the orebody, where the valuable mineral is Therefore, the hardest ore available was selected for a trial
located within it, and what mineral associations exist between blast and was followed by detailed auditing of the blasting
the ore and gangue. and processing operations. For the trial blast, RQD data
The process starts with ore characterisation to define was obtained from the geotechnical block model, and point
domains within the orebody that will behave similarly load tests were conducted on stockpile samples. These
throughout the blasting and comminution processes. The indicated that the trial blast consisted of moderately hard, but
SmartTag™ ore tracking system developed by Metso PTI reasonably low quality, jointed and fractured rock.
(La Rosa et al, 2007) is used to track the characterised ore from
the mine, through the crusher and finally into the grinding Blast modelling and simulations
mills. With the ore source and characteristics known, detailed
audits of the blasting and processing operations are used to One of the main objectives of the project was to develop
develop site-specific predictive models for each operation strategies to maximise mill throughput to maintain LOM
(blasting, comminution, separation). Using these predictive operational targets even when treating harder ores. This
models, the blast design is optimised to generate optimal can be achieved by improving ROM fragmentation through
run-of-mine (ROM) fragmentation for all ore types, and optimisation of blasting practices. A reliable model of blast
downstream processes can be adjusted accordingly. The fragmentation is required to determine the effect of changing
models also allow prediction of throughput and recovery blast parameters on ROM fragmentation.
performance for each ore domain, and when combined with The Metso PTI blast fragmentation model, which is sensitive
the mine plan can be used for forecasting, planning and to the major parameters known to affect blasting performance,
optimisation purposes. was calibrated using the ore characterisation data and design
The SmartTag™ system is also used to increase the accuracy parameters from the audited blast. Image analysis of the
of geometallurgical modelling and update the block model ROM size distribution produced by the trial blast was used
automatically. The ore is tracked from the mine through to calibrate the coarse size fractions. A limitation of image
the process with SmartTags™ and linked with the plant analysis techniques to determine ROM size distributions
control system (DCS) to provide actual plant performance is that they cannot effectively delineate fines; therefore, the
data (throughput, recovery, grade etc) for each ore type and size distribution of the primary crusher product belt cut
associated blast conditions. These data are automatically sample was used to correct the fine portion of the curve.
compared with model predictions and updated in the block A comparison between the measured and model generated
model using the SmartTag™ software. Incorporation of the ROM size distribution is shown in Figure 40. The model
actual plant data into the block model in real-time eliminates
the need for further expensive ore characterisation tests. More
accurate data in the block model improves mine planning,
and the plant receives advance notice of the ore type about to
be processed. Adjustments can then be made to blast designs
and operating conditions to optimise performance.
Ore characterisation
The optimisation methodology starts with ore characterisation
in terms of structure, strength and comminution properties.
These are used to define ore domains within the geotechnical
block model with similar blastability and fragmentation
properties.
Rock structure is determined by the in situ joints and
fractures and can be quantified with rock quality designation
(RQD), fracture frequency and joint mapping. Rock strength FIG 39 – Definition and distribution of ore domains at Phu Kham.

FIG 40 – Blast fragmentation model.

FIG 41 – Selected blast fragmentation simulation results.
predictions correlated well with measured values at both
the coarse and fine ends of the particle size distribution. This cookbook avoids excessive blasting in softer ore domains,
demonstrated that the model was quite accurate in predicting thus reducing energy consumption and costs, and preventing
ROM fragmentation, and suitable for simulation studies. the excessive production of ultra-fines that can be detrimental
Simulations were conducted using the blast fragmentation to some downstream processes. This has been successfully
model to investigate the impact of changes in spacing, applied by PTI at several other large open pit operations
burden, stemming length and blasthole diameter on ROM globally (Rybinski et al, 2011; Burger et al, 2006). The final
fragmentation. The parameters of the current blast design and definition of the blasting cookbook at Phu Kham is ongoing.
five selected scenarios (with 115 mm diameter blastholes) are
provided in Table 9. The corresponding model predictions of SmartTagTM Ore Tracking
ROM size distribution are shown in Figure 41.
To link the process performance with ore characterisation and
As expected, simulations indicated that tightening the blasting outcomes, the ore from the trial blast was tracked from
blast pattern to increase the powder factor (PF) resulted the mine through the process using SmartTagTM ore tracking.
in a significant increase in the fines generated in the blast. The SmartTag™ ore tracking system developed by Metso
Reduction of stemming length also generated more fines and PTI allows parcels of ore to be tracked from the mine, through
reduced the top size of the rock due to the increased explosive the crusher and finally into the grinding mills, as shown in
energy at the stemming horizon. These simulations indicated Figure 42. The SmartTags™ are built around robust passive
potential to increase throughput by increasing the fines and radio frequency (RFID) transponders. They do not have an
reducing the top size of the ROM fragmentation by optimising internal power source, so they can remain in stockpiles and
the blasting parameters. ROM pads for extended periods of time. Antennas to detect
Simulations were conducted for each of the nine ore domains the SmartTags™ are located at critical points in the process
(using both 115 mm and 127 mm blasthole diameters). This ahead of the milling circuit; tags can be detected a number
allows the blast design to optimised for each of the ore domains, of times and provide valuable information on material
and a ‘cookbook’ is generated which provides a ‘recipe’ (ie an movements. In particular, they make it possible to link the
optimised blast design for each ore domain). Blasting according spatial data associated with the ore in the mine to the time-
to this cookbook provides a more consistent and optimised based data of the concentrator.
feed size distribution to the downstream processes, increasing At Phu Kham, SmartTag™ antennas were installed under the
throughput, process stability and efficiency. Following the crusher product and SAG mill feed conveyors. The installation
TABLE 9
Selected blast designs.
Current design Scenario 1 Scenario 5 Scenario 9 Scenario 13 Scenario 17

RQD (%) 25
UCS (MPa) 60
Bench height (m) 10
Burden (m) 3.5 2.6 3 3.4 3.8 4.2
Spacing (m) 4 3.2 3.6 4 4.5 4.9
Subdrill (m) 1
Stemming length (m) 3 1.8 2.2 2.5 2.8 3.2
Explosive type 70% emulsion
Density (g/cc) 1.15
Powder factor (kg/m3) 0.69 1.34 0.99 0.76 0.58 0.46
Powder factor (kg/t) 0.26 0.5 0.36 0.28 0.21 0.17
Change in powder factor (%) -- 94.2 43.5 10.1 -15.9 -33.3

D Bennett et al
FIG 42 – SmartTag™ ore tracking.
on the SAG feed belt is shown in Figure 43. SmartTags™ were TABLE 10

inserted into the stemming column of every blasthole for the DWT and BWi test results.
audited blast. The origin of each SmartTag™ is saved with
its unique identification number (ID). As the SmartTags™ Ore type A b A*b ta BWi (kWh/t)
and associated orepass the antennas in the process plant, the Belt cut sample 57.8 1.06 61.3 0.67 13
system automatically records the time and tag ID, thus the
source of the ore being processed at any given time is known.
of 13 kWh/t, indicating ‘medium hardness’ with respect to
During the project at Phu Kham, this ensured that ore from the
ball milling.
trial blast was being fed to the concentrator during the plant
audits, and allowed correlations to be established between ore The survey data and measured ore parameters were used
origin and process performance. to develop and calibrate site-specific comminution circuit
models using JKSimMet. The models correlated well with
measured data and were considered suitable for simulations
Comminution modelling and simulations and throughput forecasting.
A comprehensive grinding circuit survey was successfully
conducted on 4 February 2013 when the circuit was treating
ore from the trial blast (as determined by SmartTag™ ore
Integrated analysis – opportunities
tracking). At the time of the grinding survey the plant was to increase throughput
operating under SAG mill limiting conditions, with a total The ROM size distributions generated in the blasting
mill load of 29 per cent and mill power draw of 12.0 MW. simulation study were used as inputs to the comminution
Ore samples collected during the site survey were sent for models. This allowed changes to blasting practices and
ore characterisation tests, including drop weight (DWT) and comminution circuit operation to be evaluated with respect
Bond ball mill work index (BWi) testing. These tests provide to the entire operation. This integrated approach was used to
ore parameters required for comminution modelling, and the determine effective operating strategies to increase throughput
results from the survey sample are provided in Table 10. The when processing harder ores. The results indicated that:
A*b and ta parameters are determined from the DWT. The •• Implementing a tighter blast pattern to increase PF would
A*b value for the ore puts it in the ‘moderately soft’ category improve ROM fragmentation and increase throughput.
in terms of resistance to impact breakage, and the ta value With the existing 115 mm blastholes, a tighter pattern
falls into the ‘soft’ range for abrasion resistance. The BWi test could increase PF to 1.34 kg/m3 (currently 0.69 kg/m3)
conducted at a closing sieve size of 106 µm resulted in a BWi and increase SAG throughput by four per cent. For
127 mm blastholes, the PF could be increased to 1.41 kg/m3
and increase SAG throughput by 6.5 per cent.
•• Further increases in throughput could be achieved by
combining the blasting changes with a reduction in
primary crusher closed side setting (CSS). Simulations
predicted that a 15 per cent reduction in CSS to
120 mm could increase SAG throughput by more than
eight per cent for the 127 mm blastholes.
•• The changes to blast designs increase the amount of
fine material in the SAG feed and reduce the top size of
the ROM size distribution, while reducing crusher gap
primarily affects the amount of coarse particles in the
SAG feed. Therefore, these are complementary strategies
FIG 43 – SmartTag™ antenna installation under the SAG feed belt. for increasing throughput.

Throughput forecasting analysis of historical plant data (PI data) indicated that
the variability in plant throughput to be in the order of
One of the primary objectives of the project was to develop
±10.7 per cent. The plant throughput was capped at 2600 t/h,
a throughput forecasting model based on geometallurgical
which was determined to be a conservative limit given the
modelling for long-term planning and optimisation. This was
plant variability and historical plant performance. The
achieved using the integrated site-specific models for blasting
throughput forecast until 2019 is shown in Figure 45.
and comminution with the Phu Kham LOM plan.
A further objective of the project was to evaluate whether
The LOM ore delivery plan provided by Phu Kham was
or not a secondary crushing circuit or other process changes
used along with the geotechnical block model to determine the
would be required over the LOM to maintain throughput.
ore characteristics for each period. The structure (RQD) and
The throughput forecast results indicate that the annual
strength (Is50) results over the LOM are shown in Figure 44,
throughput target can be achieved until 2018. However, in
and indicates an increase in both over time until the end of 2019
2019 when the hardest and blockiest material is scheduled
before sharply dropping off again until the end of the mine life.
to be delivered, the annual throughput is predicted to be
This corresponds to mining in the deeper, more competent significantly less.
areas of the orebody before some shallower cut-backs late in
the mine’s life. As is common with long-term mine plans, the A secondary crushing circuit represents a significant capital
resolution of the schedule reduces closer to the end of the mine investment. Therefore, firstly, the changes to blasting practices
life. Note that Is50 can be used as a proxy for UCS which is an and comminution circuit operation identified earlier in the
project were investigated to determine whether these could
important input to the blast fragmentation model.
sufficiently increase throughput. These changes included
The ore characterisation values over the LOM block model increasing blasthole diameter (as larger blastholes are capable
were used with the site-specific blast fragmentation model of generating finer fragmentation), tightening the blast
to determine the ROM size distribution for each period. pattern to increase PF and reducing the primary crusher CSS,
The standard (existing) blasting pattern and one of the high as discussed previously. Simulations indicated these changes
energy blast designs were used at this stage; designs from the alone would achieve an ore processing rate of only 1700 t/h
blasting cookbook could be included once these are finalised. in 2019. Further simulations were conducted incorporating
The resulting ROM size distributions were used in the secondary crushing into the circuit in 2019. These simulations
comminution circuit models to predict the throughput of the indicated that the annual throughput target could be achieved
circuit for each period in the ore delivery schedule. with the current blast pattern if the ore is secondary crushed
Note that the comminution circuit models also require down to a feed size F80 of around 40 mm. This suggests for the
ore parameter inputs which are normally determined from current LOM plan secondary crushing would be required for
laboratory testing. Metso PTI maintains a large database of 2019 only to achieve the target throughput.
ore characterisation data which allows the comminution ore
parameters to be determined based on correlations with the GeoMetso™ – Automated Geometallurgical modelling
strength data (Is50) available in the geotechnical block model. Geometallurgy integrates the disciplines of geology, mining
The Phu Kham ore Is50 strength values and comminution and metallurgy with the aim of developing proactive operating
parameters were determined to fit close to the Metso PTI strategies as a function of ore variability. Geometallurgical
database correlation range, providing confidence in the modelling requires a detailed understanding of the relevant
comminution parameters within the block model. ore properties, and models of how these ore properties will
The inherent variability and maximum plant capacity were affect the performance of the blasting, crushing, grinding
accounted for in the throughput forecast model. Statistical and separation stages, in terms of throughput, recovery and
FIG 44 – Average Is50 and RQD values over the mine schedule.

D Bennett et al
FIG 45 – Phu Kham throughput forecast.
product grade. This is the foundation of the predictive models

and throughput forecasting implemented at Phu Kham;
linking the plant performance with ore properties.
The predictive models rely on ore characterisation data. The
collection of ore characterisation data is labour intensive and
expensive, and therefore often sparsely distributed across
the ore deposit. Additionally, in most mining operations
it is difficult to keep track of a 1000 m3 block of ore (and its
geotechnical characteristics) from a blast, through ROM pads,
stockpiles, crushing and grinding circuits. The resulting
uncertainty in feed characteristics makes any empirical
approach to geometallurgical modelling difficult.
The SmartTag™ system allows the ore (and its characteristics)
to be tracked through the process precisely. An extension of
this system, GeoMetso™, provides a link between the ore
source (and characteristics) with plant performance data from
the process control system and automatically updates the
predictive models and block model in real-time. It is described FIG 46 – Schematic of the GeoMetso™ methodology.
in more detail by Lynch-Watson et al (2013).
More accurate geometallurgical modelling and throughput
The GeoMetso™ concept is shown schematically in Figure 46. forecasting can improve long-term mine planning, and capital
Initial ore domains are defined based on preliminary ore equipment purchases can be predicted well in advance of their
characterisation tests, and predictive models are developed requirement. In the short term, the plant receives advance
for each unit operation as per the previous methodology.
notice of the ore type/s about to be processed and adjustments
SmartTags™ are then used to continuously track ores from
made to operating conditions to optimise plant performance.
the mine through the process, and when linked with process
data from the plant DCS, provide actual plant performance The GeoMetso™ system is currently being implemented at
data (throughput, recovery, grade etc) for each ore type and Phu Kham. Due to mining operational constraints, the mining
blast conditions. These data are automatically compared with activity has predominantly been in heavily weathered areas
model predictions and updated in the block model using the with soft ore that were not included in the current system.
SmartTagTM software in real-time. Thus, the block model is Therefore, only limited data is available to date, but is
continuously updated and refined with actual plant data. This sufficient to demonstrate the sort of information the system
eliminates the need for further expensive ore characterisation will be able to automatically generate.
tests and improves the accuracy and predictive abilities of A typical screen capture of the system software is shown in
geometallurgical modelling. Figure 47. In this example, a blast and ore blocks are shown

FIG 47 – Example of GeoMetso™ system interface.
on the right (cooler colours indicate softer ore, and warmer The SAG mill throughput is inversely proportional to
colours harder ore). The mill feed trend is shown on the left feed strength and structure characteristics, with the mill
and suggests a correlation between the ore hardness and feed throughput increasing with decreasing strength (PLI) and
rate, with higher feed rates achieved for the softer ores. structure (RQD) indicators as expected. Combination of the
More detailed analysis of the data already generated by the parameters into a single ‘grindability’ index demonstrated an
GeoMetso™ system indicated several interesting correlations, even stronger correlation with an R2 value of 94 per cent.
as shown in Figure 48. The blast design was relatively
constant over data collection period, so changes in ROM size CONCLUSIONS
distribution would be a result of intrinsic ore characteristics. The Phu Kham copper-gold operation will experience harder
The data demonstrates that crusher throughput increases ore types which have the potential to limit throughput as mining
with coarser feed (a result of higher RQD), which seems progresses deeper into the pit. To evaluate how to maintain the
counter intuitive. However, Phu Kham is located in a tropical target throughputs over the LOM, Metso PTI was engaged to
zone, and the data was collected during the wet season when conduct a throughput forecasting and optimisation project.
fine clay material is known to cause problems with crusher The objectives of the project were to identify opportunities to
throughput. Therefore, this correlation may not hold true over increase throughput when treating hard ore types, develop
an entire year. This operational issue may have implications if a throughput forecasting model, and determine if and when
the PF is increased to produce more fines, and will need to be secondary crushing or other process changes would be required
considered in the definition of blasting domains. to maintain the target throughput over the LOM.
FIG 48 – Examples of correlations established from GeoMetso™ data.

D Bennett et al
The project involved ore characterisation, detailed The authors would like to particularly thank company
audits of blasting and comminution practices linked with geologists, mining engineers, metallurgists, process control
ore characterisation data using SmartTagTM ore tracking, engineers and operating staff for their work and technical
and development of site-specific models for blasting and assistance in the numerous projects to increase throughput and
comminution processes. These models were integrated to recovery at Phu Kham. The authors gratefully acknowledge
provide an optimisation tool for the overall operation, and for the support, technical reality checking and encouragement of
throughput forecast modelling. Peter Munro, Dr Bill Johnson and Katie Barns of Mineralis Pty
Opportunities to increase throughput for hard ores Ltd, Dr Greg Harbort of AMECFW for his flotation modelling
were identified using the modelling tools. These included work, and the work of John Glen and Tony Button at Burnie
increasing blasthole diameter, tightening the blast pattern Research Laboratory, and Helen Coombs at Base Metallurgical
to increase PF, and reducing the primary crusher CSS. Laboratories Ltd.
Combining these changes could increase throughput by more
than eight per cent. REFERENCES
Throughput forecasting indicates these changes may not be Baas, D, Bennett, D and Walker, P, 2014. Developing process control
sufficient to maintain the target throughput for 2019 when the standards for optimal plant performance at PanAust Limited,
hardest and blockiest material is scheduled in the mine plan. in Proceedings 12th AusIMM Mill Operators’ Conference, pp 325–
334 (The Australasian Institute of Mining and Metallurgy:
The modelling indicates that a secondary crushing circuit
Melbourne).
would be required to meet the 2019 production target.
Burger, B, McCaffery, K, Jankovic, A, Valery, W and McGaffin,
The models developed in this project are a useful tool for short I, 2006. Batu Hijau model for throughput forecast, mining and
and long-term optimisation and planning, as demonstrated by milling optimisation and expansion studies, SME 2006.
the outcomes of this project. The GeoMetsoTM system, which
Crnkovic, I, Georgiev, T, Harbort, G and Phillips, M, 2009.
is currently being implemented at Phu Kham, will improve
Commissioning and optimisation of the Phu Kham copper-
the accuracy of these models by automatically updating the gold concentrator, in Proceedings Tenth AusIMM Mill Operators’
predictive models and block model with measured plant data Conference, pp 107–116 (The Australasian Institute of Mining and
in real-time. Metallurgy: Melbourne).
Hadaway, J B and Bennett, D W, 2011. An overview of the design,
ACKNOWLEDGEMENTS construction, commissioning and early years of operation of
the SAG/Ball mill grinding circuit at Phu Kham copper, gold
This paper is a combination of three of the published papers
operation in Laos, in Proceedings Fifth International Conference
issued in AusIMM conference proceedings which describe on Autogenous and Semi-Autogenous Grinding Technology (SAG
the history of process development and improvement at Phu Conference 2011), pp 11–13.
Kham:
La Rosa, D, Valery, W, Wortley, M, Ozkocak, T and Pike, M, 2007. The
1. Bennett, D, Crnkovic, K and Walker, P, 2012. Recent use of radio frequency ID tags to track ore in mining operations,
process developments at the Phu Kham Copper-Gold in Proceedings Symposium on Computer Applications and Operations
Concentrator, Laos, in Proceedings 11th AusIMM Mill Research in the Minerals Industries (APCOM: Santiago).
Operators’ Conference, pp 257–272 (The Australasian Lynch-Watson, S, Valle, R, Duffy, K, La Rosa, D and Valery, W,
Institute of Mining and Metallurgy: Melbourne). 2013. GeoMetso™: a site-specific methodology to optimize
2. Hoyle, A, Bennett, D and Walker, P, 2015. The Increased production and efficiency over the life-of-mine, in Proceedings
Recovery Project at the Phu Kham Copper-Gold ProceMin 2013, p 507.
Operation, Laos PDR, in Proceedings MetPlant 2015, Meka, Z and Lane, G, 2010. Recent metallurgical developments
pp 246–265 (The Australasian Institute of Mining and and their impact on minerals project execution, in Proceedings
Metallurgy: Melbourne). XXV International Mineral Processing Congress (IMPC 2010),
3. Bennett, D, Tordoir, A, Walker, P, La Rosa, D, Valery, W Melbourne).
and Duffy, K, 2014. Throughput forecasting and
Rybinski, E, Ghersi, J, Davila, F, Linares, J, Valery, W, Jankovic,
optimisation at the Phu Kham Copper-Gold Operation, A, Valle, R and Dikmen, S, 2011. Optimisation and continuous
in Proceedings 12th AusIMM Mill Operators’ Conference, improvement of the Antamina Comminution Circuit, in
pp 381–390 (The Australasian Institute of Mining and Proceedings Fifth International Conference on Autogenous and Semi-
Metallurgy: Melbourne). Autogenous Grinding Technology, (SAG 2011).
The authors acknowledge Phu Bia Mining Limited and
PanAust Limited for allowing these papers to be published.

Contents
What can go wrong in comminution circuit design?

C Bailey1, G Lane2, S Morrell3 and P Staples4
ABSTRACT
The design of semi-autogenous grinding (SAG) mill based comminution circuits for
the treatment of competent ores – similar to those at Geita and Boddington – involves
the same process of sample selection, test work, data analysis and data modelling/
interpretation as that used for ‘average’ competency ores. However, over the past
couple of years a number of issues have arisen that are common to the design of
circuits for the comminution of ‘highly’ competent ores. The impact of these issues
on project viability is generally more pronounced when treating hard ores than
average to soft ores due to the greater impact on capital and operating costs.
The issues encountered have included:
•• standard test procedures have been modified
•• test equipment has worn or been damaged
•• different procedures yield different data and varying interpretation
•• modelling and empirical calculations have been based on poor benchmarks, or
used incorrectly in the case of JKSimMet, yielding misleading outcomes.
The purpose of this paper is to present, discuss and clarify some of the issues
associated with conducting test work and designing comminution circuits for the
treatment of ‘highly’ competent ores to reduce the level of conflict arising from
interpretation and application of test work data.
Specifically, the issues associated with the bond crushing (impact) and rod
work indices measurement, the various SAG mill specific energy tests, and the
interpretation of the resulting data will be discussed in the context of case studies.
INTRODUCTION
Over the past two years, several copper and gold projects have involved the processing
of competent ores.
The measurement of the level of ore competence has historically relied on the use
of bond crushing and rod mill work indices, unconfined compressive strength, point
load strength, drop tests or media competency tests such as those developed by Allis
Chalmers (Mosher and Bigg, 2002).
More recently, several other tests have been developed to suit the requirements of
autogenous grinding (AG) and SAG mill design, namely the:
•• Advanced Media Competency Test (Siddall, Henderson and Putland, 1996)
•• JK Pendulum Test, then the JK Drop Weight Test (Napier-Munn et al, 1996)
•• SMC Test® (Morrell, 2004)
•• Starkey Test, then the SPI (SAG Power Index) Test (Starkey and Dobby, 1996)
•• SAGDesign™ test (Starkey, 2006).
There are reasonable correlations between all the tests’ data that generally allow
any or all of the data to be used to design a SAG mill based grinding circuit for
‘typical’ ores of a moderate competence range.
1. MAusIMM, Principal Consultant, JKTech Pty Issues have arisen where the expected SAG mill specific energy is greater than
Ltd, Indooroopilly Qld 4068. 10 kWh/t. These issues have been associated with the way crushing and rod mill
Email: c.bailey@jktech.com.au work index tests have been conducted, issues associated with JK Drop Weight Test
2. FAusIMM, Chief Technical Officer, Ausenco machine calibration, issues associated with the relevance and interpretation of SPI
data and the way in which has be used as a design tool.
3. Managing Director, SMCC Pty Ltd, Chapel Hill A key point of context is that test work is not conducted just to gather data, test
Qld 4068. Email: steve@smccx.com work is conducted to mitigate risks associated with the selection and design of
4. MAusIMM, General Manager, Ausenco Canada a circuit and the cash flow that the project generates. Hence, a full understanding
Inc, Vancouver BC V6B 2W2, Canada. of the implications of the test work methods and data interpretation is required to
Email: paul.staples@ausenco.com effectively mitigate the risks.
235
C Bailey et al
TEST WORK METHODS As a consequence, metallurgists requesting impact work

The consistency of test work results has been recognised as an index test work should understand what equipment is being
issue in data analysis for some time. The following discussion used and how the data should be interpreted.
highlights some of the issues associated with test work used
in for designing comminution circuits. Bond rod mill work index
The Bond rod mill work index (RWi) is used by some
Historical work practitioners to determine specific energy based on empirical
Dunne and Angove (1997) conducted an audit of comminution methods.
methods across laboratories in Australia and the USA and JKTech conducts a regular program of testing for each of
concluded that: the laboratories licensed to conduct Drop Weight tests using
•• the Bond ball mill work indices determinations gave a standard rock sample. The results of this testing for the Drop
reasonably reproducible results Weight test are discussed in the next section. The test program
•• significant variation was observed in the measurement also includes the conduct of both the RWi and Bond ball mill
of the bond rod mill work index on at least one sample work index (BWi) tests on the same standard rock sample.
tested The results of the comparative testing of RWi across seven
•• large variations in abrasion index and crushing work laboratories around the world are shown in (Table 1). The
index were due to variations in the test method across standard deviation of the RWi measurements is 12.1 per cent
the laboratories. compared with value of 3.4 per cent for the BWi measurements.
Again, interpretation of the data is reliant on empirical
Bond crushing (impact) work index relationships that may be have been developed based on
In general, the impact or crushing work indices (CWi) data from one laboratory, that is not transportable without
calculated from test work carried out on machines conforming adjustment to data from another laboratory. Importantly, a
to Bond’s original design correlate well with the drop-weight large proportion of the historical data on ores in Australia was
index (DWi) which, is obtained from the SMC Test® (see generated by laboratories A and B. This data is not ‘wrong’,
blue diamond points in Figure 1). However, recently, for but models that use those data may need recalibration for
several projects the CWi values have fallen well below this data sourced from, say, North or South America.
relationship and this has led some metallurgists to claim
that the ore is not overly competent. Closer investigation TABLE 1
of the source of these data indicated that they came from a Summary of Bond rod mill work index test work.
modified design of testing machine the results from which
followed a very different relationship to the DWi as seen by Laboratory Bond rod mill work index
the red circular points in Figure 1. It is believed that these data (kWh/t)
were derived from the same (modified) design of machine
A 22.1
which were manufactured by a company in North America,
one of which ended up in a laboratory in Australia and the B 22.4
other three were bought by laboratories in North and South C 18.1
America. Part of the modifications involved the impact heads
D 17.3
which are different in shape to Bond’s original. When the
Australian laboratory started to use their new machine one E 16.8
of their clients indicated that the CWi values were far too F 19.5
low. The impact heads were then changed to match Bond’s
original and the resultant values increased by 100 per cent. G 16.2
Some comparative tests were then done with a lab which had G 18.3
one of Bond’s original design machines and similar results G 17.1
were obtained.
120
Bond design
100 Modified design
.
80
CWi*sg
60
40
20
0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
DWi
FIG 1 – Correlation of crushing work index (CWi) and drop weight index (DWi).

JK Drop Weight Test competent ores, this difference appears to be at least in part
The JK Drop Weight Test (JKDWT) data has historically been due to the way in which the data are fitted to determine the
represented in terms of the ‘A × b’ parameter. The ‘A’ is A × b values. In one example, JKTech fitted the JKDWT data
the asymptote of a plot of specific energy against T10 (per to get an A = 100 and a b = 0.2 (ie A × b = 20) for three of
cent of product below ten per cent of the feed size) and ‘b’ is the four JKDWT Tests in this example. In fitting the JKDWT
a measure of the ‘slope’, represented by T10 = A × (1-e-b.ecs). data, JKTech constrains A to its theoretical maximum value
of 100. This approach has varied over the last 15 years and
JKTech’s routine testing mentioned above has indicated that
has led to related variation in the A × b valued determined
the standard deviation of JKDWT A × b values from all the
licensed test laboratories on the same standard rock sample from a given data set. If the A value is relaxed, fitting leads
is 4.2 per cent. These results prompted an investigation by to A × b values between 22 to 23.5 for the same test data. The
JKTech into the source of the variation in JKDWT results. The latter values compare reasonably with the SMC A × b values
outcomes of the investigation are reported in Stark, Perkins of 23.2, 23.7, 23.0 and 24.4 for the comparable samples. The
and Napier-Munn (2008). The investigation involved a set of fitting methodology used to determine A and B is particularly
24 JKDWTs on homogeneous material conducted by three sensitive at the extremities of the ore competence scale and
operators. The standard deviation of the A × b values was can have significant implications for competent ore where a
5.7 per cent. The largest contribution to the variation was the 10 per cent difference in A × b has a significant impact on the
selection of particles to be tested. subsequent calculated SAG mill specific energy.
There have been reports of some issues with the JKDWT Veillette and Parker (2005) published a graph of A × b
machine if the machine is not maintained effectively. This versus SAG mill specific energy (Figure 2). The product A × b
outcome is rare but running check samples at other laboratories has no formal units although its value is inversely related to
can alleviate this type of risk. This is the reason for the JKTech ore competence, ie the lower values of A × b indicate harder
routine comparative testing at all licensed laboratories. rock. In contrast to A × b, the DWi parameter has the units
of kWh/m3 and hence tends to be more linearly related to
SMC Test® SAG specific energy for a given circuit configuration. These
Recently, Morrell has developed a simpler approach, the SMC relationships are used to indicate the potential deviation
Test®, and a DWi that is related to the A × b parameter and in SAG mill specific energy at extreme A × b values (<30).
particle SG (Morrell, 2004). The uncertainty that exists with SAG power determinations
Both the SMC Test® and JKDWT rely on being able to select for A × b values below 30 is illustrated by the pink region in
samples of competent rock or quartered core of a certain size Figure 2. Based on recent experience, with an A × b value of
in narrow size intervals in the range -63 mm + 13.2 mm. Details 25 SAG power predictions could range from 12 to 18 kWh/t,
are given on the JKTech website (JKTech Pty Ltd, 2009). a range of 50 per cent, dependent on the consultant doing the
For low competence ores, this approach places a potential evaluation and the mode of SAG mill operation.
bias on the data as the more friable component of the ore is At least some of the variation is due to feed size
not able to be tested; however, for competent ores this issue distribution, although other factors such as aspect ratio, mill
is not relevant. speed, ball load and whether pebble crushing is in-circuit
Recently, when comparing the results for the A × b values also contribute. SAG mill specific energy is sensitive to SAG
determined from the SMC Test® and JKDWT indicates mill feed size distribution. Thus, the selection of the primary
there can be a discrepancy between the two test results. The crusher and the mine blast design are critical inputs to SAG
JKDWT results can be lower (ie appear more competent) mill throughput determination which are often ignored or at
than those from the SMC Test®. Based on experience with least, underestimated.
18
Boddington
16
Geita
Area E
14
12
Ahafo
SAG kWh/t
10
Cadia
Phoenix Fimiston
6
Lone Tree
Sage
4
Batu Hijau Yanacocha

2
K Kollo
0
25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160
JK A x b
FIG 2 – Relationship between JK A × b parameters and semi-autogenous grinding (SAG) mill specific energy (modified from Veillette and Parker, 2005).

C Bailey et al
SAG Power Index test better indicates the impact of critical size on SAG mill specific
There are two aspects to the design of either single-stage energy and the feed mass (10 kg) allows a ball mill work index
SAG or two-stage SAG/ball mill circuits. First is the energy test to be conducted on the product.
required to grind the ore, taken from test work data, and On recent projects dealing with competent ores, the
second is the application of grinding equipment required SAGDesign™ approach has yielded similar outcomes to
to provide this energy, with its empirical factors to allow for those using Morrell’s approach (Morrell, 2008) and Ausenco’s
grinding efficiency using that equipment. in-house design approach based on Bond work indices, A × b
The grinding power-based tests, SAG Power Index test values and an empirical model of operating efficiency.
(SPI; circa 1991) and SAGDesignTM, were developed by John
Starkey (Starkey and Dobby, 1996; Starkey, 1997; Starkey, SPECIFIC ENERGY DETERMINATION
Hindstrom and Nadasdy, 2006). These tests are conducted All methods used for the calculation of grinding circuit
on -12 mm and -19 mm feed, respectively and interpretation specific energy are empirical, although some methods have
of the data is applied to a nominal standard feed size of a superior level of packaging (eg JKSimMet). In general there
80 per cent -150 mm (primary crushed ore) to a plant. are three types of models:
The SAGDesignTM test (discussed further below) was 1. Bond-based methods
developed, in part, to overcome some of the concerns
2. proprietary test-based models (for example, MacPherson,
regarding the ‘grindability’ of the ‘critical size’ material in the
SPI, Starkey, SMC Tests®)
SAG mill, hence the coarser feed size.
3. more sophisticated population balance/breakage
The SPI test data interpretation relies on empirical
model/classification model-based methods.
relationships to describe the impact of feed size, pebble
crushing and other operating factors on the SAG mill Using all three methods successfully requires a high
specific energy, much like most other mill specific energy level of appreciation of the limitations of the methods. Some
calculation methods. consultants use a combination of all three approaches for
Figure 3 compares the SPI and DWi data for a range of comminution specific energy determination.
projects. The DWi data incurs some scatter as the DWi For the vast majority of ores, experienced consultants find
parameter relates to the volume of the ore rather than the mass that each of the methods tends to give similar results.
(used by the SPI data) and ore specific gravity therefore has an Some specific issues related to ball mill specific energy
impact. The correlation between DWi and SPI is moderately calculation are:
good for low competence ore (SPI below 100) but diverges for
•• whether large mills are inherently more efficient than
high competence ores.
small mills, per Bond’s efficiency factor
SAGDesign™ test •• the use of the phantom cyclone to factor for ‘fines’
Starkey developed and markets the SAGDesign™ test. The produced by the SAG/AG mill.
SAGDesign™ test was created specifically for sizing of new Starkey (2006) states that:
SAG and ball mills. It was patented by Outotec who guarantee … benchmark testing at large (mill) installations shows
throughput and grind for Outotec supplied grinding mills conclusively that the diameter correction factor proposed
based on the test data. by Bond is valid and perhaps even conservative. This factor
The SAGDesign™ test has some potential advantages is: Operating Wi = bond Wi × (8/D)0.2 or 0.80 × BWi
over the SPI Test in that the feed size is coarser and arguably for a 24 ft diameter grinding mill. C Rowland proposed
16
14
12
10
D Wi (kW/m 3 )
0
0 50 100 150 200 250 300
SPI (min)
FIG 3 – Comparison of semi-autogenous grinding (SAG) Power Index test (SPI) and drop weight index (DWi) data for various ore samples.

a limiting value of 0.914 on this factor, but that value was parameters from JKDWT or SMC Test®. However, as with
based more on the size of mill available in his time rather any simulation tool, understanding the models and their
than on an empirical analysis. limitations is critical and there are several usage rules which
Starkey also sets out eight rules for design that are reasonably must be observed to get accurate predictions. The AG/SAG
consistent with general grinding circuit design practice. mill model used in JKSimMet is known as the Variable Rates
However, Starkey’s assertion concerning Bond’s diameter Model and is described in detail in Napier-Munn et al (1996).
correction factor is not supported by Morrell (2001), who Perhaps the most important limitation of this model is that
could find no relationship between energy efficiency and mill the variable rate equations were based on operating data
diameter, as inferred by Starkey’s benchmarking. from a series of mills with an average total load volume of
The authors of this paper do not use the ‘phantom cyclone’ 25 per cent. It is well known that the breakage rates vary with
methodology. That SAG mills, and more so AG mills, produce the mill load but the effect of this variation is not included in
product size distributions containing considerably more fines the model. This means that all design simulations should be
than rod mills or tertiary crushers is not in question. However, conducted at 25 per cent load.
AG/SAG mill product size distributions can also contain A second limitation arises when using this model to simulate
more coarse particles as well. This, combined with the fact modern, all pebble port, grate mills. It is essential to set the
that AG/SAG mills are good at preferentially breaking the Xm parameter (fine size) to a small number, say 1.5, the Xg
softer particles in the feed and leaving the coarse end of the parameter (nominal grate size) to a small number, say 1.5 and
product distribution harder, reduces or eliminates any effect control the discharge from the mill by specifying the pebble
of extra fines production in the primary stage of milling. port size with the pebble port open area set to 100 per cent of
Recently, the determination of specific energy for SAG mill the total open area.
based circuits treating competent ores has resulted in some A third limitation is that, in some circumstances, JKSimMet
diverging views. In the opinion of the authors, these diverging predicts excessive pebble recycle. Experience dictates that
outcomes are related to two factors: pebble recycle rarely exceeds 30 per cent of new feed, so
1. limitations of the SPI test when determining the JKSimMet predictions of greater recycle ratios should be
competence of competent ores treated with caution.
2. incorrect application of the JKSimMet package. JKMRC is currently constructing and testing an upgraded
variable rates model which is aimed at overcoming several
Limitations of the SPI approach of the limitations discussed above. This work is funded by
The SPI approach was developed by Starkey and was then JKTech.
commercialised by Minnovex (Starky and Dobby, 1996) and Random variation in the ore properties measured by
now SGS. the JKDWT or the SMC Test® can lead to variation in the
In the opinion of the authors the SPI approach to SAG predicted specific energy requirements. In association with
mill performance correlation is less applicable for the the JKDWT variation study discussed above, Stark, Perkins
determination of SAG mill specific energy for competent and Napier-Munn (2008) reported a design simulation
ores. Differences in calculated specific energy have been study involving 100 repeat simulations using correctly
observed between the SPI and DWi based approaches for distributed random values of A and b. For the particular
several large recent projects in the study phase where the A mill configuration chosen, the mean predicted throughput
× b values have been less than 30. was 339.4 t/h with 95 per cent confidence limits of ±25.0 t/h.
The SPI method has been used successfully to predict This equates to ±7.4 per cent of the throughput. This is the
performance correlations when treating soft ores, and is sort of variability that can be expected simply from random
arguably a better statistical approach to sample selection variation of A × b values.
and testing when compared with the JKDWT or SMC Test® One of the advantages or disadvantages (depending on
as the latter rely on the selection of single particles of a one’s point of view) of using JKSimMet is the requirement for
predetermined size and this in itself may create a bias in the a full feed size distribution. Crusher manufacturers publish
test method for soft ores. product size distributions which generally vary with crusher
When testing competent ores, the SPI data shows a poor gap and sometimes broadly with ore properties. However,
correlation with the DWi data. Starkey, who developed the these are almost always based on scalped feed and the
SPI method, has used a modified approach in developing scalping process can remove as much as 50 per cent of the
the SAG Design Test that appears more consistent with the feed. This will seriously underestimate the quantity of fines
JKDWT and SMC Test® based models. (<10 mm) in the AG/SAG feed.
Morrell and Morrison (1996) reported a relationship
Limitations of the JKSimMet package between AG/SAG mill new feed F80 and the abrasion
JKSimMet is a comminution and classification simulation parameter ta which is derived from a laboratory abrasion test
package based on 45 years of research at the Julius Kruttschnitt
Mineral Research Centre (JKMRC), University of Queensland. F80 = CSS - 78.7 - 28.4 # In (ta)
The package is developed and marketed by JKTech, the
technology transfer vehicle for the JKMRC. In excess of and is reported as part of the JKDWT methodology. This is a
350 JKSimMet packages are in use around the world. The function of crusher gap as shown in the relationship:
JKSimMet package is the most widely accepted population
where:
balance/breakage model/classification model and differs
from the other approaches due to the more fundamental F80 is the primary crusher discharge 80 per cent passing
rather than empirical bases of the models. size in mm (or AG/SAG mill feed)
JKSimMet is frequently used as one of the methods of CSS is the primary crusher closed side setting in mm
estimating SAG mill power requirements, using ore specific ta is the abrasion parameter from the JKDWT

C Bailey et al
The standard deviation associated with this regression that are unrealistically high. Specific mistakes that have been
found to have been made are:
F80 = 0.2 # CSS # DWi0.7
•• The incorrect setting of the grate parameter (xg) to the
actual pebble port aperture in fully ported grates. The
relationship was ten per cent of CSS. correct approach is to set xg and xm to 1.5 and to set
More recently Morrell has provided an updated relationship: the pebble port aperture parameter to the minimum
where: dimension of the actual pebble ports.
F80 is the primary crusher discharge 80 per cent •• Use of breakage rates from an existing mill which have
passing size in mm (or AG/SAG mill feed) either been fitted in an inappropriate manner or represent
operating conditions very far from those being used for
CSS is the primary crusher closed side setting in mm
design purposes. The best (most foolproof) method is to
DWi is the drop weight index from the SMC Test® rely on the variable rates model’s default rates.
Once the F80 of the comminution circuit feed is known, a •• Incorrect use of the F80 parameter and/or unrealistic feed
suitable size distribution is selected from the database and size distributions. In most cases reliable operating plant
adjusted to match the estimated F80. data indicate that AG/SAG mill feed size distributions
It is important to note that the F 8 0 predicted from both are linear in log-log space and distributions chosen
these relationships is mostly less than the primary crusher for design purposes should follow this pattern. The F80
CSS, sometimes a lot less. For the latter relationship, it is only should also be realistic and should be determined using
for DWi values of 10 kWh/m3 or higher that the F 80 value is either the DWi-F80 relationship recommended by JKTech
greater than the gap. or from an existing plant with a similar ore hardness. The
F80 parameter should be set to 100 mm regardless of the
Guidelines for JKSimMet modelling and simulation actual F80 used in cases where the default breakage rates
When used strictly according to rules provided in the are used. Where breakage rates have been taken from
JKSimMet training manual, the so-called variable rates model fitting data from an existing operation the F80 parameter
usually gives reasonably realistic throughput predictions for value for the fitting should remain unchanged.
AG and SAG mills, including mills up to 40 ft in diameter. •• The ta value, usually obtained from tests in parallel
In particular, it is important that appropriate breakage rates, with the JKDWT, should not be changed in the model
mass transfer parameters and feed size distributions are when simulating the performance of ores with different
chosen. In a number of recent projects failure to obey these A and b parameters. The ta value assists in modelling
rules has lead to throughput predictions for 40 ft SAG mills the fines production in a SAG mill but simulation of ores
FIG 4 – Example of JKSimMet model for the Cadia mill based on published data.

with different ta test values using models developed on mapping the ore types against the mine schedule may be a
other ores can lead to misleading results. reasonable strategy. Final selection of the design case needs
•• The JKSimMet models should always be checked to reflect a business case that includes operational factors that
using a power based empirical model to make sure the may relate to variation in ore characteristics, ore segregation
model outcomes are reasonable. Most comminution (critical for competent ores due to the potentially large impact
design practitioners have power based models that are on SAG mill capacity), equipment availability (such as that
used for this purpose. It is critical that these models for lower availability units such as pebble crushers) and other
have been ‘calibrated’ for the range of ore characteristics site-related factors.
being considered.
Other mistakes in setting up the model are possible either CONCLUSIONS
through lack of sufficient knowledge or simple inattention to Comminution circuit design relies on input and interpretation
detail. It is therefore recommended that the designer apply of comminution test work data. Understanding of the pitfalls
his/her model to a ‘standard’ circuit before using it in a real inherent with these tests is critical to ensure robust designs
design situation. If the model does not predict the standard are achieved in the end.
mill performance to within about ten per cent then it can be
assumed that the model is not appropriate and needs to be Comminution circuit design for average competence ores
corrected. It is suggested that the standard adopted is the is more straightforward than for very competent ores, many
’40 ft SAG mill’ at Cadia Hill (Figure 4). The reasons for this of which have been studied recently. The application of a
choice are that it is a large mill, it treats reasonably competent methodology which works well for lower competence ores
ore and has much detailed information published concerning does not guarantee accurate predictions for all ores.
its design, operating performance and ore hardness (Dunne Recently, it has been observed that engineers and consultants
et al, 1999, 2001; Dunne, Morrell and Lane, 2000; Hart et al, have, at times, underestimated the SAG mill specific energy
2001). The following data have been taken from the published for treating competent ores due to:
literature and are suggested for use as this standard:
•• lack of relevant data for comparable benchmark
Diameter inside shell 40 ft operations
Diameter inside liners 11.96 m •• lack of understanding of the test work methods used by
Length (EGL) 20 ft various laboratories
Belly length (grate, liners) 6.096 m •• poor use of packages such as JKSimMet.
Cone angle 15° This paper is aimed at assisting practitioners in
Open area 8–10 per cent understanding the potential pitfalls.
Pebble port aperture 75 mm (fully ported) The selection of the correct design point and judicious
Trommel effective 15 mm application of contingencies needs to be carefully understood
and tailored for the application in order to achieve the desired
Ore Sg 2.65 t/m3
throughput outcome an annual basis.
A 65
b 0.58 ta 0.49 REFERENCES
F80 90–100 Angove, J E and Dunne, R C, 1997. A review of standard physical
Ball load 12–14 per cent ore property determinations, in Proceedings World Gold ‘97,
Singapore, pp 139–144 (The Australasian Institute of Mining and
Total load 25–26 per cent
Speed 78 per cent of critical
Bond, F C, 1961. Crushing and Grinding Calculations, publication
Power draw 18–18.5 MW number 07R9235B, revised 2 January 1961.
Throughput 2000–2200 t/h Bond, F C, 1963. Crushing and Grinding Calculations, publication
Transfer T80 1.5–2.5 mm number 07P17001, December (Allis Chalmers).
Pebble rates 400–500 t/h Dunne, R, Crittenden, R, Lane, G and Morrell, S, 1999. The Cadia gold
copper project – exploration to start-up, SME Annual Meeting,
Other considerations Denver, Colorado, USA.
There are several other considerations that are important in Dunne, R, Morrell, S and Lane, G, 2000. Design of the biggest mill in the
determining the most effective comminution circuit design world, in Proceedings After 2000 – The Future of Mining, pp 179–185
(The Australasian Institute of Mining and Metallurgy: Melbourne).
and subsequent equipment selection. These include:
•• selecting a design point that strikes a balance between Dunne, R, Morrell, S, Lane, G, Valery, W and Hart, S, 2001. Design
of the 40 foot diameter SAG mill installed at the Cadia Gold
contingency and operational variation and cost (capital
Copper Mine, in Proceedings SAG 2001, pp 43–58 (University of
and operating)
British Columbia: Vancouver).
•• marrying the perspectives of designers and operators to
Hart, S, Valery, W, Clements, B, Reed, M, Song, M and Dunne, R, 2001.
target the best business case Optimisation of the Cadia Hill SAG mill circuit, in Proceedings
•• differentiating between instantaneous performance and SAG 2001 (University of British Columbia: Vancouver).
performance over a period (to achieve budget) that may JKTech Pty Ltd, 2009. Comminution testing information [online].
incorporate operational upsets. Available from: <http://www.jktech.com.au/Products_Services/
In the early stages of a study where limited data is Laboratory-Services/Comminution-Testing/> [Accessed: 20 July
available the design may require a suitable contingency 2017].
to reflect uncertainty due to lack of variability data. As the Morrell, S, 2001. Large diameter sag mills need large diameter ball
study proceeds and data is more ‘representative’, using the mills: what are the issues? in Proceedings SAG 2001, pp 179–193
75th percentile data point for each recognised ore type and (University of British Columbia: Vancouver).

C Bailey et al
Morrell, S, 2004. Predicting the specific energy of autogenous and Stark, S, Perkins, T and Napier-Munn, T J, 2008. JK drop weight
semi-autogenous mills from small diameter drill core samples, parameters – a statistical analysis of their accuracy and precision
Minerals Engineering, 17(3):447–451. and the effect on SAG mill comminution circuit design, in
Proceedings MetPlant 2008, pp 147–156 (The Australasian Institute
requirement of comminution circuits and assessing their energy
utilisation efficiency, Minerals Engineering, 21(3). Starkey, J H, 1997. Getting more from drill core: Preliminary SAG
design, Randol Gold Forum, Monterey.
Morell, S and Morrison, R D, 1996. AG and SAG mill circuit
selection and design by simulation, in Proceedings Autogenous Starkey, J and Dobby, G, 1996. Application of the Minnovex SAG
and Semi-Autogenous Grinding, pp 769–790 (University of British power index at five Canadian SAG plants, in Proceedings
Columbia: Vancouver). Autogenous and Semi-Autogenous Grinding, pp 345–360 (University
of British Columbia: Vancouver).
Mosher, J and Bigg, T, 2002. Bench scale and pilot plant tests for
comminution circuit, in Mineral Processing Plant Design, Practice and Starkey, J, Hindstrom, S and Nadasdy, G, 2006. SAG design
Control (eds: A Mular, D Halbe and D Barratt), pp 129–135 (Society testing: what it is and why it works, in Proceedings Autogenous and
for Mining, Metallurgy & Exploration: Englewood). Semi-Autogenous Grinding, 15 p (University of British Columbia:
Vancouver).
Napier-Munn, T J, Morrell, S, Morrison, R D and Kojovic, T, 1996.
Mineral Comminution Circuits, their Operation and Optimization, Veillette, G and Parker, B, 2005. Boddington expansion project
pp 69–94 (University of Queensland: Brisbane). comminution circuit features and testwork, in Proceedings
Randol Gold Forum.
Siddall, B, Henderson, G and Putland, B, 1996. factors influencing
sizing of SAG mills from drill core samples, in Proceedings
Autogenous and Semi-Autogenous Grinding, pp 463–473 (University
of British Columbia: Vancouver).

Contents
Do we need a gravity circuit or not?

A case study in applying best practice
A Giblett1 and K Afewu2
ABSTRACT
This paper uses a case study to demonstrate best practice techniques for predicting the
contribution of gravity concentration in a gold recovery process flow sheet. The case
study was selected as an example of an ore with a marginal gravity gold content, which
makes the value of a gravity circuit difficult to quantify. In this instance, the value
of leveraging a significant knowledge base of operating gravity circuit performance
to support the decision-making process was demonstrated. This case study analysis
serves as a useful demonstration of best practice methodologies supporting decisions
to include or exclude gravity recovery in a gold recovery process flow sheet.
INTRODUCTION
The large, low-grade deposits that dominate global gold production are typically
less generously endowed with coarse-free gold than historically significant high-
grade deposits. When the free gold contents are high or visually coarse, the value of
a gravity recovery process can be readily appreciated by plant designers, operators
and management. More commonly the designer deals with less pronounced gravity
gold values and the value of a gravity concentration step can be more difficult to
demonstrate at the study stage.
There are many examples of high-grade gold ores that produce very high gravity
gold recoveries. Where free gold is coarse and readily concentrated, gravity circuits
can account for more than 50 per cent of gold production. In these instances, the use of
gravity concentration to effectively collect this material is easily justified, as the risk of
gold losses due to inefficient downstream processing by other means is easily grasped.
The risk of gold loss due to theft, the development of undesirable inventory within the
grinding circuit, sampling issues and metallurgical accounting problems are further
factors that may support the inclusion of a gravity concentration step in such cases.
When the resource contains less substantial quantities of gold recoverable by
gravity, and the gravity gold size distribution is more moderate, the impact of the
gravity circuit on overall metal recovery will be less pronounced. This is less true
typically for flotation processes, which are more sensitive to particle shape effects and
surface condition, and more likely to suffer from the absence of a gravity recovery
stage. This is in addition to the reduced payment terms received for free gold
recovered to flotation concentrates, which often further supports the use of gravity
concentration in or ahead of flotation flow sheets. Figure 1 gives some guidance on
general classifications of gravity recoverable gold (GRG) size distributions, after
Giblett et al (2013).
The study of a number of projects and operations around the globe has led to the
generalisation that when there is an expectation of 25 per cent or more of the mill feed
gold content being recovered by gravity, it is prudent to include a gravity gold recovery
stage in the process flow sheet. This is contingent on the recovery expectation being based
on a coarse (GRG P80 >350 µm) gold particle size distribution, and high gold recoveries
(>25 per cent) under the stage 1 conditions of the GRG test. Supporting references
for this decision point include the Harmony Hill 50 operation (Minerals Engineering
International (MEI) online, 2004), where 24–30 per cent gravity recovery translated to
approximately 1.5 per cent improvement in overall recovery, and the President Steyn
mine in South Africa (Van de Steen, 2000) which reported a two per cent improvement
in total gold recovery following the installation of a gravity circuit recovering up to
1. FAusIMM, Senior Technical Advisor – 45 per cent of the gold in mill feed. This led the development of the nominal decision
Mineral Processing, Newmont Mining
matrix summarised in Table 1, based on gravity recovery expectations at plant scale, to
Services, Subiaco WA 6018.
Email: aidan.giblett@newmont.com guide early stage decision-making.
2. Chief Metallurgist, Newmont Depending again on the size distribution of the GRG, improvements in overall
Golden Ridge Limited, Accra, Ghana. plant efficiency can be expected for gravity gold contents as low as 10–20 per cent, as
Email: kodjo.afewu@newmont.com is the case with the Kalgoorlie Consolidated Gold Mine (KCGM) operation (Giblett
243
100
90
80
70
Cumulative % Passing
60
50
V. Fine
40
Fine
30
Ave
20 Coarse
10 V. Coarse
0
10 100 1000
Particle Size (µm)
FIG 1 – Gravity recoverable gold size distributions.
Table 1 Deviating from Laplante’s test conditions will reduce the

relevance of what is essentially an empirical test result. This
Nominal gravity process decision ranges.
is particularly evident when the test is conducted at feed sizes
Gravity recovery Gravity recovery Recovery coarser than 100 per cent passing 850 microns, as the test will
expectation at risk tend to understate the coarsely liberated GRG content due to
greater competition between coarse gangue particles and fine
>30% Highly feasible 2–10% to moderate GRG particle sizes. There is also some evidence
20–30% Highly feasible ~1.5% that the use of smaller GRG test sample masses tends to
10–20% Feasible (project specific decision) 0.5–1.0% inflate the GRG values (Giblett and Napier-Munn, 2015), and
this should be considered when reviewing GRG test results.
<10% Feasible (economic impact may be limited) ?
There are other techniques frequently applied to generate
an estimate of the gravity gold recovery, specifically estimates
et al, 2012), where such levels provided the basis for a 0.5– based on the total GRG content determined by the three stage
1.0 per cent improvement in overall plant recovery following or single stage GRG tests (Laplante, Woodcock and Huang,
gravity circuit installation. At low gravity gold recovery 2000; Laplante and Staunton, 2005), and population balance
levels, there are numerous factors which must be considered modelling (Laplante, Woodcock and Noaparast, 1995). Both
in addition to the GRG size distribution, such as process techniques can be prone to inflating gravity gold recovery
chemistry, the downstream recovery processes to be used, expectations, which is, (based on experience) something to be
and practical constraints such as leach circuit residence time avoided at all costs. Gravity circuits that underperform on their
and electrowinning capacity. design basis tend to become the subject of inquisitions focused
The case study presented later in this paper considers a on underperforming resource grades, gold loss, theft or
project with an expectation of plant scale gravity gold recovery metallurgical performance efficiency. When such conclusions
in the range of ten per cent. Based on the criteria in Table 1 are based on an invalid original expectation of what the gravity
this puts the project on the borderline of a gravity circuit circuit can achieve, not only are investigative resources wasted
that can add economic value and demands a more rigorous but the organisation can become frustrated with the concept of
assessment be completed before an informed decision on the gravity recovery and may avoid it in future plant designs.
value of a gravity circuit can be made.
The correlation of the total GRG value to plant scale recovery
is extremely variable, reported by Laplante (2000) to vary
A RELIABLE BASIS FOR A GRAVITY RECOVERY EXPECTATION between 25 and 85 per cent, and is highly contingent on how
It is important to qualify that Table 1 relates to a gravity much of the GRG is of a coarse particle size or coarsely liberated.
recovery expectation that is directly proportional to the The examples shown in Figure 2 reflect recent projects where
stage 1 GRG content, recovered during a three stage Laplante the correlation between total GRG and stage 1 GRG, and
GRG test. As demonstrated by comparison to plant operating therefore the correlation between total GRG and actual plant
data (Giblett, 2011) this criterion is an excellent indicator of gravity recovery, vary significantly. The clear conclusion is
plant scale gravity recovery in standard gravity recovery that it is not possible to make a reliable estimate of plant scale
circuits. The standard gravity recovery circuit is defined as gravity recovery from the total GRG value on its own.
the use of centrifugal concentration to treat 15–50 per cent of Population balance modelling has been used on several
the grinding circuit recirculating load, screened at 2–3 mm, recent projects, being a method that allows the gravity
followed by intensive cyanidation of the gravity concentrate. circuit to be studied in some detail. The response of both the
Specifically, the correlation assumes that stage 1 of the gravity recovery and the final grinding circuit product can be
GRG test has been performed in accordance with Laplante’s assessed, in response to changes in gravity circuit size, grind
specifications (Laplante, Woodcock and Huang, 2000). size and classification performance. The ability to consider the

Do we need a gravity circuit or not? A case study in applying best practice
90
80
70
Gold Distribution % of Total

60
50
40
30
20
10
0
Total Stage 1 Plant Total Stage 1 Plant Total Stage 1 Plant
GRG GRG Gravity GRG GRG Gravity GRG GRG Gravity
Site A Site B Site C
FIG 2 – Correlations between gravity recoverable gold content and plant gravity gold recovery.
coarse particle gold content of the grinding circuit product is It is noted that the stage 1 GRG content is low at 10–
particularly useful to gauge the potential impact of gravity 15 per cent, and a relationship between head grade and
recovery on overall plant recovery. Such assessments must be coarse gold content is evident in the data. While it would
based on correlations to particle leach rates using shrinking be optimistic to suggest this grade– GRG relationship is
core models, or assumptions around size by size flotation unequivocally true given the small data set, it does follow
recovery of free gold particles. However, the population similar observations made on other projects. The important
balance models require assumptions around gold particle point to note is that the average gold recovery expectation
grinding rates, classification efficiency and gravity size for the deposit should reflect the average grade of the
recovery that are not applicable to all ranges of gold particle deposit and not be distorted by high-grade composites that
purity, gravity circuit feed size and gangue specific gravity. do not reflect typical plant feed grades.
As a result, there is potential to significantly overstate the Consequently, it is a reasonable expectation that the
gravity gold recovery using this method and this has been the gravity gold recovery for the project would be of the order
authors’ experience on several projects. of 10 per cent at a design plant feed grade of 2.0–2.5 g/t Au,
While the population balance modelling is an important and may increase if higher grade ores are processed. The
tool to predict gravity circuit impacts, a degree of caution benchmarking of plant performance data against the GRG
is required when interpreting the results, particularly when test as previously discussed, suggests that it is reasonable to
assumptions regarding size recovery, grinding rates and assume that this recovery estimate will be broadly accurate
classification performance are unclear, or otherwise have not within ±5 per cent absolute. While it is possible to significantly
improve this recovery by more efficient fine GRG extraction,
been validated for the ore in question. Models of this nature
finer screening (-1 mm), significantly higher gravity circuit
often demonstrate a significant relationship between gravity
feed rates and likely higher separation centrifugal forces
circuit size and overall gravity recovery, which are not always
would be required.
evident in practice. Implicit in a relationship of this nature is a
significant increase in fine GRG recovery as the gravity circuit Based on the guidance provided in Table 1, it is estimated that
gets bigger. As has been demonstrated by Giblett (2011) for ten per cent gold recovery by gravity the impact on overall
a significant amount of operating data conflicts with this gold recovery may be as much as 0.5 per cent. This is obviously
philosophy, and implies that a gravity circuit will typically highly dependent on the gravity circuit removing gold from
allow ready removal of the amount of free gold reflected the CIL feed that is too coarse (or too coated) to leach within
by the stage 1 GRG content, even when treating very low the available residence time. To further qualify the correlation
levels of the grinding circuit recirculating load (<10 per cent). between gravity recovery and total recovery it is instructive
Circuits observed in that analysis treating significantly to compare the GRG size distributions for this study against a
suitable reference size distribution. This comparison, shown in
larger portions of the recirculating load (30–50 per cent)
Figure 3, uses the reference GRG size distributions reported for
demonstrated no quantifiable increase in efficiency, again
the KCGM operation by Giblett et al (2012).
recovering approximately the free gold content reflected by
the first stage of the GRG test. Population balance modelling of the measured GRG size
distribution can then be performed to estimate the impact of
gravity circuit operation on the free gold size distribution in the
A CASE STUDY ANALYSIS grinding circuit product. As previously discussed the model
Three GRG test results were provided for interpretation. applies measured values of GRG grinding rates and size by
An initial three stage GRG test had been performed, in size gravity recovery efficiencies to determine the impact of
addition to two further results derived from a modified gravity circuit size (feed rate) on final cyclone overflow GRG
GRG test procedure. The modified test procedure reflected size distribution. One outcome of the modelling exercise is an
smaller sample mass and a coarser stage 1 feed size than the estimate of the impact of the size of the gravity circuit on the
protocols developed by Laplante. The results of the testing are amount of gold recovered by gravity, shown in Figure 4 for
summarised in Table 2. case study data sets.

Total GRG Comparisons

60
50
GRG Content (% of total Au)

40
30
20
10
0
10 100 1000
Particle Size (mm)
Test 1 Test 2 Test 3 Reference 1 Reference 4
FIG 3 – Comparing gravity recoverable gold size distributions.
is predicted to have a limited impact on overall plant gold

Table 2 recovery.
Gravity recoverable gold (GRG) test results.
The operation further provided assay data for the cyclone
Sample ID Mass Head Stage 1 Total GRG underflow and CIL feed to allow an analysis of the amount of
gold accumulating in the grinding circuit. Operations dealing
(kg) (g/t Au) GRG (%) (%)
with coarse gold can show significant gold accumulation in
Test 1 44 2.4 9 56 the recirculating load and the comparison of gold grades in
Test 2 20 2.7 10 48 the cyclone underflow to the cyclone overflow, as a ratio,
have long been used as a metric to quantify this. With access
Test 3 20 4.2 15 44
to a large database of similar ratios for operating plants this
information can be put into context to quantify how significant
It is necessary to observe that the gravity model predicts the gold accumulation is.
20–25 per cent gold recovery by gravity for a small
Over 500 shift data points were provided for analysis, and
gravity recovery effort (ten per cent), derived by treating
demonstrate an average ratio of gold assay in the cyclone
approximately ten per cent of the recirculating load. This is
underflow against the cyclone overflow of approximately
considered a very optimistic prediction, inconsistent with the
seven. This is considered an average value with reference to
fine nature of the GRG size distribution and demonstrates
the Newmont database, and reflects some accumulation of
why the model outputs need to be reviewed critically.
gravity recoverable gold and a moderate GRG particle size. A
An added value of the population modelling approach is comparative example exists within the data set, where a gravity
the ability to estimate the impact of gravity recovery on the circuit was operated for low yield and subsequently shutdown
size of free gold particles leaving the grinding circuit and with no observed impact on overall gold recovery. In that
reporting in this case to cyanidation. In this assessment the instance the assay ratio was approximately six with the gravity
main interest is in the differences in the GRG size distribution circuit offline and three with it running. As this reduction was
in the cyclone overflow between a no gravity recovery achieved with a very low gravity gold recovery of 0–5 per cent,
scenario and a small gravity circuit treating ten per cent of the it provides useful context for the case study. Subsequently it
mill recirculating load. The smaller gravity circuit size must is expected that this moderate level of gold accumulation,
be considered due to the observed tendency of the model to comparable to the reference project without a gravity circuit
over predict the gravity recovery that might be expected at a operating, is not likely to support significant gravity gold
more typical gravity circuit feed rate, such as 25–50 per cent recovery levels, due to the fine GRG size distribution.
of the recirculating load.
The population balance model predicted that without
an operating gravity circuit, 42 per cent of the gold in the
CONCLUSIONS
cyclone overflow would be gravity recoverable. The P80 size In this particular case study, the ultimate conclusion was to
of this GRG is estimated at 67 microns. By adding a small continue to operate the grinding circuit without including
gravity circuit, recovering as much as 20 per cent of the a gravity recovery stage, while monitoring the GRG size
gold by gravity, the GRG content in the cyclone overflow distribution in the cyclone overflow on an ongoing basis. This
would reduce to 24 per cent, at a GRG P80 of 54 microns. As is an important step to ensure that the absence of a gravity
demonstrated by Figure 5, the reduction in GRG content circuit does not impact the downstream recovery process.
occurs almost exclusively (96 per cent) below 75 microns. This analysis has demonstrated the use of several tools
As such the impact of the gravity circuit is estimated to and references to support an assessment of the practical
be limited to removing material that would be expected value of including a gravity circuit in a process flow sheet.
to leach in the CIL circuit in any event, based on standard The approach described in this paper is strongly grounded
assumptions of gold dissolution rates and a typical CIL in practical applications and designed to ensure that such
circuit residence time of 24 hours. As such the gravity circuit an assessment derives a realistic expectation of plant scale

Do we need a gravity circuit or not? A case study in applying best practice
Model Predicted Gravity Gold Recovery

100
90
80
Gold Recovery by Gravity (%)

70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
% Recirculating Load to Gravity Feed
Test 1 Test 2 Test 3
FIG 4 – Gravity recovery as a function of gravity circuit capacity.
45
40
35
30
Gold Content (%)
25
20
15
10
5
0
10 100 1000 10000
Particle Size (mm)
No Gravity Gravity
FIG 5 – Cyclone overflow gravity recoverable gold size distribution.
gravity recovery, and the subsequent impact on overall ACKNOWLEDGEMENTS

project economics. The key steps in this analysis include: The authors acknowledge the permission of Newmont Mining
•• performing three-stage GRG testing using the to publish this paper.
techniques described by Laplante, or close
approximations to those methods REFERENCES
•• studying in detail the particle size distribution of the Giblett, A, Bax, A, Wardell-Johnson, G and Staunton, W, 2013.
GRG and the distribution in the first stage of the test A review of best practice in gravity circuit design and
in particular operation, in Proceedings MetPlant Conference, pp 272–284 (The
•• avoiding generalisations based on total GRG content
without qualifying the influence of GRG size distribution Giblett, A, Hillier, D, Parker, K and Ramsell, V, 2012. The impact of
gravity concentration at Kalgoorlie Consolidated Gold Mines, in
•• using population balance modelling to predict the Proceedings 11th AusIMM Mill Operators’ Conference, pp 199–206
particle size distribution of the grinding circuit product (The Australasian Institute of Mining and Metallurgy: Melbourne).
with and without gravity Giblett, A and Napier-Munn, T, 2015. Measuring the influence
•• for operating plants, studying the gold accumulation of sample size on the precision and accuracy of gravity gold
estimation, in Proceedings MetPlant Conference (The Australasian
in the recirculating load and benchmark against other
facilities
Giblett, A L, 2011. Gravity gold concentration at Newmont Mining,
•• measuring the GRG distribution by size in the grinding World Gold Conference 2011, in Proceedings 50th Conference of
circuit product as an important point of reference Metallurgists, pp 217–228 (MET SOC: Montreal).
•• at all times understanding the nature and deportment of Laplante, A R, 2000. Testing requirements and insight for gravity
gold losses from the plant. gold circuit design, Randol Gold and Silver Forum, Vancouver.

Laplante, A R and Staunton, W P, 2005. Gravity recovery of gold – Minerals Engineering Online, 2004. Plant Operation News [online].
an overview of recent developments, in Proceedings International Knelson Concentrators and ConSep ACACIA Work in Harmony.
Symposium Treatment of Gold, Calgary. Available from: <http://www.min-eng.com/ops/oz/26.html>
[Accessed: 28 July 2017].
Laplante, A R, Woodcock, F and Huang, L, 2000. Laboratory
procedure to characterize gravity-recoverable gold, Society for Van den Steen, P F, 2000. Gravity gold recovery at President Steyn,
Mining, Metallurgy and Exploration Inc, Transactions, 308:53–59. Knelson Concentrators Africa, July 2000 [online]. Available
from: <http://www.flsmidth.com/~/media/PDF%20Files/
Laplante, A R, Woodcock, F and Noaparast, M, 1995. Predicting
FLSmidth%20Knelson/Papers/20000803GravityGoldRecovery
gravity separation gold recoveries, Minerals and Metallurgical
atPresidentSteynMine.ashx> [Accessed: 28 July 2017].
Processing, pp 74–79.

Contents
Solvent extraction of uranium – towards good

practice in design, operation and management
P Bartsch1 and S Hall2
ABSTRACT
Uranium solvent extraction (USX) has been applied commercially for recovery and
concentration for over 55 years. Uranium in acidic liquor, which is prepared following
ore leaching, solid/liquid separation and clarification, can be treated through a
sequence of solvent transfer operations; extraction-scrubbing-stripping, to obtain
purified liquor. USX has dominated the primary uranium industry as the preferred
technological route to converter grade yellowcake. The practices of design and
operation of USX facilities has found renewed interest as new mines are developed
following decades of industry dormancy. This article seeks to outline principles of
design and operation from the practitioner’s perspective. The discussion also reviews
historical developments of USX applications and highlights recent innovations. This
review is hoped to provide guidance for technical personnel who wish to learn more
about good practices that lead to reliable USX performance.
INTRODUCTION
This paper provides a brief summary of practical knowledge and lessons learnt in
the field of uranium recovery by solvent extraction (SX), from test work, through to
design and production. These suggestions, clarifications and observations are offered
on the basis of the authors’ experience which are hoped to encourage constructive
discussions in the hydrometallurgical community, and thereby improve solvent
extraction performance in current and future facilities.
FOUNDATION CHEMISTRY
Uranium solvent extraction (USX) involves exchange of dissolved metals, as complex
ionic components, from an aqueous phase into an organic phase or solvent. The
organo-metal complexes may be purified by scrubbing, before they are subsequently
stripped into a more pure, concentrated aqueous phase. The soluble, metal complex, or
ion-pair, is chemically loaded by an extractant, which is dissolved in an hydrocarbon
diluent (Bosman, 1980) that together form the solvent. An exchange of metal ion-pairs
will allow a charge balance to be maintained across the transfer interface (cations or
anions will be exchanged depending on the selected extractant). Interested readers
can learn more detailed chemistry of USX in useful texts (International Atomic Energy
Agency, 1993; Ritcey and Ashbrook, 1979; Merritt, 1971; Brown et al, 1958).
SOLVENT EXTRACTION FEED CONTROL

The most important feature for good SX practice is stable streamflow, particularly
pregnant leach solution (PLS), and hence steady process chemistry. Every effort of
designers, and the eternal vigilance of operators, must be aimed at consistent rates
and composition of feed streams, and thereby internal and recycle streams. Realistic
sizing of PLS storage, as well as sensible make-up capacity for reagent streams are
essential aspects for reliable performance of a SX facility. Operators must expect that
crud, which is a generic term for dispersed solid phases that accumulate in settlers
(Ritcey, 1980) will occur. Adequate systems and procedures must be installed and
maintained to collect and treat this nuisance material so that reliable flows and hence
performance can be maintained.
The relative performance of extraction, scrubbing and stripping can be characterised
1. MAusIMM(CP), Consulting Metallurgist,
by coefficients for uranyl complexes which will shift with the overall sulfate, SO4
Alchemides Pty Ltd, Mile End SA 5031. and Cl concentrations in the respective aqueous phase (Soldenhof and Davidson,
Email: pjbartsch@gmail.com 2005). Test work for process specification must be conducted with representative
2. MAusIMM, Director and Principal Consultant, solutions’ chemistry and temperatures, and transfer results interpreted for process
Goudie Hall Consulting Pty Ltd, Mylor SA criteria. Interpretation of kinetic and equilibrium effects will be important. For
5153. Email: goudiehall@bigpond.com example, SO4 levels in PLS may rise to over 100 g/L (Thiry and Roche, 2000),
249
while few uranium leaching operations have operated with extraction of uranium in amine systems. Modifiers in cationic
‘high’ Cl concentrations, eg above 5 g/L. The Honeymoon solvents include TBP and DBBP, which can have synergistic
Project (Dobrowolski, 1983) has operated in situ recovery in influence to improve uranium extraction (Blake et al, 1958).
a saline aquifer with about 8 g/L Cl in PLS, and used solvent Appearance of a third phase is more prevalent during high
extraction for uranium purification (La Brooy et al, 2009). extractant or uranium concentration due to the potential
PLS may contain significant nitrate which can depress formation of organo-metallic salt.
uranium transfer. Nitrosyl in sulfuric acid (NOHSO4 reported Management of solvent degradation can become an
as NOx) at levels of above 50 ppm may be present in delivered operational challenge in SX circuits with high nitrate or ferric
acid that is manufactured from smelter gases (Lyne et al, concentrations in PLS. Declining extraction performance,
2002). Hydrolysis to nitrates, together with contamination or extended phase break time are indicators that solvent
from residual explosives in mined ore means nitrates can properties require improvement. Where SX facilities have
accumulate in PLS depending on the closure of the process high organic entrainment in raffinate, or include solvent bleed
water balance (Munyungano, Feather and Virnig, 2008). streams that dispose of excess crud, the degradation products
For each impurity that is partially transferred through various may not reach nuisance levels. Control of PLS chemistry is
scrubbing, or stripping stages, selective precipitation, or re- a good operating strategy, while uranium leaching rates and
leach can be investigated as potential flow sheet additions to extent are managed with acid and oxidant dosing for pH
ensure the specifications of customers (ie uranium converters) and Eh control, to reach economic levels. Alternatively, PLS
are met for marketed products (Bellino and Mackenzie, treatments to lower its chemical potential are also available,
2010). In this context, detailed quantification of PLS chemical eg dosing with metallic iron or sparging with pure SO2 gas.
speciation is interesting for process optimisation, but largely Bisulfate loading at high PLS acid level can suppress
superfluous compared to the leach chemistry in respect uranium extraction when leaching solution contains above
to selecting an economic or preferred flow sheet, unless 10 g/L free acid in SX feed. Controlled acid in PLS will
precipitate specifications are threatened. offer benefits of economic leaching and SX performance
(George and Ross, 1967). Lower acid levels for leaching are
PROCESS STAGES – EXTRACTION, SCRUBBING AND STRIPPING recommended, consistent with reaching uranium dissolution
and mass transfer targets, as well as balancing silica
Solvent extraction dissolution and accounting for downstream PLS and raffinate
or tailings treatment costs.
Cationic or neutral extractants, ie long-chain alcohol, phosphoric
derivatives, were first proposed for uranium recovery from Given the rapid kinetics of sulfate, bisulfate and uranyl
PLS (Grinstead et al, 1958) but were superseded by amines due sulfate exchange on solvent in the extraction stage, the
to greater selectivity. These reagents are no longer used alone relative loadings of U will quickly reach equilibrium. While
for uranium recovery from PLS, but are popular for yellowcake the PLS has pH between 1.5–2.5 the stage-wise extraction
purification at converter operations. Organic phosphate equilibrium is usually available within 1–2 minutes of
derivatives are commonly used for purification of uranium mixer retention, depending on mixer box design and mass
concentrates or oxides following digestion with nitric acid transfer gap (Faure and Tunley, 1971). Stage design with
where the purpose is the removal of anion impurity fractions, twin or triple sequenced mixer boxes may be cost-effective
for example silicates, sulfate and zirconia. to obtain efficient mass transfer and adequate residence time
Mixed solvent, ie cationic and anionic extractants applied distribution with low temperature PLS.
in the same transfer stage, was first described in reports from Loading and distribution of uranium on amine extractant
Colorado (Rosenbaum, Borrowman and Clemner, 1958). will generally improve with increasing acidity of PLS.
Operations at the Durango facility with mixed solvent were Adjustment of PLS with dilute sulfuric acid to lower pH
described. Application of mixed solvent, together with a to improve extraction performance has been mentioned at
neutral modifier (tri-butyl phosphate (TBP) or di-butyl-butyl Energy Resources Australia’s (ERA) Ranger operation (Nice
phosphonate (DBBP)) were investigated at Honeymoon and Banaczkowski, 2000). The solvent performance can also
for treating PLS that contains elevated Cl in process waters shift with PLS sulfate and chloride concentration. Each system
(Phillips, 2001). This project included the first application of must determine the competition of uranium and other anions
solvent extraction combined with in situ leaching for recovery in highly saline PLS with operating solvent, and allow for cost
of uranium (LaBrooy et al, 2009). of reagent inputs.
Anionic extractants such as proprietary tertiary amines
(ie organic, nitrogen compounds with three alkyl chains in Solvent scrubbing
liquid form) have become favoured for uranium recovery Solvent scrubbing can be applied advantageously between
from acidic PLS because base metal impurities are efficiently extraction and stripping to displace anionic impurity from
rejected into raffinate. Silica and halides (F, Cl, Br) can form amine based solvents, or metal cations from phosphate based
simple anions, and some transition metals form soluble solvents. Water scrubbing can wash out entrainment of PLS
complexes, eg V or Mo as oxy-anions, Co as cyanide, Bi as oxy- from the advancing solvent.
chloride, or Fe in the presence of chloride, are also extracted The number of scrub stages and make-up of scrub liquor will
by amines (Bender et al, 2010). depend on the impurity type and relative loading on solvent
Certain additives tend to inhibit third phase formation that compared to uranium. Dilute sulfuric acid, low chloride water,
can arise as semi-miscible fluids of extractant-metal salts that or strip liquor bleed, ie ammonium sulfate solution, can be
separate and accumulate at the phase interface within settlers used in two or three sequential scrub stages (Hardy, 1978).
(Kertes, 1966). Solvents that contain amines often utilise Generally, spent scrub liquor will be recycled to the PLS
modifiers such as long-chain alcohols, iso-decanol or tri-decanol, pond, or directly into the extraction feed stream, to permit
at about half the extractant concentration (per cent v/v). recovery of the contained uranium. Process stability and
Dosing the solvent with mixed polarity modifiers can entrainment are also important criteria for the choice of scrub
influence the uranium extraction coefficient, hence lowering stages and stream recycle destinations.

Solvent extraction of uranium – towards good practice in design, operation and management
Solvent stripping Stable PLS tenor is an important ally for operators to control
impurity transfer. If the loaded solvent carries high impurity
Strong ammonium or acid sulfate solutions, ca >150–300 g/L
loads then scrub liquor concentrations can be adjusted to suit
SO4 can be used for uranium stripping from tertiary amine
anion transfer rates. Progressive process start-up, shutdown
extractants (Ritcey and Ashbrook, 1979). In this process
and changes to phase ratios of each stage can be managed
step the sulfate concentration is the dominant chemical that
carefully to avoid unplanned crud displacement, and transient
displaces the uranyl sulfate into the aqueous phase, while
impurity transfer though to final product.
ammonia gas or solution is applied to regulate pH. Stripping
with strong sulfate solution is preferred above dosing with For high uranium concentration in PLS, the SX plant could be
ammonium hydroxide (aqua-ammonia) due to more stable run with alternative series-parallel extraction configuration,
operation. Sulfate strip chemistry can avoid localised, high while the raffinate is recycled to ore leach. This design option
pH that can cause premature precipitation of yellowcake in will be determined by mine plans but is unlikely to become
settlers and hence subsequent crud generation. favoured for new plants that have low ore grade and do not
recycle process waters or discharge raffinate to tails.
Stripping kinetics can be accelerated at marginally higher
pH or increased solution temperature. Stage-wise control
of pH rise across the stripping must be graduated. Tightly SOLVENT EXTRACTION EXCHANGE EQUIPMENT
regulated stripping chemistry may be countered by pH Process engineers will apply results from relevant extraction,
regulation difficulties, or the presence of excessive crud. scrubbing and stripping investigations to prepare ‘isotherms’,
Stripping with strong ammonium sulfate solution is preferred which are derived from shake-out test data, or pilot plant test
if the yellowcake precipitation process uses ammonia gas to work, and obtained with varying ratios of PLS and solvent.
make ammonium di-uranate as a marketable product. Residual The relevant test techniques and interpretation procedures are
ammonium in tailings liquor, which arises from the stripping generally known in technical literature (Gupta and Singh, 2003).
process, cannot be released off the mine site. Hence ammonia McCabe-Thiele diagrams are constructs of uranium
recycle via lime boiling, or production of ammonium sulfate distribution or concentration profiles across various ratios of
from evaporation, will be a necessary flow sheet inclusion. equilibrated aqueous and organic phases, using results from
laboratory or pilot tests run at predetermined conditions and
Strong acid stripping can be attractive for flow sheets where
constant temperature, ie graphical isotherms (International
acid, and caustic or magnesia for neutralisation, are low cost
Atomic Energy Agency, 1990). Such interpretations are used
and high quality. Acid stripping can avoid the need for a
to generate process criteria such as flow ratios of solvent to
solvent regeneration stage and potentially simplify the overall
aqueous. Design criterion can then be specified and applied
SX circuit design. Carbonate stripping will be preferred when
to equipment sizing and selection, including columns for
uranyl peroxide is the desired form of yellowcake or sulfate
extraction (Bartsch and Lawson, 2003).
stripping is undesirable for environmental reasons (Edwards,
1992). The maximum uranyl concentration in strip liquor will Initial SX investigations for prefeasibility and consequent
be determined by the extractant concentration, and impurity preliminary designs may be done with synthetic solutions.
loading, to ensure product specification or avoid third phase Experimental programs based on actual leach solutions can
formation in the strip stage. be used for confirmation of process criteria during the various
study stages, and hence detailed plant design. Fresh solvents
can be used for preliminary shake-out tests, but aged or used
PROCESS STABILITY solvent is recommended when process designs are used for
SX facilities will meet the design criteria when the feed flow final feasibility studies. The choice to run continuous mini-
and concentrations, particularly the uranium tenor, have low rig or pilot solvent extraction investigation will depend on
variability. Changes to PLS flow control set point once per the degree of innovation of process chemistry, flow sheet,
day are recommended, while efforts to minimise swings in equipment or unusual impurity occurrence. Test work must
uranium and impurity tenors, and hence mass rates, will be be conducted by experienced technicians and guided by
beneficial to overall production. Otherwise the SX operator practitioners who can translate the isotherms into design
will need to vary the PLS flow proportionally to the uranium criterion and equipment sizing.
tenor to maintain productivity, ie constant ‘metal’ feed rate. Two major types of equipment are used for extraction of
PLS, scrub or strip aqueous phase flow variations will uranium from PLS, mixer settlers and agitated columns.
induce the respective settler interface level to rise or fall Materials of construction for hydrometallurgical plants and
progressively through the sequence of settlers. A stepped equipment include lined or stainless steel or lined concrete
flow increase can displace crud from the settler interface to (Robinson, Sandoval and Cook, 2003). Fibre reinforced
the next mix box where more crud will be generated due to plastic bodies and linings have become popular for small
the turbulent intensity at the impeller tip. The transmission mixer settlers, and most columns for primary uranium
of excess crud in sequential mix boxes can promote crud extraction applications.
generation through the entire system, ie ‘crud begets crud’. Columns evolved from designs in the nuclear industry for
Large storage volume can subdue fluctuations of PLS uranium purification of numerous metals including uranium (Desson
tenor. An added benefit of thoughtful design will be suspended et al, 1983). Column size and capacity have increased in recent
solid removal, which can be available from a generously sized installations, for example Olympic Dam. Columns are suited
PLS pond. Minimisation of PLS storage volume is poor design to multiple stages of extraction in a single unit consisting of
practice that is often driven by misguided efforts of project active section with top and bottom decanters, and throughput
managers to cut facility capital costs prior to start-up, without is limited by hold-up of the dispersed phase. Columns can
accounting for overall life cycle costs of the asset. Zealous cost provide effective performance when mass transfer of uranium
cutting can also compromise equipment selections for cleaning is relatively rapid; for example, in extraction stages.
PLS such as filters or clarifiers. Dynamic process simulation Multiple columns, which operate in parallel to accommodate
can be helpful to optimised sizing of process storages for larger flows, can be installed as determined by total flux;
hydrometallurgical facilities (Smith, 2005). that is, total SX feed plus advance organic flow rates. Columns

may also be advantageous in comparison to mixer settlers due Direct dosing of ‘neat’ acid into solvent streams must be
to lower, overall solvent inventory, or when advance phase avoided due to localised over-heating and oxidation.
ratios are widely different (Movsowitz et al, 2000). Solvent scrubbing and stripping of uranium also tends
Mixer settlers were also originally developed in the to be relatively rapid, but retention may be dictated by
nuclear industry (Royston and Burwell, 1972), while design dispersion and stability of reagent dosing for pH control.
improvements and scale-up was driven by copper extraction. Equilibrium transfer of impurities during stripping can
These units have become popular for all transfer stages during require 1–5 minutes retention per stage depending on solution
the late twentieth century (Glasser et al, 1976). chemistry temperature and mass transfer gap.
For extraction mixers the turn-down ratio of organic recycle Regeneration of solvent is recommended if degradation
over a wide range is useful for maintaining organic phase products or interfering long-chain organic anions (for
continuity. Likelihood of phase inversion arises if the mixer O:A example, carboxylates) accumulate in the running solvent.
(volumetric ratio organic solvent over aqueous liquor) occurs Such contaminants can be scrubbed from solvent in a separate
below 0.9:1.0 for extended durations. The authors’ experience stage with highly alkaline liquor, ca >pH 10 (MacDonald
suggests that organic continuity in extraction mixers is essential
et al, 1978).
for management of emulsion break time and maintaining
lower solvent losses by entrainment, ie avoidance of phase The effectiveness of regeneration can vary with the recipe
inversion. The effect of phase ratios and continuity on transfer of the wash liquor, which will depend on contaminant type
of anions can be optimised with trials during commissioning and load, by systematic experimentation. Strength of the
and operation, and will depend on mixer retention time and regeneration liquor and campaign intensity will depend on
potential for unwanted transfer by entrainment. extent of solvent deterioration as measured by break time.
Overall process operation, and interface control in The Olympic Dam facility applied a three component mix
columns, can be easier than mixer settlers, particularly with (Hall and Reed, 1996). But operators will necessarily develop
respect to flow regulation and crud removal. Metal loading or procure their own recipe.
characteristics of columns in terms of approach to equilibrium A routine, laboratory, maximum-loading test can assess the
are usually rapid with changes to PLS. Operators’ control of capability limit of the working solvent. Such tests will trend
uranium loading in columns, or control of O:A hold-up, have solvent performance over extended time frames. Tracking
little influence on uranium transfer kinetics or equilibrium if phase break results from testing the plant solvent are also
the columns designs and scale-up are reasonable. useful to determine when regeneration is required.
Extraction columns may run at O:A phase ratio at up to 1:10,
while maintaining organic continuity of the hold-up emulsion, PHASE DISENGAGEMENT
depending on PLS conditions, particularly suspended solids
The specific settling rates for uranium solvent extraction plant
and temperature. Operators’ immediate response to changes
design are determined by design SX feed and solvent flow
of PLS flow or tenor in columns will be start-up or shutdown
rates, with an allowance for crud interference. Experimental
of individual units.
tests on pure fluids tend to overestimate the rate of phase
Columns can be idled while full to permit rapid re-start break. Useful methods to assess phase disengagement in
and avoid excess tank inventory of solvent. Sufficient solvent operating plants includes regular, standardised break tests,
storage is needed outside the columns to allow individual
eg separation of emulsion in a 1 Litre cylinder, on loaded and
column drainage of fluids for maintenance.
stripped solvent during each shift. A simple procedure can be
Columns for extraction will generally use more power than established for use by plant personnel.
mixer settler configurations for the same duty if agitation is
Specific rates for conventional settlers are generally set at
driven by pulse air, although direct comparison will depend
3–5 m3/m2/h depending on extractant concentration, ionic
on location and design criteria for the proposed operation
strength, solvent temperature and expectation of crud make
(Taylor, 2007). The installed power for mixer impellers is
more efficient than supply of pulsing air from blowers, but the and accumulation (Cheng, Bujalski and Schwarz, 2004). Recent
number of mixers is often more than the number of blowers. settler designs are proposed to enhance specific flow rates up
Mechanical or hydraulic agitation designs for columns are to 5–15 m3/m2/h, but potential for increased entrainment
not commercially available for PLS applications, but smaller loses are not quantified. The flow pattern of emulsion from
installations for yellowcake purification and nuclear fuel the mixer overflow to the settler will influence separation and
reprocessing are available. solvent circulation. Installation of ‘picket’ fences within the
settler box, located in proximity to the incoming flow, and
Columns are designed as closed vessels (except for pulse
spreading of organic-continuous emulsion from the mixer
air vents), with low roof space. Given good design, columns
can have greater fire suppression attributes, particularly if overflow into the solvent layer is recommended.
solvent purge or dump is arranged with due consideration Specific settling rates for pulse columns, when treating
of piping corridors. Columns will require more attention to ‘clean’ PLS can be between 20–50 m3/m2/h for the active,
piping detail during design to ensure adequate support and pulsing section. The total flux will depend on extractant
maintenance access are available. concentration, solvent condition and system temperature.
Without diligent or automated crud removal a conservative
SOLVENT CONDITION settling rate is recommended. Recent design for scrub and
A separate protonation or acidification stage is redundant strip stages have applied comparatively low settler rates due
while SX feed contains about 5–10 g/L acid. The recycle of to the prevailing high crud content. Scrub and strip settlers do
spent scrub liquor to join PLS will lower its pH, which will not generally receive suspended ore particles and silica solids,
assist process control if the upstream leach conditions are so crud management, and hence design settling rates will
variable. In-line dosing of recycle or acid streams are a design depend on precipitation of silica, zirconia or other impurities
opportunity if intermediate, sectional storages are available. as observed during pilot trials.

CRUD FORMATION AND HANDLING Careful observation and impurity tracking during pilot
testing can help designers and operators control anion transfer
Suspended solids in PLS will enter USX following poor
and set stage pH targets. Preventative measures to avoid the
clarification of PLS and are a major source of crud. Dissolved
deleterious effects of silica and crud include:
species, such as silica and zirconia, also have the potential to
generate crud during solvent extraction processes. •• Control leach temperature, retention and acidity:
Soluble silica in PLS becomes a problem when it polymerises •• extended retention time at lower temperature can
and converts to submicron colloidal, or larger particulate ensure effective uranium dissolution
form (Moyer and McDowell, 1981) which can carry into •• higher acid levels will tend to dissolve more silicate
the extraction, scrubbing and stripping sections. The silica minerals and consume more acid
colloids change in size as the silica polymerises and different •• two-stage leaching can keep final leach temperature
size colloids can cause crud generation problems in different down and conserve acid but at higher capital cost
parts of the SX circuit (Ritcey, 1980). for installation of a second leach thickener or cyclone
Slimed ore particles and floating floccules from clarification bank.
can contribute to PLS solids loading and hence crud make •• Minimise slime formation during grinding particularly
in extraction settlers. These colloids and particles will tend of ores with clays or foliated silicates.
to build-up at the settler fluid interface, which causes stable •• Operate effective PLS clarification; note that one stage
emulsion build-up, and can restrict hydraulic throughput of sedimentation will have difficulty reaching below
or lead to increased solvent losses (Saruchera, Jarvi and 50–100 ppm suspended solids in SX feed when process
Moldvan, 2010). upsets occur. A second clarifier with steady flow and
Silica content of PLS, as a quantitative forward measure deep bed operation can polish PLS to about 20–50 ppm.
of potential crud make, can be determined quantitatively by •• Provide a PLS pond with large retention volume;
laboratory filtering at defined filter pore sizes. Visual field eg one day minimum (not recommended), three days
observation of PLS solids, eg sighting in a clarity wedge, can is good practice, while one week will be preferred by
be useful for rapid operator intervention. Field operators can operators. Capability to remove settled or flocculated
collect ‘dip’ samples from each settler with a clear plastic particulates from floor of pond will provide benefits by
tube, and thereby track crud build-up. lower solvent losses.
Maintenance of organic phase continuity in all SX mixers •• Heat the clarified PLS before SX feed to remove super-
for uranium will limit crud formations and so limit crud saturation of silica and gypsum. Direct steam injection
build-up in settlers. In extreme circumstances bulk crud of PLS, or allow heating of the recycled, spent scrub
flows between settler banks can occur, otherwise known as liquor will tend to re-dissolve colloidal silica.
‘crud-runs’, which leads to higher solvent losses and poor Silica polymerisation and aggregation can be modified by
metallurgical control. anion or cation concentrations in PLS (Terry, 1983). Many
Crud handling and treatment is important to recover solvent solutes will form ion-pairs, for example Al and F, which can
and lower metal losses (Hartmann, 2010). All uranium SX influence the relative effects of dissolved species including
plants will experience crud formation and good plant designs silica. Variable solute concentrations in PLS can complicate
will accommodate its removal and treatment. The design diagnosis of silica deportment and subsequent clarification
and crud control measures.
philosophy for PLS preparation as SX feed can best focus on
prevention as compared to cure or remedial action. While less
silica is dissolved during lower intensity leaching, and if more TOWARDS THE FUTURE OF uranium SOLVENT EXTRACTION
silica is precipitated and flocculated during PLS clarification, Solvent extraction has dominated the uranium production
then less crud can be expected to accumulate in extraction industry for over 50 years, but during the hiatus in new
settlers. Crud in downstream scrub or strip settler is often uranium projects, the technology did not advance (Ritcey,
related to impurity transfer during unstable pH control. 2003). USX is currently utilised in Australia’s two largest
TABLE 1
Australian uranium extraction plants at October 2016.
Operation and location Ranger, Northern Territory Olympic Dam, Beverley, South Australia Honeymoon, South Australia Four Mile, South Australia
South Australia
Ore category Unconformity Breccia Sandstone Sandstone Sandstone
Leach chemistry Acid sulfate Acid sulfate, 3 g/L chloride Acid sulfate, 4 g/L chloride Acid sulfate, 5–8 g/L Cl Acid sulfate
Recovery technique Amine SX Amine SX Strong base IX Mixed SX Strong base IX
Nominal capacity, tU 4660 3820 850 340 1150
Operation and location Kintyre, Western Australia Wiluna, Western Australia Yeelirrie, West Australia Lake Maitland, West Australia Ranger Expansion
Ore category Unconformity Surficial/calcrete Surficial/calcrete Surficial/calcrete Stockpiles
Leach chemistry Acid sulfate Alkali carbonate Alkali carbonate Alkali carbonate Acid sulfate
Recovery technique DP DP DP DP Amine SX
Nominal capacity, tU 2000 680 3000 850 TBC
Projects with feasibility study ongoing or complete are shown in italics (after IAEA/OECD, 2010).

uranium mines, as shown in Table 1. The future of USX will Gupta, C K and Singh, H, 2003. Uranium Resource Processing,
reflect the expected trends of new projects that include lower pp 172– 175 (Springer Verlag).
grade ores, tighter controls on water supply or discharge. The Hall, S and Reed, M, 1996. Cross contamination of ODC solvent
trend of higher proportion of production via in situ recovery extraction circuits, paper presented to SME (Society for Mining,
will increase the application of ion exchange technology. To Metallurgy & Exploration) Annual Meeting, Phoenix, 11–14 March.
remain competitive USX will need to achieve greater stage Hardy, H J, 1978. The chemistry of uranium milling, Radiochemica
recovery, lower solvent losses and higher tolerance of salinity. Acta, 25:121–134.
A solvent extraction system for uranium recovery from leach
Hartmann, T, 2010. Crud treatment with 3 phase centrifuge in heap
solutions based on sea water has been investigated by CSIRO leach uranium process, in Proceedings Third International
(Zhu and Cheng, 2016), however this process remains in the Conference on Uranium, Saskatoon, pp 741–749 (Canadian
early stages of research. Institute of Mining Metallurgy and Petroleum: Quebec).
International Atomic Energy Agency, 1990. Technical report #313:
REFERENCES Manual on laboratory testing for uranium ore processing,
Bartsch, P J and Lawson, B, 2003. Confirmation of process criteria for figure 27 (IAEA: Vienna).
pulse column plant expansion, in Proceedings ALTA Conference, International Atomic Energy Agency, 1993. Uranium Extraction
Perth (ALTA Metallurgical Services: Melbourne). Technology, technical report no 359 (IAEA: Vienna).
Bellino, M and Mackenzie, J M W, 2010. Solvent extraction of IAEA/OECD, 2014. Uranium 2013 – resources, production and demand,
uranium-impurity transfer and phase separation, in Proceedings technical report, 25th edition (International Atomic Energy Agency
Third International Conference on Uranium, Saskatoon, pp 685– / Organisation for Economic Co-operation and Development).
697 (Canadian Institute of Mining Metallurgy and Petroleum:
Quebec). Kertes, A S, 1966. The chemistry of the formation and elimination of
a third phase in extraction systems, in Solvent Extraction Chemistry
Bender, J, Virnig, M, Nisbett, A, Crane, P, Mackenzie, M and
of Metals, pp 377–400 (MacMillan: London).
Dudley, K, 2010. Uranium solvent extraction circuits; operational
challenges and adjusting to unique process conditions, in LaBrooy, S, Spratford, D, Middlin, B and Otto, K, 2009. Differentiating
Proceedings Third International Conference on Uranium, the honeymoon uranium process flow sheet, in Proceedings ALTA
Saskatoon, pp 675–684 (Canadian Institute of Mining Metallurgy Conference, Perth (ALTA Metallurgical Services: Melbourne).
and Petroleum: Quebec).
Lyne, E G, Berryman, A B, Evans, C M, Sampat, S and Jensen-
Blake, C A, Baes, K B, Coleman, C F and White, J C, 1958. Solvent Holm, H, 2002. Advances in NOx removal in smelter acid plants
extraction of uranium and other metals by acidic and neutral [online], in Proceedings International Seminar New Initiatives
organic phosphorus compounds [online], in Proceedings Second Mining Sector, Chilean Copper Commission, Santiago. Available
UN International Conference on the Peaceful Uses of Atomic from: <http://www.topsoe.com/> [Accessed: 22 July 2017].
Energy, Geneva. Available from: <http://www.osti.gov/
MacDonald, J B, Mattison, P L and Mackenzie, J M W, 1978. Technical
energycitations/product.biblio.jsp?osti_id=4309218> [Accessed:
20 July 2017]. service in uranium solvent extraction, practical problem solving in
South Africa and the United States, Henkel Corporation Guidebook.
Bosman, R J, 1980. Practical considerations in diluent selection, in
Proceedings Symposium for Ion Exchange and Solvent Extraction (The Merritt, R C, 1971. The Extractive Metallurgy of Uranium, 576 p
Southern African Institute of Mining and Metallurgy: Randberg). (Colorado School of Mines Research Institute: Golden).
Brown, K B, Coleman, C F, Crouse, D J, Blake, C A and Ryon, A D, Movsowitz, R L, Kleinberger, R, Buchalter, E M and Grinbaum,
1958. Solvent extraction processing of uranium and thorium ores, B, 2000. Comparison of the performance of full scale pulsed
in Proceedings Second UN International Conference Peaceful Uses of columns vs mixer-settlers for uranium solvent extraction, in
Atomic Energy, Geneva, 3:472–487. Proceedings Uranium 2000 – International Symposium on the
Process Metallurgy of Uranium, pp 181–192 (Canadian Institute
Cheng, C Y, Bujalski, J M and Schwarz, M P, 2004. Mixer-settlers; of Mining Metallurgy and Petroleum: Quebec).
theory, design, modelling and application, report for AMIRA
Project, P706. Moyer, B A and McDowell, W J, 1981. Factors influencing phase-
disengagement rates in solvent-extraction systems employing
Desson, M, Duchamp, C, Henry, E, Jouin, J P and Michel, M, 1983.
tertiary amine extractants, Symposium on Separation Science and
Industrial experience with pulse columns in the processing of
Technology for Energy Applications, Gatlinburg.
uranium ores, AICHE Symposium Series, ISEC, 80(238):162–169.
Munyungano, B, Feather, A and Virnig, M, 2008. Degradation
Dobrowolski, H J, 1983. The Honeymoon Project – an alternative
problems with the solvent extraction organic at Rossing uranium,
method of uranium mining, in Proceedings of a symposium;
in Proceedings ISEC 2008, Tucson, pp 296–274 (International
Uranium Technology in South Australia (Australian Mineral
Solvent Extraction Conference).
Foundation: Adelaide).
Edwards, C R, 1992. Uranium extraction process alternatives, CIM Nice, R W and Banaczkowski, M, 2000. Expansion of the ore
Bulletin, 85(958):112–136. treatment plant at Ranger uranium mines, in Proceedings
Uranium 2000 – International Symposium Process Metallurgy of
Faure, A and Tunley, T H, 1971. Uranium recovery by liquid-liquid Uranium, pp 163–180 (Canadian Institute of Mining Metallurgy
extraction in South Africa, IAEA-SM-135/30 in Proceedings and Petroleum: Quebec).
Symposium on the Recovery of Uranium, Sao Paulo, Brazil, pp 241–
251. Phillips, R, 2001. Honeymoon uranium project, metallurgical testing
– laboratory and plant trials, 1998 to 2000, Southern Cross
George, D R and Ross, J R, 1967. Recovery of uranium from mine Resources Pty Ltd, May.
waters and copper ore leaching solutions, in Proceedings On Low
Grade Uranium Ore, STI/PUB/146 (International Atomic Energy Ritcey, G M, 1980. Crud in solvent extraction processing – a review of
Agency: Vienna). causes and treatment, Hydrometallurgy, 5:97–107.
Glasser, D, Arnold, D R, Bryson, A W and Vieller, A M S, 1976. Ritcey, G M, 2003. State of the art and future directions in solvent
Aspects of mixer settler design, Minerals Science and Engineering, extraction [online], in Proceedings Third International Solvent
8(1):23–45. Extraction Workshop, Digby, Nova Scotia. Available from:
<http://www.solvent-extraction.com/Proceedings/2003Proc.
Grinstead, R R, Shaw, K G and Long, R S, 1958. Solvent extraction of
pdf> [Accessed: 20 July 2017].
uranium from acid leach slurry and solution, in Proceedings UN
International Conference Peaceful Uses of Atomic Energy, Geneva, Ritcey, G M and Ashbrook, R A, 1979. Solvent Extraction: Principles
8:71–76. and Applications to Process Metallurgy (Elsevier).

Robinson, T, Sandoval, S and Cook, P, 2003. World copper solvent Soldenhoff, K and Davidson, J, 2005. Uranium recovery from highly
extraction plants: practices and design, Journal of the Minerals, saline in situ leach solutions by ion exchange, in Proceedings
Metals and Materials Society, 55(7):24–46. First Extractive Metallurgy Operators Conference, Brisbane (The
Rosenbaum, J B, Borrowman, S R and Clemmer, J B, 1958. Solvent
extraction for recovering uranium and vanadium from salt- Taylor, A, 2007. Review of mixer-settler types and other possible
roast-process solutions, in Proceedings Second UN International contactors for copper SX, in Proceedings ALTA Conference, Perth
Conference Peaceful Uses of Atomic Energy, Geneva, 3:505–509. (ALTA Metallurgical Services: Melbourne).
Royston, D and Burwell, A, 1972. The design and performance of Terry, B, 1983. The acid decomposition of silicate minerals part II,
pump-mix and gravity flow mixer settlers [online]. Available from Hydrometallurgical applications, Hydrometallurgy, 10(2):151–171.
<https://www.iaea.org/inis/collection/NCLCollectionStore/_
Thiry, J and Roche, M, 2000. Recent process developments at the
Public/05/110/5110821.pdf> (Australian Atomic Energy
Cominak uranium mill, in Proceedings Uranium 2000 – International
Commission: Lucas Heights) [Accessed: 20 July 2017].
Symposium Process Metallurgy of Uranium (Canadian Institute of
Saruchera, T, Jarvi, J and Moldovan, B, 2010. Crud separation and Mining Metallurgy and Petroleum: Quebec).
SX Optimization at Key Lake, in Proceedings Third International
Zhu, Z and Cheng, C, 2016. CSIRO – New solvent extraction system
Conference on Uranium, Saskatoon, pp 731–740 (Canadian
for uranium recovery extraction from leach solutions based on
Institute of Mining Metallurgy and Petroleum: Quebec).
seawater usage, in Proceedings ALTA Uranium and REE Conference,
Smith, H, 2005. Process simulation and modelling, in Developments in Perth (ALTA Metallurgical Services: Melbourne).
Mineral Processing (ed: M Adams), chapter 5, volume 15 (Elsevier).

Contents
Design of copper-cobalt hydrometallurgical circuits

G Miller1
ABSTRACT
Copper-Cobalt hydrometallurgy has seen a resurgence in recent years with the
development of projects in Zambia and the Democratic Republic of Congo (DRC).
Many of the flow sheets are unique in their use of hydrometallurgical techniques
to improve performance compared to the older types in current use. The emphasis
has moved from direct copper electrowinning/stripping with multiple stages of
precipitation for removal of impurities and concentration of cobalt, to innovative use
of solvent extraction (SX) and ion exchange (IX) with higher cobalt recoveries to final
product.
The poor quality direct electrowon copper has been supplanted with solvent
extraction–electrowinning, with consequent production of London Metal Exchange
(LME) ‘A’ quality metal. Many of the precipitation steps in cobalt recovery and
concentration have been replaced with solvent extraction; while the remaining
precipitation stages have been enhanced with use of oxidants to improve impurity
removal. Operating experience from two of these circuits has shown the benefits of
the new technologies and the improved overall recovery of both copper and cobalt
into higher quality products.
INTRODUCTION
Cobalt is recovered as a by/co-product of copper production and is generally treated
as a commercial bonus in the metallurgical copper circuits. As a result, it has not been
until new projects have come into the development program, that a significant focus
on cobalt hydrometallurgy has become more important. For some projects in the DRC
the potential cobalt income is now of the same order of magnitude as the copper
income (Anon, 2008). Maximising the recovery and value addition to the product is a
significant driver for hydromet process development.
It is impossible to divorce the selection of the cobalt recovery flow sheet from
the specific markets to be targeted. However, there are a number of common
unit operations that are used to remove specific groups of ions, that have all been
advanced in the most recent round of process designs. Many projects have taken
increased development times to pilot the cobalt recovery part of the process plant.
The commonly used unit operations and their relative place in the process chain are
discussed in the body of this paper.
COBALT PRODUCT SELECTION

The selection of the cobalt product to be produced on-site is vitally important in
the development of the required flow sheet. This motherhood statement is often
overlooked by the project owners who assume that the cobalt market is similar to
other metal commodities. It is, however, an extremely fragmented market, with
many different product types, purity requirements and market demands. One of the
more important drivers for recent projects of modest size has been the availability
of excess cobalt metal refining capacity in Europe and China. Both markets will take
an impure cobalt intermediate salt and refine to final metal. The available capacity
is ultimately limited, and new large projects will need to address the market directly
with finished (or semi-finished) products for direct input to customer processes. The
diversification of the BHP Yabulu refinery into cobalt chemicals is a good example
of this change in product emphasis.
The potential hydrometallurgical cobalt products are extremely wide ranging.
However, it is the major current markets that most greenfields projects are targeting.
The more significant cobalt products include:
•• cobalt salts:
•• hydroxide from either lime, magnesia or NaOH precipitation
1. FAusIMM(CP), Director, Miller Metallurgical
Services Pty Ltd, Corinda Qld 4075.
•• carbonate usually using soda ash for precipitation.
Email: gmiller@millermet.com •• value-added products:
257
G Miller
•• CoO from calcination of either hydroxide or carbonate state. SMBS is costly and is only partially utilised with side
•• sulfate from purified leach liquors and evaporative reactions producing sulfuric acid – particularly in the presence
crystallisation. of manganese ions in solution.
•• cobalt metal has many grades of product depending on A more recent development has been the use of gas from
chemical purity and physical condition. Some potential a sulfur burning acid plant (SO2 and N2) as a direct Eh
metal producers are considering the full range of post- management tool. The gas is injected into the later parts of the
production metal enhancement with: leach train to enhance the cobalt leaching. The SO2 is readily
•• hydrogen degassing furnaces soluble and reduces the cobalt while undergoing conversion
to sulfate with acid production. The large volumes of nitrogen
•• size control
(about 88 per cent v/v) can be an issue with scrubbing of the
•• surface burnishing SO2 from the solution if the gas is not dispersed properly. One
•• specific packaging for market. operation in Zambia has used a Pressure Gas Disperser to create
It is evident that there is no single target product similar ultra-fine gas dispersion using a high velocity pressurised
to LME A-grade copper. As a result, there are many possible jet of solution to entrain the gas (John, 2006). Other newer
cobalt flow sheets and processes to generate a marketable projects are considering the use of liquefied SO2 to remove
product, which is optimum for the specific operation. There the voluminous nitrogen (Grosse, September 2007, personal
are, however, a number of elements that are specifically communication). This eliminates the nitrogen scrubbing and
targeted to improve the realised value of the cobalt. The makes the gas dispersion much more controlled.
major ones include: Two projects in Zambia, at Chambishi and Nkana, calcine
•• copper a copper-cobalt-pyrite concentrate in a sulfation roast, using
•• iron fluid bed contactors (Sole, Feather and Cole, 2005). The copper
•• calcium and cobalt are in the oxide-sulfate form and are leached (along
with a lot of the iron) into solution with dilute sulfuric acid.
•• zinc Since the cobalt has been reduced in the sulfation roast to the
•• magnesium Co2+ form it is readily soluble in the acid solution. Both projects
•• manganese use recycled solution ‘barren’ in copper. This ‘preg builds’ the
•• aluminium cobalt solution levels to +10 g/L for subsequent recovery.
•• silica
•• nickel. COPPER RECOVERY AND REMOVAL
Specific products require the removal of these ions to a Primary copper removal
greater or lesser extent. Particularly with intermediate salts
The primary metal production from most projects is copper.
(for further refining) the level of penalty metals can be in the
order of 2 per cent to 4 per cent without too high a reduction The existing Zambian and DRC operations used direct
in income. On the other hand, production of super high quality electrowinning and electro-stripping to produce impure
electrowon metal needs the preparation of high purity solutions metal that generally had to be re-refined to produce a saleable
with controlled maximum amounts of all the ions mentioned. product. The electrowin and electro-strip process is also
inefficient with current efficiencies as low as 65 per cent.
Both Zambian cobalt projects have committed to changing
COBALT LEACHING the copper recovery to solvent extraction and electrowinning
Particularly in the DRC and to some extent in Zambia, (Miller and Nisbett, 2005; Kordosky, 2008; Sole, Feather and
the copper oxide ores with copper and cobalt are leached Cole, 2005). The Nkana project has already been in operation
in atmospheric systems. The dominant mineralisation is for some time with excellent results (Mwakila et al, 2008). The
malachite-azurite with accessory chrysocolla and minor copper is LME ‘A’ quality and the recovery has increased by
other secondary copper minerals. The cobalt is present as 5 per cent. The direct EW tankhouse has been reconfigured
heterogenite with cobalt in both the Co2+ and Co3+ oxidation
from starter sheets to stainless steel cathodes; and capacity
state. The cobaltic minerals are about 50 per cent of the total
increased from 14 000 t/a to 30 000 t/a in a reduced footprint
cobalt; and are not leachable at normal temperatures and
(Kordosky, 2008).
pressures. Alternate methods of enhancing the cobalt leaching
are required. The new generation of projects has gone directly for
primary copper recovery from solution by solvent extraction
The leach trains are designed to leach the copper oxides
in mildly oxidising conditions; to recover the minor cuprite, and subsequent electrowinning on to stainless steel cathodes.
chalcocite and native copper. The leach time can be extended if In conventional solvent extraction plants this will recover up
there is significant chrysocolla present (Miller, 2005). Copper to 98 per cent of the copper – still leaving 0.10 g/L to 0.20 g/L
leaching is generally taken close to completion in four to eight copper in the final raffinate; that progresses to the cobalt
hours. Those projects that do not have significant secondary recovery. This level of copper is an issue with cobalt product
copper minerals can achieve high leaching efficiencies in as quality and needs to be reduced to less than 1 ppm (Kongol,
little as two hours from the rapidly leaching malachite and Mutlae and Kalenga, 2005).
azurite (Crease, 2006).
About 50 per cent of the cobalt is leached mainly from the Secondary copper removal
Co2+ minerals. The other cobalt needs to be reduced to the 2+ Preparation of copper SX plant ‘raffinate’ for downstream
state for it to leach. This is achieved with controlled reductive cobalt metal production needs the copper to be reduced to
leaching. The main reductant used to date has been sodium less than 1 ppm for metal electrowinning. In the past, this
meta bi-sulfite (SMBS: Na2S2O5) (Mwema et al, 2002). In an was achieved by co-precipitation of the iron and any residual
acid solution, the SMBS disassociates to form SO2aq which copper at pH +4.5; and disposal of the gypsum-iron-copper
lowers the Eh in solution and reduces the cobalt oxidation solids. The copper was a direct loss from the process.

Traditional single stage copper SX will produce the raffinate •• use of High Density Sludge HDS™ style techniques
described above. This represents a significant loss of copper to improve precipitant utilisation and further reduce
that could be recovered into the primary product. Cost benefit calcium over-saturation (HGE, 2007).
analyses on three projects to date have shown that a unique The air-SO2 system has been the subject of many papers
SX Split-Circuit™ can be justified (Miller and Nisbett, 2006). in the recent past. The system is reasonably robust and can
This circuit (shown in Figure 1) takes the cobalt bleed stream reduce iron to <5 ppm with <1 ppm a common result. The
and removes the copper to consistently less than 10 ppm. The major engineering issues with the system are the very high gas
copper is recovered as electrowinning (EW) LME ‘A’. The volumes, when using acid plant feed gas, and the competing
key to the SX plant performance is to integrate it with the kinetics of gas solubility reduction and crystal growth kinetics
precipitation removal of many of the other ions. with increasing temperature. Optimum temperature is quite
The iron is precipitated at pH 3.2 with only marginal loss solution chemistry dependent but generally falls in the range
of copper. The iron free high pH solution is contacted with of 45°C to 50°C (Krause, 2007). The effect of ions in solution
standard copper SX extractant (in a second SX section – SX2) is quite marked with reduction in Fe oxidation rates of
to remove the copper quantitatively. The copper in the SX2 between 10 per cent to 25 per cent with H+, Cu2+, Co2+, Cl–
raffinate is thought to come mainly from the entrainment (Krause, 2007). As these ions are in most solutions the rate of
of EW electrolyte in the stripped organic. The copper ‘free’ Fe oxidation needs to be confirmed from laboratory test work
solution is sent on to further purification. Should the final on ‘real’ process solution.
route be to metal, the solution quality of <1 ppm copper can The gas flow issues in the precipitation tanks, are quite
be achieved with a much smaller ion exchange (IX) plant or important as they are sufficiently high to cause both SO2
less reliance placed on purification via cobalt SX. stripping from the solution and to flood impellors – even
with appropriately designed ones. The key control criterion
is the SO2/O2 ratio that can seldom be greater than 1:5. These
IRON REMOVAL factors are drivers for eliminating significant gas volume by
Iron removal has been undertaken for many years in many using liquefied SO2 from the acid plant gas. The volume
hydrometallurgical process plants. The classic method is reduction will to a large degree overcome the solubility
air oxidation to ferric and precipitation with lime and/or issues and subsequent occupational health and safety issues
limestone. All the current and previous Zambian and DRC with SO2 in the ambient air. At least two new projects are
project used this basic method. However the older style plants using this technique for enhanced iron removal.
all suffer from the usual problems of: Multiple stage reagent addition is just good chemical
•• downstream gypsum precipitation engineering to control the over concentration of calcium.
•• fouling of process equipment and pipes. Papangelakis has shown the benefits of this in the control
of calcium concentration in the final solution. He has also
Many of the new projects are using more recent process recommended that a final stage of solution maturation
developments to minimise the calcium over-saturation and without reagent addition be used to further minimise the
subsequent gypsum precipitation issues. The major focus has calcium saturation.
been on:
Higher temperature operation has also been adopted in
•• use of air-SO2 as an enhanced oxidant for the iron several projects where the benefits were required. The first of
(Krause, 2007; Ho and Ring, 2007) these was in Zambia where direct steam injection was used
•• multiple stage addition of precipitant (Papangelakis, to raise the temperature. Both Zambian roast-leach projects
2004; Demopoulos, 2004) have enhanced solution temperature (because of leaching hot
•• higher temperature operation to enhance kinetics of calcine) and their subsequent iron removal is generally good
crystal growth and kinetics (ibid) without the use of oxidants other than air.
•• recycle of seed crystals (ibid) Crystal seed recycle has been practiced for many years to allow
growth of larger crystals and minimisation of downstream
gypsum fouling. The operations at the three Australian
Nickel plants showed how important this unit operation is to
achieving high plant utilisation. None of the plants utilised all
the techniques mentioned here, in their original design. Plant
shutdowns of one to two days every three to four weeks were
common; until better control of the iron precipitation step was
achieved. Seed recycle is an integral part of this control and has
been included in all new projects.
The HDS technique (HGE, 2007) and other similar patented
processes were developed primarily to provide larger crystals
from dilute water treatment solutions. The key process is to
have large seed recycles (up to 2000 per cent from dilute
solutions) combined with a modified reagent addition
regime. The first reactor receives the recycled crystals and the
precipitant. Here the precipitant is adsorbed on to the crystal
seeds. In the second reactor(s) the process solution is added
and reacts directly on the particle surface growing there; in
preference to fresh nucleation in the solution. This technique
claims (Gabb et al, 1995) to reduce downstream calcium
precipitation and to improve solid-liquid separation rates.
Many variations on this theme are available as technology
Fig 1 – Copper-cobalt Split-Circuit™. packages. Most need to develop specific ‘recipes’ to address

G Miller
the solution chemistry to be used. However almost all the new proposed for this duty (Tinkler, Flett and Bourget, 2007) with
projects have incorporated very flexible systems of tankage, specific pH and scrubbing conditions to target zinc rather than
reagent addition, seed recycle and oxidation intensity; to cobalt. Both D2EHPA and C272 suffer from issues of calcium
allow them to adjust their plant to suit changes in the actual saturation in the strip solution; which must be addressed by
chemistry presented. large volumes of zinc stripping solution to limit the calcium to
below saturation conditions in this waste stream.
MANGANESE REMOVAL A small zinc SX was in operation at Bulong, which
Manganese removal is done mainly to achieve levels that are removed this ion before cobalt EW. It suffered severely from
suitable for the product specification. It is accomplished at a reagent poisoning with ferric whenever the iron and sulfide
higher pH than iron and has greater potential to co-precipitate precipitation plants were not operated well. No hydrochloric
cobalt. As a result, it is often done as a separate step to iron and acid regeneration was carried out.
the resultant solids recycled back to the leach to recover some
of the precipitated cobalt. This is similar to the technique used COBALT CONCENTRATION
at Bulong for recovery of co-precipitated nickel (O’Callaghan,
2003). The process used to date in Zambia and DRC has been a Solvent extraction
combined iron and manganese precipitation. This has removed At this stage, the older cobalt EW plants would direct
the copper and some zinc; but also co-precipitated significant electrowin from the cobalt solution. Some control of nickel
cobalt that has been lost. Enhanced methods of manganese would be required, generally on a side stream to reduce the
removal have been developed to minimise this loss. cost of the IX plant. Cobalt precipitation and re-leaching was
The amount of manganese that can be precipitated depends required to obtain the required water balance and to increase
on the level of Mn4+ present. But due to the reductive the cobalt concentration in the advance electrolyte. These
leaching for cobalt this is generally a small portion of the precipitation steps also lead to cobalt losses – especially from
total manganese. The air-SO2 oxidation system has been the less than perfect re-leach step.
investigated and included in several circuits to lower the More recent projects are considering the production of
manganese concentrations. The level of reduction needs to be high quality metal by EW of an SX electrolyte. The preferred
determined for the specific product. In many cases removal to reagent is Cyanex® 272 or its analogues. The cobalt is removed
1 g/L is sufficient if subsequent operation is via cobalt SX–EW from the PLS at elevated pH and stripped in a very dilute
or for production of an impure intermediate product. Some acid electrolyte. The EW of cobalt in un-divide cells is well
manganese is acceptable in cobalt EW electrolyte provided understood, and the maximum operating concentration of
that anode maintenance is regular and the deposited MnO2 acid is around 8 g/L in the spent electrolyte. This corresponds
is removed. to a cell delta cobalt of around 5 g/L.
If manganese removal is required to lower levels, then it has The engineering issues are around the selection of the pH
been found that the rate of reduction is enhanced if a two-stage control chemical – ammonia gas, ammonia solution, sodium
process is used. A primary stage to reach 1 g/L; followed by hydroxide solution or others. The lack of an ammonia gas
a solid-liquid separation step. The liquor is then treated in a infrastructure in Africa is leading many operations to consider
second oxidative step to remove manganese to ppm levels. ammonia solution (Tati Nickel) or dilute sodium hydroxide
The kinetics are slow and long residence times are required. (Sole et al, 2005). In all cases there is an issue of residue
The same issues with gypsum management and gas volumes disposal unless steps are taken to eliminate the common ions:
need to be addressed in the design of this circuit. ammonium or sodium from the effluent.
Bulong used an alternate method by precipitating the cobalt
as a sulfide and leaving the manganese in solution. The Precipitation
sulfide was re-leached in a small autoclave. This process was Cobalt precipitation is also undertaken at this stage for
complex and involved more stages of treatment. As a result, production of a saleable salt or to reject water. Precipitation
it was discontinued in favour of producing an intermediate has been conducted traditionally with lime slurry when
sulfide product. re-leaching; or sodium carbonate for a saleable product.
However, a number of projects have considered the use of
magnesium oxide instead of sodium carbonate. This is largely
ZINC REMOVAL based on the costs of the reagents and the higher utilisation of
At this point, it is possible to precipitate a reasonable quality magnesium oxide than sodium carbonate.
cobalt salt without further purification. However, if zinc levels
The precipitation is also likely to be conducted at elevated
are elevated in solution they will be elevated in the product
temperatures of up to 80°C. This is used to enhance crystal
as the zinc will precipitate with the cobalt. Zinc removal
growth rates and morphology; and to achieve extremely high
in the traditional circuits has been via lime precipitation at
recoveries from the enhanced kinetics.
elevated pH and recycle of the solids to the leach for cobalt
recovery. This has proved to be marginal at best and a
circulating load of zinc has built up. Both Chambishi and OTHER IONIC CONTROLS
Nkana use a D2EHPA SX to extract the zinc prior to cobalt Control of other ionic species is not generally practiced
metal production (Sole, Feather and Cole, 2005). The Nkana specifically unless one or another is introduced as part of the
plant is a simple 1E+1S that takes out a portion of the zinc to process reagent(s). One operation in Zambia could not use
stabilise the circulating concentration to a low enough level to local limestone as it was running 4 per cent acid soluble oxide
be acceptable. Chambishi have a more sophisticated SX plant zinc in the material. Likewise, introduction of ammonia or
with multiple stages of extraction, scrubbing, stripping and sodium carbonate/hydroxide have issues with the disposal
washing. They also include an hydrochloric acid regeneration of tails having high available nitrogen or high sulfate (as the
stage where ferric is removed from the D2EHPA. sodium salt).
Other newer projects are also considering the benefits of zinc Aluminium and silica are generally precipitated with the
SX, ahead of the cobalt purification plant. Cyanex® 272 has been higher pH reactions. At a pH of 4.5 most of the aluminium and

silica precipitate with the iron and/or manganese removal. have been run for long periods at one level of sophistication
This is non-specific removal as both co-precipitate. or another.
Use of magnesium oxide for precipitation of cobalt hydroxide
introduces magnesium into the solution. This can be removed Australian nickel operations
by lime precipitation of magnesium hydroxide but at the cost of The Australian high pressure acid leach nickel operations
the lime to force the reaction. Disposal of the calcium saturated have all suffered to some degree from significant issues of
water is now possible once it is re-acidified to pH 6.0 to 8.0. gypsum precipitation; which has resulted in lower than design
There are some issues with the disposal and management of utilisation and subsequent cash flow issues. This is a prime
the magnesium hydroxide. If it can dry and the pH drops below lesson in making sure that all the points regarding process
around 8.0, then it can partially redissolve. Similarly, if the pH optimisation are used for the iron-manganese precipitation
is reduced (by say, mixing with the acid tails stream) then the processes.
magnesium will also redissolve. For this reason, magnesium
None of the operations practice copper recovery specifically;
hydroxide precipitate is emplaced in a separate tails dam which
and copper is removed by precipitation as part of the iron
is kept covered with solution. This prevents pH 7.0 rainwater
removal process. Bulong had a final copper clean up prior to
dissolving the magnesium. The pool of solution also acts as
cobalt EW using IX resins.
a ‘maturation pond’ allowing the calcium over-saturation to
equilibrate; reducing the amount of gypsum precipitation in
the plant when this solution is recycled.
Cawse
The Cawse plant used an intermediate precipitation of mixed
Ni-Co hydroxide with re-leach in ammonia. Nickel and cobalt
PLANT MANAGEMENT were separated in ammonia solution using a LIXTM reagent.
Many of the ancillary process steps are designed to minimise The cobalt was subsequently precipitated as a salt for sale.
the effects of one process on the succeeding process(es). The Other than the gypsum issues there are no specific items that
significant number of plant designs that are using sequential can be taken to new cobalt production facilities.
solvent extraction processes, will all require the minimisation
of reagent carry over. Bulong suffered from carry over Bulong
of Cyanex® 272 from the cobalt SX into the VersaticTM 10
nickel SX (O’Callaghan, 2003). Insufficient consideration Bulong embarked on the full route to cobalt EW metal
of this interaction has lead to other operations also having production in the initial design and operation. This proved
difficult operations (Kasese, 2007). Newer plant designs are to be problematic due to the high reagent costs and large
utilising some or all the following process steps to reduce the number of operators required to keep the section running.
intermixing of SX reagents: The long train of processes to provide an electrolyte for EW
metal production was such that it created one of the bottle
•• settler designs with lower entrainment losses such as the
necks in the plant. It was subsequently modified to produce
Miller Metallurgical Services (MMS) Side-Feed™ settler
an intermediate sulfide salt. The main issues to take from
•• after-settlers and coalescors for removal of bulk this operation are the simplification of the process with an
entrained organic intermediate product, and the difficulty in operating the long
•• use of a diluent scrub stage for recovery of Cyanex® train of processes continually.
272; this was combined with a saponification process at As part of their in-house development to improve plant
Bulong to recovery the C272 in a concentrated stream utilisation Bulong undertook a six-month commercial trial
(O’Callaghan, 2003) of an anti scaling reagent (O’Callaghan, 2003). This was
•• dual media filtration to remove organics to <5 ppm successful in reducing the build-up gypsum in the process
entrained vessels and pipes; and increased their run time from three
•• carbon adsorption to reduce organics to <1 ppm total weeks to over two months. Although expensive, the reagent
entrained and dissolved. was cost justified on the basis of the reduced clean out costs
Other process steps are designed to provide higher plant and improved plant utilisation and total production.
utilisation between shutdowns for gypsum removal:
•• clean-in-place acid circulation systems for dissolving
African experiences
gypsum from heat exchangers and key process pumps
Confidential client – Democratic Republic of Congo
•• ‘spare’ mixer-settler unit for online clean out – especially
This project is planning to make a high-grade cobalt metal
for zinc D2EHPA stripping stages.
product for market. They currently operate a facility making
There is generally a need to address the issue of silica in an intermediate cobalt basic sulfate. As part of this process
the primary copper SX plant PLS. The silica will be removed the iron is precipitated with air and lime, using indirect
in the iron precipitation, but only after the primary copper steam heating. The operating plant has developed several
extraction has been done. Plant operation in organic techniques for minimising the downstream precipitation of
continuity is required; and the control of entrainment in the gypsum. A semi-commercial pilot plant has been running
loaded organic is necessary. Most new copper SX plants are for more than ten months producing 500 kg/d of high-
either including loaded organic coalescing or making design grade cobalt metal. This process uses the sequential zinc SX
provision to allow easy retrofitting of the coalescing system.
extraction with D2EHPA and cobalt concentration with a
Cyanex® 272 analogue. Nickel, magnesium and other metal
PLANT EXPERIENCE rejection is high from a solution that is essentially iron and
It is important to look at the experience of existing plants to manganese free. pH adjustment in the cobalt SX is with
try to build on the knowledge base that exists. Many errors of sodium hydroxide in dilute solution. This pilot plant is being
engineering application can be avoided if the lessons can be used to develop design criteria for a full-scale operation to
included in the new plant designs. Most of the unit operations produce 10 000 t/a cobalt metal.

G Miller
Tati paper. The client bodies are also responsible for most of the
pilot plant testing and engineering proving of the improved
LionOre (now Norilsk) have constructed and operated a
unit operations. Without this individual and collective
semi-commercial pilot plant for their Activox™ process at
commitment, none of the innovations now being put into
Tati Nickel in Botswana. The process involves initial copper
practice would have seen the light of day.
removal by single stage SX–EW, iron-copper-manganese-
silica precipitation in two separate stages at different pH,
cobalt recovery with Cyanex® 272 SX and subsequent nickel REFERENCES
recovery with VersaticTM Acid 10 (the same circuit as used Anon, 2008. Katanga, Nikanor merger completed, Engineering and
at Bulong). Cobalt is precipitated as the carbonate for sale Mining Journal, 290(1):18.
as an intermediate product. Many of the innovations in iron
Demopolous, G P, 2004. Aqueous crystallisation and precipitation
precipitation and gypsum management are included in the
research at McGill, presented to A J Parker Centre
plant design. This combined with other proprietary designs Hydrometallurgy Conference, Perth, April.
has meant that the pilot plant has run without significant
Gabb, P J, Howe, D L, Purdie, D J and Woerner, H J, 1995. The
gypsum precipitation in the cobalt or nickel SX plants.
Kennecott smelter hydrometallurgical impurities process,
presented to Copper 95 – Cobre 95, Vol III Electrorefining and
Sable Zinc Kabwe Hydrometallurgy of Copper (Canadian Institute of Mining,
The Sable Zinc Kabwe (SZK) operation was the first Split- Metallurgy and Petroleum: Montreal).
Circuit copper SX to produce very low copper concentrations HGE Corporation, 2007. HDS™ technology description, Marketing
in the cobalt plant feed (Kordosky, 2008). It has also been the material.
first to start up with SO2 reductive leaching and air-SO2 iron Ho, E and Ring, B, 2007. Oxidants for Uranium Leaching, presented
and manganese oxidative precipitation. To date the leaching to ALTA Uranium Conference, Perth.
has been successful while the precipitation has reverted to air
Kasese Cobalt Company Ltd, 2007. World’s First Cobalt Bioleach
alone due to a reduced penalty element profile for their cobalt
Plant, Company presentation.
product. The use of extended precipitation times (from the
availability of the necessary tankage) has meant that there is Krause, E, 2007. Ferrous Oxidation with SO2-air an alternative
uranium leaching system?, presented to ALTA Uranium
only slight gypsum scaling in the secondary copper SX plant.
Conference, Perth.
Nkana Kordosky, G, 2008. The copper of Africa – a renaissance in copper

hydrometallurgy, in Proceedings IV International Copper
Nkana operations have improved with the conversion of the Hydrometallurgy Workshop, pp 31–40.
copper electrowin – electro-strip operation to a Split-Circuit
Miller, G M, 2005. Analysis of leaching data – methods and
copper SX and stainless steel cathode EW plant (Mwakila et al,
interpretation, in Proceedings First Extractive Metallurgy Operator’s
2008). The copper recovery has improved 5 per cent and the Conference, Brisbane, November (The Australasian Institute of
cobalt recovery has also improved by a further +5 per cent. Mining and Metallurgy: Melbourne).
The copper product is LME ‘A’. The increased cobalt recovery
Miller, G M and Nesbitt, A, 2005. Decreasing operating costs and
is attributed to the reduction in pH required for the iron soluble loss in copper hydrometallurgy with use of innovative
precipitation stage and the subsequent gain in cobalt no longer solvent extraction circuits, in Proceedings First Extractive Metallurgy
co-precipitated with the iron. Further cobalt production has Operator’s Conference, Brisbane, November (The Australasian
been possible by adding a bleed of raffinate from the newly Institute of Mining and Metallurgy: Melbourne).
commissioned Nkana leach plant as the make up water to the Mwakila, B, Silwamba, R, Miller, G and Chisakuta, G C, 2008.
calcine leach. This has further increased cobalt production by Improvement in copper and cobalt production due to inclusion
around 10 per cent. of solvent extraction of copper at Mopani Copper Mines Plc’s
Nkana Cobalt RLE plant, presented at ISEC 2008.
CONCLUSIONS Mwema, M D, Mpoyo, M and Kafumbila, K, 2002. Use of sulphur
dioxide as reducing agent in cobalt leaching at Shituru
The selection of an appropriate cobalt recovery process is a
hydrometallurgical plant, Journal of the Southern African Institute
tortuous path threading between product specification, leach of Mining and Metallurgy, 102(1), Jan/Feb.
solution chemistry, cost/benefit analysis and operational
complexity. There is no one process selection that is appropriate O’Callaghan, J, 2003. Process improvements at Bulong Operations
Pty Ltd, presented at ALTA Nickel Conference, Perth.
to all operations – even those with similar chemistry.
Papangelakis, J F A, 2004. Aqueous scale formation and control,
Unit operations for use in cobalt circuits have undergone
presented to A J Parker Centre Hydrometallurgy Conference,
major development since the last significant cobalt plant Perth, April.
was constructed. As a result, the operations of the next
generation of cobalt plants are likely to be more stable with Sole, K C, Feather, A M and Cole, P M, 2005. Solvent Extraction in
southern Africa: An update of some recent hydrometallurgical
higher recoveries in the long-term; but fraught with the
developments, Hydrometallurgy, 78:52–78.
usual unforeseen issues in the period immediately after
commissioning. Tinkler, O, Flett, D and Bourget, C, 2007. Flowsheet considerations
for copper-cobalt projects in the DRC, presented to ALTA Nickel
Conference, Perth, May.
ACKNOWLEDGEMENTS
Miller Metallurgical Services would like to gratefully
acknowledge the permission of various clients to publish this

Contents
A review of best practice in gravity

circuit design and operation
A Giblett1, A Bax2, G Wardell-Johnson3 and W Staunton4
ABSTRACT
This paper presents several considerations for process development test work protocols
and interpretation as a reliable basis for the inclusion of gravity separation in gold
recovery flow sheets. Operating data, experiences and numerous case studies are also
discussed in support of identifying best practice principles for gravity circuit design.
INTRODUCTION
The decision to include gravity separation in gold recovery flow sheets can be the
cause of some angst during process development, due to uncertainty around the
impact of a gravity circuit on overall project economics and ultimate plant metal
recovery. Numerous case studies have been reported where retrofitting or upgrading
gravity circuits has been demonstrated to increase overall plant metal recovery. Such
examples provide ample demonstration that the processes employed to define the
potential impact of gravity recovery during the process development phase have not
always been as reliable as one should expect.
The development of the gravity recoverable gold (GRG) test at McGill University
under the guidance of Professor Andre Laplante and the subsequent adaptation of
population balance modelling for gravity circuit simulation have been significant
advances facilitating reliable gravity separation trade-off studies during process
development. The GRG test largely overcomes the poor sampling statistics associated
with coarse gold particles and small metallurgical test samples, while providing
very detailed information on the size and abundance of free gold particles. The
GRG modelling process allows simulation of gravity separation performance as
a function of several key parameters, such as classification efficiency, technology
selection, feed rate and particle size, and allows prediction of the GRG content of
the grinding circuit product to assess the impact on downstream processing. These
tools are considered the best available to base the decision to include or exclude
gravity separation from the process flow sheet.
There are then a number of principle factors that have been observed in practice to
influence the performance of a gravity circuit that must be appropriately addressed
in design. The two book ends of the gravity circuit, being feed preparation and
concentrate treatment, have frequently been observed to be undersized, resulting
in reduced gravity circuit performance either by limiting gravity circuit feed rate or
limiting gravity concentrate production. It is therefore critical to ensure that the feed
screen is designed to adequately accommodate the design flow, at a suitable screen
aperture to provide low wear rates and high separation efficiency. Selecting the
appropriate separation technology for the application, within a suitably sized gravity
circuit to achieve the desired gravity recovery are also critical considerations. The latter
must also give consideration to the mill water balance and the associated influence on
grinding efficiency, while maintaining low gold accumulation in the grinding circuit.
The ability to remove magnetic material, in particular tramp steel, from the gravity
1. FAusIMM, Senior Technical Advisor – Mineral circuit feed may further be expected to improve the effective yield and separation
Processing, Newmont Mining Corporation, efficiency of the gravity circuit. Sizing the concentrate treatment process to handle
Englewood CO 80112, USA. high primary gravity concentrate production rates is further conducive to optimum
Email: aidan.giblett@newmont.com gravity circuit performance. There is then the issue of selecting the appropriate
2. MAusIMM, Consultant, Ozmet, West Perth WA concentrate treatment process for the application, giving consideration to recovery
6872. Email: alan.bax@ozmet.com.au efficiencies, labour intensity and environmental performance criteria.
3. Senior Research Metallurgist, Gold
Technology Group, Curtin University, Perth
WA 6102.
BEST PRACTICE PRINCIPLES – PROCESS DEVELOPMENT TEST WORK
Email: g.m.wardell-johnson@curtin.edu.au When considering the optimum gravity circuit flow sheet and design basis for a new
4. MAusIMM, Adjunct Professor, Gold project there are several important considerations. First and foremost, the gravity
Technology Group, Curtin University, Perth gold recovery potential of the ore must be defined, as this becomes a key basis for the
WA 6102. Email: w.staunton@curtin.edu.au decision to include or exclude gravity processes from the metal recovery flow sheet.
263
A Giblett et al
Careful consideration must also be given to the particle size single stage GRG test is designed to reduce the sample size
distribution of the GRG as an important factor in technology requirements of the full GRG test by grinding the entire test
selection and flow sheet design. feed to P80 75 µm. A sample mass of 25 kg is used for the test
The authors are strong proponents of the application of the which as the name suggests is performed in a single stage,
GRG test to define the gravity recovery potential for a given effectively bypassing Stage 1 and Stage 2 of the full test and
project. The three-stage GRG test gives complete detail on going straight to Stage 3 conditions. The test has been shown
the size and abundance of GRG for a particular sample, and to replicate the total GRG values derived from the full three
is invaluable information to support equipment selection stage test in several instances, although this is again dependant
and to define the required gravity recovery effort. The single on the GRG size distribution being fine to moderate. Where
stage GRG test as described by Laplante and Staunton (2005) the GRG size distribution is coarse, the single stage GRG is a
can be used where the available sample mass is limited and less precise indicator of gravity recoverable gold content and
to reduce the test work cost. However, due to the final grind does not reliably represent the true GRG size distribution.
size used, the single stage test provides much less detail for Nonetheless, in many instances the single stage GRG test is
the coarse size fractions (>150 µm). The lower sample mass a very useful indicator of variability, and total GRG content.
used for the single stage test also increases the associated
sampling error, especially for coarse, low-grade ores. The Interpreting gravity recoverable gold test work
single stage GRG test is however a useful alternative to the
full stage GRG test for determining the variability of GRG Direct interpretation of the gravity recoverable gold test
contents through an orebody. There are references in the literature to what are essentially
Understanding the mineralogy of the ore and of the GRG nominal scale-up factors to predict the plant scale gravity
concentrate is critical. The presence of refractory gold alloys recovery from the three stage GRG test result. A scale-up
or minerals containing solid solution gold may have a factor of two-thirds is commonly quoted to represent the
considerable impact on the ability of the project to realise the appropriate multiplier to apply to the total GRG value to
gravity recovery potential indicated by the GRG test. account for the imperfectness of the plant separation by
comparison to the laboratory test. This scale-up or inefficiency
The application of the Laplante gravity recoverable gold test factor has elsewhere been quoted as a more broad range
The GRG test was developed at McGill University under the of 50–80 per cent which more closely reflects the range of
guidance of Professor Andre Laplante and has achieved global correlations to the GRG value observed in practice. In truth, the
acceptance as the preferred methodology for characterising plant scale realisation of the total three stage GRG test varies
the amenability of a given ore type to gold recovery by widely from ‘as low as 25 per cent to as high as 85 per cent’
gravity concentration. A large database of GRG test results (Laplante, 2000b). The wide range is no doubt a function of
has been compiled since the implementation of the test, and plant inefficiencies induced through variable design and
Newmont has on record over 50 GRG tests from its operations operating practices, but is also fundamentally related to the
and projects for reference. particle size distribution of the GRG. Given these influences,
approximating a plant scale recovery value from the total
The GRG test requires a large sample of ore (typically
GRG value without detailed consideration of the plant design
80–100 kg) to be processed through a three inch Knelson
and the nature of the GRG itself, is not recommended.
concentrator at three successive liberation sizes, known as
Stages 1–3: Newmont has observed a close relationship between the
Stage 1 GRG test result and the actual plant gravity recovery,
1. In Stage 1 the entire sample is crushed to 100 per cent
as shown in Figure 1 from Giblett, 2011. Any relationship
-850 µm and processed through the concentrator,
of this nature is certainly prone to the usual influences,
generating a Stage 1 GRG concentrate.
including the extent of GRG liberation in test preparation,
2. The Stage 1 tailings are then split to produce a smaller circuit efficiencies and gangue specific gravity. However, the
sample (25–35 kg), which is ground to 45–55 per cent relationship proposed in Figure 1 is at a minimum free of the
passing 75 µm and then run through the three inch risk of overstating gravity recovery based on exceptionally
Knelson in Stage 2 of the test, producing a Stage 2 GRG fine test feed size conditions, as present in Stage 3 of the GRG
concentrate. test. Such conditions enhance the recovery of fine GRG and are
3. Finally, the Stage 2 tailings are ground to 80 per cent a poor reflection on the typical feed particle size distribution
passing 75 µm and run through the Knelson again of the plant scale concentrator.
producing a third (Stage 3) Knelson concentrate. The
final mass balance determines the total amount of gold Gravity recoverable gold recovery modelling
recovered as GRG, and the percentage recovered in each The application of population balance models (PBM), using the
individual stage of the test. approach described by Laplante, Woodcock and Noaparast
The GRG procedure has more recently been incorporated (1995) and Laplante et al (1995), has been proven in the design
in a simulation technique that integrates both gravity and and optimisation of greenfield and operating gravity plants.
cyanidation of grinding circuit products, to predict the impact The use of PBM provides a very effective means to examine
of gravity recovery on the leaching circuit performance and the influence of varying factors such as the GRG content, GRG
overall plant gold recovery. The Integrated Liberation Leach size distribution, grind size, gravity circuit configuration and
Model is described in detail by McGrath et al (2016). gravity circuit feed rate.
It is possible to perform the test on smaller samples with In addition to providing a prediction of the gravity recovery,
reasonable reliability but this is dependent on the sample modelling can also be used to determine the GRG content
having a fine to moderate GRG particle size distribution. and size distribution in the stream reporting to downstream
This approach is compromised when the GRG particle size processes. This can then be used to evaluate the impact of
distribution is coarse and the ore grade is low. gravity gold recovery on flotation recovery (Fullam, 2010)
Developed in collaboration between Professor Laplante and leach performance (Staunton and Bax, 2010) allowing the
and the Curtin University Gold Technology Group, the benefit to overall gold recovery to be determined.

A review of best practice in gravity circuit design and operation
FIG 1 – Relationship between Stage 1 gravity recoverable gold (GRG) test result and plant gravity gold recovery for Newmont operations.
Table 1
Gravity recoverable gold (GRG) recovery efficiencies for the Newmont Tanami spiral / Wilfley table gravity circuit.
Stage Particle size (microns)

+600 +425 +300 +212 +150 +106 +75 +53 +38 +25 -25
Spirals 98% 97% 96% 99% 98% 95% 97% 96% 87% 74% 27%
Tables 97% 98% 99% 98% 97% 94% 98% 98% 98% 98% 89%
Total 95% 94% 95% 97% 95% 90% 95% 94% 85% 72% 24%
Nominal decision points Depending again on the size distribution of the GRG,
improvements in overall plant efficiency can be expected for
Where the gravity gold particle size distribution is coarse,
and relatively abundant, there is clear potential for gravity gravity gold contents as low as 10–20 per cent, as is the case
recovery to have a positive impact on overall recovery circuit with the KCGM operation (Giblett et al, 2012), where such
performance, in particular overall metal recovery. This has levels provided the basis for a 0.5–1.0 per cent improvement
been demonstrated at KCGM (Giblett et al, 2012) amongst in overall plant recovery. At low gravity gold recovery
other operations and is consistent with Professor Laplante’s levels there are numerous factors which must be considered,
(2000a) observation that ‘The highest economic impact including the GRG size distribution, process chemistry, the
for gravity recovery has been observed when the gravity downstream recovery processes and system constraints
recoverable content (GRG) is high and coarse’. Interpretation including leach residence time.
of the particle size distribution of the GRG and the Stage 1 For example, Newmont would consider the installation of
result of the GRG test offers essential guidance on the merits a gravity circuit to remove 10–15 per cent of the gold from
of gravity gold recovery. the feed to a flotation process, particularly when the GRG
Study of several projects and operations around the globe size distribution is suitably coarse to present a challenge to
has led Newmont to the generalisation that where the plant flotation. The gravity circuit in this instance would be given
gravity gold recovery expectation is 25 per cent or more of more thorough consideration if the plant is without the safety-
the mill feed gold content, the requirement for gravity gold net of a cyanidation circuit to treat the flotation tails. Where
recovery in the process flow sheet is difficult to ignore. This cyanidation is to be employed at typical residence time levels
is contingent on the recovery expectation being based on a of approximately 24 hours, a gravity circuit would be given
coarse (GRG P80 >350 µm) gold particle size distribution, consideration when the gravity gold recovery expectation is
and high gold recoveries (>25 per cent) under the Stage 1 15–20 per cent or more. This is contingent on the gravity gold
conditions of the GRG test. Figure 2 gives some guidance on recovery expectation being based on interpretation of the
general classifications of GRG size distributions. Supporting Stage 1 GRG content determined by the standard Laplante
references for this decision point include the Harmony Hill 50 GRG test.
operation (Minerals Engineering International (MEI) online, Where the gravity gold recovery expectation is lower, and/
2004), where 24–30 per cent gravity recovery translated to or the GRG size distribution finer, the potential value add from
approximately 1.5 per cent improvement in overall recovery, gravity gold concentration may be less clear. The opportunity
and the President Steyn mine in South Africa (Van de Steen, to increase cash flow by removing free gold values ahead of
2000) which reported a 2 per cent improvement in total flotation is one example of where gravity gold concentration
gold recovery following the installation of a gravity circuit can have economic value at lower total gravity recovery levels
recovering up to 45 per cent of the gold in mill feed. or where finer GRG dominates.

A Giblett et al
FIG 2 – Classifications of GRG size distributions.
The use and interpretation of small-scale gravity tests report the gravity gold content of the sample by artificially
Small-scale tests are frequently conducted as a part of the enhancing the separation of gold from fine non-GRG. This
recovery flow sheet for flotation or cyanidation recovery tests will additionally be affected by the quality of hand panning
on 500 g to 1000 g samples. The samples are often ground to and the final mass of concentrate submitted for assay. The
final product size for the recovery process in question, and extent of grinding of the test feed and the fineness of the GRG
passed through a batch gravity concentrator to remove any distribution will have a large influence on the scale-up to
coarse gold remaining in the sample following grinding. The plant performance. The low sample mass used in these tests
gravity concentrate may then be hand-panned to reject any further introduces sampling errors that conspire to produce
non-GRG or gold bearing minerals present, also reducing variable gravity recovery results.
the concentrate to an appropriate mass with reference to the One example of the qualitative nature of the small-scale
original test feed. batch test is shown in Figure 3, for the Phoenix gold project.
These small-scale batch tests are a useful qualitative indicator The laboratory gravity recovery test data shows a large degree
of the presence of gravity gold values in the sample. Due to of scatter (influenced by variable coarse gold content within
the fine feed size often used for the test, the high mass pull the Phoenix ores but primarily the poor sampling statistics
and the slow grinding rates of gold, the test is likely to over associated with small sample mass) in the laboratory data
FIG 3 – Comparison of gold recovery to gravity concentrate for laboratory tests and plant data.

against actual plant performance. The black bars show actual Preparation of the feed to the gravity circuit – screening
gravity gold production which over the course of several The capacity of the gravity circuit feed screen is an important
years typically ranges from 10–20 per cent of gold in feed. consideration as an undersized screen very quickly becomes
The laboratory testing results in grey bars average 35 per cent a limitation to gravity circuit performance. Gravity recovery
gravity gold recovery and range widely from 10–60 per cent. can often be increased by increasing the feed rate to the gravity
Taking these small-scale test results as a direct indicator of circuit. The ability of the screen to accommodate high flows
plant scale gravity recovery is clearly a risky proposition. is therefore a critical requirement. Benchmarking performed
As proposed by Laplante (2000b), the relevance of batch in support of the KCGM gravity circuit design (Figure 4)
recovery test work would be materially improved by preparing identified a number of undersized gravity circuit screens that
a larger sample for gravity testing, and splitting the tailings were limiting gravity circuit feed rate. This benchmarking
sample to provide feed for cyanidation or flotation test work. established 55 m3/m2/h as a suitable upper limit design flux
for the gravity screen, based on low viscosity ore and a screen
Plant design considerations size aperture of 2–3 mm. Based on interpretation of Figure 4
there is somewhat of a grey area between 45 and 55 m3/m2/h
Having determined that the project requires a gravity gold which would make 45 m3/m2/h a more conservative design
recovery flow sheet of some description, there are several basis for screen flux.
design considerations that are important to the effectiveness
of the gravity circuit. Screen aperture is another important consideration, as larger
screen apertures will facilitate higher gravity feed flows, but
also may induce higher wear rates and reduced separation
Selection of gravity circuit feed stream efficiency. As stated by Staunton and Bax (2011), typically
Gravity circuits commonly operate on bleed streams of either few gravity gold particles coarser than 1 mm can be expected
cyclone underflow (CUF) or cyclone feed-ball mill discharge. hence it is preferred to screen the gravity feed at no coarser
The selection of the optimum gravity feed stream has been the than 1 mm. However, many installations can be considered to
topic of deliberation in several publications, with a thorough be operating effectively with screen apertures in the 2–3 mm
consideration of relative merits given recently by Staunton range and designing for a screen aperture of 2 mm in many
and Bax (2011). In this assessment a gravity circuit being fed instances can be considered an acceptable compromise. Finer
by the cyclone feed or ball mill discharge has the advantage of screening is specifically recommended where the GRG size
lower screen surface area requirement due to a finer particle distribution is fine and the gangue mineral specific gravity is
size distribution, lower capital costs, higher gravity recovery high, in order to realise high gravity gold recovery efficiencies.
at high feed rates, lower wear rates and lower absorbed power.
Observations of Newmont’s gravity circuits has identified Preparation of the feed to the gravity circuit – magnetic separation
the KCGM installation, discussed in detail in Giblett et al A frequent concern in the design and operation of gravity
(2012), as representing best practice in gravity circuit feed circuits is the presence of magnetics, particularly tramp
control. The use of dedicated, variable speed gravity circuit steel from worn grinding media, in the gravity concentrate.
feed pumps directing mill discharge slurry to the gravity The Goldex operation (Robichaud, Belzil and Fullam, 2012)
screens allows greater control over the gravity circuit feed rate reports magnetics levels in Knelson concentrate between 38
than is achieved when cyclone underflow slurry is diverted to and 65 per cent. This quantity of steel reduces the volume of
the gravity circuit. The KCGM Mt Charlotte gravity circuit, useful gravity concentrate that can be processed by a leach
where the gravity circuit feed is drawn from a dedicated reactor, and the presence of large quantities of iron will also
hopper isolated from the main cyclone feed hopper, to which reduce leach kinetics. The presence of high specific gravity
the gravity circuit tailings are distributed, further represents magnetics restricts separation efficiency when primary gravity
best practice design to prevent the bypass of gravity circuit concentrates are cleaned on shaking tables. Cloutt (1995) gives
tailings back to gravity circuit feed. an account of such issues as encountered at St Ives, where the
FIG 4 – Gravity screen performance benchmarking.

A Giblett et al
Knelson concentrate contained up to 20 per cent by weight the GRG size distribution is fine and there is economic value
as magnetics. This led to the design of a gravity circuit where in removing these values ahead of downstream processing.
a Knelson concentrate was cleaned by magnetic separation Flash flotation has found application at the Cadia, Telfer,
ahead of tabling. Boddington, Kanowna Belle, Bulyanhulu and Phoenix
Removal of magnetic material from the primary gravity operations either as a primary GRG recovery mechanism
concentrate is often considered an unattractive option due or to scalp fine GRG from gravity concentrator tailings. The
to the tendency of gold particles to become entrained in the configuration of these flash-gravity circuits is best derived
magnetic fraction, generating a product too high-grade to from a detailed understanding of the relative abundance
discard. Heroux and Hart (2005) describe the operation of of coarse and fine GRG in the ore, as installing either flash
the Cadia gravity circuit, where magnetite is removed from flotation or a conventional gravity concentrator as the
the gravity concentrate, with 4 per cent of the gravity gold primary recovery device to the exclusion of the other may
reporting to the magnetic fraction. Robichaud et al (2012), have a detrimental impact on overall gravity gold recovery.
also reported a magnetic fraction gold grade of 535 g/t, Modelling of flash flotation circuit performance as described
representing 12 per cent of the gold sent to the gold room earlier in this paper can give valuable insight into the optimal
trapped in the magnetic concentrate. circuit configuration.
Alternatives to magnetic treatment of the gravity concentrate Where the gravity gold is particularly coarse, spirals have
are treating the entire gravity circuit feed as is currently proven to be an effective technology and an example is
practiced at St Ives (Figure 5), or treating the tailings stream provided in Table 1 for the Newmont Tanami Operations
from the gravity concentrate treatment process. The latter has spiral circuit performance. This rougher-cleaner spiral and
the advantage of requiring much smaller equipment capacities, cleaner-cleaner scavenger shaking table circuit configuration
and due to the low-grade nature of the feed material the
was analysed by Wardell-Johnson and Bax (2012), and
resulting magnetic concentrate could potentially be of suitably
observed to achieve GRG recovery efficiencies at moderate
low gold grade to allow direct disposal of the magnetics. Over
to fine sizes exceeding the measured efficiencies for typical
time this would be expected to reduce the amount of magnetic
centrifugal concentrator installations.
material in the gravity circuit feed, by drawing down the
recirculating load of magnetics within the mill. There may be instances where intensive cyanidation of the
gravity concentrate is either undesirable for social reasons
Flow sheet selection or not viable due to environmental constraints. Under these
There are numerous examples of what has become today’s conditions the cleaning of the gravity concentrates by shaking
‘conventional’ gravity circuit, where centrifugal gravity tables and direct smelting is the logical alternative. A good
devices are used to remove gravity gold from the mill example of such a flow sheet designed for high recovery
recirculating load and the gravity concentrate is processed by efficiency is given by Robichaud, Belzil and Fullam (2012) for
intensive cyanidation and electrowinning prior to smelting the Goldex plant as shown in Figure 7.
to bullion. Such flow sheets have been demonstrated to be In the Goldex flow sheet primary gravity concentrates are
robust and suitable for processing a range of GRG contents tabled, with the table tailings scavenged by a 12 inch Knelson
and ore types. The KCGM circuit (Figure 6) is a good example prior to processing on a dedicated cleaner scavenger table.
of such a circuit. Importantly the feed to the 12 inch Knelson is prescreened
There are however instances when a variation on the to remove coarse gangue thereby enhancing the recovery
standard theme is required. This is frequently the case where efficiency of fine GRG present in the primary table tailings.
FIG 5 – St Ives gravity circuit flow sheet (SAG – semi-autogenous grinding).

FIG 6 – Kalgoorlie Consolidated Gold Mine gravity circuit flow sheet (SAG – semi-autogenous grinding).
Determination of gravity circuit feed rate accommodating design density and volumetric flow rates
feeding the separation devices, in addition to any fluidisation
Where the GRG content is fine it is clear that the mass split to
gravity circuit feed needs to be high in order to realise a high water requirements as with centrifugal concentrators. These
recovery rate of GRG. At the finer sizes the gold is less prone specifications are important to the optimal performance of
to remain in the recirculating load and more readily reduced the gravity concentrator, and equipment must be carefully
to non-GRG by further grinding. sized to ensure that these specifications can be met without
negatively impacting grinding and classification efficiency.
Where the GRG content is moderate-to-coarse it has been
observed that the gravity circuit recovery is less sensitive Concentrate treatment practices
to mass split due to the higher retention and slow grinding
rates facilitating the preservation of coarse GRG. However, Intensive cyanidation has become the technology of choice for
where the mass split to gravity is low, the recirculating gold processing gravity concentrates prior to smelting to bullion
grades tend to be higher, representing a greater risk of coarse due to the high gold recovery (Gray and Katsikaros, 1999;
gold carryover to cyclone overflow (COF). It is primarily Lethlean and Smith, 2000) of these devices. Both the Gekko
for this reason that while the gravity circuit performance In-Line Leach Reactor and the Consep Acacia Reactor are
will be comparable at lower feed rates, it is recommended popular and effective technologies for this purpose. In some
that high mass splits to gravity (40–50 per cent) are applied instances the reactor capacity has been under designed to the
where possible to manage the recirculating gold values. extent that the reactor becomes a bottleneck and limitation on
This is particularly important where substantial levels of gravity gold production, hence it is important to ensure this
coarse gold are present in the mill feed. Data that has been part of the circuit is not undersized.
compiled relating to recirculating gold grade ratios, defined In instances where very coarse gold particles are present in
by the cyclone underflow gold grade divided by the cyclone the ore it may become impractical to completely dissolve these
overflow gold grade, is shown in Figure 8. The high circulating particles during the available intensive cyanidation cycle time.
gold grades observed for the Jundee and Tanami operations In such instances it may be advisable to include a coarse gold
are of note, with both circuits recovering high levels of gold recovery step either ahead of intensive cyanidation, such as
by gravity (40–70 per cent) at moderate mass splits to gravity employed at the Teck-Pogo operation described by Konigsman
(nominally 10–20 per cent). Contrast this with the KCGM et al (2007), or following intensive cyanidation to recover
circuit which recovers 10–20 per cent of gold by gravity at a unleached particles. The coarse particles can be directly smelted
high mass split (40–50 per cent) and the Phoenix circuit which to avoid complicating the operation of the leach reactor.
recovers a similar level of gold by gravity to KCGM, albeit at Prior to the widespread adaptation of intensive cyanidation
a much lower mass split (<10 per cent) at the time of survey. for processing gravity concentrates, the use of tabling
to upgrade gravity concentrates prior to smelting was
Water balance commonplace. Tabling is still applied today in a range of
The impact of the water requirements of the gravity circuit applications, including upgrading bulk gravity concentrates
on the water balance for the grinding circuit should be given or scalping coarse gold prior to intensive cyanidation, and
specific consideration during design, such that the circuit remains the preferred technology in the event that intensive
is adequately designed to allow the equipment to operate cyanidation is not an option for political, environmental or
as intended. This is important both from the perspective of social reasons.

A Giblett et al
Gold recovery by tabling has been reported to vary widely CONCLUSIONS

from 20–90 per cent (Watson, 2006), and is particularly
Detailed guidance has been provided on performing and
challenged where the gravity gold is fine and non-GRG
mineralisation has a high specific gravity. Figure 9 compares interpreting reliable test work programs in support of gravity
gravity gold recoveries at sites employing intensive gold recovery circuit selection and design. A number of best
cyanidation or tabling to treat primary gravity concentrates, practice examples have been provided to support an effective
sourced from a 2012 industry survey performed by the design basis for the gravity gold recovery circuit, drawing
AMIRA P420E project team. The data indicates the dominance heavily from the experiences of numerous practitioners over
of intensive cyanidation and a higher average total gold many years and installations.
recovery associated with the method.
Scavenging of table tailings using centrifugal concentrators ACKNOWLEDGEMENTS
as described by Giblett et al (1998), Laplante (2000b) and
Robichaud et al (2012), can be successfully employed. Under The authors would like to acknowledge Newmont Mining
such conditions the use of screening to assist with the and Rob Dunne in particular for supporting this effort. The
efficiency of gravity separation in the gold room has been Curtin University Gold Technology Group and the AMIRA
demonstrated and is recommended best practice. P420 project team are also acknowledged for their efforts and
FIG 7 – Goldex grinding and gravity circuit flow sheet (SAG – semi-autogenous grinding).
FIG 8 – Recirculating gold ratios for gravity circuits.

FIG 9 – Total gold recovery by gravity as a function of concentrate treatment method.
substantial contribution to the collective knowledge base of Laplante, A R, Woodcock, F and Noaparast, M, 1995. Predicting
gravity gold separations. gravity separation gold recoveries, Minerals and Metallurgical
Processing, May, pp 74–79.
Lethlean, W and Smith, L, 2000. Leaching gravity concentrates
REFERENCES using the ACACIA Reactor, Randol Gold and Silver Forum,
Cloutt, B, 1995. Gravity concentration at WMC’s St Ives Gold Mines, Vancouver, 2000.
Randol Gold Forum.
McGrath, T D H, Staunton, W P, Byleveld, R and Eksteen, J J, 2016.
Fullam, M, 2010. Predicting the benefit of gravity recovery prior Application of an integrated liberation leach model for process
to flotation, in Proceedings Gravity Gold, Ballarat, pp 97–99 (The optimisation, in Proceedings 13th AusIMM Mill Operators’ Conference,
Australasian Institute of Mining and Metallurgy: Melbourne). pp 173–178 (The Australasian Institute of Mining and Metallurgy:
Giblett, A, 2011. Gravity gold concentration at Newmont Mining, Melbourne).
in Proceedings World Gold Conference 2011, pp 217–228 (Canadian Minerals Engineering Online, Plant Operation News, 2004. Knelson
Institute of Mining, Metallurgy and Petroleum: Montreal). Concentrators and ConSep ACACIA Work in Harmony, February
Giblett, A, Hillier, D, Parker, K and Ramsell, V, 2012. The impact of 2004 [online]. Available from: <http://www.min-eng.com/ops/
gravity concentration at Kalgoorlie Consolidated Gold Mines, in oz/26.html> [Accessed: 27 July 2017].
Proceedings 11th AusIMM Mill Operators Conference, Hobart (The Robichaud, F, Belzil, G and Fullam, M, 2012. Goldex, The GRG Jewel,
Australasian Institute of Mining and Metallurgy: Melbourne). in Proceedings 44th Canadian Mineral Processors Conference.
Giblett, A, Shuttleworth, J, Dunne, R and Fenton, K, 1998. An Staunton, W P and Bax, A R, 2010. Modelling the effect of gravity
evaluation to improve recovery in the gold room at the Telfer gold recovery on leaching performance, in Proceedings Gravity
gold mine, Randol Gold and Silver Forum, 1998. Gold, Ballarat, pp 155–158 (The Australasian Institute of Mining
Gray, A H and Katsikaros, N, 1999. The inline leach reactor – the and Metallurgy: Melbourne).
new art in intensive cyanidation of high grade centrifugal gold Staunton, W P and Bax, A R, 2011. Optimising circuit design for
concentrates, Randol Gold and Silver Forum, Denver, May. gravity gold recovery, in Proceedings World Gold Conference
Heroux, P and Hart, S, 2005. Gravity Circuit Performance at Cadia 2011, pp 229–240 (Canadian Institute of Mining, Metallurgy and
Valley Operations, Randol Forum, Perth. Petroleum: Montreal).
Konigsmann, E, Abols, J, Pyecha, J and Boudreau, T, 2007. The inline Van den Steen, P F, 2000. Gravity Gold Recovery at President Steyn,
leach reactor installation at Teck-Pogo Inc Pogo mine, Austmine Knelson Concentrators Africa, July 2000, [online]. Available
Conference, Perth, February. from: <http://www.flsmidth.com/~/media/PDF%20Files/
FLSmidth%20Knelson/Papers/20000803GravityGoldRecoveryat
Laplante, A R, 2000a. Ten do’s and don’ts of gold gravity recovery,
PresidentSteynMine.ashx> [Accessed: 27 July 2017].
Randol Gold and Silver Forum 2000, Vancouver.
Wardell-Johnson, G and Bax, A, 2012. Newmont Tanami (granites)
Laplante, A R, 2000b. Testing requirements and insight for gravity gold
gold mine gravity survey report (confidential to AMIRA P420D
circuit design, Randol Gold and Silver Forum 2000, Vancouver.
project sponsors) Curtin University Gold Technology Group.
Laplante, A R and Staunton, W P, 2005. Gravity Recovery of gold –
Watson, B, 2006. Gravity leaching with the ConSep ACACIA –
An overview of recent developments, in Proceedings International
predicting recovery benefits, Minerals Engineering International,
Symposium on the Treatment of Gold, Calgary.
Gravity Concentration.
Laplante, A R, Vincent, F, Noaparast, M, Woodcock, F, Boulep, A,
Dune, G and Robitaille, J, 1995. Predicting gold recovery by
gravity, in Proceedings International Mineral Processing Congress
XIX, San Francisco, pp 19–25.

Contents
Leach residue and pregnant liquor

separation – process and capital comparison
of counter-current decantation and counter-
current washing with vacuum filtration
R Klepper1 and P McCurdie2
ABSTRACT
Separation of leach residue suspended solids from pregnant liquor solution is an
important step in hydrometallurgy flow sheets. Alternative flow sheet options are
counter-current decantation (CCD) thickeners or counter-current washing (CCW)
horizontal belt vacuum filters. Extensive sedimentation and vacuum filtration
simulations have been completed by FLSmidth on leach discharge slurries to identify
equipment capacity and performance.
These data were used to compare capital expenditure (CAPEX) and operating
expenditure (OPEX) for CCD versus CCW alternatives. In addition, CAPEX and
OPEX for large horizontal belt vacuum filters with filtration areas greater than 250 m2
are compared to current state-of-the-art sizes of 150 m2. This paper presents this
information and discusses various advantages and disadvantages.
INTRODUCTION
Filtration with washing can recover soluble metals using less water than thickening
and washing. However, the overall filtration rate must be at a rate where the capital
cost and operating cost is less than thickening and washing in a CCD circuit.
Solid-liquid separation equipment selection for separating leach residue suspended
solids from pregnant liquor solution (PLS) in leach discharge slurry is controlled
predominately by the particle size distribution of the suspended solids. The quantity
of particles less than 37 µm diameter controls the rate of gravity separation and
permeability of the packed bed of suspended solids or filter cake formed during
filtration. The contents of the paper will identify the parameters and physical
properties that must be considered during feasibility evaluations and in the detailed
design of hydrometallurgical plants.
A review of the history of gravity thickeners and vacuum filtration is also presented
to better understand existing plant design compared to future plant designs now
being proposed. Examples of capital cost and operating costs will also be presented
as a basis for various options.
THICKENER HISTORY
One item has dominated thickener development – flocculation! Figure 1 (Klepper,
2009a) is a pictorial timeline of thickener development. The development of polymer
chemistry has led to the reduction of the diameter of thickeners and allowed
throughput to increase significantly. Originally thickener performance was controlled
by gravity and the settling rate of the smallest particle. These thickeners are defined
as conventional thickeners.
In the 1970s flocculants were added into thickeners at various locations and collected
the finest particles into larger groups or floccules increasing the gravitational settling
rate. Some design philosophies proposed the feed slurry exit the feedwell at a position
under suspended solids within the thickener to enhance capture of unflocculated
particles. These smaller diameter thickeners are defined as high rate thickeners (HRT).
In the 1980s knowledge of optimising the use of flocculants was learned along
1. Global Manager Hydromet, FLSmidth, with combining gravity and compression forces created by the height of flocculated
Midvale Utah, USA. Email: rkl-us@fls.com suspended solids within the thickener. These smaller and taller thickeners are defined
2. Filtration Process Manager, FLSmidth, as deep cone paste or high density thickeners (HDT) depending on the rheological
Midvale Utah, USA. Email: pmc-us@fls.com properties of the underflow slurry.
273
FIG 1 – Counter-current decantation (CCD) thickener development history timeline.
Optimising flocculation was achieved by dilution of the •• sidewall height is increased to create more compressive
feed slurry into the thickener feedwell to some optimum force, add slurry residence time for liquor release and
suspended solids concentration that allows flocculants to be creates resistance to short circuiting of liquor into the
evenly dispersed into the feed slurry. Another phenomenon underflow discharge
that occurs at the optimum feedwell suspended solids
•• rotating vertical members or pickets are used to alter the
concentration is larger particles become part of the floccules
acting as ballast. Consequences are a measurable maximum permeability of the slurry and compress floccules within
settling flux as identified in Figure 2 and a homogeneous the thickener allowing more rapid liquor release
particle size distribution within the thickener contents. •• rake geometry is changed to minimise cross-sectional
Additionally, computational fluid dynamics (CFD) area reducing resistance or torque development when
modelling of various feedwell designs, pioneered in the the rake rotates
AMIRA International (2011) project ‘P266 Improving •• tank floor geometry is changed to a steeper slope and
Thickener Technology’, identified feedwell geometries a discharge cylinder is included to allow discharge
that created more shear than other geometries. Shear or of a homogeneous slurry exhibiting non-Newtonian
turbulence breaks floccules in the feedwell resulting in lower rheological properties such as an apparent yield stress
settling velocities and lower suspended solids concentrations
of >300 Pa
in the underflow slurry. Figure 3 illustrates a CFD model
of the consequence of thickener feedwell design changes •• greater installed unit torque in the rake drive is used
minimising shear and floccule destruction. so the thickener can operate at higher suspended solids
Paste and HDT are also designed with physical features, concentrations in the underflow slurry without stopping.
which can be seen in Figure 4 (Klepper, 2009b), to enhance Figure 5 illustrates the impact on thickener diameter and
liquor release from the slurry within the thickener: CCD footprint at the same throughput for conventional, HRT
FIG 2 – Settling flux versus suspended solid concentration.
FIG 3 – CFD modelling of feedwell development.

Leach residue and pregnant liquor separation
FIG 4 – High density thickener features.
and HDT thickeners determined for a hydrometallurgical •• filter cake washing zone to displace the residual PLS
copper project. from the pores of the filter cake
In summary, HDT or paste thickeners can produce underflow •• filter cake dewatering or drying zone where air is drawn
slurries with higher suspended solids concentrations in through the pores of the filter cake
smaller volumes that can be controlled automatically. •• filter cake discharge zone before the sequence is repeated
with the next revolution.
VACUUM FILTER HISTORY Again, polymer chemistry played an important role in
the expansion of vacuum filtration capabilities leading into
Vacuum drum filters were initially used for solid-liquid
rubber belt technology advancements and as a result, the
separation and washing of various leach discharge slurries first horizontal belt vacuum filter (HBVF) was invented in
because soluble metal recovery with these can use much less the 1960s. The advantages of the HBVFs over drum filters
water than in CCD thickener circuits. However, the capacity are complete flexibility of the duration of all segments of
per filter is limited by the size (maximum area 140 m2) and the filtration cycle and greater capacity per filter. The rubber
geometry of the drum thus restricting the size of each filtration carrier belt and filter cloth rotate around two large rolls and
zone. Figure 6 illustrates a vacuum drum filter and maximum the time the filtration area is under vacuum is variable. The
size of filtration zones during one rotation is also listed below: feed slurry does not require mechanical suspension because
the slurry is poured on top of the filter cloth. Advances in
•• filter cake formation zone composed of leach residue
understanding the mechanical design of HBVF components
suspended solids captured on the filter cloth as the PLS
(Figure 7) such as rubber carrier belts has allowed the total
flows through filtration area to increase to a maximum of 250 m2/filter with
a carrier belt width of 6.2 m. The capacity per filter is therefore
increased.
There are five filtration zones in a HBVF filter cycle:
1. filter cake formation zone or time (θf) it takes to form a
given weight per area (w)
2. filter cake wash zone or time (θw) it takes for wash to
pass through the filter cake
3. filter cake dewatering or drying zone time (θd) it takes to
obtain minimum residual moisture content of filter cake
4. filter cake discharge zone or time (θdc) to discharge filter
FIG 5 – Comparison of thickener diameter by thickener type. cake from the filter cloth
FIG 6 – Drum filter and filtration cycle zones.

FIG 7 – Horizontal belt vacuum filter geometry.
5. filter cloth washing zone where residual suspended the filter cake thus creating an interfering thin layer of low
solids and/or precipitated solids are removed from the permeability. The cake permeability must also be maintained
filter cloth to maintain permeability. and not allowed to shrink and crack between cake formation
The overall filtration rate is the weight per area of the dry and cake wash or the cake liquor will bypass the filter cake
suspended solids in the filter cake (w) divided by the sum of pores in favour of going through the cracks.
filter cake form time (θf) plus the filter cake wash time (θw) Filter cloth permeability must be maintained by providing
plus filter cake dry time (θd) plus cake discharge time (θdc). continuous filter belt washing to remove suspended
w solids remaining on the belt after filter cake discharge. In
Overall filtration rate = addition, intermittent acid washing of the filter cloth with
i f + i w + id + idc
dilute hydrochloric acid must be provided to remove an
The overall filtration rate is controlled by several variables,
accumulation of precipitated solids deposited on the filter
which are illustrated in general in Figure 8:
cloth as a consequence of a change in solubility. The liquor
•• pressure differential or vacuum level immediately flowing through the filter cloth evaporates water
•• permeability of the filter cake or suspended solids due to the lower pressure on the vacuum side of the filter
particle size distribution cake. Deposition of small quantities of precipitated solids
•• permeability of the filter cloth onto the filter cloth is inevitable reducing the open area of the
filter cloth or permeability so that more pressure drop occurs
•• quantity of liquor to be removed from the feed slurry to
across the filter cloth than the filter cake thus reducing flow of
form the filter cake
liquor over time.
•• quantity of wash liquor used to recover soluble metals.
The quantity of pregnant liquor in the filter feed slurry must
Pressure differential is a variable created by the atmospheric be minimised by using a thickener upstream of the HBVF to
pressure at the plant elevation. At sea level there is a concentrate the suspended solids sufficiently so that the feed
theoretical 101 kPa pressure differential available that reduces slurry can be evenly distributed across the width of the filter.
to about 84 kPa at 1500 m elevation. The installed vacuum
The quantity of wash liquor can be optimised by using
pump capacity controls how close to theoretical pressure
CCW with multiple stages. Barren wash liquor is applied at
differential can be achieved. The quantity of air the vacuum
the back end of the HBVF passing through the filter cake and
pump must remove is controlled by the permeability of the
collected as filtrate and reapplied to the top of the filter cake
filter cake during the cake drying portion of the filter cycle.
repeatedly. The reason for CCW is the fact that the removal
The permeability of the filter cake can increase significantly
if the filter cake shrinks and cracks during the dry portion of of liquor containing soluble metals is a combination of
the filter cycle. displacement and diffusion of soluble metals from non-linear
channels inside the filter cake. The amount of diffusion is a
Filter cake permeability can be maximised by concentrating direct function of the quantity of flocculated fines or particles
the suspended solids in the feed slurry or by using flocculants less than 10 µm in the filter cake.
to prevent fine particles from accumulating on the top of
THICKENER COUNTER-CURRENT DECANTATION SCOPE OF SUPPLY

Now that a brief history and review of operating parameters
has been completed, the scope of supply associated with
the cost to install CCD thickeners will be discussed.
Figure 9 illustrates a typical CCD circuit used for a copper
hydrometallurgy flow sheet. Figure 10 illustrates a typical
piping and instrumentation diagram (P&ID) for a CCD
thickener.
The CCD circuit system includes multiple thickeners in
series including the following components:
•• thickener (including tank, feed system and rake
mechanism)
•• flocculant preparation and control of distribution system
to each thickener
•• flocculant dilution and distribution system into each
thickener feedwell
•• underflow pumps and control
•• overflow tank and pumps (optional versus gravity flow-
FIG 8 – Parameters controlling overall filtration rates. through the CCD circuit)

Flocculant Mix Wash

Leach Residue Mix Liquor
Low Grade PLS
High Grade PLS
CCDn
CCD2 Leach Residue
CCD1
FIG 9 – Counter-current decantation (CCD) circuit process flow diagram (PFD).
•• interstage mix tank and agitator print rather than the current state-of-the-art HDT with their
•• wash liquor supply. reduced footprint.
The dominant variables affecting the installed cost of any In hydrometallurgical applications, the next item affecting
capital cost is the materials of construction of the thickener
CCD thickener circuit are the number of stages and thickener
rakes and tank to provide corrosion resistance to the acidic
diameter. These two variables define the land area of the plant
PLS. The choices are suitable alloy metals or carbon steel
or footprint required for the equipment. Thickeners smaller
or concrete lined with one of several suitable non-metallic
than 50 m diameter can be designed as elevated tanks with
linings or coatings.
a bridge spanning the tank diameter with the rake drive
mechanism mounted on the bridge. Thickeners larger than Selecting the proper material of construction requires
defining the physical chemical conditions in the CCD circuit
50 m are on ground tanks with a central column supporting
while keeping in mind that due to washing, the conditions
the rake drive mechanism and a bridge spanning from the
are most severe in CCD1 and less severe in CCDn. This means
perimeter to the rake drive for operation and maintenance
there is potential to change materials of construction at CCD3
access. or CCD4 where the concentrations of corrosive components
Leak detection and secondary containment of thickener in the liquor are lower. The physical chemical variables are:
contents with an elevated tank is positive and relatively easy •• temperature
to design. However, leak detection or secondary containment
•• acid concentration (g/l H2SO4 or HCl)
of thickener contents for an on ground tank requires a design
solution specific to the materials of construction. •• pH
The smallest diameter thickeners available for a CCD circuit •• redox potential
should be based on optimised flocculation occurring at some •• ionic chemical composition (g/l Al+3, Ca, Cl-1, Co+2, Cu+2,
optimum suspended solids concentration of slurry into the Fe+2&+3, Mg+2, Mn+?, Ni+2, Si+6, Zn+2)
feedwell. The current state-of-the-art uses HDT to obtain the •• organic content (g/l).
smallest foot print. However, many major mining companies Temperature is a fundamental parameter to define suitable
go through a conservative risk versus current state-of-the-art alloy metals or non-metallic linings. Temperatures greater
analysis and may not know of differences between HRT and than about 90°C require consideration of both personnel
HDT. Because of this lack of understanding, consequence can safety and corrosion resistance. Personnel safety can be
be using older ‘proven technology’ in HRT with a larger foot achieved by installing a roof over the thickener to contain
FIG 10 – Counter-current decantation (CCD) thickener piping and instrumentation diagram (P&ID).

water and acid vapours and to prevent injury, insulating solution at a concentration of 0.5–2 wt per cent. Once the
tanks and pipe wherever contact is a possibility. Most process flocculant is totally dissolved, the flocculant solution can be
designs recover heat after leach and lower the temperature metered to each thickener using a variable speed positive
of the leach discharge slurry to the CCD circuit to 50–70°C. displacement pump at a dose sufficient to properly flocculate
However, if the temperature is 100–110°C then brick or the suspended solids.
fibreglass reinforced plastic lining of the thickener tanks The neat flocculant solution is viscous and to allow
are some options available. Brick acts as an insulator and is efficient blending of the flocculant with the feedwell slurry
actually placed over a rubber membrane that provides the
the viscosity must be reduced. Viscosity reduction of neat
corrosion protection.
flocculant solution is achieved by blending overflow liquor
Generally, at temperatures less than 90°C suitable non- with the neat flocculant solution in a static mixer. Overflow
metallic lining systems can be rubber sheets, fibreglass liquor is used rather than water to avoid diluting the pregnant
reinforced plastic panels, flaked glass filled resin coatings, liquor and to conserve freshwater usage. The overflow liquor
or thermal plastic sheets. The effectiveness of almost all can be pumped from within the thickener using a submersible
non-metallic lining systems is dependent on proper surface pump or the overflow liquor can be pumped external to the
preparation of the tank structural material and application thickener using a horizontal pump. Taking the dilution liquor
of the lining under environmental conditions such that the from within the thickener minimises the actual overflow rate
properly prepared surfaces do not degrade. and carryover of suspended solids in the overflow liquor.
By partially neutralising the residual acid in the leach Centrifugal underflow slurry pumps with variable
discharge liquor, some process designs alter the leach
frequency drives are typically used to control the inventory
discharge liquor chemistry favourably for less costly materials
or residence time of suspended solids in each thickener and
of construction to be used.
transfer the underflow slurry to the next CCD interstage
The choice of metallic materials of construction for thickener mix tank. The inventory of suspended solids is measured
rakes and tanks are stainless steels, duplex stainless steels, indirectly by measuring the hydrostatic pressure at the bottom
lean duplex stainless steels, super duplex stainless steels, of the thickener and/or by measuring the height of the bed
austenitic stainless steels and other specialty alloys. Selection of suspended solids. In some CCD circuits the plant water
should be based on measured corrosion rates at the physical balance requires minimising the quantity of PLS dilution and
chemical conditions existing in the PLS recovery system. double mechanical seals may be used rather than packing
A lesser cost but paramount to optimising the flocculation gland seals. The quantity of seal water can be reduced to zero
on minimising thickener diameter is the feed dilution system. if the quality and quantity of underflow slurry allows the use
Several feed dilution options exist with varying degrees of peristaltic hose pumps. The quality depends on rheology
of flexibility to maintain the optimum suspended solids and particle size distribution.
concentration in the feedwell slurry. There are:
Overflow liquor can be pumped using centrifugal pumps
•• self-diluting feedwells designed based on relatively with variable frequency drives or flow by gravity from each
constant differential densities of feed slurry, feedwell thickener into the interstage mix tank for blending with
slurry and overflow liquor producing a relatively underflow slurry. The plant topography and geotechnical
constant dilution rate properties define the cost of civil excavation for the CCD
•• systems that use kinetic energy of the feed slurry to circuit and whether gravity flow is feasible.
create an eductor affect drawing thickener overflow The fundamental design basis of a CCD circuit is mixing or
liquor into the feed slurry as the diluent at a relatively blending the liquor in the underflow slurry of the upstream
constant dilution rate CCD stage with the overflow or wash liquor from the
•• low-head pumps installed in the feed pipe with variable downstream CCD stage so a homogeneous soluble metal
speed drives that pump the appropriate volume of concentration is obtained. The interstage mix tank is where
overflow diluent into the feed slurry to obtain practically this occurs. The mixing task in the interstage mix tank is
any optimum suspended solids concentration in the defined as blending, but the blending efficiency or degree of
feedwell for all types of ore. homogeneity cannot be 100 per cent because there is always
Thickener diameter also is an important parameter affecting some short circuiting in a single tank. It is more difficult
the selection of the rake drive and the torque required to blend streams with great differences in rheology. HDT
to rotate the rakes. HDT will have the smallest diameter underflow slurry has greater suspended solids concentrations
compared to HRT or conventional thickeners. Torque can be exhibiting non-Newtonian rheological characteristics than
defined as T = K×D2 where K is unit torque factor (N/m) and produced by HRTs. Therefore the interstage mixing system
D is thickener rake diameter (m). Minimising the diameter must be designed to provide adequate shear to disperse
of the thickener means less torque is required to rotate the the underflow slurry into the overflow liquor so that and
rakes. HDT are designed to produce higher suspended solids efficiency or degree of homogeneity of about 95 per cent can
concentrations in the underflow slurry resulting in more be achieved.
resistance for the rakes to rotate through the slurry. For the
The wash liquor rate or wash ratio, defined as mass of wash
thickener rakes to continue to rotate in more dense underflow
liquor per mass of leach residue suspended solids, should be
slurry a drive with greater K value must be utilised than
controlled at a rate proportional to the suspended solids feed
would be necessary in a HRT.
rate into the CCD circuit. The wash liquor quality used in the
Flocculant preparation for CCD circuits is required to supply CCD circuit should have minimal dissolved pay metals to
sufficient flocculant for each thickener in the CCD circuit to ensure maximum recovery of the pay metals.
function correctly. The flocculant preparation system consists
of a storage area to store un-hydrated flocculant keeping it
dry prior to use. A conveying system is usually used to add HORIZONTAL BELT VACUUM FILTER SCOPE OF SUPPLY
a quantity of flocculant into an agitated tank of water so the The scope of supply of CCW vacuum filter systems will be
long flocculant molecules can totally dissolve into a neat discussed next.

Filtering leach discharge slurry requires the use of at least vacuum pan, a feed distribution box, wash liquor distribution
two filters in parallel to be able to maintain production during boxes and cloth cleaning mechanisms. The rubber carrier belt
maintenance or to be able to have adequate filter area to filter is driven by either one or two variable speed drives so the
total production. The filtration system includes multiple cake thickness or filtration rate can be controlled to match the
HBVFs in parallel and each HBVF includes the following physical/chemical characteristics of the solids and liquid in
components: the feed slurry.
•• filter (including rubber carrier belt with roller deck, filter A filter feed slurry thickener is used to reduce the liquor
cloth support rolls, vacuum pan with filtrate collection content minimising the time it takes to form the filter cake.
pipes and belt drive as seen in Figure 11) Filter cake form time is a major component of overall filtration
•• filter feed slurry thickener (required to minimise liquor rate. An additional benefit of thickening the feed slurry
concentration in leach slurry and minimise separation of is elimination or minimised segregation of fine particles
from coarse particles which would inevitably lower the
fine particles during filter cake formation)
permeability of the filter cake. For operation control it is
•• feed slurry storage tank to allow independent thickener very important to have a filter feed tank between thickener
control and distribution system underflow discharge and filter feed pump. The storage tank
•• flocculant preparation and distribution system (optional) provides surge capacity for the thickener underflow slurry
•• filtrate receivers and pumps (single wash or counter- discharge rate to be independent of the filter feed slurry rate.
current wash scheme) The storage tank also provides time to stop either piece of
equipment for short periods of time without stopping both or
•• vacuum pumps, seal water and moisture traps
even total plant production.
•• cloth wash pump for suspended solids removal
Filter feed slurries can also be flocculated also to eliminate
•• cloth wash pump for scale removal segregation of fine particles from coarse particles and so
•• filter cake discharge handling system. avoiding a reduction in filter cake permeability created by
HBVFs can be designed with filtration areas under vacuum the layer of interfering fine particles. However, the cost of
ranging from 11 m2 up to 250 m2 as illustrated in Figure 12. flocculant can produce an uneconomical OPEX.
The filter consists of a frame to mount numerous rollers The feed slurry distribution system is important as it is
that support, guide and drive the rubber carrier belt during necessary to have equal filter cake thickness across the width
filtration and return to the front of the filter. Additional rollers of the filter so wash liquor flows through the filter cake evenly
are used to support and guide the filter cloth at discharge and and does not short circuit through thinner filter cake areas
return to the front of the filter. Also attached to the frame is with less resistance.
a vacuum pan to collect filtrate and air, rubber wear belts Filtrate receivers as illustrated in Figure 13 collect filtrate
to produce a seal between the rubber carrier belt and the from the manifold connecting the filter cake formation zone,
FIG 11 – Horizontal belt vacuum filter (HBVF) components.
FIG 12 – Largest and smallest horizontal belt vacuum filter (HBVF).

FIG 13 – Horizontal belt vacuum filter (HBVF) layout.
filter cake wash zones and filter cake dry zone, and are used liquor system must use suitable materials of construction for
for preliminary separation of liquid from air. The filtrate from corrosion resistance. On an occupational health and safety
the filter cake form zone is PLS and the filtrate from the wash (OHS) basis and to prevent injury the wash spray nozzles are
zone is combined with the PLS or if there are multiple wash inside an enclosure that the filter cloth passes through.
stages can be reapplied to the filter cake. As an example large tonnage of leach residue require
Double mechanical seals are used on some filtrate pumps multiple filters. There are two rubber carrier belt widths
to minimise PLS dilution. These filtrate pumps require a (4.2 m and 6.5 m) available to determine the total number
barometric leg equal in hydrostatic height to the vacuum used of filters. When there are more than four filters required, a
in the filter to prevent flashing of PLS that may lead to seal savings of about 15 per cent lower CAPEX can be achieved by
damage or scale formation. In processes where the dissolved using fewer larger filters. Figure 14 illustrates a layout of eight
solids concentration is not near saturation in the filtrate and 170 m2 HBVF with 4.2 m rubber carrier belts providing a total
dilution of the PLS is not a concern, barometric legs are not of 1360 m2 of total filter area. Figure 15 illustrates a layout of
required as there are open filtrate pumps that operate at five 273 m2 HBVF’s with 6.5 m rubber carrier belts providing
reduced suction head. a total filtration area of 1365 m2.
Vacuum pumps evacuate air from within the filter creating It should be noted that HBVF development has kept pace
the pressure differential for flow of liquor through the filter with other minerals processing equipment development into
cloth and formed filter cake, flow of wash liquor through the larger size to take advantage of economy of scale.
filter cake and flow of air through the filter cake to remove
residual moisture or dry the filter cake. A moisture trap as CAPEX AND OPEX COMPARISON FOR CCD THICKENERS
illustrated in Figure 12 must be installed between the filtrate
receivers and the vacuum pump to ensure no process liquid
AND HORIZONTAL BELT VACUUM FILTERS
goes through the pump causing damage. Vacuum pumps The outlines below lists the items to consider for CAPEX and
require seal water for the shaft seal, temperature control and OPEX for a CCD thickener circuit and a HBVF installation.
to lubricate the rotors. Temperature of the PLS must also be
considered with regard to the amount of water evaporation CCD Thickeners CAPEX and OPEX
that will occur at the reduced pressures. Additional water 1. CAPEX
may be required to condense water vapour so the pumps 1.1. Civil site preparation
remove more non-condensable air rather than water vapour. 1.1.1. Sufficient area for all tanks
Cloth washing is required to maintain permeability of the 1.1.2. Potential terracing for gravity flow of overflow
filter cloth, prevent suspended solids from accumulating liquid from stage to stage
under the filter as the filter cloth returns to the front of the
filter and to keep the filter cloth support rolls clean so the 1.1.3. Spill containment ponds
cloth remains in the desired location. In an effort to maintain 1.2. Equipment
a water balance in the plant and conserve freshwater use, seal 1.2.1. CCD Thickener
water from the vacuum pumps can be used for cloth wash 1.2.1.1. Number of thickeners to recover pay
water and ultimately as filter cake wash water. Seal water metals
from the vacuum pumps can be collected in a tank. A pump is
1.2.1.2. Type and size of thickeners to produce
used to create pressure for the cloth wash water to go through
maximum underflow suspended solids
spray nozzles positioned across the width of the filter cloth on
concentration at plant production
both sides of the cloth after the filter cake has been discharged.
1.2.1.3. Materials of construction for submerged
In almost all hydrometallurgical applications scale
components and tank suitable for
forms over time on the individual fibres of the filter cloth
corrosion resistance and potential scale
progressively restricting the flow of liquor until ultimately the
removal
filter cloth is totally blinded or impermeable, restricting the
flow of liquor. At or before the filter cloth is ‘blinded’, the filter 1.2.1.4. Leak detection
must be taken out of service and a separate acid wash applied 1.2.1.5. Rake drive with sufficient torque
to dissolve the scale to restore the filter cloth to near original suitable for underflow slurry physical
permeability. This cloth wash liquor will contain sufficient properties, variable speed and torque
acid concentration to dissolve the scale therefore the wash transmitter

FIG 14 – 136 m2 total filtration area using eight 170 m2 filters 4.2 m wide.
FIG 15 – 1365 m2 total filtration area using five 273 m2 filters 6.5 m wide.
1.2.1.6. Rake lift to recover from abnormal 1.2.2.3. Underflow slurry pumps variable
operation with rake height transmitter speed, operating and standby
1.2.1.7. Feedwell slurry dilution system for 1.2.2.3.1. Mechanical seal water supply
efficient flocculation (low-head dilution system or peristaltic pumps
liquor pump optional) 1.2.2.3.2. Underflow slurry piping
1.2.1.8. Flocculant dilution system 1.2.2.3.3. Underflow slurry density
1.2.1.8.1. Dilution liquor pump and transmitter
shaft seal 1.2.2.3.4. Underflow slurry flow
1.2.1.8.2. Dilution liquor/flocculant transmitter
static mixer 1.2.2.3.5. Underflow slurry density
1.2.1.8.3. Diluted flocculant distribution control
piping and valves 1.2.2.3.6. Suspended solids inventory
1.2.1.9. Suspended solids bed level transmitter control
1.2.1.10. Suspended solids bed hydrostatic bed 1.2.2.4. Interstage mixing system tank and
mass transmitter agitator
1.2.2. CCD thickener ancillaries 1.2.2.5. Overflow liquor pump variable speed,
1.2.2.1. Flocculant supply system operating and standby (optional
1.2.2.1.1. Flocculant dry storage without gravity flow)
1.2.2.1.2. Flocculant dissolution tank 1.2.2.5.1. Overflow liquor pump sump
and mixer 1.2.2.5.2. Pump sump level transmitter
1.2.2.1.3. Flocculant solution storage 1.2.2.6. Interconnecting pipe and pipe rack steel
tank 1.2.2.7. Stairs and platforms between and
1.2.2.1.4. Flocculant metering pump around thickener tanks
1.2.2.1.5. Flocculant dose control 2. OPEX
1.2.2.2. Feed slurry mass measurement 2.1. Power
1.2.2.2.1. Feed slurry density 2.1.1. Rake drive motors
transmitter 2.1.2. Rake lift motors (intermittent and emergency
1.2.2.2.2. Feed slurry flow transmitter power)

2.1.3. Feedwell dilution liquor low-head pumps 1.2.4.2.4. Flocculant metering pump
(optional) 1.2.4.2.5. Flocculant dose control
2.1.4. Flocculant metering pumps 1.2.4.3. Feed slurry mass measurement
2.1.5. Flocculant dilution liquor pumps 1.2.4.3.1. Feed slurry density
2.1.6. Underflow slurry pumps transmitter one out of filter
2.1.7. Overflow liquor pumps (optional without feed storage tank
gravity flow) 1.2.4.3.2. Feed slurry flow transmitter
2.1.8. Interstage mixers one per filter
2.1.9. All controls 1.2.4.4. Vacuum system
2.2. Consumables 1.2.4.4.1. Vacuum pumps operating
2.2.1. Flocculant (100–400 g/t total) and standby
2.2.2. Process water 1.2.4.4.2. Moisture traps
2.2.2.1. Wash liquor 1.2.4.5. Filtrate system
2.2.2.2. Flocculant dissolution 1.2.4.5.1. Filtrate receives
2.2.2.3. Pump seal water 1.2.4.5.2. Filtrate pumps
2.2.2.3.1. Underflow slurry pump 1.2.4.6. Cloth wash water system to remove
suspended solids
2.2.2.3.2. Overflow liquor pump
(optional without gravity 1.2.4.6.1. Wash water storage tank
flow) 1.2.4.6.2. Wash water pump
2.2.2.3.3. Flocculant dilution pump 1.2.4.7. Cloth acid wash system to remove scale
2.3. Maintenance (intermittent)
2.3.1. Gearbox oil 1.2.4.7.1. Acid wash storage tank
2.3.1.1. Rake drive 1.2.4.7.2. Acid wash pump
2.3.1.2. Feedwell dilution liquor low-head 1.2.4.8. Interconnecting pipe
pump (optional) 1.2.4.9. Stairs and platforms between and
2.3.1.3. Interstage mixer around filters
2.3.2. Hydraulic oil and filters with hydraulic rake 1.2.4.10. Filter cake conveyer
drive (optional) 2. OPEX
2.3.3. Descaling periodically 2.1. Power
2.3.4. Lining or coating repair if damaged during 2.1.1. Filter feed slurry tank mixer
abnormal operation or descaling
2.1.2. Filter feed slurry pumps
2.3.5. Pumps and pump shaft seals
2.1.3. Flocculant metering pumps
Horizontal belt vacuum filters 2.1.4. Rubber drainage belt drive motors
1. CAPEX 2.1.5. Vacuum pumps
1.1. Civil Site Preparation to accommodate total filtration 2.1.6. Filtrate pumps
area 2.1.7. Cloth wash water pumps
1.1.1. Filtrate splash or drip containment 2.1.8. Acid wash pump (intermittent)
1.2. Equipment 2.1.9. Filter cake conveyors
1.2.1. Leach discharge or filter feed thickener 2.1.10. All controls
1.2.2. Filter feed slurry storage tank with mixer 2.2. Consumables
1.2.3. Horizontal belt vacuum filter 2.2.1. Flocculant (20–200 g/t range)
1.2.3.1. Number of filters to recover pay metals 2.2.2. Filter cloths (two per annum per filter)
1.2.3.2. Size of filters to produce minimum 2.2.3. Vacuum pan wear belts (two sets per filter per
liquor content at plant production annum)
1.2.3.3. Materials of construction for corrosion 2.2.4. Rubber drainage belts (one per ten years per
resistance during production and filter filter)
cloth acid wash to remove scale
2.2.5. Roller deck rollers and bearings
1.2.3.4. Cake thickness transmitter
2.3. Process water
1.2.4. Filter ancillaries
2.3.1. Vacuum pump seal water
1.2.4.1. Filter feed slurry pumps variable speed
2.3.2. Filtrate pump seal water
1.2.4.2. Flocculant supply system (optional)
2.3.3. Cloth wash water
1.2.4.2.1. Flocculant dry storage
2.3.4. Acid wash
1.2.4.2.2. Flocculant dissolution tank
and mixer 2.3.5. Vacuum pan wear belt lubricant water
1.2.4.2.3. Flocculant solution storage 2.4. Maintenance
tank 2.4.1. Bearing lubrication

2.4.2. Cloth replacement consider the range of physical chemical properties of leach
2.4.3. Vacuum pan wear belt replacement discharge slurry and select the appropriate solid/liquid
separation rates for thickeners and filter using the suitable
2.4.4. Roller and bearing replacement
materials of construction. Obtaining costs specific to the plant
2.4.5. Gearbox oil location for the CAPEX and OPEX items identified in the paper
2.4.5.1. Filter feed slurry tank mixer will provide a meaningful economic comparison to be made.
2.4.5.2. Rubber drainage belt drive
2.4.6. Rubber drainage belt replacement ACKNOWLEDGEMENTS
The comparison of the total CAPEX and OPEX for either The authors wish to acknowledge Jared Quilter for his
CCD thickeners and HBVF can be made using solid/liquid expertise in preparing some of the graphics used in this paper
separation rates and physical properties measured in lab scale and other peers within FLSmidth who have contributed
simulations. The most cost-effective pregnant liquor recovery expertise in preparing this paper.
option can then be selected.
REFERENCES
CONCLUSIONS AMIRA International, 2011. Improving thickener technology
[online]. Available from: <http://www.amira.com.au/>
Equipment development continues to keep pace with the
demand to process larger capacities in the hydrometallurgical Klepper, R P, 2009a. The evolution of thickeners, in Proceedings ALTA
process market to recover copper, cobalt, gold, nickel, silver, 2009 Nickel/Cobalt Conference, Perth, 25–27 May.
uranium and zinc. Competing process developments require Klepper, R P, 2009b. High density thickeners in CCD circuits: case
consideration of the feasibility to recovering PLS using either study, in Proceedings International Symposium, Hydrometallurgy
a CCD thickener circuit or a number of HBVFs. One must of Nickel and Cobalt 2009, 39th Annual Hydrometallurgical Meeting
(held in conjunction with) Nickel and Cobalt 2009, pp 195–207.

Contents
Filtration test work – extracting the

whole story for studies and design
G Bickert1 and B Länger2
ABSTRACT
Australia has developed and is developing and engineering a range of specialised
processing options for rare earth (RE) (Lynas, 2011; Lacaze, 2016; MacDonald and
Chalmers, 2014; Alkane Resources, 2016), zirconia (MacDonald and Chalmers, 2014),
lithium (Galaxy Resources, 2008; Tianqi Lithium, 2016), high purity alumina (Altech,
2016) and other special minerals and metals required in high-tech applications.
The processing of battery grade lithium, high purity alumina, zirconium and RE
minerals is based on complex leaching processes which require extensive liquor
clarification, solid liquid separation and solids washing. The filtration properties
for those leach liquors, residues and precipitates are not well understood and differ
drastically according to the process route applied. Thus, filtration test work is essential
for equipment selection, sizing and design. Similar types of slurries are filtered in
laterite nickel, TiO2 and other chemical oriented processes, while more experience is
available here as those processes are applied widely for several decades. The increasing
need for dry stacking of tailings requires advanced tailings filtration. Tailings differ
drastically in filtration properties which makes filtration test work essential. More
traditional mineral processes such as flotation commonly apply filtration. These and
other cases require proper evaluation of filtration test work in particular to evaluate
vacuum versus pressure filtration and the need for cake pressing and/or air blowing
for filter presses. Bench scale test work and/or pilot plant operation is common
to determine filtration rates, achievable cake moisture and wash efficiencies and
sometimes filtrate clarity. While those values are important, there is much more valid
information to be obtained by filtration test work.
This paper discusses the intricacy of filter selection and sizing based on driving
forces (vacuum and pressure air, compression by mechanical means and slurry)
and parameters (cake thickness, forces and times) and different processes utilised in
common filtration equipment. It will discuss filtration test work design, planning, data
gathering, processing, reporting, evaluation and filter sizing from recent examples of
bench and pilot scale filtration test work for the above applications, together with
methods of predicting filtration performance where test work is lacking. One focus
will be the realistic extrapolation of test work results towards different parameters
utilised in full-scale equipment and plants. The conclusion of the paper will contain a
checklist for filtration test work.
INTRODUCTION
Filtration (Chen, 1997; ANZAPLAN GmbH, 2011) is an important unit operation in
almost all metallurgical plants and flow sheets. Flotation concentrates, leach residues
and precipitated or crystallised valuable intermediates or products are dewatered by
vacuum or pressure filtration (Townsend, 2003). Clarifier overflows, pregnant liquor
or other filtrates are polished by candle or pressure leaf type filters.
The processing of lithium (Galaxy Resources, 2008; Savarovsky, 2000; Tianqi
Lithium, 2016; Enirgi Group Resources, 2016), high purity alumina (Altech, 2016),
zirconium (MacDonald and Chalmers, 2014; Alkane Resources, 2016) and RE
minerals (Lynas, 2011; Lacaze, 2016) for example is based on partly new and complex
leaching processes which require extensive liquor clarification, solid liquid separation
and solids washing. The filtration properties for those leach liquors, residues and
precipitates are not well understood and differ drastically according to process
1. MAusIMM(CP), Principal Consultant, route applied. Thus, filtration test work is essential for equipment selection, sizing
GBL Process Pty Ltd, Federal NSW 2480. and design. Similar slurries are filtered in laterite nickel, TiO2 and other chemical
Email: gb@gblprocess.com oriented processes, while more experience is available as those processes are applied
2. MAusIMM, Principal Consultant, widely. The increasing need for dry stacking of tailings requires tailings filtration
GBL Process Pty Ltd, Federal NSW 2480. with filtration properties differing drastically depending on tailings properties and
Email: bernd.laenger@germany.org.au requires filtration test work. More traditional mineral processing for example of base
285
metals commonly applies filtration of flotation concentrate. more systematic approach to gather and evaluate specific test
These and other cases require proper filtration test work in work data holistically by covering all the required variable
particular evaluation of the need for and quantity of cake process data for selecting, sizing and design.
pressing and/or air blowing. Filtration theory is useful in extrapolating test work results
Sizing and designing filtration equipment is more often towards different parameters utilised in full-scale equipment
done by limited bench scale test work rather than pilot and plants when test work data is lacking.
scale test work. The filtration test work might already be
done during concept studies, should be part of a feasibility CAKE FILTRATION THEORY
study (Anderson, 2000) and is sometimes only undertaken
Following is a summary of the filtration theory applicable for
during detail design in boom times for highly profitable and
filter testing, sizing and optimisation. It is based on Darcy’s
thus, fast-tracked mine developments. Test work results
Law (Anlauf and Stahl, 1996), the basic filtration theory
for filtration equipment selection and sizing often includes
describing the specific filtrate flow as a function of resistance,
only a few tests with limited results which do not reflect
cake height and liquid viscosity is given below but it is only
the full-scale filter conditions or else detailed pilot plant
valid for cake formation and cake washing. The second part
studies with extensive quantity of test results which partly
of this only considers the total cake resistance through specific
lack data gathering or focus only on recording filtration
cake resistance multiplied by cake thickness plus the filter
sizing parameters. Filtration test work is mostly done by
media resistance:
original equipment manufacturers (OEM) or metallurgical
laboratories. Limited scientific work (peer reviewed journal Op Op m3
papers) is published on filtration test work recoverable by q = dV 1 =
dt A rc n1H
=
n1 (rc H + Rm) = G (1)
m 2s
searching the keywords ‘filtration’ and ‘test work’ bringing
up a little information from all major scientific journals.
where:
Non-peer reviewed information, in particular company
information on web pages is plentiful, partly very useful and q is specific filtrate volume/m3/m2/s
partly advertising only. Kram (2002) discusses filtration test V is filtrate volume accumulated/m3
work do’s and don’ts and provides detailed requirements A is filter area/m2
regarding production plant and slurry (solids and liquids)
rc is specific cake resistance/1/m2
material properties important for filtration test work and
discusses filter test scale and appropriate filter types. H is cake height/m
The literature on filtration covering theory, filter selection, µ l is liquid viscosity/Pas
sizing and experimental work is extensive (Anlauf and ∆p is applied filtration pressure
Sorrentino, 2004; Anlauf and Stahl, 1996; Anthony et al, Rm is filter media resistance/1/m
2010; Svarovsky, 2000; Sutherland, 2005; Wakeman and
The cake resistance rc [1/m2] is used, while α [m/kg] is also
Tarleton, 1999, 2005a, 2005b; Chen, 1997) however practical
commonly used. Both are directly related according to:
test work often does not apply all this available knowledge.
Haekkinen introduces statistical design of experiments to
aw = acV = rc H = H 8 m B
1 (2)
reduce the number of filtration tests required to optimise A k
pumping, pressing (squeezing) and drying time and pressure where:
(Haekkinen et al, 2008). However, not many practical
w is mass of solids per cake height/kg/m
guidelines for reporting and utilising filtration test work data
for filter selection, sizing and designs are published. α is specific cake resistance/m/kg
Selecting, sizing and designing of filtration equipment Equation 1 can be directly used for cake washing ie a steady
for common duties, such as base metal concentrates is state one phase flow of fluid through a particulate bed (the
straightforward due to extensive experiences, plant examples filter cake). Only the liquid viscosity might change while the
and historical full-scale data, which resides with OEM’s, mother liquor is progressively replaced by wash liquor, if
engineering firms and consultants. More complicated filtration both liquids have different viscosities.
duties in hydrometallurgical flow sheets with leaching, As the cake height is not known during filtration test work
precipitation, crystallisation or difficult to filter tailings but the filtrate volume can be measured, the cake height is
filtration duties require case by case evaluation of well planned, replaced by the filtrate volume through utilising the filter feed
executed, evaluated and reported filtration test work. solids concentration κ which can be calculated from volume
Specifying filtration equipment for plant design often includes or weight concentration if solids and liquid density and cake
only limited data, filtration rate (dry solids production rate per porosity are known:
filter area in kg/m2/h) and required cake moisture. However,
l= H CV CW t1
this limited data is only suitable for equipment sizing if the V and l= = 6-@ (3)
parameter under which the filtration rate was determined are 1 - f - CV (1–f) tS–fCW t1
A
identical to full-scale parameters, including feed density, cake
thickness, driving force applied and others. where
This paper discusses the intricacy of filter selection and ε is cake porosity (pore volume/total volume)/-
sizing based on driving forces (vacuum and pressure air, Cw, CV is solids concentration/wt/wt, vol/vol
compression by mechanical means and slurry pressure) ρ is density of solids (s) or liquid (l)/kg/m3
and parameters (feed density, cake thickness, dewatering
forces and times) and different processes utilised in common For constant pressure filtration, which is the case for vacuum
filtration equipment. It provides guidelines and examples filtration, the Equation 3 results in:
for filtration test work design, planning, and data gathering n rl nR
t s
and processing on recent examples of bench and pilot scale = 1c V+ 1 m ; E
V 2A 2 Op A Op m3
filtration test work for various minerals. This should lead to a (4)

Filtration test work – extracting the whole story for studies and design
Thus, if filtration test work has recorded the filtrate difference and cake thickness) has to be determined
accumulated over time, plotting the time versus filtrate volume experimentally by one test recording the filtrate volume over
(or mass) over filtrate volume should give a linear curve. The time during drying. This could be the same test used for
specific cake resistance (or specific cake resistance multiplied determining cake and media resistance. For disc filters, the
by viscosity if latter is unknown) can be determined from drying time is almost fixed through disc filter speed and filter
the slope of the line and the filter media resistance from the design and thus, the correct ratio between cake formation time
intercept of the y-axis (EN ISO 6145–5:2010). Only one filter and desaturation time (about 0.8 to 2) needs to be applied.
test is required for this, which also could be used to determine
the desaturation kinetics.
VACUUM BELT, DISC, DRUM AND PAN FILTER TESTING
The filtration time for neglectable media resistance (Rm= 0)
Buchner funnel test work is common to simulate continuous
is a function of filtration parameters such as cake height,
vacuum filters such as horizontal vacuum belt, rotary vacuum
dewatering driving pressure, feed density and specific cake
disc, drum and pan filters. The Buchner funnel set-up could
resistance according to:
be of various designs and complexity. Bourgeois proposed
n1rc hc2 a standard benchmark filter test for vacuum filtration,
t= 6 s@ (5) simulating bottom-fed and top-fed filtration for coal, but the
2lOp
method was not adopted (Bourgeois et al, 1995). A detailed
‘European norm’ exists for the determination of filter cake
This formula also shows that the cake height increases resistance (EN ISO 6145–5:2010) which is partly considered
according to a square root proportionality with filtration time, but still, not well-known. This norm uses over-pressure
so that the specific solids throughput of continuous vacuum filtration (hyperbaric) for determining filtration parameters
filters is proportional to the square root of the filter speed. such as cake resistance as shown in Figure 1 (EN ISO 6145–
For rotary filters, where the cake formation ratio to cake 5:2010). Hyperbaric filtration is even used for sizing vacuum
drying ratio is constant, the rotation speed n is applicable filters (Ehrfeld et al, 2008) which assumes that filtration
and the solids throughput rate is dependent on filter speed is similar for similar pressure difference, be it vacuum or
n according to:
over-pressure. All these batch tests simulate the continuous
2lOpn 2fOpn vacuum filter by applying the filter stages across time (batch
kg
mo s \
n1rc
=
n1rc (1 - f - CV ; E (6) filtration), while the full-scale continuous filters have similar
s
stages across the position on the filter (continuous filtration).
Thus it is important to considering the different filter stages.
where:
n is filter rotation speed in 1/s Continuous vacuum filter stages
mo s is dry solids throughput/kg/s All continuous vacuum (and hyperbaric) filters apply the
Thus, plotting the specific solids throughput rate as function following stages while cake washing might or might not be
of square root of filter speed should provide a linear curve, necessary:
which is confirmed by Ehrfeld, Bott and Lageloh (2008).
The required filter area for cake formation can be calculated
from Equation 5 with the cake density according to:
mo s n1rc hc
A= 6m 2@ (7)
2lOpt s (1 - f)
While extensive work has been undertaken to predict
cake desaturation, this is not achieved unless extensive
experimentation and analyses are performed to determine
to determine their broad size distribution and complicated
microscopic structure (Anlauf and Sorrentino, 2004).
However, the change in cake saturation with time is
proportional to the desaturation driving force and the
cake height to the power of minus two. The driving force
for desaturation is the pressure difference between applied
pressure and the capillary entrance pressure at the particular
cake saturation.
Op - Opcap(S) Op - Opcap(S)
dS \
8 s B
. 1 (8)
dt (H + HcE) 2
(H) 2
where:
∆pcap(S) capillary entrance pressure at the particular cake
saturation in Pa
HcE equivalent cake height (for filter media)
With the saturation or moisture reducing over time, the
moisture drop gradually decreases to zero if no thermal
drying occurs.
As the cake desaturation cannot be sufficiently predicted
theoretically for practical filter duties, the achieved cake Fig 1 – Pressure filter used to determine specific cake and
moisture for various drying times (for applicable pressure filter media resistance (EN ISO 6145–5:2010).

•• cake formation – one phase (liquid = mother liquor) predict and optimise filtration. Nevertheless, it is only applied
flow-through the filter media to transform a slurry into in very limited practical filtration test work cases directly.
a filter cake
•• cake washing – one phase (liquid = wash liquor) flow- Example for filtration theory and filtration testing
through the filter cake to replace the mother liquor for vacuum and hyperbaric filtration
within the cake with wash liquor The above theoretical considerations will be used for
•• cake desaturation – two phase (liquid and gas) flow to extrapolating filtration and washing from an example of leach
replace the liquor within the cake with air to reduce its residue where only limited test work could be performed. The
cake moisture. This is also called deliquoring or drying. following graphs in Figure 2 show an example of applying this
theory for one vacuum filtration test, where the filtrate volume
It is important to record the transition from one stage
gathered over time is first plotted over time and then according
to the next during test work and use different theoretical
to Equation 4 as time over filtrate volume versus filtrate volume.
approaches for those stages during test work evaluation and The filtrate volume can be replaced if preferred by filtrate mass
for extrapolation. for known liquid density. In this example, the cake formation
time was not recorded and only one filtration test without cake
Common vacuum filter test work washing was undertaken while applied filtration parameters
Most commonly, vacuum filtration is tested directly by (vacuum pressure and cake thickness) were different from the
applying reasonable parameters for: anticipated full-scale parameters.
•• slurry feed density It is essential to record the cake formation time
•• cake formation time in ratio to cake desaturation time (transformation from slurry to cake) during bench scale
batch filtration test work and then apply filtration theory
•• cake thickness individually for cake formation and cake washing and
•• vacuum pressure applied. consider cake drying separately. This is particular important
Varying the ratio of cake formation to cake desaturation for testing of horizontal vacuum belt filter applications, as
time is limited for rotary vacuum filters due to their design the cake formation to cake desaturation ratio is not fixed but
of pick-up zone and drying zone and it is mostly fixed by depends on filter feed conditions and could vary drastically.
design and not adjustable during operation. In particular for This is not as critical for filtration test work done properly for
disc filters, this ratio is only changeable by reducing the cake rotary filter applications (ie applying the correct and constant
formation (pick-up) zone in length and thus in time. Here, ratio between cake formation and desaturation).
no difference is made between disc filters using textile cloth The cake formation time in above example was not recorded
or mesh or ceramic material as filter media as this applies to but it can roughly be estimated from the following graph
both. Drum filters have a slightly bigger ratio variation. In (Figure 3) which shows again the t/V over V curve. The
contrast, the ratio between cake formation and desaturation t/V curve is linear to the last red square data point (six data
develops freely on horizontal vacuum filters such as belt and points) so that the cake is formed between this value and the
pan filters and thus, is not a controllable design parameter. next, which is in this example between 70 and 120 s. Visual
observation during test work would have been significantly
The filtration test work result, specific solids throughput more accurate.
rate (kg/m2/h – commonly referred to as filtration rate) and
Here, very limited filtrate data points are available as the
cake moisture is valid only for those selected parameters.
filtrate was recorded manually. The first two data points have
Extrapolating the results for different feed densities, cake
to be neglected for data analysis, due to experimental start-
thicknesses or cake formation to cake desaturation time ratio
up errors. These start-up errors are typically experienced, and
is only valid through experience (and applying filtration
most likely due to, no instantaneous measurement of the filtrate
theory indirectly) or by directly applying the filtration theory.
being taken as it exits the filter media (for example where the
More commonly, the effect of those parameters is directly filtrate must be collected within a volumetric flask or on a
tested, thus requiring more test work. scale). The following graph demonstrates that by applying
The well-known filtration theory allows determining Equation 4 across different sets of data points (or excluding
filtration parameters such as the specific cake resistance (or certain data points on the fringes of cake formation) results
cake permeability) which then could be applied to extrapolate, in different linear approximations and thus, different values
A B
Fig 2 – Test work data (filtrate over time) plotted in (A) and as per filtration theory Equation 4 in (B) as t over V versus V.

Fig 3 – Test work data plotted as t over V versus V with linear approximations (according to Equation 4 to determine specific
cake and media resistance) depending on selection of data points shown with linear formula shown.
for specific cake resistance and filter media resistance. Thus, filter operation at different operating parameters or to predict
having electronic data recording in short intervals and cake washing kinetics as this could not be tested experimentally.
recording cake formation time is helpful and improves the
accuracy of the test work results and their application. Cake washing
However, with those parameters determined from plot Once the specific cake resistance and the filter media
above, the required filter area for cake formation can be resistance is determined from one filtration test, the required
universally determined for various vacuum pressures, cake filter area for cake washing depending on the wash liquor
thicknesses and feed solids concentrations. The following amount can be determined theoretically. However, if possible
Table 1 gives some examples the of required filter area it is still recommended to confirm cake washing kinetics by
depending on selected cake height and vacuum pressure by experiment, while it is essential to determine wash efficiency.
pure theoretical approach according to Equation 4. Theory allows the determination of the required filter area
for varied wash amounts and various wash stages such
The feed solids affect the required filter area according
as multistage counter current cake washing on horizontal
to Equation 5 with κ replaced through Equation 3. The
vacuum belt filters.
area increases with more diluted feed while the increase is
dependent on solid and liquid density and cake porosity. The Equation 2 can directly be used to calculate the filter
required filter area can be estimated for feed solids which have area required for cake washing. It depends on wash liquor
not been tested if those material and cake properties are known. flow, viscosity, pressure difference, cake height and filter
media resistance. The dependence for those parameters for
Applying filtration theory therefore eliminates the need
neglectable filter media resistance is similar as in the case of
to test various feed solids concentration, slurry volumes or
cake formation. Thus, the required filter area for cake washing
cake formation time, vacuum pressures and cake heights in
is twofold for half the pressure or double the cake thickness.
particular for cases where very limited slurry for testing is
available and can assist in optimising filtration parameters.
If sufficient slurry is available it is still recommended to
Cake desaturation
confirm that the filtration theory which is based on certain The cake desaturation cannot be sufficiently predicted
assumptions (in particular homogeneous filter cake and theoretically for practical filter duties, so that the achieved
no cake cracking) is valid across the parameters tested cake moisture for various drying times (for applicable
by undertaken filtration tests with the end points of the pressure difference and cake thickness) has to be determined
parameters altered. Filtration theory assumes a homogeneous experimentally by one test recording the filtrate over
time during drying. This could be the same test used for
filter cake. This is not valid if segregation occurs. If feed solids
determining cake and media resistance. For disc filters, the
concentration is reduced significantly and the slurry changes
drying time is almost fixed through disc filter speed and filter
in regards to sedimentation from hindered to swarm settling
design and thus, the correct ratio between cake formation time
(Bickert, 2010) then significant segregation suddenly could
and desaturation time (about 0.8 to 2) needs to be applied.
occur on top-fed filters due to gravity settling in particular for
long cake formation times. Following graph (Figure 4) shows the extrapolated moisture
drop over time during desaturation time, commencing after
Filtration theory allows also to extrapolate expected filtration
cake formation time.
performance beyond parameters tested or to predict full-scale
The appropriate drying time can be determined from
Table 1 Figure 4 with the knowledge of the required cake moisture.
Calculated required cake formation filter area for different For example is 60 s drying time sufficient to achieve 30 per cent
moisture while 220 s are required to achieve 28 per cent
combinations of vacuum pressure and cake height.
cake moisture. For sizing, additional limitations have to be
Cake height m 0.01 0.02 0.015 0.02 0.02 considered which are:
•• the drying time on rotary filters limitation due to limited
Vacuum pressure kPa 4000 2400 4000 3200 4000
ratio between cake formation and drying
Relative filter area required % 30% 100% 45% 75% 60% •• to provide sufficient filter length on top-fed (belt and
for cake formation pan) filters to avoid slurry flowing into cake discharge

Fig 4 – Test work data for moisture over desaturation time.
area and to allow for uneven cake distribution across the •• technical time for opening, cake discharge, cloth
filter width. washing, closing and other equipment requirements.
For cake desaturation, belt and pan filters utilise typically •• candle, leaf and Kelly type filters applying some of the
30 to 70 per cent of the filter area required for cake formation. above cycles
This should be above 10–15 per cent if multiple stage cake •• belt press filters and screw presses.
washing is done on the filter.
Filter presses are predominantly solids focused, producing
The total effective filter area required is the sum of filter a dry cake, while the other pressure filters are predominantly
areas for cake formation, washing and desaturation, all for a liquid focused, producing a clear filtrate. However, versatile
particular cake thickness, feed solids and pressure. hybrids such as the Fundabac® candle filter can apply all the
above filter press cycles excluding squeezing but with optional
PRESSURE FILTER, CHAMBER, MEMBRANE AND TOWER FILTER steam drying within the pressure vessel with automatic cake
PRESS AND CANDLE, LEAF AND KELLY FILTER TESTING discharge. Thus, filter presses are often specified by using a
‘filtration rate’ better expressed as specific solids throughput
Overview of filtration principles rate (kg/m2/h), while candle, leaf and Kelly type filters use
‘filtration rate’ which is better expressed as specific filtrate
Continuous pressure filters such as hyperbaric disc filters
throughput rate (m3/m2/h).
are treated similarly to continuous vacuum disc filters and
the above details apply with the appropriate pressure. Batch
operated filters have different duties, designs, cycles and
Common filter press test work
requirements. The pressure filtration occurs due to different Small-scale filter press test work (Mayer, Scholtyssek and
driving forces: Reiber, 2009) is commonly undertaken with a range of filter
cells of various sizes, with most OEM’s having their own
•• Slurry feeding under pressure (pumping) results in
design system. The important thing here is (Palmer, 2008) that
initial filtration and compaction of a cake which is
the laboratory filter reflects the design of the full-scale filter
completely saturated.
as closely as possible with same and variable chamber depth
•• Squeezing by mechanical means (‘membrane’ or better and drainage area. Palmer (2008) recommends at least 0.1 m
diaphragm bladder squeezing through fluid behind it diameter as the error increases below this. At bench scale,
(mostly either air or water) or flexible chambers through test work simulating tower filter presses should preferably
rubber seals and hydraulic plate compaction). The cake is be done with a horizontal chamber on bench scale while
still fully saturated but at lower moisture if compressible. chamber/membrane filter presses require a vertical chamber
•• Compressed air pressure replaces the liquid in the orientation with the latter including a pressing device such as
cake with air and thus, reduces the saturation and cake membrane or piston.
moisture, if the air pressure is sufficient to overcome the The laboratory filter press should be able to simulate all
capillary forces holding the water within the cake. process cycles, including filling/filtration by either pump or
•• Shear combined with mechanical pressure such as in pressurised slurry vessel, squeezing by membrane or other
belt press filters or screw presses results in cake moisture means, cake washing if required and cake compressed air
reduction for compressible cakes which benefit from blowing.
shearing. The available filter area should reflect full-size, either one-
Simplistically, the main pressure filters used in metallurgical sided or two sided, with the correct drainage area used for
plants are: squeezing and air blowing. One-sided laboratory test work for
•• filter presses (chamber, membrane, tower, tube presses) two sided filter presses simulates only half the cake thickness,
with following cycles being all or partly applied: while cake squeezing and, in particular, cake blowing is not
simulated properly.
•• feeding and filtration (also referred to as pumping)
Most commonly, pressure filtration is tested directly, by
•• squeezing by membrane pressing or other means (also
applying reasonable times and pressures for feeding and
referred to as pressing)
filtration (pumping), squeezing and compressed air drying.
•• cake washing (optional) However, the result is sometimes only reported as ‘filtration
•• cake drying by compressed air blowing rate’ meaning the specific solids throughput rate in kg/m2/h.

This throughput rate might include the technical time or not. the filtrate time in minutes and not the filtrate volume) for
Thus, the throughput rate cannot be used directly for filter three tests undertaken, varying only chamber thickness at 25,
press sizing, unless all individual full-scale filter press cycle 41 and 57 mm. This requires the recording of filtrate weight
times are the same as in the laboratory test work and the correct (or volume) for the duration of filtration.
technical time, which depends on the filter press design, is The tests show a linear increase over time for the initial
already correctly incorporated in the throughput rate. feeding, till the chamber is filled. Test T3 is parallel (indicating
Pilot scale filter presses are more truly scale-downed full- a similar specific cake resistance to the other tests) but is
scale filters with the pilot scale reflecting the OEM’s design parallelly shifted upwards. This indicates that the filter media
particulars such as horizontal plate orientation for tower filter resistance is larger for this test, so the test used either a less
presses. Thus, their scale-up is more straightforward and less permeable filter cloth or it was blocked, the latter being the
risky. case here. The thinner the chamber thickness, the earlier (at
Filter press sizing is based on two requirements: lower filtrate weight collected) does the slope of the curve
increase drastically, which indicates that the chamber is full.
1. sufficient filter area for filtering the slurry within the
used feeding cycle time, taking into account the feeding This is visible for test T2 in Figure 5 and allows optimising the
pump characteristics during chamber filling and the filling time, which is about 9 min in this example.
filtration characteristics (filtration rate) during filtration Equation 3 provides the relationship between filtration
2. sufficient cake volume that is discharged every cycle to time and cake (here chamber) thickness, so that the optimal
provide the required solids throughput, with taking the filtration time can be calculated from this for changes
dry cake density into account. in chamber thickness. Doubling the chamber thickness
increases the filtration time by four fold. Table 2 shows the
The second requirement is often the limiting factor for filter test conditions, applied filtration times and the optimised
sizing ie considering the filter press as a batch volume that is filter time for test T2 based on end of the linear curve in
discharged certain times an hour and dropping a certain cake Figure 5 above and approximating for tests T1 and T3. Those
volume and weight, resulting in the total solids throughput.
approximated values are shown in above Figure 5 by solid
It takes the total cycle time into account, while the filtration
symbols. The optimal filtration time of test T1 and T3 could
cycle is determined by the first requirement of filtration rate.
thus have approximated purely from filtration theory and so
Thus, proper reporting of filter press test work requires are quite similar to the actual optimum. The underestimation
all data, including filtrate gained during squeezing and air for the thinnest chamber (T1) is thought to be due to neglecting
blowing cycle, the times and pressures applied for feeding, the chamber filling time (prior to any filtrate flow).
squeezing and drying as well as the final cake weight, its wet
Thus, filtration theory can be used to estimate filtration
or dry cake density and its compression due to squeezing.
times if test work has not been performed for different
chamber thicknesses. Again this implies certain assumptions
Examples for filtration theory and filtration testing for filter presses so that if possible, test work should at least be used to confirm
An example of test work for filter presses to dry stack tailings the major trends.
is presented. The test work simulated the filter press process
of feeding, squeezing and air blowing for various chamber Table 2
thicknesses as well as with and without flocculant. Measuring Filter press test conditions and calculated optimal times.
the filtrate flow over time for all cycles is beneficial to extract
further information out of the test work, for the filtration, Test Unit T1 T2 T3
squeezing and drying kinetics which than could be used to
optimise cycle times theoretically. The assumption of constant Chamber thickness mm 25 41 57
filtration pressure is not valid during the complete feeding/ Feeding / filtration time (test) min 6 10 17.5
filtration period, as the feeding pressure is increased initially Optimal feeding / filtration time from min 4.5 9 17
to a maximum pressure. From then on the assumption is Figure 5
valid and the filtration theory for constant pressure discussed
previously is applicable. Figure 5 shows the plot of ‘t over m Optimal feeding / filtration time from min 3.3 17.4
versus m’ (in this case we use the filtrate weight in grams and theory (extrapolated from T2)
Fig 5 – Test work data for filtration time over filtrate weight versus filtrate weight for filter press laboratory scale test work at various chamber
thicknesses with and without flocculant. Presented is the theoretically approximated filtration time (CFT) from 41 to 25 and 57 mm.

After feeding/filtration, cake squeezing and air blowing and air blown at 6 bar for another six minutes. The feeding
are usually applied. For selecting and designing the most cycle produces a cake within the chamber of 15.3 per cent
suitable filter press it is important to know the efficiency of moisture which is reduced in moisture by squeezing only
the squeezing and the air blowing cycle for cake moisture to 14.9 per cent as the base metal concentrate is virtually
reduction. Thus, it is necessary to record the filtrate collected incompressible if compacted sufficiently during feeding.
(best over time) for all cycles including feeding/filtration, Longer squeezing times do not reduce the moisture further.
squeezing and the air blowing cycle. This allows the The main moisture reduction occurs during the air blowing
extrapolation backwards in time of cake moisture at any time with the reduction in moisture leveling off with increasing
during the filtration cycle. air blowing time, which is obvious from the larger
Figure 6 shows the cake moisture for the test with 41 mm downwards slope when starting cake blowing compared to
chamber during squeezing at two different pressures and when starting squeezing.
air blowing for this tailing material. Significant moisture
Thus, the effectiveness (moisture reduction over time) of
reduction is achieved with squeezing at 6 bar while increasing
cake squeezing and air blowing has to be tested experimentally
the squeeze pressure to 15 bar increases the desaturation
rate (larger slope downwards). Thus, the tailings filter cake while recording the filtrate over time and back-extrapolating
is compressible. Cake blowing in contrast does not lead to for the moisture provides moisture reduction kinetics and
significant moisture reduction and thus, is not considered allows to select optimal squeezing/drying time even if only a
useful after squeezing. Air blowing is only effective, if the air test with longer time was undertaken.
pressure is larger than the capillary entrance pressure within The filter press size (filter area, chamber thickness and thus,
the cake and 6 bar air pressure does obviously not exceed the cake volume) is selected and sized based on first selecting
capillary entrance pressure of this cake formed of ultra-fine the best suited chamber thickness based on the test work.
tailings with partly plate-shaped particles. Then follows the determination of the total process cycle
Such a plot as in Figure 6 allows the selection of the appropriate time (feeding/filtering, squeezing plus air blowing) plus the
squeezing time depending on cake moisture requirements. technical cycle time (opening, cake discharge, cloth wash if
Figure 7 is another example of the effect of squeezing and required and closing) for the particular filter press design
air blowing on the moisture of a fine base metal concentrate selected. After this the required total filter press volume can
filtered (fed) at up to 6 bar, squeezed for one minute at 15 bar be calculated from the solids duty, the total cycle time and
Fig 6 – Cake moisture as a function of filtration time (excluding feeding) of squeezing and air blowing.
Fig 7 – Cake moisture as a function of total filtration time for feeding, squeezing and air blowing.

the dry cake density while accounting for the cake volume particular, the shape of the filter element (leaf or candle) and
reduction during squeezing. the sealing of the filter cloth, both of which are important
for liquor polishing. For example, this is achieved in DrM’s
Common candle, pressure leaf and Kelly filter test work Thorough Scale Down (TSD), a laboratory candle type filter
Pressure filters utilising a pressure vessel filled with slurry simulating the DrM Fundabac® candle type filter (Chen,
and filtration elements such as candle, pressure leaf and Kelly 1997), which is suitable for liquor clarification, wet or dry cake
filters are used for liquor clarification of thickener overflow, discharge, cake washing, cake drying through compressed
filtrate from vacuum or pressure filters. Direct filtration of air and/or steam and automatic cake discharge by snap-
slurries in these pressure filters can be applied if thickening is blow (for dry cake discharge) or flush-back (for wet cake
problematic or cake production and liquor clarification within discharge). Figures 8 and 9 show the schematics and pictures
one filter is needed or beneficial. They are also fully enclosed of the candle design, bench scale and pilot scale filters.
and allow very corrosive applications with plastic internals. As the bench scale filter has the same design compared to
Those filters, in particular the candle type Fundabac® filter the full-scale filter, only smaller candle diameter and length,
which is extensively used in rare earth, battery grade lithium it reflects the design scale-down well and even cake discharge
plants, become more important for liquor clarity with high to properties (cake stickiness) can be tested. The laboratory
very high material purity and thus, requirements to remove scale filter provides a worst-case for cake discharge due to
impurities from pregnant leach liquors. Determining candle, the smaller candle and thus, lower cake weight and tighter
pressure leaf and Kelly filter sizing is quite similar to that for radii and so provides only a conservative estimation of
filter press sizing and is based on the same two requirements, appropriateness for automatic cake discharge.
only this time the feeding pumps usually do not provide a Similarly, pressure leaf filters should be ideally simulated by
restriction and the specific liquid filtration rate is relevant as pressure vessels with a leaf of similar design inside. However,
the main filter requirement is commonly production of clear filtration test work using either vacuum Buchner funnel or
filtrate. This might require precoat, body feed or the (usually pressure cells for pressure leaf and candle filter design is
experimental) selection of a suitable filter media. better than no test work and are reasonable for determination
Again, the bench scale pressure filter should reflect the of cloth selection and liquid filtration rate but they do not
design of the full-scale filter as closely as possible and, in reflect cake discharge properties.
Fig 8 – Schematic design of the Fundabac® candle filter (left) and candle design for full and bench scale with cake discharge principal (right).
Fig 9 – Picture of the bench scale Thorough Scale Down (TSD) candle filter in polyvinylidene difluoride (PVDF) (left),
cake pictures (centre top with candle and bottom cake only) and pilot scale Fundabac® unit (right).

Examples for filtration theory and filtration testing filter aid could be due to the less clear filtrate and thus, a cake
formed with less ultra-fines; ie higher permeability.
for candle filters, pressure leaf and Kelly filters
The main filtration within these filters for cake formation The filtration rate could be extrapolated for longer filtration
occurs due to slurry pumping into the pressure vessel times (and more filtrate cumulated) which is shown in the left-
containing the candles or leafs. The specific liquid filtration hand graph in Figure 11 for the test with precoat and body
rate during this filtration cycle is described by Equation 2. feed using Equation 4 and solving the quadratic equation.
If constant pressure is applied, Equation 3 is valid. This is This is useful if not enough liquor is available for test work or
the case for the second part of the filtration from when the if the cake formed in the test was not sufficiently thick to be
maximal feed pressure is reached. This relationship is shown discharged. According to filtration theory, it takes twice the
in Figure 10 for an example of Bayer liquor clarification time to form a cake that has double the thickness and generates
(Bickert and Schumacher, 2008). double the filtrate which is what is being experienced here. As
Applications of filtration for liquor clarification in nickel or per the examples above, the filtration can be extrapolated and
lithium (ANZPLAN, 2011) processing, precipitate removal in predicted for changed operational parameters from filtration
a particular zirconia process (MacDonald and Chalmers, 2014) theory which reduces the required number of tests required.
or lithium process (Galaxy Resources, 2008) or in rare earth The optimal cake thickness and overall cycle time can then be
flow sheets (SGS) partly require a precoat and/or body feed selected using additional cycle times for vessel filling, cloth
so the requirements for effective liquor clarification filtration cleaning, vessel emptying, cake washing, cake discharge and
cannot be predicted easily and test work is required. Figure 11 any other cycle steps.
shows an example for a precipitate removal step when using a
very tight filter cloth, using precoat filter aid or both a precoat
and body feed filter aid.
CONCLUSIONS (CHECKLIST FOR TEST WORK AND SIZING)
The following checklist summarises the various theoretical
Using precoat and body feed results in initially the highest
and practical evaluations and considerations discussed.
filtration rates, using only precoat shows initially average
filtration rates while the test without filter aid and a less
permeable filter cloth results in lower filtration rates from the General
start. The filtration rates are similar from a filtration time of There are a number of general rules that apply to all filtration
approximately 20 min. The higher values for the test without testing and the acquisition of design data as follows:
Fig 10 – Specific liquid filtration rate as a function of filtration time measured in lab scale, expressed by Equation 3. Typical plant flow and pressure is indicated.
Fig 11 – Specific liquid filtration rate as function of filtration time measured in laboratory scale without filter and with filter aid as precoat only
and precoat plus body feed (left) and Equation 3 applied for predicting filtration and cake formation rate for longer filtration times.

•• A true reflection of the full-scale filtration step is in combination with the cake moisture achieved and the
required covering all slurry, solids and liquid properties, cake thickness used, as the specific solids throughput
including but not limited to particle size, feed solids (filtration) rate drops significantly with increasing cake
content, temperature, residence time after chemical thickness and reduced cake moisture.
processes such as precipitation and crystallisation, •• Critical for filter press sizing is the total cycle time, the
entrapment of air from flotation and other items. total filter press chamber volume, the dry cake density
•• Bench or pilot scale filtration equipment should be – which is not the bulk density of the solids material –
similar to the full-scale equipment in design and (or wet cake density in combination with cake moisture)
process. Filter cells with cake squeezing and air blowing and the cake compression, so that test work (and data
for example should be used for sizing membrane filter sheets for press sizing) should report those together
presses with an air blowing cycle. with the specific solids throughput rate to determine the
•• Filtration theory can be used to extrapolate filtration required filter area.
results based on specific cake resistance and filter
media resistance where no test work covers those Pressure filtration by candle, leaf and Kelly filters
process parameters. It is important to consider all •• Cake formation is the main filter cycle and the sizing is
assumptions, apply the appropriate equation (constant similar as for filter presses.
pressure or constant volume filtration) and account for •• Cake discharge is important and needs to be tested in a
cake compressibility. scaled-down version of the full-scale filter.
•• The effect of flocculants, coagulants or filter aids used as •• Filtration theory is suitable to extrapolate filtration after
precoat or body feed has to be evaluated experimentally. the maximum filtration pressure is applied and constant
pressure filtration occurs.
Vacuum filtration and hyperbaric continuous pressure filtration
•• Cake washing and drying is possible in some filters and
•• It is essential to record the cake formation time and apply
needs to be tested in the same fashion as on full-scale,
different theoretical approaches to cake formation, cake
such as compressed air or steam drying as well as spray
washing and cake desaturation/drying.
washing or displacement washing.
•• Filtration theory for constant pressure is suitable to
predict filtration performance once the specific cake
resistance is measured for both cake formation and ACKNOWLEDGEMENTS
cake washing. In particular the effect of variation of The authors would like to thank all companies and
process parameters such as feed solids, cake thickness, organisations for which they could undertake filtration test
filter speed and vacuum pressure applied can be work in the past, which allowed us to gain, test and apply the
predicted without further test work. However, filtration presented considerations and understanding.
test work, if sufficient sample is available, should be
applied to confirm the material behaves according to the REFERENCES
assumptions made in filtration theory. Alkane Resources Ltd, 2016. The development of the Dubbo
•• True to the full-scale equipment, bench scale test work Zirconia project – a new strategic source of critical metals, 5th
for rotary filters such as disc and drum filters require Annual Cleantech and Technology Metals Summit, Toronto [online].
fixed ratios between cake formation and drying, while Available from: <http://www.alkane.com.au/images/pdf/
the cake formation time on top-fed filters such as belt asx/2016/20160511.pdf> [Accessed: 26 July 2017].
and pan filters is dependent on slurry properties. Altech Chemicals Limited, n/d. Altech’s Processing Technology
[online]. Available from: <http://www.altechchemicals.com/
Pressure filtration by filter presses high-purity-alumina-hpa-project#processing-technology>
•• Filtration test work data for filter presses is required [Accessed: 26 July 2017].
for general equipment sizing due to the batch nature Anderson, W, 2000. The scheduling, costing and importance of
of the operation of these presses thus needing data on metallurgical testwork programs, in Process Plant Feasibility
different cycle times in regards to feeding and filtration, Studies, MINPREX 2000 Melbourne, 11–13 September, pp 79–89.
squeezing, cake washing, cake blowing and technical Anlauf, H and Sorrentino, J A, 2004. Laboratory settling analysis to
times such as opening, closing, cloth washing etc. predict separation in solid bowl centrifuges for ultrafine particles,
Essential are the detailed cycle times and pressures Filtration, 10(4):266–276.
applied in each of these steps of the press operation. Anlauf, H and Stahl, W, 1996. Dewatering of filter cake by vacuum,
•• ‘Filtration rate’ or specific solids throughput rate can be pressure and pressure/vacuum filtration, 4th World Filtration
either reported for: Congress, Orlando, April, Poster Session 5.1–5.8.
•• the cake formation time only Anthony, D, Stickland, A D, Kretser, R G and de Scales, P J, 2010.
One-dimensional model of vacuum filtration of compressible
•• incorporating the process times (feeding, squeezing,
flocculated suspensions, AIChE Journal, 56(10):2622–2631.
cake blowing) only
ANZAPLAN GmbH, 2011. Conversion of spodumene to lithium
•• the complete full-scale filter times including technical
chemicals [online]. Available from: <http://www.anzaplan.
times (longer feeding due to pump capacity restrictions, com/strategic-minerals/lithium/download/> [Accessed: 1
full-scale squeezing and cake blowing times and non- March 2011].
process/technical times).
Bickert, G, 2010. The influence of particle collective characteristics
•• Thus it is important to specify how the value was on cake filtration results, Chem Eng Technol, 27(10):1080–1084.
derived (and on which basis) and it is best to provide
Bickert, G and Schumacher, I, 2008. Effective bayer liquor clarification
the cycle time split for each of these activities.
– Retrofit or new with DrM Fundabac® filter internals, in
•• Cake thickness and final cake moisture effect filtration Proceedings 8th International Alumina Quality Workshop 2008, 7–12
cycles and thus, filtration rate figures should be stated September, Darwin, pp 130–135.

Bourgeois, F, Wightman, E, Clarkson, C Z, Rui, Z, Hilden, M and MacDonald, A and Chalmers, I, 2014. A new source of zirconium
Davis, J J, 1995. A new benchmark single leaf filtration test for fine materials and other critical metals and oxides – Alkane Resources
coal dewatering, in Proceedings Seventh Australian Coal Preparation [online], presented at Industrial Minerals International Congress
Conference (ed: J Smitham), Paper D2, pp 177–199. and Exhibition, Vancouver. Available from: <http://www.
alkane.com.au/images/pdf/presentations/20140402.pdf>
Chen, W, 1997. Solid-liquid separation by filtration, Chemical
Engineering, pp 66–72.
Mayer, E, Scholtyssek, P L and Rieber, R S, 2009. A new laboratory
Ehrfeld, E, Bott, R and Lageloh, T, 2008. Layout of rotary filters on
device for both standard recess and membrane filter press
the basis of laboratory results, in Proceedings 10th World Filtration
dewatering, AFSS 22nd Tech Conf and Expo, Minneapolis, USA,
Congress, 14–18 April, Leipzig, pp 190–195.
4–7 May.
Enirgi Group Resources, 2016. Watch Our Lithium Technology
Palmer, J, 2008. Study of the scalability of pressure filtration in
Video [online]. Available from: <https://www.youtube.com/
pilot and bench scale test equipment, in Proceedings 10th World
watch?v=jPP1MopWx0k> [Accessed: 26 July 2017].
Filtration Congress, 14–18 April, Leipzig, pp 86–90.
EN ISO, 6145–5:2010. Filtering properties of suspensions – The
Savarovsky, L, 2000. Solid-Liquid Separation, 464 p (Butterworth
determination of filter cake resistance, VDI-Gesellschaft
Heinemann: Oxford).
Verfahrenstechnik und Chemieingenieurwesen, Dusseldorf,
Germany. SGS, 2013. Rare Earth Ore Processing, T3 SGS 302, 10-2013 [online].
Available from: <http://www.sgsgroup.com.ar/~/media/
Galaxy Resources, 2008. Flow diagram – Jiangsu lithium carbonate
Global/Documents/Flyers%20and%20Leaflets/SGS-MIN-
plant [online]. Available from: <https://www.scribd.com/
WA368-Rare-Earth-Ore-Processing-EN-11.pdf> [Accessed: 30
document/237275051/Jiangsu-Lithium-Carbonate-Plant-docx>
July 2017].
Sutherland, K S, 2005. Solid/Liquid Separation Equipment (Wiley-VCH:
Haekkinen, A, Huhtanen, M, Ekberg, B and Kalls, J, 2008. Utilization
Weinheim).
of statistical design of experiments for improving the efficiency of
test filtration tasks, in Proceedings 10th World Filtration Congress, Tianqi Lithium, 2016. Tianqi Lithium Australia receives board
14–18 April, Leipzig, pp 81–85. approval to construct $400 million lithium plant in Western
Australia [online]. Available from: <http://www.tianqilithium.
Kram, T, 2002. Filtration Testwork DOs and DON’Ts, Outokumpu’s
com/en/base-detail.aspx?ContentID=6&t=59> [Accessed: 26
electronic newsletter Output, Issue 4, December 2002 pp 3–6.
July 2017].
Lynas, 2011. Rare earth we touch them every day, Investor
Townsend, I, 2003. Automatic pressure filtration in mining and
Presentation March [online]. Available from: <http://www.
metallurgy, Minerals Engineering, 16(2):165–173.
lynascorp.com/Shared Documents/Investors and media/
Reporting Centre/Presentations/2011/Investor_Presentation_ Wakeman, R J and Tarleton, E S, 1999. Filtration: equipment selection,
May_2011.pdf> [Accessed: 26 July 2017]. modelling and process simulation, Elsevier Advanced Technology,
Oxford, UK 1st edition.
Lacaze, A, 2016. Breaking the barriers, the Lynas story, Diggers
and Dealers Mining Forum [online]. Available from: <https:// Wakeman, R J and Tarleton, E S, 2005a. Solid/liquid separation:
www.lynascorp.com/Shared Documents/Investors and media/ principles of industrial filtration (Elsevier: Oxford).
Announcements and media/2016/Diggers and Dealers Lynas
Wakeman, R J and Tarleton, E S, 2005b. Solid/liquid separation:
Presentation 020816 ASX 1578169.pdf> [Accessed: 26 July 2017].
scale-up of industrial filtration (Elsevier: Oxford).

Contents
Considerations for effective

gold process development
A Giblett1, D Appelhans2 and R Dunne3
ABSTRACT
There are several hurdles that must be negotiated to design an effective gold ore
treatment facility and to ensure the basis of project economics are valid and can be
realised in practice. Selecting samples based on geometallurgical context, performing
and interpreting test work relevant to the selected process flow sheet are essential
considerations towards achieving this aim. This paper discusses recent experiences in
process development for free milling gold ores, considering frequently encountered
challenges in these areas.
INTRODUCTION
Free milling gold ores are a reasonably straightforward proposition for the process
engineer when it comes to designing a process plant for metal recovery. Since the
widespread adaptation of cyanide-in-pulp technology in the 1980s the processes
applied in gold recovery flow sheets have remained largely unchanged. The
technology essentially consists of grind-gravity-leach, with gold recovery performed
by Merrill Crowe precipitation, carbon-in-pulp or carbon-in-leach adsorption and
elution. Ore characterisation testing still commonly revolves around the Bond ball mill
work index and bottle roll leach tests, while numerous ‘newer’ tests exist for assessing
semi-autogenous grinding (SAG) mill amenability and the gravity recoverable gold
(GRG) test has become a standard basis for gravity recovery assessments.
Despite the relative simplicity of the task of developing a design basis for a gold
process plant there are several important considerations that will influence the success
of the endeavour, as measured by the treatment plants ability to meet ‘nameplate’
(life-of-mine average) throughput and metal recovery expectations. Recent project
development experiences have highlighted several areas where the process would
benefit from more consistent application of best practices. A number of these practices
associated with the selection of samples, determining the optimum number of samples,
developing the design basis and throughput estimates for the comminution circuit,
application and interpretation of the GRG test and the interpretation of metallurgical
test performance data are discussed in detail in this paper.
SAMPLE SELECTION
During the earliest stages of project development, it is necessary to select samples
based on the best classification of the resource available at the time. This may initially
be as simple as developing a matrix of ore grades and major ore types (individual
deposits or major lithology). As the geological definition of the resource improves,
metallurgical domains should be classified based on major lithology and alteration
zones. The number of metallurgical domains may increase or decrease by consolidation
as the evaluation progresses, based on the metallurgical characteristics observed in
1. FAusIMM, Senior Technical Advisor – Mineral
testing, or the basis of domain definition may be completely reworked to better reflect
Processing, Newmont Mining Corporation, metallurgical properties of the ore. The number of samples required for each domain
Englewood CO 80112, USA. is determined by the current level of study, the distribution of mill feed tonnes or
Email: aidan.giblett@newmont.com contained metal in the domains, and the nature of the deposit. Deposit styles known
2. Manager Process/Metallurgy, Newmont to reflect more challenging metallurgy (eg polymetallics) should be assigned a higher
Mining Corporation, Englewood CO 80112. density of metallurgical samples per tonne of ore to ensure metallurgical variability
Email: dennis.appelhans@newmont.com
is fully addressed. The focus of metallurgical characterisation will further vary from
3. MAusIMM, Metallurgical Manager Technical
Services, Newmont Mining Corporation,
high level baseline assessments to support early stage process design, to seeking more
Englewood CO 80112, USA. granularity in support of life-of-mine planning, in turn influencing the number of
Email: robert.dunne@newmont.com samples required at any given stage of study.
297
COMMINUTION PROCESS DEVELOPMENT average hardness values for each ore domain. This analysis
can also indicate when reclassification of metallurgical ore
There are several important considerations that influence
types is required, when excessive variability in comminution
the selection of the comminution circuit flow sheet and
properties is observed.
major equipment sizes, including the type and number of
characterisation tests performed, extent of geometallurgical In one recent project only four drop weight index (DWi)
classification of the resource, methods for computing values for primary ore were available as the basis for a SAG
comminution power requirements and the design basis for mill sizing, which needed to be selected to meet the project
major equipment sizes. The following discussion provides a development schedule. These DWi values averaged 6.96 kWh/
perspective on how to approach each of these concerns based m3 with a standard deviation of 2.86, calculating to a 90 per cent
on recent experiences in project development and operations. confidence level of ±3.4 kWh/m3 on the average drop weight
Alternative approaches certainly exist and in many cases are value, or 48 per cent relative. This represents a wide band
well documented; however, the discussion in this paper will of confidence and in most cases would be considered an
focus on an approach that has been determined to be effective unacceptable design basis. Few enlightened project managers
from firsthand experience. would accept an error margin of ±48 per cent in their project
revenue calculations, and this level of error is inconsistent with
In most instances for early stage projects the samples
the requirements of all but the earliest of scoping studies. To
available for comminution test work are determined based on
reduce this confidence range to ±1 kWh/m3, or 14 per cent
the number of drill intersections required for metal recovery
relative, another 45 samples would be required based on
testing. To preserve spatial appreciation of comminution
the observed standard deviation, utilising equation 3.15
characteristics it is preferred not to composite across different
from Napier-Munn (2015). For reference, a target 90 per cent
drill intervals and to keep the comminution tests relevant to
confidence interval of ±10 per cent relative for the DWi is
discrete metallurgical domains (defined by alteration and
defined by the authors a functional indication of a well-defined
lithology). Consequently, comminution samples must be
data set. Simple statistical analysis of confidence limits was
identified prior to any compositing of samples for metallurgical
therefore practically applied to define the level of uncertainty
recovery test work.
in the design assumptions and to provide clear guidance on
Having selected the samples based on the best available how much additional work should be undertaken.
geometallurgical classification of the resource, and having
In this example, another 22 samples were tested and
access to as many individual drill sections as are available
interpreted based on classification of the resource into six
for recovery test work, Newmont then applies a standard
major ore domains. By this more appropriate geometallurgical
approach to determine comminution characteristics. All
classification of the resource the variability in DWi was
samples are subjected to the SMC Test® (Morrell, 2004), Bond
substantially constrained. This in combination with the
ball mill work index and Bond abrasion index determinations.
additional tests saw the confidence limits in the average
More specialised testing to determine power requirements
DWi reduce to 10–25 per cent by domain. Clearly this
for stirred milling and high pressure grinding roll (HPGR)
effort substantially improved the reliability of subsequent
equipment will follow as and if required. This information
calculations of plant throughput rates that are a critical
permits a sample-by-sample power-based analysis of
component of project economics. The reliability of the
comminution power requirements based on the Morrell
same project economics would be (and in fact were) further
methodology (2009) for conventional comminution circuit
improved by performing the additional tests required to
flow sheets – SAG/ball mill circuits, with or without pebble
reduce the confidence limits on the average DWi to less than
crushing (SAB/C), three stage crush/ball milling, HPGR/
15 per cent as previously calculated.
ball milling. This has been demonstrated to be a robust
and flexible methodology to support flow sheet trade-offs, The discussion of confidence limits around hardness
equipment sizing and mine planning efforts. The use of the parameters, and similarly metal recovery estimates, needs
Morrell method (2006) for SAB/C circuits allows reliable to be put into perspective by considering the overall impact
estimation of the required power split between SAG and ball on project economics. How much difference does it make to
milling, and can also be coupled with population balance the confidence in the metal production plan estimates if the
modelling to ensure pebble crushing circuits are adequately confidence limits on a key hardness parameter such as DWi
designed. This and other comparable methods derived from are reduced?
extensive industrial benchmarking can therefore be applied to There is clearly a balance between too much test work and
avoid common constraints on grinding circuit capacity. not enough, as even though the costs of test work are small
compared to the costs of an underperforming process plant;
Interpretation of comminution test work data at some point the costs and schedule impacts of testing will
Armed with sufficient results it is possible to perform a outweigh the benefits. Part of the solution must always be better
statistical analysis to determine if the extent of testing geometallurgical classification of the resource. As was the case
performed is adequate to support the required level of in this example, by constraining variability to well-defined
confidence in the project estimates. This is contingent on metallurgical domains, the number of samples required for a
having enough samples in the first place to reliably estimate given level of confidence can be substantially reduced.
the variability in key hardness parameters, as performing While statistical analysis can be very useful to define
statistical analysis on small data sets has the potential to additional testing requirements, the analysis of comminution
misrepresent the inherent variability in the ore domain. There properties should not be limited to desktop mathematical
is also a need to ensure the data represented by the tests has study. The concept of confidence limits can be misleading
some relevance to the distribution of tonnes in the resource. when applied to large data sets that display significant
Based on the guidance of Morrell (2011) a critical requirement variability, and in such instances will give a false indication of
is not less than ten samples per ore type to form the basis of a a well behaved data set. Visual analysis of the distribution of
meaningful assessment. Determination of confidence limits on the data can be useful, as in Figure 1, to define bimodal data
such data sets can be useful to determine if additional sampling sets that may display acceptable statistical properties based
is required to achieve the desired level of confidence in the on a large testing database. For the data set shown in Figure 1,

Considerations for effective gold process development
Fig 1 – Sample distribution of SMC Test® results.
17 SMC Tests® gave an average DWi of 9.3 kWh/m3 with a However, the data included three very similar results (5.0,
90 per cent confidence interval of 8 per cent relative. In terms 5.7 and 5.9 kWh/t) and one high result (11.2 kWh/t). Two
of raw numbers this indicates a well-defined data set; however approaches to sizing the mill were proposed: take the average
the visual interpretation of the results as a distribution of the data, or use the 85th percentile of the data (8.8 kWh/t)
demonstrates bimodal properties, and a concentration of to recognise the risk of the hard component of the deposit
results around the drop weight 10–11 kWh/m3 range. As a being under sampled. Perhaps with the popular motivation
result, more investigation is required to determine how the of minimising the capital spend associated with a larger mill,
resource tonnes are distributed between competent and several theories were proposed to support the exclusion of
very competent material, to avoid misrepresentation of the the high DWi value from consideration. Two assertions were
resource properties by consolidating a wide spread of results ‘it’s an outlier’ and ‘that material is outside the current pit
into a single design value. design’. Indeed, the high value would have failed a statistical
Comminution data must always be considered in spatial test of outliers as described by Napier-Munn (2015), which
terms, to confirm the data set is truly representative of is a useful demonstration of the problems with performing
the resource in spatial context. Ensuring appropriate statistical analysis on very small data sets. As the project
geometallurgical classification of the resource, based on a developed under favourable metal prices the ore represented
database providing good spatial coverage of the deposit, is by the high DWi rapidly fell within an optimised pit design,
an essential component of effective process design. This will underlining the importance of considering the characteristics
allow practical optimisation of the project design and the of subgrade material in all such assessments with a view to
mining schedule. future processing.
Predictions of grinding media consumption from Bond An example is provided in Figure 2 where initial process
abrasion index test results should consider the issues observed development test work was sufficient to predict an increase in
with the accuracy of wear predictions from the models of Bond ore hardness after six years of mine life although insufficient
(1961) as discussed by Radziszewski (2002) and Giblett and to accurately estimate the magnitude of this increase.
Seidel (2011). More current models for predicting grinding Subsequently plant throughput decreased well below design
media wear in ball mills have been proposed by Giblett and levels in later years of the mine life. Adequate quantification
Seidel (2011), Guzman and Rabanal (2014) and Doll, Barratt of ore properties by hardness testing for the entire resource
and Godoy (2010), and for SAG mills by Giblett and Seidel is essential to ensure that plant capacity does not deviate
(2011) and Conger and Schick (2014). significantly from expectations in the latter years of the mine
life. Ensuring good spatial coverage of the deposit in the
Selection of comminution circuit design basis hardness database is an important means to this end.
A common point of discussion during project development is In the first example discussed in this section, the mill was
the criteria to be used for mill sizing based on the available ore sized based on the 85th percentile and an order placed,
characteristics. Opinions vary from using the average values although additional testing was initiated to gain a better
of the hardness database to using the hardest values recorded, appreciation of the spread of ore characteristics. This testing,
with reasoning based on anything from extreme optimism supplemented by improved geometallurgical classification
to extreme conservatism on a case-by-case basis. There of the deposit, determined that there were significant
are many examples where grinding circuits have had to be concentrations of very competent ore throughout the
upgraded or expanded due to an inability to meet nameplate resource. The average DWi for the deposit increased from
throughput, and in certain cases this is directly due to under 7.0 to 8.0 kWh/m3; 14 per cent more than the initial data
sampling the resource during the project study phase. Basing average, and 9 per cent less than the P85 design point selected.
the mill design on some conservative representation of ore This is a good example of the value of being conservative
characteristics is one approach to mitigating the risk of under when sizing mills based on limited information, certainly
sampling the deposit, which if not recognised and addressed where very hard components of the orebody are observed
may result in an undersized comminution circuit that will and/or geometallurgical classification of the resource is not
significantly erode the economic value of the project. sufficiently developed.
Referring to the example discussed earlier where a mill The P85 design point has become Newmont’s standard
sizing was required based on four hardness results. The approach when dealing with early stage projects. Validation of
four drop weight test results gave an average of 7.0 kWh/t. this approach was provided by Seidel (2011), who performed

Fig 2 – Normalised actual throughput over life-of-mine against predicted ore hardness.
a look back analysis of several recent Newmont projects. milling power between an average and a P85 design basis was
One SABC circuit was noted to be operating consistently at approximately 50 MW of grinding power.
ten per cent below nameplate capacity over several years,
having been designed based on the average DWi of the Predicting mill throughput for variable ore blends
database at the time. During this period of assessment, the mill Problems can be encountered when computing the mill capacity
feed was back calculated to average close to the 80th percentile for blended ores based on the methodology used and the
of DWi values for the database. A second example is cited in the extent of the difference in ore characteristics. Overly simplistic
analysis by Seidel where another SABC circuit was observed to methods such as a weighted average of the calculated milling
consistently operate at a SAG mill specific energy comparable rate for each ore type can readily be shown to overestimate
with that expected for the 85th percentile of the drop weight mill throughput, as demonstrated in Table 1. This becomes
database for that project. Both examples provide practical particularly pronounced when the ore types are very different,
validation of the conservative P85 approach to greenfield mill as in the case of a blend of highly competent fresh rock and
sizing, particularly when faced with a limited budget for test
friable saprolite feeding a SAG mill. Fundamentally a weighted
work and/or when variable ore characteristics are observed.
average of the specific energy for each ore type is accurate
In the case of brownfield mill expansions or where the and possibly the simplest approach to this issue, although
understanding of a greenfield resource is significantly Newmont has applied two additional approaches to simplify
advanced by high quality ore domain definition and large this calculation for mine planning purposes.
sets of ore hardness data, a more aggressive approach to
The use of a ‘grind index’, being a ratio of the calculated
mill sizing may be justified. In a recent early stage project
milling rate for an ore divided by the nameplate capacity of
the comminution test database was already at a total of
the plant, is one method that has been used by Newmont
80 samples. The DWi was observed to vary widely from
on recent projects as an attempt to simplify the inputs into
<1 to 10 kWh/ m3 for the deposit as a whole, however the
block modelling. As long as the grind index is based on a
ore was well characterised into major geometallurgical
domains consisting of 5–20 individual hardness results per valid methodology for computing throughput, such as ore
domain. Additionally the deposit geology, averages and type specific energy, it will give the right answer. However,
distribution of hardness values bore strong similarity to a as with the specific energy approach in a two-stage circuit,
nearby operation where several years of operating data were such as SAB based circuits, two inputs are required (one for
available. In this instance it was deemed that using the P85 as each milling stage), if the models are to consider transitions
the initial design basis was too strong a penalty for the project, between SAG and ball mill limiting conditions.
and that there was sufficient information available to support A simplified example of the grind index method is as
a more aggressive approach. There was as always strong follows, considering three ore types of varying hardness. The
economic incentive for a more aggressive design approach, grind index for ore types 1, 2 and 3 is based on a nominal plant
due to the large-scale of the project the difference in installed capacity of 7 Mt/a, with values of 1.2, 1.1 and 0.8 respectively.
Table 1
Comparison of methods to calculate blended ore treatment rate.
Blend composition Tonnage-based average Specific energy basis Grind index basis Linear model
(% primary) (Mt/y) (Mt/y) (Mt/y) (Mt/y)
0% 12.4 12.4 12.4 10.6
30% 10.6 9.8 9.8 9.4
50% 9.5 8.5 8.5 8.6
70% 8.3 7.6 7.6 7.7
100% 6.5 6.5 6.5 6.5

As such ore type 1 is the softest and ore type 3 is the hardest. influence of the fundamental sampling error on small-scale
For a feed blend consisting of 20 per cent, 30 per cent and recovery test work.
50 per cent of these ores, the weighted grind index of the One recent illustration was encountered for a low-grade
blend is then: gold resource containing appreciable levels of coarse GRG.
The low sample masses (20–25 kg) used for GRG testing,
1 combined with the low-grade but coarse gold content of the
= 0.94
0.2 0.3 0.5 sample conspired to produce very high confidence intervals in
+ +
1.2 1.1 0.8
the GRG test result. This is shown in Figure 3 which presents
The plant capacity on this particular blend is then the several GRG test results from Newmont’s database and the
reference capacity of 7 Mt/a multiplied by the weighted grind calculated 90 per cent confidence limits based on the observed
index of 0.94, giving an annualised capacity of 6.58 Mt/a. gold particle size and concentration.
Where a simpler approach is required, linear regression This analysis provides early indication of the difficulties
equations can be used with effect to describe the influence of that might be expected in trying to reliably determine grind
feed blend under well-defined conditions. As the relationship sensitivity, leach kinetics and ultimate gold extraction
between blend composition and mill capacity is non-linear, one from small (~1 kg) scale test composites. The presence of
linear model cannot be used to describe the entire response; coarse GRG in a low-grade ore makes reliable subsampling
however, it is possible that two equations can do so adequately. a considerable challenge if the GRG is not removed prior
As in the example shown in Table 1, the model describes the to splitting. In this particular case, a failure to address the
relationship between blend composition and capacity very fundamental sampling error by performing a coarse gravity
recovery step indeed influenced the metal recovery test work
well over the range of 50–100 per cent primary ore. However, at
results and compromised the interpretation.
compositions of less than 50 per cent primary ore the accuracy
diminishes and a second relationship is required to describe With respect to the reliability of the gravity recovery
that range of blends. This can readily be accomplished and estimate derived from this GRG test program the impact of
utilised to good effect, although the possibility of an inflection small sample sizes used for the GRG test is less critical. Having
in the relationship as the mill alternates from a SAG to ball mill determined that there is a reasonable concentration of coarse
limiting condition must be considered. GRG in this ore it is very clear that the operation requires a
gravity recovery step in the flow sheet. From a process design
perspective, it is less critical to define with precision what the
GRAVITY PROCESS DEVELOPMENT gravity gold recovery number will be, it is simply enough to
Newmont consistently applies the Laplante GRG test to define know if the process is required or not in early stages of study.
the proportion of gold recoverable by conventional gravity The presence of appreciable levels of coarse GRG will mean
concentration technologies. This method has proven to be a that the operation risks a substantial recovery penalty if this
reliable basis for estimating plant scale gravity recovery when GRG is not removed ahead of conventional cyanidation.
performed and interpreted correctly and providing a valid However, it is important that the influence of gravity recovery
basis for deciding when a gravity circuit is required (Giblett on total metal recovery estimates derived from the laboratory
et al, 2013). The detailed size-assay outputs of the GRG test is consistent with the expectations derived from GRG testing,
additionally allow an assessment of the error in the test result due to the propensity for small-scale gravity tests to overstate
by calculation of the fundamental sampling error associated the amount of gravity gold that would be recovered in the
with the gold particle size (Giblett and Napier-Munn, 2015). plant. This will in turn lead to an over estimation of total
In this way, it is possible to define the confidence in the gold recovery that may be expected by gravity followed by
gravity recovery estimate derived from the test, to estimate conventional cyanidation of the gravity tail. For example;
the sample mass required to reduce these confidence limits a laboratory test that achieves 65 per cent gold recovery by
through further testing, and importantly to consider the gravity, and a total gold recovery of 95 per cent following
Fig 3 – Stage 1 gravity recoverable gold (GRG) test results with 95 per cent confidence intervals.

cyanidation or flotation of the gravity tail, has likely grossly Using the single stage gravity recoverable
overestimated the total recovery possible if the plant will
only achieve 30 per cent gravity gold recovery as indicated
gold test to demonstrate variability
by the GRG test. The small-scale test has in this case assumed The single stage GRG test proposed by Laplante and Staunton
100 per cent metal recovery (by fire assay) of a large portion (2005) has been demonstrated to give a reliable approximation
of the contained gold that will in fact demonstrate a lower of the total GRG derived from the three stage GRG test and
recovery due to incomplete leach extraction in practice. In this subsequently is a useful tool to demonstrate variability in
case if we expect 30 per cent gravity recovery at plant scale and GRG levels through a deposit. The correlation between total
85.7 per cent (30 ÷ 35) leach extraction, then our best estimate GRG and actual plant gravity gold recovery varies widely,
of total gold recovery is 30 + 70 × 85.7 per cent = 90 per cent, however a correlation between Stage 1 GRG and actual plant
some five per cent lower than the original test which was gravity gold recovery has been observed by Giblett (2011).
heavily biased due to the inflated gravity recovery. The bias is If the ratio of Stage 1 GRG to total GRG can be estimated,
heavily attributed to the fine particle size distributions fed to then the total GRG measured by the single stage test can be
the laboratory gravity stage. In these instances, it is important approximated by multiplying the test result by the ratio of
to correct the metal recovery predictions based on a realistic Stage 1 GRG to total GRG measured in the three stage GRG
estimate of plant scale gravity recovery. test. An example is shown in Table 2 for a recent project
where one full three stage GRG test and five single stage GRG
Where coarse GRG is present it is often difficult to achieve
tests were available for the major ore domains. The Stage 1
reproducible metal recovery performance in conventional
GRG content for each of the single stage tests could then be
small-scale testing, even if good practices are followed.
approximated using the method described. Bulk samples of
Subsamples will have variable coarse GRG contents and the
each ore domain were also processed at pilot plant scale to
efficiency of small-scale gravity recovery testing will vary
demonstrate the recovery flow sheet, allowing a larger scale
significantly in the efficiency of hand-panning primary gravity
measurement of the gravity gold content of the ore.
concentrates. Flattening of GRG particles during laboratory
grinding can also impede the gravity separation stage. This Table 2 shows acceptable correlation between the estimated
can have a substantial influence on the total gold recovery Stage 1 GRG content of the ore domains and the actual gravity
observed when gravity concentration and cyanidation are gold recovered from the pilot plant runs. While the ratio
combined at the laboratory scale. Under such circumstances of Stage 1 GRG to total GRG should not be assumed to be
a variation of the test work approach proposed by Laplante constant for all ore deposits, in this instance the assumption
(2000) will give superior reproducibility in the laboratory, in has proven reliable and a useful means to quantify the
addition to also providing better estimations of plant scale variability in gravity gold levels throughout the deposit.
gravity recovery. This method involves performing gravity An example is provided in Figure 4 from Giblett (2011) of
separation on the bulk sample, under similar conditions to the influence that sample grade can also have on the amount
the first stage of the GRG test (avoiding overgrinding and of Stage 1 GRG present and subsequent estimates of plant
excessive feed size reduction), prior to splitting the gravity
stage tail into leach or flotation test charges that are then Table 2
ground (in the absence of coarse GRG) to the target grind Correlation between Stage 1 gravity recoverable
size. This prevents excessive gravity gold production as well gold (GRG) and pilot plant gold recovery.
as preventing flattened coarse gold particles reporting to the
gravity tail. Composite no Estimated stage 1 GRG Observed pilot plant GRG
In all gravity gold recovery test work it is essential that a 1 3% 3%
visual inspection of the gravity concentrate be performed to
determine the nature of the gold occurrences. The extent scale 2 8% 11%
up of the laboratory result to plant gravity gold production 3 9% 7%
will vary depending on whether the gold is free, present as 4 9% 11%
alloys (for example, electrum, tellurides) or as fine inclusions
in minerals such as pyrite. 5 10% 14%
Fig 4 – Example correlation between head grade and Stage 1 gravity recoverable gold.

gravity recovery and then total gold recovery. In this particular If data availability is limited or correlations are poor then
case, a very strong correlation between Stage 1 GRG and head circumstances do not support the use of recovery models at
grade is observed. This relationship highlights the potentially all. A fixed recovery assumption will often be a better option
significant impact of performing test work on samples that and less prone to error. This is particularly the case when
don’t reflect the average grade of the resource, particularly a recovery model would otherwise be extrapolated over a
on samples higher in grade than the resource average. The wide range of ore grades. Improving the geometallurgical
lesser risk is a significant overstatement of gravity gold understanding of the resource and implementing focused
content, while the more significant risk is that overall gold metallurgical testing would be the logical next steps.
recovery assumptions will be biased high due to the high- A recent project evaluation generated a reasonable amount
grade and the elevated gravity gold content of the samples. of gold recovery data for a free milling ore, based on a
It is certainly important to ensure metallurgical test work standard test workflow sheet consisting of gravity recovery
programs consider very high and very low-grade samples via a three inch Knelson concentrator followed by hand-
to better define performance variability. However, the panning and bottle roll cyanidation of the gravity tail. As
design assumptions must adequately reflect the distribution per standard practice, small samples of 500 to 1000 g were
of grade as a function of resource size and ideally not be used for individual recovery tests. Due to a combination
disproportionately skewed towards higher or lower grades. of test protocols, small sample size and moderate to coarse
Generating realistic assumptions of plant scale gravity GRG size distribution the gravity gold recovery varied
recovery is therefore contingent on multiple factors. widely, from 2 per cent to 79 per cent, without any apparent
Performing the correct test work (Laplante GRG test), on relationship to feed grade. Overall gold recovery similarly
appropriate samples and performing the correct interpretation varied significantly as shown in Figure 5, with the scatter in
of the test work are all critical steps. Appropriate samples are results not supporting attempts to model gold recovery as a
defined by their origin, as well as the amount and grade of the function of head grade.
sample used in testing. Good guidance is to ensure that, as a It is expected at least in part that the higher gold recoveries
minimum, Laplante GRG tests are performed as a minimum are likely due to over representation of the gravity gold
on each major ore type, on large samples (70–100 kg), at or content, both as a function of sampling errors and the impact
near the average grade of the ore type. of the gravity extraction procedure followed. Specifically
grinding a small sample mass to final grind size before gravity
GOLD RECOVERY RELATIONSHIPS separation in a batch laboratory Knelson concentrator will
Numerous examples have been observed where an obsession dramatically enhance the efficiency of the gravity separation,
with regression modelling has led to some questionable both as a function of the narrow feed size distribution and the
interpretations of test work data. In some instances, complex high mass pull to concentrate. Hand-panning of the Knelson
functions have been selected to represent a small number of concentrate is performed to mitigate the effect of the high
data points on a feed versus tail grade plot as the R2 value is Knelson mass pull, although the efficiency of that process is
a couple of points higher than a simple linear model. In many highly operator dependent and subsequently variable.
cases models are applied to describe data without consideration Low recoveries were similarly a function of the presence of
to the confidence in the model prediction, leading to the use of coarse gold, via a combination of inefficient gravity recovery
models that predict with very poor precision. Extreme cases due to hand-panning losses and insufficient leach residence
are observed where exotic model functions are provided to time, even at 30 hours. Hand-panning of the higher-grade
describe metal recoveries that provide impractical predictions bottle roll tailings confirmed the presence of free gold particles
at extremes of feed grade. One recent observation during a up to 150 µm in size (Figure 6), demonstrating the inefficient
mine plan optimisation exercise noted that gold recovery was nature of the gravity recovery process applied. Again, this
predicted to exceed 100 per cent once the grade dropped below was influenced by fine grinding of the gravity test feed to
a threshold value that was well outside the range of the original 80 per cent passing 106 µm, flattening of gold particles and
test data on which the model was based. In all cases recovery reduced gravity separation efficiency. The potential to both
models should be constrained to predict only over the range overstate (sampling and gravity effects) and understate the
of feed grades from which they were developed. A pragmatic gold recovery (laboratory gravity stage inefficiency) that will
approach is required to recovery modelling and the KISS be achieved in the plant based on testing protocols is therefore
principle (keep it simple, stupid) is strongly recommended. clearly apparent.
Fig 5 – Gold recovery data from laboratory test work.

but insufficient precision for a feasibility study. This is a good

example where more test results and more robust testing
protocols, particularly large test sample masses, would
improve confidence in metal recovery projections.
CONCLUSIONS
This paper has discussed several important considerations
that will influence the effectiveness of gold plant design.
From sample selection and test work design, through to data
interpretation and design criteria development, numerous
examples have been provided to demonstrate a series of
approaches to ensure that the plant design is based on
solid foundations. The main considerations presented are
summarised as follows:
•• comminution circuit design should be based on testing of
well-defined geometallurgical domains that demonstrate
reasonably consistent metallurgical properties
Fig 6 – Free gold in bottle roll tailings, •• statistical analysis can be useful to assist with the
post-gravity separation and 30-hour leach. definition of metallurgical ore types, to guide testing
programs and to define confidence levels for predictions
The occurrence of coarse gold particles in the leach tail at
of plant scale performance
the laboratory scale serves as a strong warning to ensure that
the gravity recovery circuit is adequately sized and operated, •• the use of conservative design values based on safety
and also that the plant sampling and analytical practices must factors or database percentiles can mitigate the risk of
be robust enough to detect the presence of coarse gold losses. inaccurate early stage capital estimates and suboptimal
The sampling and analytical protocols for plant tailings are plant performance
too often a shortcoming of modern gold process plants.
•• appropriate throughput modelling techniques should be
Consideration should also be given to the average grade of applied to support mine planning, ensuring the impacts
the test samples against the average grade reported for the of blending ore types is appropriately accounted for and
resource, as this may lead to biased estimates of recovery
that circuit constraints (SAG or ball mill limitations) can
where strong grade-recovery relationships exist. To better
define the dependence of gold recovery on feed grade, the be accounted for
calculated leach feed grade for each cyanidation test after •• metallurgical testing programs should be designed
gravity separation was plotted against the leach residue grade to account for the impacts of free gold and utilise best
yielding the relationship seen in Figure 7. The models appear practices such as GRG test procedures to ensure plant
to give a sensible representation of the data, however, the recovery estimates are practical
scatter is sufficient to generate wide 95 per cent confidence
limits on any prediction of the model. For example, at a •• the use of regression models to predict plant performance
leach feed grade of 2 g/t Au, the predicted residue grade should always be constrained to the range of feed grades
is approximately 0.15 ± 0.10 g/t Au. This translates to a used in the original test work, and the confidence limits
95 per cent confidence band of ±5 per cent on the prediction of the model predictions should be calculated and
of leach extraction, which may be fine for early stage studies communicated.
Fig 7 – Leach extraction model fit.

ACKNOWLEDGEMENTS Guzman, L and Rabanal, C, 2014. Updated Benavente correlation for

estimating grinding media consumption rates, Proceedings XXVII
The authors would like to acknowledge Newmont Mining International Mineral Processing Congress, Chapter 8, pp 22–33.
Corporation for permission to publish this paper. The expert
Laplante, A R, 2000. Testing requirements and insight for gravity
guidance of Tim Napier-Munn with respect to various aspects
gold circuit design, in Proceedings Randol Gold and Silver Forum.
of data analysis is also gratefully acknowledged.
Laplante, A R and Staunton, W P, 2005. Gravity recovery of gold –
an overview of recent developments, in Proceedings International
REFERENCES Symposium on the Treatment of Gold.
Conger, W and Schick, G, 2014. Industrial SAG media consumption Morrell, S, 2004. Predicting the specific energy of autogenous and
study: definition of consumption and relationship with SAG semiautogenous mills from small diameter drill core samples,
specific grinding energy, in Proceedings XXVII International Minerals Engineering, 17(3):447–451.
Mineral Processing Congress, Chapter 8, pp 159–168.
Morrell, S, 2006. Design of AG/SAG mill circuits using the SMC Test®,
Doll, A, Barratt, D and Godoy, R, 2010. Microcracking versus in Proceedings International Autogenous and Semiautogenous
the Phantom cyclone; comparing SAG mills and HPGR on a Grinding Technology Conference, Volume I, pp 279–298.
consistent basis, in Proceedings Procemin 7th International Mineral
Processing Seminar, pp 73–82. Morrell, S, 2009. Predicting the overall specific energy requirement of
crushing, high pressure grinding roll and tumbling mill circuits,
Giblett, A, 2011. Gravity gold concentration at Newmont Mining, Minerals Engineering, 22(6):544–549.
World Gold Conference 2011, in Proceedings 50th Conference of
Metallurgists, pp 217–228. Morrell, S, 2011. Mapping orebody hardness variability for AG/
SAG/crushing and HPGR circuits, in Proceedings International
Giblett, A, Bax, A, Wardell-Johnson, G and Staunton, W, 2013. A Autogenous Grinding, Semiautogenous Grinding and High Pressure
review of best practice in gravity circuit design and operation, Grinding Roll Technology.
in Proceedings MetPlant, pp 272–284 (The Australasian Institute of
Mining and Metallurgy: Melbourne). Napier-Munn, T J, 2015. Statistical Methods for Minerals Engineers, How
to design experiments and analyse data, Julius Kruttschnitt Mineral
Giblett, A and Napier-Munn, T J, 2015. Measuring the influence Research Centre Monograph Series in Mining and Mineral
of sample size on the precision and accuracy of gravity gold Processing: No 5, (Sustainable Minerals Institute – JKMRC:
estimation, in Proceedings MetPlant, pp 367–372 (The Australasian Brisbane).
Seidel, J F, 2011. Percentile criteria for design data, Newmont Mining
Giblett, A and Seidel, J, 2011. Measuring, predicting and managing Corporation.
grinding media wear, in Proceedings International Autogenous
Grinding, Semiautogenous Grinding and High pressure Grinding Roll
Technology Conference.

Contents
Energy efficient ball mill circuit –

equipment sizing considerations
A Jankovic1 and W Valery2
ABSTRACT
The ball mill in closed circuit with hydrocyclones is an industry standard, and well
known methodologies exist for equipment sizing, selection and design; however,
the potential to improve the capacity of ball mill circuits by replacing cyclones with
screens has been recognised for many years and recently has been successfully
realised, thanks to developments in fine screening technology. Proper methodology
for equipment sizing and selection is required to capitalise on this opportunity, but
there is currently insufficient information and few guidelines available on how to
design such a circuit. This paper discusses the differences between ball mill circuits
closed with cyclones and with screens. A methodology is described for equipment
sizing and selection for a circuit closed with screens and the laboratory testing
requirements are described for low-grade base metal ores and magnetite ores. An
economic evaluation is also presented for a case study demonstrating the potential
benefits of closing a ball mill circuit with screens.
INTRODUCTION
The ball mill in closed circuit with cyclones is currently the most common grinding
circuit in the mining industry, and has been so for over half a century. Due to its
compact design, simple operation and low capital cost, cyclones have ‘squeezed out’
other types of classifiers such as spiral and rake classifiers, which these days are only
found in specific applications and old plant designs, mostly in Nordic countries.
The practical advantages of cyclones outweighed the sacrifice in classification
efficiency, especially when operating at relatively high-solids content. Figure 1 shows
how cyclone classification efficiency deteriorates rapidly with increasing cyclone
feed per cent solids. In the range of 45–65 per cent solids, typically used in ball mill
circuits, cyclone classification efficiency is commonly lower than 50 per cent. This has
an adverse effect on ball mill circuit capacity as shown by work conducted by Hukki
and his co-authors (1965, 1968, 1973, 1979). Extensive work has been done in the past
to improve cyclone efficiency (two-stage cycloning, cyclone design modifications etc);
however, to date these efforts have not delivered large improvements in practice.
Screens are known to be superior classification devices compared to cyclones and
other hydraulic classifiers. The effect of using screens instead of cyclones to close ball
mill grinding circuits has also been long understood. The problem with screens is
that for the fine size separations, below say 1 mm, they become impractical due to
the requirement for excessive working surface area. However, recent developments
in fine screening technology by Derrick Corporation have overcome many of the
problems previously encountered in continuous fine screening, such as blinding
and low capacity. Some of the early applications to coal and mineral slurries have
been described by Kirk (1984) and Rogers and Brame (1985). Their results showed
that fine screening can provide sharp classification of some slurries with very little
bypass of fines. The recent developments in fine screening technology are discussed
by Barkhuysen (2009). Improved plant performance at several sites, with reduced
circulating loads and improved throughputs is described. Details of several fine
screens applications have been discussed by Dundar et al (2014). Metso has also
introduced UltraFine Screen (UFS) technology in 2016.
The energy saving resulting from the replacement of cyclones with screens in a
1. MAusIMM, Principal Consultant / Mineral
closed ball mill circuit was estimated to be in the range of 10 to 25 per cent (Jankovic
Processing, HATCH, Brisbane Qld 4000.
Email: alex.jankovic@hatch.com and Valery, 2012a). An analysis of the economic benefits of replacing cyclones with
fine screens for a small-scale (1.5 Mt/a) gold ore operation has concluded that in
2. FAusIMM, Global Director – Consulting and
Technology, Hatch – Mining and Minerals countries such as Australia where electricity prices are high and expected to rise, fine
Processing, Brisbane Qld 4000. screening should be considered for any new comminution circuit or expansion, and
Email: walter.valery@hatch.com should be evaluated during engineering studies (Jankovic and Valery, 2012b).
307
The first step is considered as standard practice and it is not

further discussed in the paper. The second step requires further
clarification and it is discussed in the following sections.
Ball mill capacity

When using screens in a ball mill circuit, high classification
efficiency can be obtained which would result in increased
circuit capacity compared to the same circuit utilising
low efficiency classifiers such as cyclones (Hukki, 1979;
Jankovic and Valery, 2012a). If an appropriate model can be
developed, this simple model or simulation could be used to
determine the increase in capacity obtained by using screens
rather than cyclones.
The general form of the equation, or circulating load and
efficiency (CE) model, for the relative capacity of a closed
ball mill cyclone circuit as function of circulating load and
classification efficiency (Magdalinovic, 1991) is:
FIG 1 – Hydrocyclone classification efficiency. (1 + C1) c1.5 + C2 - 1 m

Q2 E2
KQ = = (1)
Q1 (1 C ) 1.5 C - 1
E1 m
BALL MILL CIRCUIT SIZING METHODS + 2c + 1
The basis for ball mill circuit sizing is still Bond’s methodology
(Bond, 1962). The Bond ball and rod mill tests are used to where:
determine specific energy consumption (kWh/t) to grind KQ relative capacity of the milling circuit (fraction)
from a given feed 80 per cent passing size (F80) to 80 per cent Q1, Q2 milling circuit capacity under different circulating
passing size product (P80). Variations in the original Bond load and classification efficiency
methodology are used to account for the effect of feed
C circulating load (fraction)
size distribution when feed to the ball mill is coming from
equipment other than three stage crushing circuit or rod mill. E classification efficiency (fraction)
Morrell’s methodology (Morrell, 2004a, 2004b, 2008) does For a standard ball mill circuit the circulating load is usually
not use the Bond work index directly; however, it uses the 250 per cent, as established by Bond (1962). The classification
Bond ball mill test as basis for calculation of the energy efficiency depends directly on circulating load (Hukki and
required to grind material from an F80 of 0.7 mm to the Eklund, 1965) although it can vary depending on operating
required product size. It therefore may be considered as a conditions. Figure 2 presents industrial data on classification
variation of the Bond methodology. efficiency as a function of circulating load. An average
classification efficiency of 50 per cent can be considered for the
The above methods are considered ‘power-based’ as they
design purposes. Note: classification efficiency is with respect
calculate the required power for the duty. The basis for to ~95 per cent cumulative passing size of the classifier fine
development of these methods were the databases of existing product, as defined by Hukki and Eklund (1965).
operations which had ball mill circuits closed with hydraulic
classifiers, with the vast majority being cyclones. Therefore, Manipulating the CE model (Equation 1) and assuming
these calculations may not be directly applicable to ball mill average operating conditions, the ball mill circuit capacity
increase that can be achieved by replacing cyclones with
circuits closed with screens.
screens can be calculated using Equation 2:
When the appropriate pilot or operational ball mill data are
available, modelling and simulation may be used to size the Q2 = 0.875 : Q1 : (2.5 - 1 ) (2)
ball mill circuit. This approach is not as straightforward as the E2
power-based methods as it requires good understanding of where:
the models and their limitations, and it should only be used
Q1 is circuit capacity with cyclones
by experienced professionals. The simulation outcomes are
always compared to the power-based method results as a way Q2 is circuit capacity with screens
of cross-checking. The convenience of this methodology is that C1 is 2.5, circulating load with cyclones (250 per cent)
it may be used to evaluate the effects of different operating C2 is 1.0, circulating load with screens (100 per cent),
and design changes which may include replacement of common in practice
cyclones with screens.
E2 is classification efficiency with screens
E1 is 0.5, classification efficiency with cyclones
SIZING BALL MILL IN CLOSE CIRCUIT WITH SCREEN
A classification efficiency of E2 = 0.9 (ie 90 per cent) can be
Two steps are proposed to evaluate the ball mill size for achieved with screens, and in this case the circuit capacity
operation in closed circuit with screens: increase would be 1.215 (that is, 21.5 per cent more compared
1. A power-based method is first utilised to determine the to a circuit closed with cyclones). It should be emphasised
required ball mill power consumption to achieve the that this increase is with respect to ~95 per cent passing size
required product P80. This would give the required ball material and not 80 per cent passing size (or P80) as defined in
mill power in closed circuit with cyclones operating at Bond’s method.
250 per cent circulating load. The accuracy of the CE model was tested by completing
2. Adjustment of the ball mill power to account for effect of several closed circuit laboratory grinding tests (Jankovic et al,
higher classification efficiency of screens. 2013). More than 20 tests were conducted under a number

Energy efficient ball mill circuit – equipment sizing considerations
FIG 2 – The relationship between classification efficiency and circulating load (PTI – Process Technology and Innovation).
of circulating load and classification efficiency conditions predictions had errors larger than 20 per cent and the largest
as well as various closing screen sizes and using different error was greater than 60 per cent. The largest errors are
materials/ores. It can be seen in Table 1 and Figure 3, which with high- and low-circulating loads, which indicate that the
present the data for tests done with a lead-zinc ore and a current ball mill model cannot accurately simulate conditions
0.3 mm closing screen, that the CE model provides reasonable that are very different from the base-case. Work is underway
prediction of the experimental results. The capacity for
to address this issue in more detail.
several combinations of circulating load and classification
efficiency was predicted relative to the base-case condition
of 248 per cent circulating load and 50 per cent classification Ball mill size and power selection
efficiency. The maximum error in predicting the relative As discussed previously, ball mill circuit capacity could be
capacity was lower than 20 per cent. increased by around 20 per cent if cyclones are replaced
It is obvious from the data that the material specific properties, with screens operating at high classification efficiency of
size distribution and hardness, as well as milling conditions, 90 per cent and low circulating load of 100 per cent. This
affect the mill capacity and therefore limit the accuracy of means that a ball mill selected in Step 1 of the design/selection
the CE model. For this reason more sophisticated modelling process would have 20 per cent higher capacity than required
and simulations were carried out using JKSimMet. The test by design criteria when designed to operate in closed circuit
with 248 per cent circulating load and 50 per cent efficiency with screens. Depending on the particular project situation,
was taken as the base-case for modelling. The perfect mixing the designer has the following major options:
ball mill and efficiency curve models were used to develop
the base-case model, which was used to simulate the results •• Reduce the mill size and installed motor power by
from other experiments with different circulating loads and 20 per cent. To follow this option would require high
classification efficiencies. Table 1 shows the summary of the confidence in, and understanding of issues related to
simulated results in terms of relative throughput increase. operation with screens. This could be a very attractive
Surprisingly (or not?), the throughput predictions obtained option for ball mill capital expenditure (CAPEX)
from the JKSimMet simulations were rather poor. Half of the reduction.
FIG 3 – Relative capacity comparison: experimental versus model.

Table 1
Comparison of test work results and predictions using JKSimMet and the simple CE model.
Test conditions Simulation Relative capacity (%)

Classification efficiency (%) Circulating load (%) Exp t/h JKSimMet t/h Experiment CE model JKSimMet
50 248 7.12 7.12 100.0 100.0 100.0
100 52 7.29 5.3 105.0 117.9 74.4
75 100 6.82 5.25 96.4 102.5 73.7
90 102 7.76 7.45 105.0 122.6 104.6
75 150 7.81 7.9 124.2 117.2 111.0
85 151 8.35 9.3 124.2 128.4 130.6
100 242 9.71 14.3 132.5 150.1 200.8
50 400 9.27 11 120.4 123.0 154.5
50 587 8.85 14.3 127.9 137.4 200.8
•• Keep the same mill size and installed motor power difficult, if not impossible, unless information from similar
and recommend mill operation at lower ball charge/ applications is available.
power draw. This is a low-risk option in case there are Table 2 provides a summary of some published data
uncertainties related to screen performance and effects regarding the capacity of high-frequency screens that may be
of grinding circuit product properties (size distribution) used for preliminary estimates. The screen specific capacity
on the downstream flotation process. data shown in Table 2 ranges from 16–38 t/hm2 which
•• Reduce the mill size and installed motor power by less means that care should be taken when estimating screening
than 20 per cent. This is basically a compromise between capacity. The most comprehensive information regarding
the first two, with some risk lowering as well as CAPEX industrial applications can be found in Valine, Wheeler and
reduction potential. Alburquerque (2013). The total screening area and therefore
The operating data suggest that with efficient screen number of screens can be estimated based on the screen feed
operation the circulating load drops to around 100 per cent. rate and estimated screen specific capacity.
For screens, this circulating load may be considered as
optimum in respect to gains in grinding capacity and Comparison of products from ball mill circuits
required screening capacity. It is possible to run screens closed with screen and cyclones
at higher circulating loads and achieve higher grinding One should be aware that product from the mill in closed
capacity; however screening capacity would have to increase circuit with screens was coarser than from the circuit closed
at least proportionally to the circulating load increase, which with cyclones for all industrial applications where cyclones
would further increase the already high screen CAPEX and were replaced by screens (Valine, Wheeler and Alburquerque,
operating expenditure (OPEX). Figure 4 shows the increase 2013). The major difference is lower content of fines in the
in mill capacity when cyclones are replaced by screens. The screen undersize due to reduced overgrinding in the mill.
cyclone circulating load is 250 per cent and classification The difference is illustrated in Figure 5 which shows the
efficiency 50 per cent as discussed earlier. The relative product size distribution from the closed circuit ball mill test
capacity increases with better screening (lower percentage of with a 150 micron screen at two different circulating loads
fines in screen oversize) and higher circulating load; but the and classification efficiencies. The product from the test with
gains are not proportional because the classification efficiency 250 per cent circulating load and 50 per cent classification
decreases with increased circulating load (Hukki and Eklund, efficiency, emulating operation with cyclones, has about
1965) at unchanged screen performance (percentage of final 8 per cent more -38 micron material and a P80 of 96 microns
product in the screen oversize). compared to a P80 of 110 microns for test at 100 per cent
circulating load and 100 per cent efficiency emulating operation
Screen selection with screens. There is also more coarse material in the product
While the screen selection for coarser applications is from the test emulating operation with screens. Therefore, it
usually done using the various versions of ‘screen capacity should be made clear that the increased throughput for the
calculations’ from Allis Chalmers, screen selection below operations with screens has, therefore, a coarser product due
1 mm cut size is not ‘routine’ and usually would require to reduced production of fines. The reduction of fines may be
experimental tests. Screening below 1 mm is traditionally beneficial for flotation circuits as the loss of fine and liberated
considered ‘tough’ and requires a large screening area using valuable minerals would be reduced, while loss of locked-up
conventional screening technologies. The big improvement valuable minerals in coarse fractions may not be affected as
in fine screening is achieved by high-frequency screens that significantly. For gold leaching operations, however, reduced
utilise different vibration frequencies and ‘flexible’ screening recovery should be expected as the leaching process benefits
media for a range of screening duties. These screens are from very fine material.
currently the most practical option for ball mill circuits with
capacities larger than 100 t/h. Screen separation size selection
To obtain an accurate estimate of the required screening The screen size should be selected based on the expected
capacity for a particular application requires extensive test 95 per cent passing size for the cyclone overflow. Figure 6
work necessitating a large amount of sample. This is usually demonstrates the relationship between cyclone overflow
not available at the feasibility study stage which makes design P95 and P80 for several operating plants from the authors’

Table 2
High-frequency screening capacities.
Commodity Slot size (mm) Feed top size (mm)a Feed % passing 74 μm Indicative solids SG t/m3 Screen specific capacity t/hm2
Coal 0.29 3.0 40–50 1.4 32
Coal 0.21 3.0 40–50 1.4 24
Cu ore 1 0.30 3.0 30 2.8 28
Cu ore 2 0.23 3.0 40 2.8 29
Cu ore 2 0.30 3.0 40 2.8 30
Cu ore 2 0.35 3.0 40 2.8 32
Cu ore 2 0.43 3.0 40 2.8 38
Cu/Pb/Zn ore 0.23 1.0 30 3.4 30
Cu/Pb/Zn/Au ore 0.5 3.0 38 3.4 17
P2O5 ore 0.5 6.0 35 3.1 24
Apatite ore 0.4 3.0 20 2.9 16
Magnetite ore 1 0.1 0.5 40 4.2 38
Magnetite ore 2 0.063 0.1 95 4.2 32
a. Estimate; SG – specific gravity.
FIG 4 – Relative capacity of screens compared to cyclones.
database. This can be used to estimate screen slot size, as the

screen slot size should be similar to the cyclone 95 per cent
passing size.
It should be mentioned that due to the screening surface
material flexibility, a relatively small percentage of particles
coarser than the screen slot size reports to screen undersize.
MAGNETITE MILLING
Metso PTI investigations into the effect of classification
efficiency on ball mill circuit capacity identified opportunities
to reduce the energy consumption required for ball milling
by around 20 per cent by replacing cyclones with fine screens
(Jankovic and Valery, 2012a, 2012b). The potential for process
improvement and energy savings by application of fine
screening is even higher for magnetite ores. This is due to
the large density differential between magnetite and gangue
which results in poor size separation when using cyclones
FIG 5 – Grinding product size distribution for different and consequently overgrinding of magnetite. A reduction in
circulating load and classification efficiency. energy consumption of the order of 50 per cent was reported

FIG 6 – P95 versus P80 relationship for industrial grinding circuit from the authors’ database.
in a magnetite concentrator in Russia when cyclones were selected for both options. The estimated capital and installation
replaced with screens (Valine, Wheeler and Alburquerque, cost for screens was $US 2.5 M compared to $US 0.35 M for
2013). While the effect of density differential on cyclone cyclones. The analysis also accounted for increased operating
classification efficiency is widely understood and accepted, costs associated with screen maintenance and spare parts
research has not been conducted to quantify the magnitude of (panels), and this was estimated to be $US 0.235 M/a. The
the effect for magnetite ores. energy saving resulting from the replacement of cyclones
with screens in a closed ball mill circuit was estimated to be
ECONOMIC EVALUATION in the range of 10 to 25 per cent (Jankovic and Valery, 2012a).
An analysis of the economic benefits of replacing cyclones Based on the conditions and assumptions described
with fine screens was conducted for a small-scale (1.5 Mt/a) above, a high level net present value (NPV) comparison of
gold ore operation (Jankovic and Valery, 2012b). The basis comminution circuit costs has been undertaken to evaluate
for comparison is a semi-autogenous ball mill crushing the key differences between the circuit with cyclones and
(SABC) circuit with ball mill in closed circuit with cyclones. screens. Results of the ten-year schedule using an 8 per cent
The alternative circuit uses high-frequency screens instead discount rate for the two options is summarised in Figure 7.
of cyclones with the same size semi-autogenous grinding This analysis indicates the price of energy and energy savings
(SAG) and the ball mill. The capacity of the grinding circuit in the ball mill circuit required to justify the additional
evaluated is 187.5 t/h and the circuit is required to deliver a costs associated with the use of high-frequency screens for
final product with a P80 of 75 µm. the circuit in question. It can be concluded from the graph
Data from Minera Condestable SA (Wheeler and Packer, that at energy costs of around $US160/MWh or higher,
2011) were used to estimate the required number of screens high-frequency screens can be justified if energy savings of
and the associated capital cost. It was estimated that six Stack 15 per cent or greater are obtained in the ball mill circuit.
Sizer screens would be required to achieve a classification Considering that 15 per cent is at the lower end of the range
efficiency of 80–90 per cent with 0.12–0.15 mm screen panels of energy savings estimated, high-frequency screens are likely
in order to deliver the required product size (P80 of 75 µm). The to be justified in Australia where energy costs are typically
same size ball mill (and same capital and installation cost) was around $150/MWh.
FIG 7 – Net present value (NPV) sensitivity to energy saving and energy price (BM – ball mill).

This analysis was based on a relatively small capacity REFERENCES

circuit (1.5 Mt/a) and very hard ore, and the conclusions Barkhuysen, N J, 2009. Implementing strategies to improve mill
drawn are only relevant for similar conditions. For example, capacity and efficiency through classification by particle size only,
for cases with much softer ore, the cost of energy would have with case studies, in Proceedings Base Metals Conf, pp 101–113 (The
significantly less effect on NPV, and therefore fine screening South African Institute of Mining and Metallurgy: Johannesburg).
would be more difficult to justify. For larger capacity circuits, Bond, F C, 1962. Crushing and Grinding Calculations – April 1962
the increased capital cost for fine screening compared to Additions and Revisions, Allis-Chalmers Manufacturing Co,
cyclones would be much higher than for small circuits, and Milwaukee, Wisconsin.
the application of fine screening may be only justified for Dundar, H, Kalugin, A, Delgado, M, Palomino, A, Turkistanili, A,
electricity prices higher than typically experienced at present. Aquino, B and Lynch, A, 2014. Screens and cyclones in closed
Nevertheless, it is recommended that fine screening should be grinding circuits, in Proceedings XXVII IMPC, Santiago, 16:67–77.
considered for any new comminution circuit or expansion in Hukki, R T, 1979. Fundamentals of the closed grinding circuit,
countries with high electricity prices, such as Australia. Engineering Mining Journal, April, pp 102–109.
Hukki, R T and Allenius, H, 1968. A quantitative investigation of
CONCLUSION the closed grinding circuit, Transactions of the Society for Mining,
Ball mills in closed circuit with cyclones are the industry Metallurgy & Exploration, 241:482–487.
standard and dominate grinding circuit design. However, Hukki, R T and Eklund, H, 1965. The relationship between sharpness
the application of screens in closed circuit with ball mills can of classification and circulating load in closed grinding circuits,
increase the circuit capacity in the order of 20 per cent due to Transactions of the Society for Mining, Metallurgy & Exploration,
the favourable effect of high classification efficiency. September, pp 265–268.
A ball mill circuit in closed circuit with screens can be Hukki, R T and Heinonen, P, 1973. An experimental investigation of
the closed fine grinding circuit, Zement-Kalk-Gips, 26:303.
designed using power-based methods, similarly to circuits
closed with cyclones. These methodologies calculate the Kirk, T A, 1984. Closed circuit grinding with vibrating screens, in
required mill specific power to achieve certain product Proceedings 16th Canadian Mineral Processors Conference, Ottawa,
80 per cent passing size, ie P80. Using a simple model a factor (Canadian Institute of Mining, Metallurgy and Petroleum:
Quebec).
can be calculated describing the increase of ball mill relative
capacity when using screens, and this may be used to reduce Jankovic, A and Valery, W, 2012a. Closed circuit ball mill –
the selected ball mill size and power. basics revisited [online], Minerals Engineering, April, 43–
44:148–153. Available from: <http://dx.doi.org/10.1016/j.
Selection of screens for fine applications is usually based mineng.2012.11.006> [Accessed: 25 July 2017].
on extensive test work done by the manufacturer, which
Jankovic, A and Valery, W, 2012b. The impact of classification on the
requires a large amount of sample. It is often not possible to
energy efficiency of grinding circuits – the hidden opportunity,
carry out this test work at the feasibility stage. Therefore, the in Proceedings AusIMM 11th Mill Operators Conference (The
designer needs to rely on available published or unpublished Australasian Institute of Mining and Metallurgy: Melbourne).
information regarding screen capacity, and the screening
Jankovic, A, Valery, W, Lee, D, Peres, H and Jeston, S, 2013. Validation
capacity estimate is based on 100 per cent circulating load for of a closed circuit ball mill model, Journal of Mining and Metallurgy
screen feed and the specific screening capacity (t/hm2). For Section A: Mining, 49(1):37–43.
the design P80 size, the screen slot size should be selected close
Magdalinovic, N, 1991. Grinding and Classification of Raw Materials (in
to the cyclone 95 per cent passing size. One should be aware
Serbian) ISBN 86-23-45006-X (Scientific Books: Belgrade).
that product size distribution from the mill in closed circuit
with screens is coarser than from the circuit closed with Morrell, S, 2004a. Predicting the specific energy of autogenous and
semi-autogenous mills from small diameter drill core samples,
cyclones due to reduced fines content.
Minerals Engineering, 17(3):447–451.
It is recommended that fine screening should be considered
Morrell, S, 2004b. An alternative energy-size relationship to that
for any new small to medium size comminution circuit or proposed by Bond for the design and optimisation of grinding
expansion in countries with high electricity prices, such as circuits, Int J Miner Process, 74:133–141.
Australia. In these cases it could result in significant reduction
in energy consumption and have a positive effect on the
requirement of comminution circuits and assessing their energy
project economics. utilisation efficiency, Minerals Engineering, 21(3):224–233.
The benefits associated with replacing cyclones with screens Rogers, R S C and Brame, K A, 1985. An analysis of the high frequency
are higher in magnetite milling circuits due to the large density screening of fine slurries, Powder Technology, 42:297–304.
differential between magnetite and gangue which results in
Valine, S B, Wheeler, J E and Albuquerque, L G, 2013. Application of
poor size separation when using cyclones and consequently
fine screens in grinding circuits, in Proceedings ‑11th Mill Operators
overgrinding of magnetite. Energy reduction of the order of Conference (The Australasian Institute of Mining and Metallurgy:
50 per cent was reported for grinding of magnetite ore when Melbourne).
cyclones were replaced by high-frequency screens.
Wheeler, J and Packer, B, 2011. Increasing capacity and efficiency
of grinding circuits with high frequency screens, in Proceedings
MetPlant, pp 107–114 (The Australasian Institute of Mining and

Contents
Advances in dense medium cyclone plant design

T J Napier-Munn1, G Gibson2 and B Bessen3
ABSTRACT
In coal preparation, large dense medium cyclones (DMC) at low pressure heads
(typically 9D where D = cyclone diameter) are the norm. This contrasts with high-
density mineral applications where, largely for historical reasons, smaller cyclones
at high heads are the standard. There are advantages in preferring large diameter
cyclones at low heads, assuming that metallurgical performance does not suffer
significantly. These include simpler plant layouts, easier maintenance, smaller spares
inventories, reduced wear, reduced medium pumping demands, smaller buildings
and overall lower capital and operating costs. In 2008, therefore, Xstrata Zinc decided
to trial a single Dutch State Mines (DSM) design 800 mm cyclone in one of its four
DMC lines at Mount Isa, replacing the four 400 mm cyclones previously installed.
Circumstances allowed a low head of 9.5 D to be trialled, compared with the existing
400 mm head of 21.5 D.
This paper describes the trial, including tests using painted ore particles as
tracers, estimates of yield and metal recovery and wear monitoring. The larger unit
performed well. There was no evidence that metallurgical performance in terms of
metal recovery and waste reject rate suffered, and there was some evidence that it
may have improved. Cyclone life increased, and preliminary data suggest that the
replacement cost of the 800 mm cyclones is less than half that of the 400 mm units,
and this is expected to improve. The plant layout and operability have also improved
significantly. Xstrata has therefore decided to convert the whole plant to four low
head 800 mm units, replacing the original 16 high head 400 mm units.
The paper includes an explanation of the effect of static head on DMC separating
force. It also reviews earlier research on diamond applications, and applies a general
DMC process model to the Xstrata case, both of which confirm that large, low head
cyclones are appropriate in mineral applications, contrary to conventional practice.
Such units could significantly reduce both the capital and operating costs of DMC
plants in mineral applications.
INTRODUCTION AND BACKGROUND

Xstrata Zinc (now Glencore) operates a DMC plant at its Mount Isa operations to
preconcentrate lead-zinc ore prior to grinding and flotation (Munro et al, 1982). The
plant comprises two circuits each of two lines with four 400 mm DMCs per line.
Ferrosilicon is used as the dense medium. The key performance criteria are the
recovery of metal and the reject rate of waste.
The DMC was originally invented by DSM for coal preparation and all the early
development work was done in this context. Later, the technology was extended to the
beneficiation of minerals at higher medium densities. In Australia DMCs are widely
used in coal preparation, and also for the beneficiation of iron ore in the Pilbara, in
the recovery of diamonds at Argyle, and in the preconcentration of lead-zinc ore at
Mount Isa, the subject of this paper.
There are two key differences in industrial practice between coal and minerals:
in coal, the cyclones tend to be larger, and the feed pressure (the static head in the
case of a gravity-fed unit) tends to be lower. Cyclones greater than 1 m diameter are
now common in coal, whereas the standard DMC for Australian mineral plants has
1. FAusIMM(CP), Emeritus Professor, Julius been 350–400 mm. The standard head on coal preparation DMCs is 9 D (D = cyclone
Kruttschnitt Mineral Research Centre, diameter), whereas the recommended head for mineral applications is in the range
Sustainable Minerals Institute, The University 15–40 D, and the usual Australian practice is around 20–25 D.
of Queensland, Indooroopilly Qld 4068.
Email: t.napier-munn@uq.edu.au The use of larger, lower head DMCs potentially confers several advantages over
2. Formerly Shift Supervisor – Zinc Lead smaller, high head units, including simpler plant layouts (for example no distributor
Concentrator, Xstrata Zinc, Mount Isa Qld is required for single large DMCs, and greatly reduced pipework), easier maintenance,
4825. Email: ggibson2@xstratazinc.com.au smaller spares inventories, reduced wear, reduced medium pumping demands,
3. Process Superintendent, Fortescue and overall lower capital and operating costs. Therefore, in early 2008, Xstrata Zinc
Metals Group (formerly Xstrata Zinc). decided to trial an 800 mm diameter dense medium cyclone in place of four existing
Email: bbessen@fmgl.com.au 400 mm DSM cyclones in one of the four parallel lines in its Mount Isa plant. The main
315
purpose of the trial was to determine whether metallurgical evidence to suggest that there is little if any benefit in using
performance could be maintained (and even improved) larger heads. The head used in the Mount Isa 800 mm test
whilst enjoying the operating advantages outlined above. The installation is 9.5 D, slightly higher than the coal standard but
trial was known to carry risk, as few if any DMCs as large much lower than the norm for metalliferous applications.
as 800 mm have been used in metalliferous preconcentration, Other aspects of the installation, such as the choice of
and the cyclone was installed at a head significantly lower ferrosilicon medium and ore particle size, are conventional.
than conventionally used in non-coal applications. The trial The upper particle size is generally chosen to be less than
took place over most of the remainder of 2008. This paper one-third of the cyclone inlet diameter to avoid blockages,
describes the results, and reviews some relevant previous ie 53 mm for the 800 mm unit. The nominal top size fed to the
work which supports the conclusions of the trials. plant is actually 23 mm, well within normal practice, though
coarser material is sometimes encountered. The diameter of
THE TRIAL IN THE CONTEXT OF DENSE the inlet of the 800 mm unit is twice that of the 400 mm units,
MEDIUM SEPARATION PRACTICE and this appears to have already led to a reduced incidence of
inlet blockage. The lower particle size selected (by choice of
As noted earlier, the two key novelties of the Xstrata
feed preparation screen aperture) is governed by washability,
800 mm trial installation relative to conventional dense
and the fact that the performance declines rapidly at the fine
medium separation (DMS) practice in metallurgical (non-
sizes, generally below 3–4 mm, the ‘breakaway size’ which as
coal) applications are the relatively large size of the cyclone
mentioned earlier increases with cyclone diameter. Typical
(800 mm) and the low static head adopted (7.63 m, or 9.5 D).
lower sizes adopted in practice are in the range 0.5–3 mm.
The standard cyclone size for non-coal applications in Again, the current size of 2.5 mm is therefore in the typical
Australia is 350–400 mm. The iron ore beneficiation plants range. The coarser the lower size is made, the sharper the
in the Pilbara, the Argyle Diamond Mines DMS plant and of overall separation and the lower the medium consumption,
course the Mount Isa heavy medium plant (HMP) all have but of course there is no beneficiation of sizes below the cut-off.
gravity-fed units in this size range. However, the standard
cyclone for diamond applications in Southern Africa has been
610 mm for many decades, and in coal applications cyclones THE TRIAL AND ITS RESULTS
over 1 m in diameter are now common. The conventional
wisdom is that the cyclone size needs to be matched to the The test cyclone
particle size being treated. Mengelers (1982) pointed out that The 800 mm test cyclone was manufactured by Minco
fine particles needed higher centrifugal acceleration and thus Tech Australia Pty Ltd in Ni-hard steel to a standard DSM
smaller cyclones. More recently the concept of a ‘breakaway configuration (Figure 1). The key dimensions of the 800 mm
size’ has been proposed, which is the size below which Ep and 400 mm units and other installation information are
(note: the Ep (in specific gravity (SG) units) is the width of shown in Table 1.
the central part of the partition curve (eg Figure 10), and is The 800 mm unit was installed in Line 4 of the plant, and
a measure of the inefficiency of the separation, higher Eps its performance compared with Line 3 (with newly installed
indicating a poorer separation) increases rapidly, effectively 400 mm DMCs) and Line 2.
the lower size for efficient treatment with a particular cyclone
diameter (Bosman, 2005). The breakaway size for a 400 mm Tracer tests
cyclone is around 1 mm and for an 800 mm unit around 3 mm.
One of the best methods of assessing the performance of
However, the experimental evidence is somewhat a gravity concentration process such as dense medium
equivocal. For example, work has been done to demonstrate separation is to determine the partition curve for the process.
no detectable decline in performance with the large coal This records the per cent of material of a given density which
DMCs in current use. Clarkson et al (2002) reviewed several reports to (say) the dense product (Wills and Napier-Munn,
plant surveys of coal DMCs and concluded that ‘There is no 2006). The shape and position of the curve is a complete
significant effect on Ep for all size fractions [+1 mm] using description of the concentrating features of the process.
cyclones 700–1300 mm diameter’. Additional modelling of
However, it requires representative samples of the feed and
Clarkson et al (2002) data by the present authors confirms
products, and a full size-by-size densimetric analysis of these
that there was no effect of head on either Ep or cut-point
(separation density), no effect of cyclone diameter on Ep, and
only a small effect of cyclone diameter on cut-point.
The question of head is also moot. The appendix gives a
simple explanation for why cyclone diameter is used as the
unit of head (eg 9 D); it defines the separating power. The
selection of head is usually governed by the rules laid down
by the inventors of the DMC process, DSM, in the mid-20th
century, as listed in the widely-available DSM Handbook. For
coal applications, a head of 9 D is standard. For metallurgical
(non-coal) applications the handbook recommends heads
in the range 15–40 D, and the non-coal plants in Australia
mentioned above comply with this standard. However,
there is no experimental evidence known to the present
authors specifically in support of the much larger heads
suggested for non-coal applications. A lower head will of
course result in lower pumping costs and is likely to result
in lower wear of wearing parts. A possible historical reason
for the DSM recommendation is suggested in the appendix.
Under ‘previous work’ below we present earlier experimental FIG 1 – The 800 mm Ni-hard trial cyclone.

TABLE 1 loss of metal. This can be attributed to the higher operating

Features of 400 mm and 800 mm cyclones. medium density.
Tests were conducted at different times and operating
Nominal property 400 mm cyclones 800 mm cyclones densities over a period of some weeks. The observations
Cylinder inside diameter (mm) 400 800 showed that there was no significant deterioration apparent in
Inlet diameter (mm) 80 160 the performance of the 800 mm unit compared with that of the
400 mm, in terms of misplacement of product or reject material.
Vortex finder diameter (mm) 172 344
It is also worth pointing out that although this method has
Spigot diameter (mm) 130 260 not given quantitative results, it is a useful visual indicator,
Head (m) 8.63 (21.6 D) 7.63 (9.5 D) and could give more quantitative results if the coloured
Desired feed rate per cyclone (t/h) 48 192 products were collected, separated into colours and weighed,
or appropriate images analysed by image analysis.
Max feed rate per cyclone (t/h) 61.2 245
Nominal feed size (mm) -23 + 2.5 -23 + 2.5 Recovery and reject rate – sampling surveys
FeSi medium 65 D (Washington 65 D (Washington In assessing the performance of the new cyclone, it was
Mills) Mills) important to confirm quantitatively that there had been no
deterioration in metallurgical performance, that is, reject rate
samples using heavy liquids, followed by assaying, which and metal recoveries. This is best done by a comprehensive
is expensive and time-consuming. An alternative, which has sampling and measurement program to allow a full mass
been employed for many years particularly in coal preparation, balance to be constructed across the 800 mm unit and one or
is the use of specially manufactured tracers which are added more of the 400 mm lines so that a comparison of performance
to the feed and then recovered (usually manually) from the could be made under the same operating conditions, over
products. The colour-coded tracers are sorted by density and time. However, this was precluded by a lack of line-specific
the partition curve can then be constructed directly (Napier- mass flow measurement facilities and the time and costs
Munn, 1985). However, commercial tracers are expensive, required for a comprehensive survey and sample analysis.
and difficult to recover from a full-scale operating process. Instead, short duration (approximately one hour) parallel
As an alternative to these methods, the Xstrata Zinc staff samples were taken of the feed, product and reject screen
devised a simpler tracer technique to give a visual indication oversize for both Line 4 (the 800 mm unit) and either Line 2 or
of performance. Samples of product and reject screen oversize 3 (with 400 mm units), on several days in the period 23 June to
were taken from a 400 mm circuit running at a known medium 14 September 2008. The operating medium specific gravities
density. The solids were cleaned, dried and painted a bright (SGs) were in the range 2.55–2.75. The samples were sized and
colour, yellow for reject and white for product. Checks assayed for lead and zinc. Mass balancing methods were then
on the integrity of these ‘tracers’ were made by rerunning used to estimate the prevailing reject rate (percentage mass
them through the 400 mm circuit at the same medium to reject) and the lead and zinc recoveries. This provided
density to ensure that they reported to the same product, nine true pairs of results comparing the 800 mm and 400 mm
by measurement of the true density of the material, and by reject rates and recoveries for identical sampling periods and
assaying. The tracers were then used to test the performance conditions (Table 2).
of Line 4 (the 800 mm cyclone) by adding them to the cyclone These data can be rigorously compared statistically using
feed and observing the reject and product screens, visually the paired t-test. The results are shown in Table 3.
and with photographs. The 800 mm cyclone appears to produce a substantial increase
Figure 2 shows the result of a trial with Line 4 running at a in reject rate of 14 per cent with only small associated decreases
medium density of 2.8 using tracers collected from a 400 mm in metal recovery. The P-values imply 98.6 per cent confidence
circuit at a medium density of 2.6. It is clear that the product in the conclusion that reject rate has increased, whereas the
(B) is largely clean whereas there is some contamination of the apparent changes in recovery are not statistically significant.
reject (A), suggesting a higher reject rate and therefore some This suggests that the 800 mm cyclone has actually improved
A B
FIG 2 – Line 4 reject screen (A) and product screen (B), 17 June 2008, medium SG 2.8. Material collected
from 400 mm cyclones operating at 2.6 SG. White – product; yellow – reject.

TABLE 2
Reject rates and recoveries for 400 mm and 800 mm cyclone comparisons.
Date Medium SG Reject rate Pb recovery % Zn recovery %

Lines 2/3 Line 4 Lines 2/3 Line 4 Lines 2/3 Line 4
23 Jun 08 2.60 30.4 57.2 89.18 82.94 90.61 84.66
23 Jun 08 2.60 33.1 54.6 90.09 90.01 87.71 87.52
07 Jul 08 2.70 46.7 63.1 93.22 94.19 95.46 97.55
07 Jul 08 2.70 48.8 62.2 92.64 94.44 95.08 97.63
15 Jul 08 2.65 51.1 65.4 91.27 89.35 92.40 89.56
03 Aug 08 2.65 28.5 65.1 96.73 93.38 98.40 92.67
14 Aug 08 2.55 39.7 41.1 91.10 95.08 93.13 95.29
14 Aug 08 2.65 43.4 44.9 91.58 92.28 95.51 93.24
14 Aug 08 2.75 58.2 52.5 88.98 89.68 94.37 94.95
TABLE 3 Lines 2, 3 and 4. This suggests that using the 800 mm unit is
Summary of 400 mm and 800 mm comparisons. unlikely to lead to any significant change in metallurgy.
Quantity Reject Pb recovery Zn recovery The effect of medium density

rate (%) (%) (%) Figure 4 shows a time series plot of the recovery and reject
Lines 2/3 mean (400 mm) 42.2 91.6 93.6 rate data.
Line 4 mean (800 mm) 56.2 91.3 92.6 The surveys conducted on 14 August were specifically
intended to assess the effect of changing the operating
Difference (800 mm–400 mm) 14.0 -0.4 -1.1
medium density. Figure 5 shows the effect of the change
Number of pairs 9 9 9 in density from 2.55 to 2.75 on the reject rate and metal
P-value for t-test a
0.014 0.72 0.36 recoveries of Line 4 (the 800 mm cyclone). As expected the
a – The P-value measures the risk of being wrong in concluding that the observed difference is reject rate increases with density, by about 11 per cent, with
real (significant). P-values <0.05 are generally accepted as being statistically significant. P = 0.05 corresponding small declines in recovery. Similar trends were
implies that we are 95 per cent confident that the observed difference is real. A conservative obtained for the 400 mm units.
2-sided P is used here as we have no a priori assumption as to the direction of any change.
Recovery and reject rate – plant data
metallurgical performance over the 400 mm units. There is
Another way of assessing how the 800 mm DMC is operating
certainly no evidence that performance has deteriorated.
is to compare plant performance before and after the
Figure 3 shows the relationship between metal recovery installation. This is difficult as no formal experimental design
and reject rate for the two units using the data in Table 2. has been followed, and conditions change with time, in
One would normally expect an inverse relationship akin to a particular the ore blend, medium density and other variables,
conventional grade-recovery curve, with recovery declining which would tend to confuse the effect of the cyclone change.
with increased reject rate. Statistical tests show in this case that It would therefore be misleading simply to compare the mean
there is no such relationship, that is, the lines are essentially values before and after the change. Also the data, calculated
horizontal and the recovery is approximately constant across routinely by the plant, is based on limited sampling and
all reject rates, for both metals. This probably has more to measurement, and must be treated with caution. In addition,
do with the relatively large uncertainties in the data and the the comparison is not a clean one between 400 mm and
scatter apparent in Figure 3 than whether there actually is 800 mm DMC, because the effect of the single 800 mm line
a relationship. However, Figure 3 shows that the data does will tend to be diluted by the dominant effect of the remaining
not support a conclusion that there is any difference between three lines of 400 mm units.
A B
FIG 3 – Metal recovery versus reject rate for Pb (A) and Zn (B) for the sampling surveys.

100
Reject rate & metal recoveries (%)

90
80
70 Pb recovery (%)
Before change After change Zn recovery (%)
60
Reject rate (%)
50
40
30
20
10
0
01-Jun-08 11-Jun-08 21-Jun-08 01-Jul-08 11-Jul-08 21-Jul-08 31-Jul-08
Date
FIG 4 – Plant reject rate and metal recovery, 1/6–31/7 (shutdown 24/6–2/7).
Figure 5 suggests that it would be unwise to perform a Clearly there is a lot of scatter, due to the errors in determining
simple comparison before and after the change in cyclone, recovery and reject rate, and to the real effects of the several
as there were other changes going on (as normal in any variables listed above. Also there is a tendency for the ‘after’
concentrator) which are not due to the cyclone, as noted condition to show higher metal recoveries and correspondingly
earlier. In particular the reject rate shows a sharp decline lower reject rates. However, there is a general negative trend
just before the shutdown on 24 June with a slow recovery as expected (high reject rates being associated with lower
after the plant restarted, which cannot be attributed to the recoveries, akin to a conventional grade-recovery curve), and
cyclone. The two metal recoveries follow each other closely the data seem to be on the same recovery-reject rate curve for
as would be expected. the two conditions – before and after – for both metals.
Probably the best way to obtain the most from the To confirm this, trend lines were fitted to the data. It would
information is to compare the reject rate – metal recovery be unwise to fit straight lines as these would not extrapolate
relationship for the two conditions, before and after the correctly to the point at which recovery = 100 per cent for
800 mm unit was installed. Figure 6 shows the comparison. zero reject rate, nor to high reject rates where recovery would
be expected to decline rapidly. A model used to fit grade-
100 recovery curves, the so-called AREV function (Vera, Franzidis
90 and Manlapig, 2000), was therefore used:
Reject rate/recovery (%)
80
Recovery = 100 - a sinh (b [RR - 1]) % (1)
70
60 where:
50
a, b parameters estimated from the data using non-linear
40
Pb recovery (%) least squares regression
30
20 Zn recovery (%) RR reject rate (per cent)
10 Reject rate (%) These lines are shown on Figure 6. Although they look
0 straight over the range of the data, they do have the required
2.50 2.55 2.60 2.65 2.70 2.75 2.80 properties of a theoretical recovery-reject rate curve.
Medium SG The trends before and after the change are absolutely
identical for lead (Pb) recovery and only slightly different (not
FIG 5 – Effect of medium density on 800 mm cyclone performance. significantly so) for zinc (Zn) recovery. This suggests that the
A B
FIG 6 – Plant recovery-reject rate relationship for Pb (A) and Zn (B) before and after 800 mm dense medium cyclone installation.

recovery-reject rate relationships for the two metals are not

SG 2.7, 90 t/h
significantly different with and without the 800 mm unit. The
800 mm period seems to be operating at a higher point on the SG 2.7, 120 t/h
0.08 SG 3.0, 90 t/h
curve (higher recovery, lower reject rate) but it is not clear
whether this is due to the cyclone or some other factor. If the 0.07 SG 3.0, 120 t/h
former, one can probably move along the curve if desired by 0.06
changing the operating conditions, eg by increasing medium
Ep (SG units)
0.05
density to increase reject rate.
0.04
Wear, downtime and plant design 0.03
Preliminary experience to May 2009 suggested that the life of 0.02
the 800 mm at the low head (9.5 D) is significantly greater 0.01
than that of the 400 mm units running at the higher head
0.00
(21.6 D) – over 20 weeks compared with 14 weeks for the
5.5 6.0 6.5 7.0 7.5 8.0 8.5
400 mm cyclones. This would be expected, as wear is a strong
function of the velocity of the slurry entering the chamber, Head (m)
and this would be less for the lower head. The 800 mm unit
is also more than 30 per cent cheaper than 4 × 400 mm units FIG 7 – Ep versus head for tests with a 610 mm dense
plus distributor. With the additional life, preliminary cost- medium cyclone in diamond applications.
benefit analysis suggests that the annual spares cost for the
4 × 800 mm units will be less than half that of the 16 × 400 mm clear that there is a tendency for the Ep to rise (ie performance
units. Additional savings can be expected in reduced change to decline), possibly due to increased turbulence. It should be
out time, and increases in 800 mm life are also expected. noted that the lower head (9.7 D) is very similar that used for
the 800 mm unit in the Mount Isa tests (9.5 D), and the Ep
In addition there has been a reported reduction in cyclone
values are relatively low at this head. This is circumstantial
blockages and therefore plant downtime, presumably due to
evidence that there is little risk in running the 800 mm unit at
the larger cyclone inlet.
the low head. Note that the 610 mm unit is commonly used
An important element of the simplification of the circuit is in diamond applications, at a standard head of 8 m or 13 D.
the elimination of the distributors needed to distribute feed
between the four 400 mm DMCs in each line. The distributor JKMRC dense medium cyclone model prediction of the Xstrata
contributes to operating cost due to wear, is difficult to change
out, and earlier work on coal has shown that distribution 800 mm performance
contributes to process inefficiency (Davis, Wood and Lyman,
In 1999 Michael Dunglison published his JKMRC PhD thesis
1985). These disadvantages are all overcome with the single
on the modelling of the dense medium cyclone, using data
800 mm unit per line.
from a variety of both coal and metalliferous applications at
full and pilot plant scales (Dunglison, 1999; Dunglison and
PREVIOUS WORK Napier-Munn, 1996, 2000). He also implemented the model
To complement the evidence of the plant trial of the 800 mm in an Excel spreadsheet, allowing process simulations to be
unit, two other historical developments have been reviewed. carried out for any given cyclone installation and operating
The first was a program of experiments with a 610 mm DMC conditions. This spreadsheet has been used here to simulate
at a range of static heads, in diamond recovery applications. the performance of the Mount Isa 400 mm and 800 mm units
The second is a mathematical model of the DMC process over a range of heads including those used in the plant.
developed from both coal and mineral applications, which is The model requires details of the cyclone geometry,
used here to predict the likely outcome of applying a larger head, ore feed rate, medium density, medium grade (size
cyclone and low head to the Xstrata situation. specification), and the ore characteristics (size distribution,
and size-by-size density distribution and metal assays). In
Study of a 610 mm dense medium cyclone running at a lower head the present case there was no current ore information other
A comprehensive study was conducted of a 610 mm gravity- than size distribution, which was taken as the average of the
fed DMC treating diamondiferous ores. The details are given samples taken for assay in the present work. Accordingly
in Napier-Munn (1977). The cyclone was of DSM proportions some information was obtained from an earlier JKMRC report
and thus similar to the Mount Isa 800 mm unit, though on surveys conducted in the Mount Isa HMP in the 1980s
smaller. It was rated at 100 t/h but was run in the test work up (which Dunglison included in his database) (Scott, 1987). The
to 147 t/h. The objective of the work was to explore the effect data are given in Table 4.
of static head, feed density and feed rate on cut-point density Note that this washability is very unlikely to be the same
and Ep-value, for different ore types and ferrosilicon media as that currently being treated in the HMP, not least because
grades. The cut-point and Ep were determined by sampling
the current ore is a variable blend of Blackstar and George
and densimetric analysis.
Fisher. Moreover, the densimetric data were only available
Figure 7 shows a plot of Ep versus static head for four for the finer sizes (3 mm and less), and no size-by-size-by-
groups of tests at different feed densities and feed rates. The density data were available for the metal assay. In every
medium grade was 100 D (somewhat finer than the 65 D sense therefore this washability is a simplification of the truth
used at Mount Isa, but probably cleaner and thus of similar and may be significantly different to that currently being
viscosity), and the ore type a kimberlite in the nominal size treated. In addition the Dunglison model, though based
range -12+0.4 mm. The heads were in the range 5.9–8.1 m on a comprehensive data set, has not been independently
(9.7–13.3 D for the 610 mm cyclone). validated and its predictions should therefore be treated
The error in determining the Ep is quite large and thus one with caution. Also in simulating the 800 mm unit in a high-
should not over-interpret the trends in Figure 7. However, it is density application the model is being used outside the limits

Table 4
95
Mount Isa heavy media plant feed ore washability used in
simulations (from Scott, 1987; quoted as ‘average ore’). 90
Zn recovery (%)
Float Specific Gravity Weight % Zinc % 85
Current heads
2.75 6 0.2
80
2.90 29 0.6
800mm cyclone
3.05 12 2.5 75
400mm cyclone
3.20 8 5.1 70
3.35 7 8.3 30 40 50 60 70
3.70 12 14.6 Reject rate (%)
4.05 10 15.0 FIG 9 – Zn recovery versus reject rate for the 400 mm and
+4.05 15 12.1 800 mm cyclones (simulations using Dunglison model).
density would be required to obtain a higher reject rate and

of its database. The simulation predictions discussed below
thus lower metal recovery. For example, the model suggests
therefore should not be taken as a literal prediction of current
that at a feed medium SG of 2.90 the 800 mm reject rate would
performance. The purpose is to obtain a relative qualitative
rise to 56 per cent at a Zn recovery of 81 per cent, close to that
comparison of the performance of the 400 mm and 800 mm
of the current 400 mm simulated performance. If the 800 mm
cyclones operating at different heads, based on some plausible
head were raised 1 m and the medium were converted from
assumptions.
65 D to the coarser (and cheaper) 48 D with a feed medium
The simulations were conducted over a range of heads for SG of 2.85, then the 800 mm would maintain a Zn recovery of
the two cyclones, including the actual heads currently used. 81 per cent but at a slightly higher reject rate of 57 per cent (for
The feed medium SG was assumed to be 2.65, with 65 D this particular ore).
ferrosilicon plus ten per cent contamination at a temperature
In passing, it is interesting to note that the Dunglison model
of 35°C. The predicted reject rates and corresponding Zn
gives the following prediction for the prevailing cut-point,
recoveries are shown in Figure 8.
ρ50, and Ep-value as functions of the differential (u/f – o/f
It is clear that reject rate increases and metal recovery medium SG), for both cyclones:
decreases as head increases in both cyclones. If the predictions
are of the right order, then it suggests that the 800 mm t50 = 2.83 + 0.42 diff (2)
unit at a head of 9.5 D will deliver a lower reject rate and
correspondingly higher recovery than the 400 mm units at a Ep = 0.030 + 0.034/diff (3)
head of 21.6 D (the current heads). A plot of Zn recovery as a
function of reject rate (Figure 9) shows that both cyclones are Thus if the observed differential is 0.4, the cut-point is
essentially operating on the same reject rate – recovery curve, predicted to be 3.00 and the Ep 0.115.
which confirms the general conclusion of Figure 6. This suggests that, if these predictions are correct, then
Interestingly, despite the necessary caveats attached to these the calculated ρ50 and Ep can be inserted into the following
simulations, it turns out that the recovery-reject rate curve of equation (note: based on a derivation of Rong and Lyman
Figure 9 predicts a Zn recovery of 92 per cent at a reject rate of (1985), Scott showed that for separations without short-
41 per cent, almost exactly the same as that obtained from the circuiting, the partition curve could be described by this
curve fitted to the actual production data (Figure 6). equation (Scott and Napier-Munn, 1992)) to predict the full
The simulations predict that the current reject rate of the partition curve of the prevailing separation:
400 mm units is 58 per cent at a Zn recovery of 82 per cent, 100
and that of the 800 mm unit 41 per cent at a Zn recovery of Pi = % (4)
ln 3 (t 50 - t i)
92 per cent. In order for the 800 mm unit to approach the same 1 + exp = G
Ep
operating point as the current 400 mm units (if this is desired),
other variables can be used to move performance along this where:
curve, particularly the medium density. A higher medium Pi is the partition number for density ρi
FIG 8 – Predictions of reject rate and Zn recovery as a function of head, for 400 mm and 800 mm cyclones (simulations using Dunglison model).

100 (for new plants), and thus reduced capital costs.

Operating advantages include reduced pumping costs,
80 less operational problems (including poor distribution
and inlet blockages) and reduced operating costs. A
% to product
60 particular advantage is no requirement for distributors

which introduce wear costs, maintenance difficulties
ρ50 = 3.00
40 and process inefficiency.
Ep = 0.115
•• Painted tracer ore particles taken from the process are a
20 useful way of testing performance.
The trial has therefore shown that there is no metallurgical
0 disadvantage in using the larger cyclone in the Xstrata
2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 application, and there are significant advantages in terms
of operating cost, plant design and possibly metallurgical
Density (SG)
performance.
FIG 10 – Simulated partition curve for dense medium Clearly these conclusions have implications for the design
cyclone separation with differential = 0.40. of future DMC plants for metalliferous applications. It seems
that such plants should follow their coal cousins more closely
Figure 10 shows the predicted partition curve for an – employing large cyclones at low heads.
observed differential of 0.4.
Equations 2–4 will only apply to the conditions simulated, ACKNOWLEDGEMENTS
but they do suggest a simple way of monitoring the complete The authors thank Sandra Ferraz and Victoria Butten for their
cyclone performance. Differential will of course change with assistance in the sampling and laboratory work and Vince
medium density, medium grade, head and other variables. It Garner who supported the trial and provided resources. The
may be worth conducting test work in the plant to calibrate paper is published by kind permission of Xstrata Zinc.
this predictor (ie Equations 2 and 3) to the actual operating
conditions in the plant.
REFERENCES
Bosman, J, 2005. Dense medium cyclones – sizing and selection,
CONCLUSIONS in Proceedings Ninth International DMS Symposium, Paarl
An 800 mm DSM cyclone has been trialled at a low head in (DMS Powders).
the Xstrata Zinc lead-zinc preconcentration DMS plant. The Clarkson, C J, Edward, D J, Davidson, J and Lahey, A E, 2002. Analysis
metallurgical performance was monitored using painted ore of large diameter cyclone plant performance, in Proceedings Ninth
particles as tracers, and by estimating metal recovery and Australian Coal Preparation, pp 68–89 (Australian Coal Preparation
reject rate from assayed samples and from production data. Society: Dangar and Indooroopilly).
Cyclone life was also monitored, and relevant previous work Davis, J J, Wood, C J and Lyman, G J, 1985. Effects of operating
was evaluated. The following conclusions can be drawn: variables on dense medium cyclone operation, in Proceedings
Third Australian Coal Preparation, pp 107–124 (Australian Coal
•• There is no evidence that the 800 mm cyclone exhibited
Preparation Society: Dangar and Indooroopilly).
poorer metallurgical performance than the 400 mm
units. Most of the data showed that they are very similar. Dunglison, M E, 1999. A general model of the dense medium cyclone,
The sampling suggested that the 800 mm was superior, PhD thesis (unpublished), University of Queensland (Julius
Kruttschnitt Mineral Research Centre).
operating at a higher reject rate and unchanged recovery.
Other data suggested that they operated on different Dunglison, M E and Napier-Munn, T J, 1996. Progress in modelling
parts of the same reject rate – recovery curve and simple the dense medium cyclone, in Proceedings Hydrocyclones ’96
(eds: D Claxton, L Svarovsky and M Thew), pp 203–213
action can be taken to move along that curve if desired.
(Mechanical Engineering Publications: London).
•• Previous research on diamond applications indicated
Dunglison, M E and Napier-Munn, T J, 2000. The application of a
that DMCs can run at lower heads than conventionally new model of the dense medium cyclone, in Proceedings Dense
used without compromising performance, confirming Media 2000 – 7th Samancor Symposium on Dense Media Separation,
the results of the Mount Isa trial. pp 119–135 (Samancor Limited: Benmore).
•• A DMC process model also confirmed that there is no Mengelers, J, 1982. The influence of cyclone diameter on separating
penalty in running large cyclones at low heads in such performance and economy, in Proceedings Tenth International
applications. In particular it suggested that the 400 mm Coal Preparation Congress, pp B4.1–B4.10 (Central Coalfield Press:
and 800 mm units would run on the same reject rate – INDE).
recovery curve, as observed in practice. Munro, P D, Schache, I S, Park, W G and Watsford, R M S, 1982. The
•• Reject rate increased and metal recovery decreased with design, construction and commissioning of a heavy medium
plant for silver-lead-zinc ore treatment – Mount Isa Limited,
increase in medium density in the 800 mm DMC, as
in Proceedings 14th International Mineral Proceedings Congress
would be expected. (Canadian Institute of Mining and Metallurgy: Montreal).
•• Preliminary wear monitoring has suggested that the life Napier-Munn, T J, 1977. Dense medium cyclones in diamond recovery,
of the 800 mm unit is at least 50 per cent longer than that MSc thesis (unpublished), University of the Witwatersrand,
of the 400 mm unit, leading to annual spares costs across Johannesburg.
the plant less than half those of the smaller cyclones.
Napier-Munn, T J, 1985. Use of density tracers for determination of
There has also been a reported decrease in the incidence the Tromp curve for gravity separation processes, 94:C47–C53
of inlet blockage. Both these effects will contribute to (Institute of Mining and Metallurgy: London).
reduced plant downtime.
Rong, R X and Lyman, G J, 1985. Computational techniques for coal
•• Plant design advantages include simpler circuitry, washery optimisation – parallel gravity and flotation separation,
smaller footprint, potentially lower building height International Journal of Coal Preparation and Utilization, 2:51–67.

Scott, I A, 1987. Metallurgical data from surveys of the Mount Isa DMS experience different centrifugal forces. Denser particles will
preconcentrator and pilot plant 1983–1986, Julius Kruttschnitt experience a larger outward force than less dense particles.
Mineral Research Centre report to Mount Isa. The centrifugal force is opposed by a fluid drag force which
Scott, I A and Napier-Munn, T J, 1992. A dense medium cyclone depends on the properties of the medium and the prevailing
model based on the pivot phenomenon, Transactions of the flow regime, both of which are complex in DMC, not least
Institutions of Mining and Metallurgy, Mineral Processing and because the medium is unstable and segregates in the cyclone,
Extractive Metallurgy, 101:C61–C76. so the medium density and viscosity vary in different parts of
Vera, M A, Franzidis, J P and Manlapig, E V, 2000. The locus of flotation the cyclone. However, the fate of a given particle will depend
performance, in Proceedings XXI International Mineral Processing in large part on the balance of these two forces. In general,
Congress (ed: P Massacci), pp 74–82 (Elsevier Science: Chatswood). particles denser than the medium will move to the periphery
Wills, B A and Napier-Munn, T J (ed), 2006. Wills’ Mineral Processing and report to the apex, and less dense particles will be carried
Technology: An Introduction to the Practical Aspects of Ore Treatment by the medium toward the axis and report to the vortex finder.
and Mineral Recovery, seventh edition, 456 p (Butterworth-
Heinemann).
Equation A1.1 in some senses describes a relative
‘separating force’. For a given particle entering the cyclone at
the periphery, this force is proportional to Vi2/R, or 2Vi2/D
APPENDIX where D is the cyclone diameter.
As noted earlier, this force is driven by the static head on
The influence of static head on separating force the cyclone, H. The steady head system can be simplified to
Note: The following treatment is simplified. It ignores friction losses that of a tank (the steady head) with an exit orifice near the
in pipes and the cyclone, the way pressure energy is used in the DMC, bottom (the cyclone inlet). Simple fluid mechanics tells us
medium rheology and instability, the true mechanism of separation in that the velocity of fluid exiting such an orifice (ie entering the
the DMC, detailed ore particle properties and other factors. cyclone) is given by:
The separation in a DMC is driven by the kinetic energy
available from the flow of medium, reflected in the pressure Vi = Cv 2gH (A1.3)
drop across the cyclone. This energy is delivered either by a
where:
pump, or by a steady head arrangement. The latter is usually
a mixing box in which the feed ore is mixed with the medium, g acceleration due to gravity
connected to the cyclone inlet via a pipe with a vertical drop H static head
of the desired magnitude. In the latter case, as in Mount Cv velocity coefficient, which here would include a
Isa’s gravity-fed DMCs, the amount of energy available is resistance to flow caused by the cyclone and its load of
determined by the static head, less friction losses in the pipe medium, ie the orifice is not open to atmosphere
and losses associated with the cyclone inlet.
Combining Equations A1.1 and A1.3 gives:
The separating effect is largely determined by the centrifugal
force acting on an ore particle: Fc = 4mgCv 2 ` H j (A1.4)
D
2
Fc = mv (A1.1)
r Equation A1.4 shows that the separating force is proportional
to H/D, the ratio of head to cyclone diameter. It is for this
where:
reason that DMC heads are often quoted in ‘numbers of D’
Fc centrifugal force which gives some indication of relative separating force,
m net mass of the particle and normalises head to the chosen cyclone size. The cyclone
v local tangential velocity of the particle size is generally chosen for capacity reasons, and sometimes
to match the prevailing particle size, smaller cyclones being
r radius of rotation used for finer material.
At the entry to the cyclone, v = Vi, the inlet velocity, and The difference between the coal and mineral conventional
r = R, the radius of the cylindrical section of the cyclone heads referred to earlier in the paper may be an accident of
(assuming the particle is at the periphery). The net mass of history. The DMC was originally developed by the DSM
the particle is the true mass less the mass of fluid displaced organisation in Holland around the time of World War
(the buoyancy effect). For a spherical particle therefore: II, and the patents vested in its wholly-owned subsidiary
2 Stamicarbon. It was comprehensively researched in their
m = r d (t s - t f ) domestic coal applications, which led to the adoption of the
6 (A1.2)
9 D head. Subsequent experience around the world confirmed
where: that this was a satisfactory standard. However, when it
d particle diameter came to applying DMC to mineral separations there was no
previous experience, and it is believed that a safety factor
ρs particle (solid) density
may have been built in to allow for the higher densities of the
ρf fluid (medium) density particles to be separated, leading to a recommendation that
Equations A1.1 and A1.2 demonstrate that particles of the head should be in the range 15–40 D. This factor has remained,
same size but different density (ie different masses) will despite evidence that it may be unnecessary.

Sampling, metallurgical
accounting and control
Contents
Using metallurgical data to drive

continuous improvement
R Dunne1
ABSTRACT
Continuous improvement on metallurgical treatment plants is possible only when good
data and analysis of the data are available. Good data implies appropriate and accurate
mass measurement coupled with good sampling practices and accurate chemical
analysis, especially when carrying out monthly metallurgical metal reconciliations.
Adequate characterisation of key samples (for example, tailings, feed, concentrates
and others) at timely intervals is an important part of a good continuous improvement
program. The data from this will help identify opportunities to improve operation
performance and thus profitability. Examples are provided of projects that have led to
improved plant performance based on good metallurgical data collection and analysis.
INTRODUCTION
Taggart (1947) in his handbook provides insight to the method of material/metal
balancing at that time. Taggart referred to these as ‘milling calculations’ and they
were mostly used to determine the performance of a metallurgical treatment plant, the
performance of individual unit operations and to gauge the accuracy of smelter returns.
He provided formulas for calculating weight splits and mineral recoveries based on
chemical assays, particle size distribution or water to solids ratios using the now well-
known two product, three product and n-product approach. The reason for developing
and using them was attributed to the fact that most metallurgical plant operations were
continuous rather than batch, that large quantities of solids were treated and that water
was added during the treatment process making if difficult or impossible to weigh
the different slurry streams (that is, there was no mass flow measurement available
at that time). Taggart stressed that although the formulas are theoretically correct, the
accuracy of the answers are wholly dependent upon the accuracy of the sampling and
assaying systems and methodology. He commented that monthly balances based on
automatically weighed feed with accurately measured moisture content and appropriate
assays of tailings and concentrate should be within reasonably close limits to the dry
weights of products dispatched from site. Furthermore, that over a year the cumulative
balance should reconcile very closely and that differences on a monthly basis should be
both positive and negative. The main reason for this is that balances over longer time
frames obviate errors in measuring holdups (inventory) in the plant. Taggart makes the
point that biases in the material balances that are continually either positive or negative
need to be closely investigated.
So what is the situation today? We have better and more appropriate sampling and
accurate analytical systems, better and more accurate weightometers to measure feed,
online moisture measurement equipment and mass flow equipment to measure the
mass of solids in slurry streams. A Metallurgical Accounting Code of Practice (AMIRA,
2007) was published to aid practitioners in implementing best practice systems on their
plants (Gaylard et al, 2009). More specific details of the most appropriate sampling
systems and mass measurement equipment including calibration methodology
are provided by Holmes (2004) and Wortley (2009) respectively. Finally, extensive
work has been done over the years to determine the errors in the measurements
required for a metallurgical balance and also how to appropriately handle these in
the application of the metallurgical balance to provide a better final outcome (Bazin
et al, 2005; Lachance and Flament, 2011; Lachance, Leroux and Gariepy, 2012).
COLLECTION OF METALLURGICAL DATA AND REPORTING

Metallurgical data on treatment plants are collected for different reasons and at
different times of the day, month and during the year. Examples of the type of data
1. MAusIMM, Principal, Rob Dunne Consulting, collected are provided below and it is not an exhaustive list. Reporting of the data
Kalamunda WA 6076. may be on a daily, weekly, monthly, quarterly and annual basis and each reporting
Email: rob.dunne13@gmail.com system has a different purpose:
327
R Dunne
•• Hourly and shift data collection and samples are composite samples, diagnostic leach and a batch gravity
taken on many plants and these are usually process test on the composite tailings sample. For gold plants,
control orientated. They may include particle size thermogravimetric analysis and inductively coupled
analysis, pH, reagent addition rates, slurry/pulp plasma analysis of elements are performed on activated
densities plus many others for flotation plants. These carbon from the elution and regeneration circuits. All of
may be measurement continuously or as individual the data provides a more detailed picture of subtle or step
measurements. Continuous measurement systems are changes that may be occurring on the plant over time.
usually coupled with control systems while individual Lastly, reagent and consumable inventories are taken at
measurements imply manual control. On gold leaching month-end to provide consumption rates on a tonnes
plants, it is usual to measure particle size, pH, cyanide treated basis. Detailed monthly reports are provided to
concentrations in the first and last leach tanks, gold in mine management and head office. Production figures
solution both around the electrowinning circuit and the are compared to budget and reasons for reasonable
cyanide leach tailings stream and carbon concentrations differences require explanation.
in each of the adsorption tanks. On many gold plants the •• The quarterly production report is a consolidation of
carbon activity is also measured on a regular basis after three monthly plant data and is provided to the stock
carbon regeneration. exchange as part of the required financial reporting cycle
•• Daily samples and mass measurements are primarily for of a mining company. It is available to shareholders and
metallurgical accounting purposes, but they also serve as the general public. Production data such as tonnes milled,
quality assurance data. They demonstrate at a base level head grades, overall metal recovery, metal produced
if the process is on track. For copper or other base metal and metal sold are provided. Some companies also give
treatment plants this would include samples of feed, cost data on a metal unit basis. Comparisons are made
concentrate and tailings streams for metal assays, the to the last quarter and the same reporting quarter for the
feed tonnage milled for the day as well as the quantity previous year. Reasons for large differences are explained.
of concentrate produced. For gold plants samples of the •• Annual production report is a consolidation of monthly
feed, tailings (solid and solution), loaded and eluted production data for the financial reporting year of a
carbon, electrowinning feed and tailings solution mining company. The reporting format is similar to the
samples are taken for gold assay. Tonnes milled for the quarterly reports and the information is available to
shift and day are measured together with the quantity of shareholders and the public.
carbon eluted and the volume of solution treated in the
electrowinning circuit. The data is reported daily to the
mill manager and a summary is provided to the general VERACITY OF METALLURGICAL BALANCES AND RECONCILIATION
manager, mine and geological managers plus any other The monthly metallurgical balance is usually in ‘mathematical’
interest parties. terms expressed as follows:
•• Weekly data on many plants are consolidated to better
reconcile daily data that has reasonable variance. On
many gold plants it is usual to smelt the weekly gold
produced from the electrowinning circuit. The smelted
gold, in the form of doré bars, is weighed and sampled Most metallurgical balances do not show the last term in the
prior to dispatch on the same day to a refinery. The brackets. Taggart (1947) included the term to account for dust
weekly gold production estimates that are determined losses containing metal. For flotation concentrate transport
from gold leach solids (feed and tailings), gold eluted and shipping, a dust loss estimate is usually included in
from carbon and gold electrowon (solution in and out) are the balance for reconciling concentrate tonnage transported
compared to actual gold produced. Large discrepancies from the mine site and received by a smelter. Another form
are an alert for potential theft. Weekly gold production of loss can be associated with theft, this is an important
reports are usually also sent to the corporate office who consideration on gold treatment plants (Wunderlich, 1998).
manage gold sales. Another potential loss, though it may be argued that it is part
•• There are two parts to monthly samples and data analysis. of the inventory, relates to metal contained in spillage from
In the first instance samples at month-end are taken to operating equipment that seeps into the ground and concrete
determine the metal inventory in the plant to allow and is not collected during spillage clean-up. For gold
for a complete metallurgical accounting balance to be treatment plants a large quantity of gold is usually recover
performed. The monthly metal accounting system accrues during the clean-up when dismantling the plant after final
data for metal into the plant (based on daily milled tonnes closure of the mine (de Ruijter, 1996; Wardell-Johnson, 2003).
and grade) and metal out of the plant (concentrate taken Unfortunately, there is a dearth of information as usually
off-site or gold bars dispatched) as well as accounting the plant is sold and the new owner will recover the gold
for metal inventory changes and metal shortfalls on the during dismantling the equipment. Consequently, the gold in
plant and additions/shortfalls based on data received spillage loss is never recorded.
back from the smelter or refinery. The metallurgical Interestingly there are very few papers that provide
balance prior to any adjustment provides actual versus information on the comparison of the measured (theoretical)
theoretical metal production. One can either report this versus actual reconciliation for monthly plant metallurgical
data or make adjustments to get the two numbers to balancing. This being the difference in the metal measured
reconcile to a zero balance. On many plants it is usual to in the feed to the plant versus the metal as final product
generate monthly composite samples, made from small dispatched from site plus metal in the tailing and
portions removed from the daily production samples, accounting for differences in metal inventory in the circuit.
for detailed characterisation. This usually includes At the Chimney Creek mine, Nevada USA it was reported
size by size metal assays on the composite samples of (Marsden and Arnold, 1990) that the cumulative deviation
the feed, concentrate and tailings samples, detailed between actual production and theoretical after three years
mineralogy/deportment on concentrate and tailings of operation was +0.42 per cent with a standard deviation

Using metallurgical data to drive continuous improvement
monthly of 8.7 per cent. The large monthly variations were that was one the original objectives. The payback period was
attributed to the inherent difficulty of estimating monthly some 24 months.
gold in plant inventories. A recent paper by Giblett, Dunne
and McCaffery (2012) provides a good overview of the KCGM coarse gold recovery project
plant gold reconciliation differences for ten of Newmont’s The Kalgoorlie Consolidated gold mine (KCGM) suffered from
operating mines. For eight of the plants the variance is large negative reconciliation biases of up to seven per cent
within a band of ±5 per cent while two of the plants had over a four period the during the early 2000s (Giblett et al,
much larger positive variances of between six per cent and 2012). Despite the installation of best practice, multistage cross
18 per cent. The large positive differences in reconciliation stream samplers to improve the reliability of the measured
for these two plants were attributed to the presence of coarse mill feed grade, the bias remained. It was postulated that the
nuggety gold in the feed. Plants that were in the ±5 per cent presence of coarse gold in the feed and intermittent losses to
and lower variance range tended to have a predominance of tailings was mainly to blame for the biased reconciliation.
finely disseminated gold in the feed. Extensive mineralogical examination and continuous gravity
Typically, metallurgical accounting balances as mentioned test work on the tailings showed coarse free gold equal
above are finalised monthly, representing a practical to around one per cent of the gold in feed being present in
and convenient short-term interval to reliably measure the final tailings stream. The installation of a gravity circuit
and report actual production information. Preliminary to treat the ball mill circulation load and the inclusion of an
balances are frequently done weekly; however, as pointed intensive cyanidation circuit to treat the gravity concentrate
resulted in a 0.45 per cent increase in overall gold recovery.
out by Taggart in his 1947 handbook the imprecision in
The implementation of the gravity circuit also decreased the
inventory measurements is too great for appropriate weekly
reconciliation bias.
reconciliations to be accepted as conclusive. The month-end
metal balancing or reconciliation process is the accepted
standard in the mining industry.
Cadia rougher-scavenger composite treatment circuit
Long-term monthly tailings characterisation and detailed
Metallurgical accounting systems as recommended by
gold deportment studies on composite Cadia flotation
Giblett, Dunne and McCaffery (2012) should be fit for
tailings samples showed that up to 50 per cent (0.10 g/t) of
purpose and preferably engineered into the original design. the gold was present in binary/ternary composites associated
Woollacott and Stange (1987) also provide useful guidelines with sulfide minerals. Obviously, this finding demonstrated
for obtaining reliable metal/material balances from plant that the greatest potential for improving gold recovery
data. On most operating plants, retrofitting of appropriate resides with treating the gold bearing composites. Following
sampling systems and mass measurement instruments extensive laboratory and pilot plant test work a flow sheet
requires large and costly engineering modifications, thus was developed that was shown to economically recover and
demanding a proper and compelling business case for the treat the tailings gold composites (Cesnik et al, 2005). The flow
expenditure. The recommendation made by Giblett, Dunne sheet consisted of recovering the coarse grained composites
and McCaffery (2012) is based on the observation that on from the rougher-scavenger circuit using a selective collector
a number of plants in the Newmont group very good in combination with appropriate froth crowding and a
reconciliation is ‘seemingly achieved in violation of some of flotation booster-impeller to enhance the suspension of
the major principles of the Metallurgical Accounting Code’. coarse material without creating excessive turbulence. The
However, they do stress there should be no compromise scavenger concentrate was cleaned and dewatered in the
on providing a high-quality tailings sample collection cleaner-scavenger flotation circuit. The cleaner-scavenger
system. For new plants there is no reason not to install concentrate was then cycloned and the overflow was
the appropriate equipment in the appropriate location as recycled with the cleaner tailing back to join the feed of the
recommended in the Metallurgical Accounting Code of rougher circuit. The cyclone underflow containing the coarse
Practice (AMIRA, 2007). composites was reground in the available Vertimill® that had
excess capacity. To improve the copper grade of the final
EXAMPLES OF CONTINUOUS IMPROVEMENT PROJECTS concentrate, the coarse cleaner concentrate generated from
treating the flash flotation concentrate was also reground in
Continuous improvement is possible only when good data the Vertimill® before being sent to the recleaner circuit. In the
and analysis are available and the following examples show ‘old’ circuit the coarse cleaner concentrate was combined with
the value of this approach. the recleaner concentrate to make a final combined copper
concentrate. The implementation of the circuit modifications
New celebration tailing treatment project led to a seven per cent increase in gold recovery (73 per cent
At the New Celebration gold mine the advent of treating to 80 per cent) and three per cent increase in copper recovery
harder primary gold bearing sulfide ores at depth resulted (85 per cent to 88 per cent). Payback period was around
in decreased gold recoveries from around 92 per cent down 20 months.
to 87 per cent. The fall was partially due the coarser grind
required to achieve similar treatment rates to those possible Cyclone and mill stabilisation project at Copper Mountain
for the previously treated oxide ores. After extensive test The cyclones in a milling circuit are important to the effective
work, and with the requirement for a solution having low operation of a comminution circuit and provides the necessary
capital and operating cost, a flow sheet was developed to classification to ensure that adequately liberated minerals are
satisfy these criteria. The flow sheet comprised a separate sent to the flotation circuit or leaching circuit. To stabilise and
gravity treatment circuit to recover gold bearing sulfides improve cyclone performance the Copper Mountain mine in
followed by fine grinding of the concentrate before recycling Canada recently implementation detection systems in both
the concentrate back to the carbon-in-leach circuit (Martins, the overflow and underflow sections of a cyclone to identify
Dunne and Delahey, 1993). The tailings retreatment circuit plugging and roping of cyclones (Westendorf et al, 2015).
restored gold recoveries to the original level of around The cyclone detection system, coupled with a new cyclone
92 per cent (that is, increased recovery by about four per cent) pump box controller, has provided a more constant feed to

R Dunne
the mill cyclones. Consequently, the volumetric flows back concentrate (Potorieko, Duran and Cuttriss, 2010). While at
to the milling circuit and to the downstream flotation circuit the Dome mine, Ontario, Canada installation of a gravity
have fewer perturbations, resulting in both better milling and circuit decrease gold solution losses in the tailings increasing
flotation performance. overall gold recovery by 0.2–0.4 per cent (Folinsbee, Chong
There are many more examples in the literature of and Fullam, 2005).
successfully implemented continuous improvement projects
that were initiated from the routine collection of samples PROCESS MEASUREMENTS – CURRENT STATUS AND THE FUTURE
with appropriate analysis of the data generated from In summary, data collection and analysis can be subdivided
treating these samples. Generally, once the opportunity for simplicity in terms of timescales and include the following:
for improvement was identified it was followed by a
•• Short-term – these are mainly for process control
comprehensive test work program and engineering study to
purposes to stabilise the operation. This leads to
show the economic benefits of implementing the change to
more consistent performance resulting in additional
the process or flow sheet.
throughput, additional recovery, less reagent/
consumable consumption and improved quality of the
Additional referenced examples final product. However, the gains are ‘relatively’ small
For those interested in obtaining more information on the and are dependent on how well the plant was run prior
different types of continuous improvement projects reported to the implementation of the control system. These
in the literature, the following references with a short outcomes can be further enhanced, depending on the
explanation of the improvement project (implementation of application, by incorporating expert systems that use
control or a specific optimisation) are provided. online visual measurement (eg particle size; Miranda
Smith et al (2004) provides a comprehensive overview of et al, 2012), colour and flow as in froth flotation (Miranda
applying process control to optimise milling and flotation et al, 2012) and/or analytical measurements. Real-time
plants. A detailed description of the control systems belt elemental analytical systems (for example, prompt
installed at the Bagdad copper concentrator, Arizona, USA gamma neuron activation analysis, or PGNAA), rear
to support the application of an expert system are provided infrared (NIR) and laser induced fluorescence are
by Amelunxen and Lockhart (2005). A model based control presently mostly used for blending feed or concentrates
system at the Aitik copper concentrator in Sweden increased to and from a processing facility; for example, iron
copper recovery by 0.5 per cent for the same final copper ore (Vorster and Flatman, 2001), cement, coal, copper
concentrate grade (Johansson et al, 1999). The installation concentrates (Cottle, 2007; Broicher, 2005). PGNAA has
of control and expert systems at the East Marlartic copper also been applied to improve the performance at a coal
concentrator in Quebec, Canada led to an increase of plant by lowering the ash content in the final product
one per cent in final copper concentrate grade while copper (Khattri et al, 2012). The application of PGNAA, NIR
recovery was unchanged (McMullen et al, 2002). Online and magnetic resonance (Schwitter, Bennett and Miljak,
cyanide control has shown to lead to substantial savings of 2010) to better characterise feeds to base metal mineral
cyanide (Matthews and du Toit, 2011) during gold leaching processing plants and combine this with an appropriate
and reduced cyanide destruction costs (Mulpeter and control system that will lead to improved performance,
Kotzen, 2007; Voster and Flatman, 2001). is still in its infancy. The main impediments are the
Many flotation optimisation projects have provided relative high cost of implementation and identifying and
positive outcomes. As examples at the Elura zinc flotation quantifying the benefits. The gold industry on the other
plant in New South Wales, Australia the fine grinding of hand suffers from not having a reliable and cost-effective
recycled cleaner tailings lead to a seven per cent increase online analytical system to measure gold in the feed
in zinc recovery (Frew et al, 1993). While at the Telfer mine, and tailings. As a consequence, control systems on gold
Western Australia test work resulted in the installation of plants are based solely on monitoring and controlling
a fine grinding circuit to improve both copper concentrate the particle grind size, pulp density, oxygen, lime and
recovery and grade (Burns et al, 2012). The addition of cyanide concentrations to and in the leach circuit.
an IsaMill® for fine grinding followed by cleaning in a •• Mid-term – these data relate to weekly, monthly
Jameson flotation cell at the Prominent Hill copper mine and quarterly production, performance and cost
in South Australia resulted in reduced silica and fluorine reconciliations. The main focus is to generate reliable
contamination of the final concentrate and improved copper data to determine the feed grade, final product quantity
recovery (Barns, Colbert and Munro, 2009). Lowering the and quality, recovery performance and consumable
pH by one unit during copper flotation at the Ernest Henry consumption information that are not impaired by the
copper mine, Queensland resulted in a 0.6 per cent increase inaccuracies of measurement of inventories/holdup in the
in copper recover at the same final copper concentrate grade. process (Muza, 2005). Reliability of data for reconciliation
A savings of A$700 k/a was realised also due to lower is very much a function of proper and accurate (good
lime consumption at the lower pH of flotation (Kirkwood calibration) measurement of mass flows including
et al, 2013). Improvements to the column froth washing sampling and assays of appropriate products around
system at the Miduk copper concentrator, Iran, decreased a circuit. All this information plus cost, maintenance,
the silica content of the final concentrate by 2.7 per cent safety and environmental data provides the foundation
while the final copper recovery increased by 1.4 per cent to better understand and manage the enterprise (mining,
(Massinaci et al, 2007). The addition of a flotation circuit to production, sales, financial, safety and environmental)
recover copper and gold tellurides at the Barrick Buzwagi with the aim of becoming more profitable.
gold mine, Tanzania, increased overall gold recovery by •• Long-term – these data collection (includes test
six per cent (Gorain et al, 2007). Including a gravity circuit work) and analysis is more to do with identifying
in the milling circuit at the El Toqui copper-zinc-gold mine, opportunities to make larger plant improvements
southern Chile, led to an overall payable gold of around (for example, throughput and recovery) compared
six per cent mainly by reducing the gold content of the zinc to those achieved by process control systems. The

Using metallurgical data to drive continuous improvement
larger improvements, however, come with additional de Ruijter, M A, 1996. The clean-up of the plant at Bracken gold
capital and operating cost expenditures, as well as long mine, Journal of South African Institute of Mining and Metallurgy,
implementation times. Because of these impediments, October, pp 215–220.
it is important that ore changes during and after Folinsbee, I A, Chong, T Y and Fullam, M, 2005. Continuous
implementation do not change significantly to impact improvements of Porcupine Joint Venture (PN) Dome Mine’s
the financial outcome of the project. gravity circuit, in Proceeding Treatment of Gold Ores, 44th Conference
The current trend in the mineral industry is to generate of Metallurgists of CIM, pp 65–79 (Canadian Institute of Mining,
more data with the expectation that this will lead to a better Metallurgy and Petroleum: Montreal).
understanding of each unit of the enterprise resulting in an Frew, J A, Davey, K J, Glen, R M and Orrock, G I, 1993. Performance
improved overall performance across the enterprise. One of Pasminco Mining Elura zinc flotation circuit, SME Special
problem with this approach is the lack of people and systems Symposium – Flotation Plants are they Optimized, pp 95–103
to properly analyse the data and provide direction on whether (Society for Mining, Metallurgy and Exploration: Denver).
new opportunities exist and how to pursue these. In addition, Gaylard, P G, Morrison, R D, Randolph, N G, Wortley, C M G and
there is little effort on many operations to determine what Beck, R D, 2009. Extending the application of the AMIRA P754
measured data are redundant. Data measurement comes at a code of practice for metal accounting, in Proceedings Base Metals
cost and any information not used should be eliminated. Conference, pp 15–38 (The Southern African Institute of Mining
and Metallurgy: Johannesburg).
CONCLUSIONS Giblett, A, Dunne, R and McCaffery, K, 2012. Defining practical
A prerequisite for ongoing continuous improvement on metallurgical accounting discrepancy limits for gold operations, in
Proceedings 11th AusIMM Mill Operators’ Conference, pp 1–10 (The
a plant is the availability of reliable data that can then be
analysed to identify opportunities that can be pursued. The
process starts with good sampling, mass measurement and Giblett, A, Hillier, D, Parker, K and Ramsell, V, 2012. The impact of
analytical practices that firstly provides the platform for an gravity gold recovery at Kalgoorlie Consolidated Gold Mines, in
appropriate metal accounting system, as well as adequate Proceedings 11th AusIMM Mill Operators’ Conference, pp 199–206
characterisation (for example, mineralogy, size-by-size metal (The Australasian Institute of Mining and Metallurgy: Melbourne).
assays etc) of key samples (tailings, feed, concentrates and Gorain, B K, Beaudoin, P, Kondos, P, McMullen, J and Shuttleworth, J,
others) at a timely interval (composites over a month). When 2007. The impact of flotation in improving the economics of
an opportunity is identified, the next step is to develop and Barrick’s Buzwagi project, in Proceedings Cu2007, vol II, Mineral
implement an appropriate test work program to provide Processing, pp 73–85 (Canadian Institute of Mining, Metallurgy
the necessary information to build and present a compelling and Petroleum: Montreal).
business case that will convert the opportunity to a value- Holmes, R J, 2004. Correct sampling and measurement – the
added project for the company. As shown in this paper the foundation of accurate metallurgical accounting, Chemometrics
percentage improvements in metallurgical performance (that and Intelligent Laboratory Systems, 74(1):71–83.
is, recovery, grade, throughput) can vary widely from less Johansson, B, Bergmark, B, Guyot, O, Bouche, C and Broussaud, A,
than one per cent to an upper limit of around seven per cent 1999. Model-based control of Aitik bulk flotation, Journal of
for optimisation projects that apply control systems or Minerals and Metallurgical Processing, 16(2):41–45.
changes to flow sheets and operating parameters.
Khattri, P, Krishna, C G, Srinivassa Rao, P, Pratim Banerjee, P and
Mahana Rao, S, 2012. Gammametry – a technological intervention
REFERENCES for optimization and control at coal washeries, in Proceedings
Amelunxen, P and Lockhart, C, 2005. Bagdad concentrator process XXVI International Mineral Processing Conference, pp 2403–2411
control update, in Proceedings SME Annual Meeting, Salt Lake (The Indian Institute of Metals: Bidhannagar).
City, preprint 05–106 (Society for Mining, Metallurgy and
Kirkwood, B, Kinal, J, Siliezar, J and Sofatzis, H, 2013. Low pH trial
Exploration: Denver).
in the Ernest Henry Mining Copper rougher flotation circuit, in
AMIRA International, 2007. P754: Metal Accounting Code of Practice Proceedings MetPlant 2013, pp 371–377 (The Australasian Institute
and Guidelines: Release 3, February. of Mining and Metallurgy: Melbourne).
Barns, K E, Colbert, P J and Munro, P D, 2009. Designing the optimal Lachance, L and Flament, F, 2011. A new ERA tool for state of the
flotation circuit – the Prominent Hill case, in Proceedings Tenth art metallurgical balance calculations, in Proceedings 2011 – 43rd
Mill Operators’ Conference, pp 173–182 (The Australasian Institute
Annual Meeting of the Canadian Mineral Processors (Canadian
Institute of Mining, Metallurgy and Petroleum: Montreal).
Bazin, C, El-Ouassiti, K, Hodouin, D M and Zouadi, D, 2005. Data
Lachance, L, Leroux, D and Gariepy, S, 2012. Implementing
reconciliation in hydrometallurgy – applications to leaching of
clay and copper solvent extraction, in Computational Analysis in best practices of metal accounting at the Strathcona mill, in
Hydrometallurgy, pp 175–190. Proceedings 2012 – 44th Annual Meeting of the Canadian Mineral
Processors (Canadian Institute of Mining, Metallurgy and
Broicher, H F, 2005. Dilution control-online and real-time recognition Petroleum: Montreal).
of ore and waste by LIF, SME Annual Meeting, preprint 05–66.
Marsden, J O and Arnold, J R, 1990. Metallurgical accounting gold
Burns, F, Seaman, D R, Peng, Y and Bradshaw, D J, 2012. Development
fields cripple creek mine, in Proceedings Control ‘90, 13:109–120.
of a pyrite regrind circuit at Telfer gold mine, in Proceedings 11th
Mill Operators’ Conference, pp 45–52 (The Australasian Institute of Martins, V, Dunne, R and Delahey, G, 1993. New celebration tailings
Mining and Metallurgy: Melbourne). treatment plant – 18 months later, in Proceedings XVIII International
Cesnik, F, Hart, S, Dioses, J, Clements, B and Dunne, R, 2005. Low- Mineral Processing Congress, pp 1215–1222.
grade concentrator regrind circuit upgrade at Newcrest’s Cadia Massinaei, M, Oliazadeh, M, Kolahdoozan, M, Noaparas, M
Valley Operations, in Proceedings Centenary of Flotation, pp 815–822 and Oliazadeh, A, 2007. Optimization of the froth washing
(The Australasian Institute of Mining and Metallurgy: Melbourne). system of columns in the rougher circuit of the Miduk Copper
Cottle, I, 2007. Novel online analysis applications in the iron ore Concentrator, in Proceedings Cu2007, vol II, Mineral Processing,
industry, in Proceedings Iron Ore Conference, pp 941–944 (The pp 233–240 (Canadian Institute of Mining, Metallurgy and
Australasian Institute of Mining and Metallurgy: Melbourne). Petroleum: Montreal).

R Dunne
Matthews, D and du Toit, T, 2011. Real-time online analysis of iron Schwitter, B, Bennett, D and Miljak, D, 2010. A new online detection
ore, validation of material stockpiles and roll out for overall method for arsenic minerals in process streams, in Proceedings
elemental balance as observed in the Khumani iron ore mine, XXV International Mineral Processing Conference, pp 3009–3014 (The
South Africa, in Proceedings Iron Ore Conference, pp 297–305 (The Australasian Institute of Mining and Metallurgy: Melbourne).
Australasian Institute of Mining and Metallurgy, Melbourne).
Smith, G C, Jordaan, L, Singh, A, Vandayar, V, Smith, V C, Muller,
McMullen, J, Pelletier, P, Breau, Y and Robichaud, F, 2002. Gold- B and Hulbert, D G, 2004. Innovative process control technology
copper ores processing / flotation process optimization by the for milling and flotation circuit operations, Journal of South African
development of an online expert system, in Proceedings SME Institute of Mining and Metallurgy, July, pp 353–365.
Annual Meeting, Phoenix, Preprint 02–010 (Society for Mining,
Taggart, A F, 1947. Handbook of Mineral Dressing (John Wiley and
Metallurgy and Exploration: Denver).
Sons: New York).
Miranda, R C, Martins, M A S, Gontijo, M M and Nogueira, A, 2012a.
Vorster, B J and Flatman, S R, 2001. Cyanide control in the metallurgical
Crushing and autogenous grinding advanced control assisted
process of gold extraction in AngloGold (SA), Journal of the South
by image analysis, in Proceedings XXVI International Mineral
African Institute of Mining and Metallurgy, October, pp 359–366.
Processing Conference, pp 3433–3436 (The Indian Institute of
Metals: Bidhannagar). Wardell-Johnson, G, 2003. Report on post closure gold recovery
activities at Boddington gold mine, A J Parker Cooperative
Miranda, R C, Martins, M A S, Gontijo, H M and Vieira, M G, 2012b.
Research Centre for Hydrometallurgy, pp 1–9.
Non ferrous metal ore flotation advanced control assisted
by image analysis, in Proceedings XXVI International Mineral Westendorf, M, Tinney, D, Banerjee, T, Kaushik, S and Schaffer, M,
Processing Conference, pp 3428–3432 (The Indian Institute of 2015. Managing cyclones: a valuable asset, the Copper Mountain
Metals: Bidhannagar). case study, SME Annual Meeting, preprint 15–140.
Mulpeter, T and Kotzen, D, 2007. Online cyanide analysers optimise Woollacott, L C and Stange, W, 1987. Guidelines for the derivation of
cyanidation and ensure cyanide code compliance for plant reliable material balances from plant data, Journal of South African
tailings, in Proceedings World Gold Conference, pp 129–133 (The Institute of Mining and Metallurgy, July, pp 207–217.
Australasian Institute of Mining and Metallurgy: Melbourne). Wortley, C M G, 2009. Mass measurement for metal accounting –
Muza, P, 2005. Process information management system at Mount Isa, principles, practice, in Proceedings Fourth World Conference on
in Proceedings First Extractive Metallurgy Conference, pp 179–187 Sampling and Blending, pp 121–128 (South African Institute of
(The Australasian Institute of Mining and Metallurgy: Melbourne). Mining and Metallurgy: Johannesburg).
Potorieko, C A, Duran, A and Cuttriss, R H, 2010. Maximising Wunderlich, A D, 1998. Theft prevention in gold mining, Minerals
payable gold in base metal, in Proceedings Gravity Gold Engineering Journal, October, pp 41–44.
Conference, pp 141–146 (The Australasian Institute of Mining

Contents
The importance of sampling in the mineral industry

R J Holmes1
ABSTRACT
An accurate knowledge of the chemical, mineralogical and physical characteristics of
ores and mineral products that are mined, processed and traded, is crucial to effective
resource utilisation, plant design, process optimisation and finally quality control to
ensure that contract specifications are met. However, all too often the responsibility
for sampling is entrusted to managers and personnel who do not fully appreciate
the significance and importance of collecting representative samples for analysis, and
everyone seems satisfied as long as some material is collected and returned to the
laboratory for analysis. Furthermore, sample stations incorporated into the design
of new plants are often the first casualties when cost cutting exercises commence.
This needs to change, because there is little point in accurately analysing samples
if they are not representative in the first place. The resultant analyses can be next to
meaningless and the effort taken to analyse such samples is, to a large extent, a waste
of time and resources.
Ultimately, a poor knowledge of the characteristics of ores and mineral products
limits the ability of mining companies to optimise the life of ore resources and plant
performance, and failure to meet final product specifications can have disastrous
consequences, including price penalties and more seriously loss of sales contracts. It is
high time that sampling is given the attention it deserves as the first critical step in the
quality measurement chain.
INTRODUCTION
Samples are taken from a wide range of locations in the mineral industry, including
diamond and percussion drill holes, blastholes, feed and product streams, conveyor
belts, trucks, railway wagons and stockpiles, for optimising resource utilisation, process
control, grade control, metallurgical accounting and ultimately maximising profitability
(Holmes, 2004; Holmes, 2010). However, it is surprising how frequently poor sampling
practices are used in this technologically advanced day and age. Sampling is often left
to personnel who do not understand its critical importance in generating representative
samples and meaningful analyses that can be relied upon to make correct resource
and grade control decisions. Sampling is where it all begins and it is not good enough
just to collect some material and send it back to the laboratory for analysis. The whole
measurement chain is corrupted at the outset if the sample is not representative, and
accurate analysis of the samples submitted to the laboratory can be a waste of time and
lead to suboptimal recovery in processing plans, reduced mine life and loss of sales
revenue. It is therefore critical to ensure that samples are free of significant bias and that
the overall precision of the final analysis is appropriate for the required task.
The golden rule for correct sampling is that ‘all parts of the material being sampled
must have an equal probability of being collected and becoming part of the final
sample for analysis’ (Gy, 1982a, 1982b; Pitard, 1993, 2005; Holmes, 2004, 2005, 2010).
If this golden rule is respected at the outset, then extraction of representative samples
is largely assured. However, if this rule is not respected, then sample bias is easily
introduced, which is particularly serious because no amount of replicate sampling and
analysis will reduce bias once it is present (Holmes, 2004). The main errors that need
to be eliminated (Gy, 1982a) include incorrect delimitation of increments (incorrect
cutter/increment geometry), incomplete extraction of increments, preferential
exclusion of specific size fractions, sample loss and sample contamination. However,
these requirements are frequently ignored in the design of sample stations to reduce
capital costs, which is false economy as the samples taken are likely to be seriously
biased, the precision may be compromised, and the subsequent cost of retrofitting a
correct sampling system can be large. It is also very difficult if not impossible to take
representative samples from some of the sampling locations listed above and should
be avoided, eg in situ sampling from a large stockpile. The design of subsequent
1. FAusIMM(CP), Chief Research Scientist, CSIRO sampling stages is also very important, particularly in terms of the relationship
Mineral Resources, Clayton South Vic 3169. between particle size and the sample mass that needs to be retained to achieve
Email: ralph.holmes@csiro.au acceptable precision.
333
R J Holmes
SAMPLING BASICS different in composition from the rest of the sample, so bias is
introduced into the crushed and/or divided sample.
While samples are taken from many locations in a mining
operation, this paper is focused on sampling in and around
mineral processing plants. By far the best method is to sample Delimitation and extraction errors
a moving stream either at a transfer point between two Increment delimitation (or delineation) and extraction errors are
conveyor belts or at the end of a slurry pipe (Holmes, 2004). especially relevant in the design of sample cutters and sample
Here the process stream can be intersected at random or plants. Delimitation errors arise when the geometry of the
regular times or tonnages, and ‘increments’ can be collected increment to be extracted is incorrectly defined, while increment
by taking a full cross-section of the stream with a sample cutter extraction errors occur when increments are not fully extracted.
and subsequently combined into a representative sample, Consequently, when sampling from moving streams, all parts
thereby satisfying the golden rule for correct sampling. of the stream cross-section must be diverted by the sample
cutter for the same length of time (correct delimitation) and the
Examples of poor plant sampling practices include taking a
increment must be completely extracted without any material
sample from the top layers of ore or concentrate on a conveyor
‘hanging up’ in the cutter or being lost from the cutter (correct
belt, taking cuts from only part of a falling ore stream, taking
extraction). Hence, the delimitation error is eliminated if the
cuts from a fixed location within a launder or taking a ‘bleed’
cutter geometry is correct, eg selection of a parallel section of
from the side of a slurry pipe. There will almost certainly
ore on a conveyor belt or radial cutter lips on a rotating Vezin
be segregation both vertically and horizontally across a
cutter, while the extraction error is eliminated if increments
conveyor belt due to the action of the idlers and the way the
are completely extracted without any sample loss, ie no reflux
material is fed onto the conveyor, and particles suspended in
or loss of sample from the cutter aperture.
a slurry will inevitably segregate under the effects of gravity
and centrifugal forces, so partial stream cuts will not provide
representative samples. Weighting error
Weighting errors occur when increment masses are not
Before discussing the requirements for correct sampling in
proportional to the mass of material they represent and
more detail, it is useful to to understand the sampling errors
are caused by rapid variations inflow rate while sampling
that can occur (Gy, 1982a; Pitard, 1993) to better understand
moving streams. However, the magnitude of the error can
how to eliminate or minimise them. The key components of
be reduced considerably by regulating the flow rate prior to
sampling error are as follows:
sampling (for time-basis sampling) or using a variable speed
•• fundamental, grouping and segregation errors cutter programed to take increments of constant mass (for
•• long-range quality fluctuation error mass-basis sampling).
•• periodic quality fluctuation error
•• weighting error
Periodic quality fluctuation error
Periodic variations in quality can introduce serious errors
•• increment delimitation error
when the sampling interval approaches that of the periodic
•• increment extraction error variation and increments are taken in phase with this variation,
•• accessory errors. the classic examples being periodic variations introduced by
Of these errors, the last four components require special bucket wheel reclaimers and centrifugal pumps. In the case of
attention as they can introduce serious bias, but fortunately bucket wheel reclaimers, the centre and edges of a stockpile
they can be controlled and in most cases eliminated by using have different grades and particle size distributions due to
correct sampling equipment and procedures. Minimising or segregation during stacking, so if increments are taken at
even better eliminating bias is critical, because bias cannot intervals that correspond with the reclaimer being at the edge
be eliminated once it is present. There is no point in being or the middle of the stockpile, the resulting sample will be
‘precisely incorrect’. Sources of bias that can be eliminated biased. Consequently, such sampling situations need to be
include incorrect delimitation and extraction of increments, avoided unless stratified random sampling is used to break
sample spillage and sample contamination, while sources of the phase relationship between the periodic variations and
bias that need to be minimised include changes in moisture increment extraction.
content and dust losses. In contrast, the other components of
sampling error introduce random errors that can be reduced to Long-range quality fluctuation error
acceptable levels by intelligent design of the sampling regime, The long-range quality fluctuation error results from
but they can never be completely eliminated. variations in quality of the process stream as a function of
time and how well the increments that are extracted follow
Accessory errors these variations in quality. Hence, this error component can
Accessory errors are associated with non-selective operations be reduced by increasing the number of increments taken so
where there is no change of mass, eg crushing, grinding, that the variations in quality can be followed more closely.
mixing, sample transfer, drying etc. These errors include In practice, the quality variation of the material needs to
sample contamination, sample spillage, particle degradation, be characterised (eg via the increment variance), so that
dust loss, moisture loss, operator mistakes, fraud and the number of increments required to achieve the required
sabotage, all of which can be eliminated or minimised by using sampling precision can be calculated.
correct sampling practices and equipment, as well as operator
training. Examples of operator mistakes include mixing up Grouping, segregation and fundamental errors
sample labels or tipping out part of the sample because it is The grouping, segregation and fundamental errors are
too heavy, while an example of fraud is collecting a ‘timed’ all due to very short-range variations in quality, largely
sample well in advance of the time at which it was scheduled between individual particles. Of these the fundamental error
to be taken. An error that often goes unnoticed is loss of the is particularly important and results from the differences in
finer particles in crushers or dividers due to excessive airflow composition between the individual particles in the material
in dust extraction systems. The finer particles are usually being sampled. It can be reduced by taking more particles,

ie increasing the sample mass, or reducing the particle size, •• The cutter aperture must be designed so that the cutting
thereby enabling more particles to be taken for a given time at each point in the stream is equal, which ensures
sample mass. correct increment delimitation. Consequently, the cutter
A major consequence of the fundamental error is that, for a lips must be parallel for linear-path cutters, while the
cutter lips must be radial for cutters travelling in an arc,
given particle size, there is a minimum mass of sample that
eg Vezin cutters. Flap or diverter type cutters that divert
must be retained after division to obtain the desired division
one side of the stream for a longer period of time than
precision, which varies depending on the element of interest
the other are not acceptable.
and its concentration. This is a very important sampling rule,
which unfortunately is often ignored in the design of sample •• The sample cutter must be non-restrictive and self-
clearing, discharging completely each increment without
regimes to reduce the sample mass that sampling personnel
any reflux, overflow or hang-up, thereby ensuring correct
need to take back to the sample preparation laboratory. While
increment extraction. This is particularly important for
this is admirable from the safety perspective, unfortunately
‘reverse spoon’ type cutters, where the material being
it can seriously compromise the integrity of the sample. The
sampled has to change direction as it strikes the back
correct approach is either to crush the sample to a smaller of the cutter body, which can result in sample reflux at
particle size prior to sample division, thereby enabling the high flow rates (see Figure 1) if the cutter does not have
sample mass to be reduced, or to provide mechanical lifting sufficient capacity. In addition, damp material has a
aids. tendency to hang-up in the cutter aperture. These issues
The minimum sample mass can be calculated using the can be overcome by utilising generously large cutter
equations and methods described by Gy (1982a), Pitard (1993), bodies and chutes as well as designing the back of the
François-Bongarçon (2005) and Holmes (2005), or alternatively cutter to deflect material striking the cutter away from
it can be determined experimentally by analysing replicate the incoming material, thereby avoiding sample reflux
samples as a function of sample mass and determining the and overflow from the cutter aperture (see Figure 2).
variance observed in practice. Minimum sample masses are For sticky materials, steep chute angles (>60°) and
also specified in the relevant Australian and ISO sampling stainless steel or polythene chute linings are often used
standards, eg ISO 3082 (2009) for sampling iron ores (Table 1). to reduce adhesion, and the cutter aperture may need to
be enlarged to prevent bridging of the aperture.
•• The cutter should intersect the stream either in a plane
PLANT SAMPLING normal to, or along an arc normal to, the mean trajectory
of the stream.
Sampling particulate material
•• The plane of the cutter aperture must not be vertical
As already indicated above, the best location for sampling a
or near vertical, because particles that strike the inside
process stream in a mineral processing plant is the discharge edge of the cutter aperture and which should therefore
point of a conveyor belt or chute where the complete stream end up in the sample are deflected away from the cutter
can be intersected at regular or random intervals. One of the aperture by gravity into the reject stream, resulting in
most important considerations is the design of the sample sample loss (Pitard, 1993; Holmes, 2004, 2010).
cutter, which must satisfy the following requirements to
•• The cutter must travel through the stream at a uniform
eliminate increment delimitation and extraction errors: speed, accelerating up to its cutting speed before entering
•• The sample cutter must take a complete cross-section of the stream and then decelerating to a stop after leaving
the process stream, ensuring that both the leading and the stream. The cutter drive must also have sufficient
trailing edges of the cutter completely clear the stream at power to ensure that the cutter does not slow down as it
the end of each traverse. Furthermore, the length of the enters the stream and speed up as it leaves the stream.
cutter aperture must be large enough to intercept all the Electric cutter drives are best for ensuring uniform cutter
material in the stream, including particles that bounce speed, although hydraulic drives can also be satisfactory
off the inside edges of the cutter aperture. if they are well maintained. Pneumatic drives are not
recommended, because it is usually impossible to
Table 1 adequately control cutter speed.
Examples of minimum mass of divided sample for moisture determination
and/or chemical analysis of iron ore for division precisions (σD) of
0.1 per cent and 0.05 per cent Fe (from ISO 3082, 2009).
Nominal top size Minimum mass of divided gross sample (kg)

(mm) σD = 0.1% Fe σD = 0.05% Fe
40 325 1,300
31.5 180 710
22.4 75 300
10 10 40
6.3 3.2 13
2.8 0.5 1.7
1.4 0.5 0.5
0.50 0.5 0.5
0.25 0.5 0.5 FIG 1 – Example of sample reflux from a cutter aperture at high flow rates.

R J Holmes
for two principal reasons as follows, particularly for high

capacity streams:
1. Cross-belt cutters tend to leave a layer of material on the
conveyor belt if the skirts at the bottom of the cutter are
not correctly adjusted as they gradually wear out or if
the wear is not uniform leaving gaps, so the increment
extraction is incorrect. Furthermore, maintenance staff
often increase the gap between the cutter skirts and
the conveyor for fear of damaging the conveyor belt.
Consequently, cross-belt cutters can be seriously biased,
because the material on the bottom of the belt can be
different in grade from the bulk of the material on the
conveyor belt.
2. It is virtually impossible to check visually whether a
cross-belt cutter is performing correctly in terms of
correct increment delimitation and increment extraction.
A typical example of a cross-belt cutter is shown in Figure 3
FIG 2 – Cross-stream cutter with a generously large cutter body to avoid where the cutter is fully enclosed, so it is impossible to visually
reflux and overflow from the cutter aperture at high flow rates. check its operation. Figure 4 shows an ore stream after taking a
cross-belt sample cut using a similar cutter to that in Figure 3,
•• The cutter aperture must be at least three times the
which indicates that the cutter probably did not remove a full
nominal top size (d) of the material being sampled, ie 3×d,
cross-section of ore from the conveyor belt, while Figure 5 is
to prevent preferential loss of the larger particles, subject
a photograph of an actual cross-belt cutter showing the poor
to a minimum of 10 mm for fine dry solids (Gy, 1982a).
condition of the rubber skirt at the bottom of the cutter. The
•• The cutter speed must not exceed 0.6 m/s unless the sample cut shown in Figure 4 is clearly unsatisfactory, so
cutter aperture exceeds 3×d (Gy and Marin, 1978),
because the ‘effective’ cutter aperture decreases as the
cutter speed increases, which can preferentially exclude
the coarser particles and thereby introduce bias. When
the cutter aperture (w) is increased above the minimum
cutter aperture w0 (ie 3×d or 10 mm, whichever is the
greater), the maximum cutter speed (vc) can be increased
as follows (Gy and Marin, 1978), subject to an absolute
maximum of 1.2 m/s:
vc = 0.3 (1 + w )
w0
•• Bucket-type cutters must have sufficient capacity to
accommodate the increment mass obtained at the
maximum flow rate of the stream without any reflux or
overflow of sample from the cutter aperture, ensuring
that the gate on the bottom of the cutter bucket does not
jam in the open or closed position and that no sample is
lost from the bucket.
•• No materials other than the sample must be introduced FIG 3 – Fully enclosed cross-belt sampler installation.
into the cutter or the sample delivery chute and there
must be no change in quality of the sample. If necessary,
the cutter needs to be covered in the parked position
between increments to prevent ingress of dust or spillage.
•• If a belt scraper is required to remove material adhering
to the belt, the scraped material must fall within the
area traversed by the cutter, although in many instances
the amount of material removed by belt scrapers is
negligible, particularly for dry materials.
The cutters discussed so far have been cross-stream cutters
where the cutter passes through a falling stream of ore or
concentrate. Provided there is suitable access, it is reasonably
straightforward to visually check that the cutter intercepts
the complete stream and that increment delimitation and
extraction is correct. However, cross-belt cutters which
take samples directly off conveyor belts are also used in the
mineral industry, but it is very difficult to check visually
whether they are operating correctly and remove a complete
and correctly delimited cross-section of material from the
conveyor belt. Hence, while they may be less expensive than FIG 4 – Example of a cross-belt sample cut indicating that the cutter
cross-stream cutters, cross-belt cutters are not recommended probably did not remove a full cross-section of ore from the conveyor belt.

The only acceptable way of extracting representative samples

in situ from vessels is full depth sampling where a number of full
vertical columns are extracted from the vessel. Alternatively,
the concentrate can be sampled from a moving stream as the
vessel is filled or emptied. In the latter case, dry concentrates are
sometimes sampled while they are being conveyed via feeder
pipes into vessels, but once again this is only satisfactory if the
cutter extracts a full cross-section of the concentrate stream.
Consequently, using a pneumatic sampling device attached to
the side of the feeder pipe such as that shown in Figure 8 is
not satisfactory, because the full cross-section of the stream is
not sampled. Furthermore, the suction action of the sampling
device will preferentially collect the smaller less dense particles
resulting in a seriously biased sample.
Sampling slurries
As for sampling particulate materials and dry concentrates,
FIG 5 – Cross-belt sample cutter showing the poor condition the best location for sampling a slurry is at a transfer point
where a cross-stream cutter can gain access to the full slurry
of the rubber skirt at the bottom of the cutter.
stream and take a complete cross-section of the stream at
cross-stream cutters are recommended in preference to cross- uniform speed, thereby providing a representative sample.
belt cutters to obtain representative samples. The cutter must have sufficient capacity to accommodate
the entire increment at the maximum flow rate of the stream
Manual sampling from the top of conveyor belts (Figure 6) without any reflux or overflow from the cutter aperture. One
is also totally unacceptable for several reasons. Firstly, there key difference with slurries is the need to ensure that ‘dribbles’
are serious safety issues with this practice and, secondly, it is from underneath pipes and launders are also intercepted by
not possible to take a complete cross-section of the material on the sample cutter. When sampling slurries, the cutter aperture
the conveyor belt. Consequently, the sample collected will not should be a minimum of 10 mm (Gy, 1982a) and to reduce bias
be representative. the cutter lips must be sharp and the sample cutter streamlined
to minimise turbulence as it cuts the slurry stream (Bartlett,
Sampling dry concentrates 2002). A suitable cross-stream cutter for sampling a slurry
The requirements for sampling dry concentrates are essentially stream is shown in Figure 9. The turbulence introduced by the
the same as those for sampling particulate material, except cutter is minimal and there is no evidence of reflux from the
that the particle size is much smaller. Hence, a cutter that cutter aperture. On the other hand, sample cutters that collect
takes a complete cross-section of the concentrate stream at a only part of the slurry stream on a continuous or periodic
transfer point is required, the recommended minimum cutter basis, such as the ‘shark fin’ cutters shown in Figure 10, are
aperture being 10 mm. Consequently, sampling practices unacceptable for taking representative samples.
such as in situ ‘dip’ sampling from holding tanks and Manual sampling of slurry streams is quite acceptable
vessels as shown in Figure 7 is not recommended, because provided access is available to the full slurry stream and a
the concentrate will almost certainly segregate in the vessel complete cross-section of the stream is taken at uniform speed
resulting in the extraction of a non-representative sample. with a manual cutter or ladle. As for mechanical sampling,
FIG 6 – Manual sampling from the top of a conveyor belt is poor sampling practice and raises serious
safety concerns. In any event it does not provide a representative sample.

R J Holmes
FIG 7 – Manual sampling of vessels using a ladle is not recommended due to segregation of the material in the vessel.
the cutter must have sufficient capacity to accommodate the

entire increment at the maximum flow rate of the stream
without any reflux or overflow from the cutter aperture and
the cutter aperture should be a minimum of 10 mm to reduce
bias and provide a representative sample. In contrast, taking
manual samples using a ladle that does not intercept the full
slurry stream (Figure 11) is not acceptable and the aperture on
the manual cutter shown in Figure 12 is far too small, so the
samples collected are unlikely to be representative.
While it is common practice in industry, sampling of
slurries via ‘taps’ on the side of pipes such as shown in
Figure 13 is not satisfactory for extracting representative
samples, because segregation and laminar flow of slurries in
pipes is common. Furthermore, pressure pipe samplers used
for extracting samples for online analysers (Figure 14) do not
extract a full cross-section of the slurry stream, so they are
FIG 8 – Sampling dry concentrates pneumatically from the prone to bias as well.
side of a feeder pipe does not extract a full cross-section of the
concentrate stream and generates a biased sample. Sampling trucks, railway wagons and stockpiles
Sampling of stationary ore and concentrate in trucks, railway
wagons and stockpiles is quite problematic, because in the
majority of cases it is not possible to respect the golden rule of
sampling, ie that all parts of the material being sampled have
an equal probability of being collected and becoming part of
the final sample for analysis. This is particularly true for large
stockpiles. Apart from sampling the ore or concentrate from
a moving stream while it is being transferred to or from the
truck, railway wagon or stockpile, the only way of obtaining
representative samples is full depth sampling by extracting
a number of full vertical columns. In principle this can be
achieved by driving an auger or spear down to the bottom
of the truck, railway wagon or stockpile and then extracting
the column of ore or concentrate without any sample loss.
However, it is difficult to drive the sampling implement
down to the bottom and equally difficult to then extract the
sample without any loss of material. Furthermore, collecting
material from the tops of trucks or railway wagons (Figure 15)
or from the side of a stockpile (Figure 16) will not provide
FIG 9 – Cross-stream cutter for sampling a slurry stream. representative samples either.

FIG 10 – ‘Shark fin’ cutters are unsatisfactory, because they FIG 12 – Manual sample cutter with an aperture
do not take a full cross-section of the process stream. that is far too small for sampling slurries.
FIG 11 – Manual sampling of a slurry stream using a ladle that is too small
for the stream being sampled will not provide a representative sample. FIG 13 – Taps from the side of slurry pipes do not extract a full
cross-section of the slurry stream, so the samples will not be representative.
PERFORMANCE VERIFICATION
Verification of the performance of sampling systems is an •• worn and/or missing cutter lips
important part of initial and ongoing quality assurance and •• build-up and/or blockages in cutter apertures and chutes
comprehensive checklists are available (Docherty, 2005), •• reflux from cutter apertures
including in several ISO standards, eg for iron ore and coal
•• ingress of extraneous material when the cutter is parked
and coke. Consequently, for mechanical sampling systems,
large and readily accessible inspection ports are required •• increment mass
to enable inspection of sample cutters to ensure that they •• number of primary, secondary and tertiary cuts
intercept the whole stream and are in good condition and depending on the number of sampling stages
free of build-up and blockages (Pitard, 2005). Safety concerns •• holes in chutes and bins resulting in sample loss
associated with inspection ports can be overcome by welding
•• crusher performance
steel mesh over the inspection port to prevent physical access
while at the same time allowing inspection of cutters as they •• condition of vibratory feeders
intercept the stream to validate correct operation. Inspection •• sample mass as a function of particle size to ensure that
ports should also be provided for checking chutes, crushers it conforms to minimum sample mass requirements.
and sample dividers for blockages and maintained condition.
An example of a partially blocked primary cutter chute is SAMPLE PREPARATION
shown in Figure 17, which was exacerbating sample reflux
While use of correct methods for collecting primary increments
from the cutter aperture. The ability to monitor sampling ratio
is critical to obtaining representative samples, the subsequent
and increment mass/extraction ratio also provides valuable
preparation of samples in the sample preparation laboratory
information. The key items that need to be checked when
prior to chemical analysis is equally important in ensuring
verifying system performance include:
that the final sample submitted for analysis is representative.
•• size and geometry of cutter apertures, including Unfortunately, many instances can be found where the veracity
checking that cutters take a complete cross-section of the of samples delivered to the sample preparation laboratory is
stream being sampled destroyed in the preparation process. Examples include use of
•• cutter speed and uniformity while cutting the ore stream poor sample division methods such as coning and quartering

R J Holmes
Other sample preparation problems that need to be

eliminated include:
•• discarding part of the sample because it is too heavy
•• sample loss in dust extraction systems
•• cross contamination between samples.
Robotic sample preparation systems are now widely used
in the mineral industry to overcome occupational health and
safety issues, reduce overall costs and address the vagaries
of manual sample preparation. However, they still need to
respect all the requirements for correct sample preparation,
which must not be compromised simply to increase sample
throughput through the robotic system. Key items to take
into account in the system design and monitor on an ongoing
basis include:
•• particle size after crushing
•• minimum sample mass requirements as a function of
particle size
•• elimination of sample loss in rotary driers due to
excessive airflow
•• elimination of cross contamination between samples.
OVERALL PRECISION
The overall precision of sampling, sample preparation
FIG 14 – Pressure pipe samplers do not extract a full cross-section and analysis must be appropriate for the required task and
of the slurry stream, so the samples taken are unreliable. decided at the outset. An appropriate sampling regime can
then be designed, noting that it is impossible to control
and incorrect use of riffles and rotary sample dividers, as well plant, stockpile or shipment grades to high precision if the
as dividing samples down to masses that are too small for overall precision of measurement is poor. Examples of plant
the nominal top size of the ore or concentrate sample being operators responding to apparent changes in grade that are
prepared, ie the minimum sample mass requirements are not no more than quality measurement ‘noise’, ie imprecision, are
observed as the particle size of the sample is reduced during not uncommon. Furthermore, if the overall precision of grade
sample preparation. A classic illustration of this last problem measurements is high, then target grades can be moved closer
is shown in Figure 18, where the mass of an iron ore sample to contract specifications without incurring penalties, thereby
has been divided down to about 300 g for pulverisation. significantly improving resource utilisation.
However, the sample is far too coarse (~10 mm) for division The actual precision achieved in practice can be determined
down to this sample mass due to poor performance of the jaw via duplicate ‘interleaved’ sampling, where alternate primary
crusher in the sample preparation laboratory, resulting in poor increments are directed to duplicate samples A and B (see
division precision. The sample should have been crushed to a ISO 3085, 2002), which are subsequently prepared and
much smaller particle size (~3 mm) prior to division down to anaysed in duplicate. This enables the separate precisions of
around 300 g (see Table 1). sampling, sample preparation and analysis to be determined.
FIG 15 – Sampling from the top of railway wagons will not provide representative samples.

FIG 18 – Sample for pulverisation, but the sample

mass is far too small for its particle size.
FIG 16 – Sampling from the side of stockpiles will

not provide representative samples.
FIG 19 – Provision for duplicate sampling in a sample station.
standards on correct sampling practices, the responsibility

for sampling is often entrusted to managers and personnel
who do not fully appreciate the significance and importance
of collecting representative samples for analysis. Cost is
often the main driving force rather than whether the samples
collected are meaningful, which seriously undermines
the reliability of the final analyses. Hence, sample stations
incorporated into the design of new plants are often the first
casualties to save capital costs, but there is little point in
accurately analysing samples if they are not representative in
the first place. Company management needs to recognise this
and act accordingly to ensure that sampling systems are well
designed and provide representative samples for analysis,
thereby enabling better resource utilisation and reduction
of financial risks. It is high time that sampling is given the
attention it deserves as the first critical step in the quality
FIG 17 – Partially blocked primary cutter chute measurement chain.
identified during a sample station audit.
Consequently, duplicate sampling facilities should be REFERENCES

incorporated into sample stations at the outset (Figure 19) so Bartlett, H E, 2002. Design of primary samplers for slurries in
that performance in terms of precision achieved in practice concentrators and statistical methods for measuring components
of variance in sampling, Journal of the South African Institute of
can be determined and monitored on an ongoing basis as
Mining and Metallurgy, 102(8):485–490.
part of routine quality assurance / quality control (QA/QC)
procedures. Docherty, J, 2005. Mechanical sample plants, in Proceedings Second World
Conference on Sampling and Blending, pp 83–93 (The Australasian
CONCLUSION François-Bongarçon, D, 2005. The modelling of the liberation factor
An accurate knowledge of the quality characteristics of ores and its calibration, in Proceedings Second World Conference on
and subsequent products that are mined, processed and Sampling and Blending, pp 11–13 (The Australasian Institute of
traded is crucial to cost-effective resource utilisation, plant Mining and Metallurgy: Melbourne).
design, process optimisation and quality control. However, Gy, P M, 1982a. Sampling of Particulate Materials – Theory and Practice,
despite the availability of training courses, conferences and 2nd Edition (Elsevier: Amsterdam).

R J Holmes
Gy, P M, 1982b. Sampling from high capacity streams, in Proceedings Holmes, R J, 2010. Sampling mineral commodities – the good, the
First Australian International Bulk Materials Conference, pp 407–423. bad, and the ugly, Journal of the Southern African Institute or Mining
and Metallurgy, 110:1–8.
Gy, P M and Marin, L, 1978. Unbiased sampling from a falling
stream of particulate material, International Journal of Mineral ISO 3082, 2009. Iron ores – Sampling and sample preparation procedures
Processing, 5:297–315. (ISO: Geneva).
Holmes, R J, 2004. Correct sampling and measurement – The ISO 3085, 2002. Iron ores – Experimental methods for checking the precision
foundation of metallurgical accounting, Chemometrics and of sampling (ISO: Geneva).
Intelligent Laboratory Systems, 74:71–83.
Pitard, F F, 1993. Pierre Gy’s Sampling Theory and Sampling Practice,
Holmes, R J, 2005. Design of sample plants – Getting it right first 2nd Edition (CRC Press Inc: Florida).
time, in Proceedings Second World Conference on Sampling and
Pitard, F F, 2005. Sampling correctness – A comprehensive guideline,
Blending, pp 103–110 (The Australasian Institute of Mining and
in Proceedings Second World Conference on Sampling and Blending,
Melbourne).

Contents
A progressive, iterative approach leading to reliable

metallurgical accounting sampling systems
F F Pitard1
ABSTRACT
The optimisation of metallurgical processes requires knowledge of the heterogeneity
carried by constituents of interest. The common scenario, almost an industry standard,
is to rely on the engineering firm who in turn relies on favourite manufacturers of
sampling equipment to provide the appropriate sampling systems in the plant
to perform accurate metallurgical accounting. This sequence is flawed from the
beginning to the end because everything should be based on a careful understanding
of the Theory of Sampling (TOS) by the company project team and their expert who
invariably would suggest various approaches and different sampling systems.
The necessary progressive and iterative approach should be that the company’s
project team (mining or other industries) would do its home work to have a good
understanding of the geometallurgy, mineralogy and texture of the ore in each
geological unit so consultants and engineers would have the best information to start
with. Then a reputable sampling expert would recommend which sampling systems
are needed, so the company would require engineering firms to follow pre-established,
non-negotiable requirements, and finally, select the sampling equipment existing on the
market today that is most likely to lead to accurate and precise metallurgical accounting.
It is the author’s experience that when this necessary progressive and iterative
approach is bypassed, sampling systems installed in the plant are likely to transgress
the most elementary rules of sampling correctness. Invariably the argument is that
the company cannot afford correct sampling systems for metallurgical accounting as
suggested in the TOS. There is nothing that is farther from the truth. Later when
problems develop, and inevitable reconciliation difficulties occur between miners and
geologists on a daily basis, the sampling expert may be called in as a last resort to
perform a post-mortem analysis.
The progressive and iterative approach suggested in this paper should be a due-
diligence requirement for any bank providing a loan for an important new mining
project. Anything less than this is likely to generate unpleasant financial losses.
Historical, well-known examples are referred to in this paper to illustrate this.
INTRODUCTION
The use of automation and attempts at cost minimisation are responsible for
companies and engineering firms taking unfortunate short cuts when the time comes
to install necessary monitoring systems in a plant. This trend is further emphasised
by manufacturers’ effective marketing to suggest simplistic online sampling systems
both for process control and metallurgical accounting. Existing online monitoring
systems for metallurgical accounting are far from having reached sufficient maturity
as they are all based on non-probabilistic sampling systems. Usually, X-ray or
Gamma analysers have reached an attractive level of reliability and accuracy; it is
not the case for the simplistic non-probabilistic sampling systems that are used to
feed them. The resulting effect is always the same: it is impossible to have a reliable
reference anywhere when reconciliation problems take place between the mine
and the plant. The staggering costs of these situations have been well addressed by
Carrasco, Carrasco and Jara (2003) and Carrasco et al (2007). The author witnesses
such situations every day in his worldwide consulting work.
SCOPE
The objective of this paper is to address the enormous risk companies are taking when
accepting that online analysers fed with material collected with non-probabilistic
1. President, Francis Pitard Sampling in-stream stationary probes or stationary cutters are sufficient to achieve reliable
Consultants, LLC, Broomfield Colorado USA. metallurgical accounting. These systems are all based on the intolerable assumption
Email: fpsc@aol.com that the stream is not segregated and are necessarily biased. It has been well-known
343
F F Pitard
for half a century that segregation in a stream is a transient converge toward the unknown truth. The problem is that such
phenomenon and changing all the time, so sampling biases convergence cannot and will not take place if very stringent
introduced by non-probabilistic sampling systems are not rules are not respected to ensure every part of the stream
constant. Variable biases create confusion and the misleading has the same chance of being sampled. This shortcoming is
information generated leads management to pursue solutions always the case with non-probabilistic sampling systems
remote from the actual source of their problems. A preventive such as grab samples, pressure samplers, stationary probes
strategy is offered to help companies avoid falling into the trap or stationary cutters located at a fixed point within a flowing
of manufacturers recommending the use of non-probabilistic stream. Such shortcomings may be somewhat acceptable
sampling systems for metallurgical accounting purposes. It when the metallurgist wants a rough idea of where the
is fair to say that such a strategy can reduce the subsequent
process is at any given time. However, it is most certainly
cost of managing operations where finger pointing becomes
completely unacceptable for the manager trying to accurately
the rule and true, effective accountability is nearly impossible
quantify metal recovery or reconcile the plant production
because causes of problems elude everyone.
estimates with those apparently delivered by the mine. An
inbuilt incapability to determine who is right, the plant or the
GEOMETALLURGY, MINERALOGY AND TEXTURE OF THE ORE mine, can be responsible for staggering financial losses and
A famous geostatistician said once ‘good geology always a misdirection of the use of valuable professional resources.
comes first’ (David, 1977). Actually, it is important to realise
that geology is sitting between two other important disciplines
which are geochemistry and geometallurgy. Structural SELECTION OF SAMPLING SYSTEMS AT FEASIBILITY TIME
geology knowledge is not sufficient for a valid feasibility Regardless of statistical tests that may demonstrate that
study; the heterogeneity of the constituents of interest must be some non-probabilistic sampling systems used for feeding
understood. For example, it is not enough to understand where online analysers can sometimes do a ‘good-enough’ job for
the gold is situated. It is equally important to understand why metallurgical accounting, the risk taken by the company
it is where it is and this is why geochemistry is important. The accepting the use of such systems for metallurgical accounting
author experienced similar difficulties for nickel laterites in is real, substantial and not worth taking in the long run. This
New Caledonia where the nickel-cobalt ore would respond is a critical point: any non-probabilistic sampling system
very differently in a process depending on the level of cannot and will not provide information that can sufficiently
peridotite alteration in the plant feed. Also, as the ultimate goal converge toward the truth; it is mathematically impossible.
is the effective recovery of valuable constituents, it is critically Therefore, it is of the utmost importance that an equi-
important to define clear objectives so the metallurgists may probabilistic sampling philosophy for accounting purposes
perform the most appropriate tests prior to building pilot be selected and enforced by the company prior to completing
plants and ultimately full-size plants. For this task to progress
a feasibility study; it should be considered a due-diligence
smoothly it is important that the metallurgists effectively
basic requirement. The only valid options left are cross-stream
communicate with the geologists, so geometallurgy becomes
sampling systems installed under the discharge of a conveyor
an essential ingredient of feasibility studies. For example it is
critically important to know all the associations of gold with belt or under the discharge of a liquid flowing stream loaded
other minerals, so it becomes possible to predict what kind of with suspended solids such as the ‘feed’ to a flotation plant,
recovery we can expect from a given process. All this allows ‘concentrate’ going to a filter plant or a ‘final tail’. Even cross-
sampling experts to make sound recommendations about belt samplers installed on the top of a conveyor belt are
sampling systems, sampling protocols and finally the creation essentially non-probabilistic because fine particles located at
of a reliable data bank necessary to optimise the various the bottom of the stream are not collected most of the time, and
economic units at operations. even if they were bouncing material coming back to the belt
or going somewhere else and contamination from material
VIEW OF AN OBSERVER VERSUS REALITY OF A PARTICIPANT dragged in on the upstream side of the bucket remain huge
problems with unsatisfactory solutions. The following steps
There is a major misconception on the part of sampling
are recommended:
practitioners and engineers that, regardless of whether the
principles recommended by the TOS are applied, it is always •• Geochemists, geologists and geometallurgists provide
possible to test and quantify sampling precision and accuracy information to the sampling expert who suggests an
after the fact using basic statistical tests. As a result, it is the acceptable sampling protocol.
viewpoint of some uncertainty measurement experts that •• Using further information provided by metallurgists, the
in many instances the principles developed by the TOS are sampling expert designs appropriate sampling stations
somewhat irrelevant. Such a viewpoint is not correct because at important points where samples must be collected for
there is an immense difference between what a qualified and the sole purpose of metallurgical accounting, material
diligent observer would think of the material to be sampled,
balance and reconciliation between mine and plant. Note
and what a participant making an important decision or
that a large majority of existing cross-stream sampling
even a sampling operator would do, when the time comes to
systems will never do the job because of design and
actually take samples.
because they are obsolete technology. This does not
The TOS according to Gy (1967) and Matheron (1967) make things easy for the unaware metallurgist.
unmistakably demonstrates that if sample increments
are correctly taken they can generate a level of sampling •• At the same time, and this is too often a forgotten point,
uncertainty that converges toward the true (but unknown) the location for the official weightometer used by the
content for the constituent of interest as sample mass and the plant must be selected carefully. Very stringent, non-
number of increments increase. The reason is that if increments negotiable rules must be provided to the engineering
are correctly taken they can be accumulated to give a perfectly firm. Most of the time these are ignored or forgotten and
random representation of the whole unit. Therefore as more the sampling expert must address this matter as well as
increments are taken the estimates generated by the sample describing the best locations for the sampling stations.

A progressive, iterative approach leading to reliable metallurgical accounting sampling systems
FROM NON-NEGOTIABLE WEIGHING AND •• Guidelines must emphasise the necessity to have total
existing data transparency: the idea is that if someone
SAMPLING SYSTEMS TO THE ENGINEERING cannot take advantage of existing data, someone else
After all this necessary preparation has been done, it becomes at a very high level in the company most certainly will.
possible to require engineering firms to comply with basic
There must be a self-auditing data management system
requirements; then it is beneficial for the company to make
to optimise the recovery of your natural resources: in
sure all proposals obey the same rules and there is no chance
the 21st century it is hoped that new operations should
for engineering to cut corners on required sampling systems
be able to avoid the old and naïve tonnage philosophy
and weighing systems as they often do to win bids. It is
important to be aware that many engineering firms take the coupled with the improper use of data as a tool of power.
least resistance approach, something along the following •• Decide where to install weightometers, ensuring
lines: ‘if the project team did not make up its mind about the that a critical weightometer reading can be directly
many necessary details on how to build sampling systems compared with another weightometer exposed to the
for material balance and process control, weightometers, same tonnages at all times. They must be installed on
flowmeters, sample preparation facilities and laboratories, it horizontal belts, on a concrete floor, in the shade, and the
will be given low standard, cheap systems, so the proposal belts should be less than 100 m long. Hendrik (1975), a
from the engineering firm can be as low as possible in order well-known expert on weighing systems, even suggests
to win the contract’. Although this approach leads to a low a maximum of 75 m. The reason is that the calibration is
capital cost, invariably it has devastating consequences for performed on an empty belt and excessive tonnage on
the project as for its operating life it will struggle to optimise the belt will result in excessive tension and possible false
the recovery of natural resources at a minimum cost. This readings regardless of calibration.
is a typical capital cost over operating cost trade-off and
the engineer is not always to blame. ‘Value engineering’ is •• Most procurement processes require three competitive
a beloved term of owners who believe that 10 per cent or bids and this is an engineering company function in
20 per cent can be cut out of any capital cost. An engineer will most instances. If the weightometer specification and
often specify acceptable (though likely not the best) sampling sampling systems are correct and the procurement
systems, but this is often where the costs are cut. Rarely is process follows its required rules of bid compliance,
the operating cost considered and metallurgical accounting is then this is the most appropriate pathway to ensuring
certainly not a priority of ‘value engineers’. the correct units are installed in the correct manner.
The prevention of such a prevalent design process anomaly
is achieved through implementing a careful strategy that Engineering firm proposals
leads to well developed and agreed in-house guidelines, and Though no details will be available at the proposal stage, it is
foremost, is the product of firm leadership. The following of the utmost importance to take a few precautions. In a large
sequence is suggested. majority of cases specifications have been put together by an
engineering firm in the prefeasibility study (PFS) or feasibility
Necessary homework in conjunction with engineering firms study (FS) phases on behalf of the owner. It is the engineer in
•• Divide the deposit into its basic geological units. these phases, together with the owner, that set the scene as to
•• A geometallurgical analysis needs to be made of the what is to be built, how much geometallurgy is performed,
proposed mine planning strategy to determine if the what design principles are applied etc. There is a distinction
planned ore delivery system is optimal for the process made between the study team (owner plus PFS or FS engineer)
and the project economics. Ideally the mine plan and the engineer selected to construct the project (who is often,
(number of concurrent ore faces, truck size, mining but not always, different than the study engineer). The study/
method and stockpiling strategy) is altered in light of the implementation interface is where lots of money is ‘saved’
processing needs of the project. Often the geology, the and it is where actual procurement happens. If the sampling
mine plan and flow sheet change during the engineering rules are established correctly and ‘part of the furniture’ in the
study phases. It is essential that the engineer and the study phases then it is more than likely they will survive the
owner agree on principles of sampling and apply these implementation phase ‘unscathed’.
religiously as a team while the project evolves.
•• Perform heterogeneity tests for each geological unit or Decision-making for a given proposal
group of units. Under no circumstances should a decision be made to
•• Optimise sampling protocols. Naturally the three select a given engineering firm until all the above has been
former steps are moving targets and not necessarily in thoroughly investigated. If the project is to be implemented as
chronological order. a fixed price, lump sum-turn key etc then the owner has ‘dealt
•• Design sampling systems for material balance at the themselves out of the poker game’ once they award the project
plant: allow for 30 per cent streamflow over design. to an engineer. If the specifications are not correct at that point
They must all be cross-stream samplers installed under it is too late because the engineer has full control. It is better
the discharge of the streams. These systems must be for the engineer and owner to work in partnership through to
reasonably proportional, an almost always important commissioning and handover while changes are still possible.
property of a reliable sampling system. It would be unfortunate to have a situation where the owner
•• Design sampling systems for process control. cannot change the rules because all variations are outside of
•• Make the Process Information system easily accessible, the original contract.
so anyone within the company can have easy access to An exercise is suggested in the Appendix to help a project
existing data; usually people do not take full advantage team acquire the necessary information in a new project prior
of existing data. to a FS; it may provide a convenient checklist.

F F Pitard
THE MYSTERIOUS VIRTUES OF MANUAL GRAB SAMPLING there is no doubt in the mind of this author that the TOS
and Measurement of Uncertainty are complementary worlds
There is another disease worth mentioning in this paper
rather than competing worlds; perhaps it is just a matter of
because it is widely spread in many industries and especially
working on agreed notations and definitions which should be
the mining industry. For 40 years sampling professionals
achievable with cooperation.
have been teaching industry that grab samples collected by
operators every hour and sometimes every two hours have It is essential that companies and engineering firms adopt
zero value to manage an operation properly. Despite many principles at the beginning of projects in such a way that
courses being presented on this issue, grab sampling still they do not become the target of manufacturers who often
occurs at many operations. transgress the most elementary rules of sampling correctness
in order to sell their products.
Grab samples collected every hour or every two hours
cannot be representative for the following reasons:
•• They are non-probabilistic and therefore introduce ACKNOWLEDGEMENTS
delimitation biases, extraction biases and weighting This author is thankful to Anne Hollonds formerly with
biases. Independent Process Analysis Inc whose keynote paper at
•• They are operator dependent, and therefore MetPlant 2006 inspired him to create the exercise suggested
unpredictable. in the Appendix. One of her case studies gave us the ground
to adapt it to a more integrated exercise combining quality
•• Most of all, they introduce a massive interpolation error assurance / quality control requirements with those suggested
because they are collected at very large time intervals, by the TOS. The author also wishes to thank David Pollard
when they should be collected every ten minutes. and Dean David who kindly helped to improve this paper.
•• Sample results from single increments are affected by
a huge uncertainty that may have nothing to do with
the process. In these instances, the operators are better
REFERENCES
Carrasco, P, Carrasco, P and Jara, E, 2003. The economic impact of
off leaving the process alone than making modifications
correct sampling and analysis practices in the copper mining
that may be detrimental. industry, in a special issue by Chemometrics and Intelligent
It cannot be emphasised too strongly that for process control Laboratory Systems: 50 years of Pierre Gy’s Theory of Sampling,
and metallurgical accounting such irresponsible practices in Proceedings First World Conference on Sampling and Blending
should be forbidden as they transgress the most elementary (WCSB1), pp 209–231 (Elsevier).
rules of due-diligence. Carrasco, P, Wilke, A, Jara, E and Suarez, E, 2007. Some applications
of statistics and geostatistics to the sampling of a big copper
tail stream, Proceedings Third World Conference on Sampling and
CONCLUSIONS Blending (WCSB3), Porto Alegre, Brazil, Fundação Luiz Englert.
Companies and engineering firms vastly undervalue the David, M, 1977. Geostatistical Ore Reserve Estimation, 384 p (Elsevier:
skills of qualified and experienced sampling experts and Amsterdam, Oxford, New York).
should engage their help at the beginning of a project and
Gy, P M, 1967. L’Echantillonage des Minerais en Vrac (Sampling of
during feasibility studies. Reliance on the principles taught
particulate materials), Volume 1, Revue de l’Industrie Minerale,
by the TOS is necessary because it is the only way to prevent Special issue, January 15 (St Etienne: France).
the inevitable major reconciliation problems that occur
between the mine and the plant. There is criticism from many Hendrik, C, 1975. Weighing and Proportioning of Bulk Solids, series
on Bulk Materials Handling, volume 1 (1975/77), Trans Tech
working in the world of ‘Measurement Uncertainty’, that
Publications.
the error terms listed in the TOS are largely not quantifiable.
Such a statement completely misses the point that the TOS Matheron, G, 1967. Comparison between samples with constant
mass and samples with constant fragment population size
is a preventive tool to minimise uncertainty bias. When
(and calculations of their sampling variances) (translated by
sampling is reviewed during plant operation it is most likely D François-Bongarçon and F F Pitard), in Proceedings World
the measurement of uncertainty would generate acceptable Conference on Sampling and Blending WCSB7 (IMP).
outcomes rather than generate bad news. To be positive,

A progressive, iterative approach leading to reliable metallurgical accounting sampling systems
Appendix
Sampling and analytical quality assurance and quality control requirements for a new project
The following data validation characteristics are necessary for the further validation of almost everything done in a feasibility study. It is an exercise to be performed by
the working team of a new project. It is a quick and qualitative checklist.
1. Please define the overall confidence in the sampling protocols used in the project, from early exploration to later feasibility:
Excellent Moderate Poor
1 2 3 4 5
□ □ □ □ □
2. Please define the overall confidence in the correctness of the sampling equipment used to implement the various sampling protocols:
1 2 3 4 5
□ □ □ □ □
3. Describe how the preliminary mineralogical study of major, minor, and trace commodities in the various geological units was performed, with respect to the
acquisition of necessary information to define sampling characteristics:
Yes No
Finely disseminated and invisible to the naked eye □ □
Finely disseminated, but visible to the naked eye □ □
Clustering? □ □
If clustering, clustering where? Describe:
Did you log the size of clusters during exploration? □ □
‘Nuggeting’? □ □
Did you log the size of nuggets during exploration? □ □
4. Describe how the heterogeneity of major, minor, and trace commodities in the various geological units was quantified:
Yes No
Were sampling constants determined? □ □
Were liberation models determined? □ □
5. Please define the heterogeneity of major, minor, and trace commodities in the various geological units:
Little Moderate Much
1 2 3 4 5
Major commodities □ □ □ □ □
Minor commodities □ □ □ □ □
Trace commodities □ □ □ □ □
6. Please define the overall confidence in the determination and use of the heterogeneity characteristics of major, minor, and trace commodities to optimise the
following areas/issues:
Areas/Issues Excellent Good Moderate Poor Bad
Data quality objectives □ □ □ □ □
Selection of drilling diameter □ □ □ □ □
Selection of diamond core drilling □ □ □ □ □
Selection of reverse circulation drilling □ □ □ □ □
Validity of cutting core in halves □ □ □ □ □
Determination of necessary sample mass: □ □ □ □ □
• to send to the laboratory □ □ □ □ □
• for subsampling stages □ □ □ □ □
• to be assayed □ □ □ □ □
• for blastholes □ □ □ □ □
• for underground sampling □ □ □ □ □
• for metallurgical testing □ □ □ □ □
• from run-of-mine feeding the plant □ □ □ □ □
Design of sampling stations for the plant □ □ □ □ □
Selection of the overall sampling equipment □ □ □ □ □

F F Pitard
7. Please describe if the following practices are used in the project: 14. Please describe if blanks and reference materials were used and how
Yes No many:
Sampling of stockpiles □ □ Yes No
• If no, describe alternatives: Blanks □ □
Sampling of trucks or railroad cars □ □ >5 3–5 1–2 None

• If no, describe alternatives: Reference materials □ □ □ □
Sampling of underground drawpoints □ □ 15. Please describe who will be in charge of operating, maintaining and
• If no, describe alternatives: cleaning sampling systems during the operations:
Yes No
Quality assurance/quality □ □
8. Please describe if the following pilot plant sampling systems and
control department
methodology were used for sampling feed, tail and concentrates:
Production departments □ □
Yes No
Cross-belt samplers □ □ 16. What budget, expressed in % of total operating cost, will be given to
Stream discharge cross-stream samplers □ □ sampling and assaying resources in general (ie, capital, operation and
Manual sampling □ □ preventative maintenance):
>2% 1–2% <1%
9. Please describe if the following pilot plant sampling systems and
□ □ □
methodology were used for process control sampling:
Yes No 17. Please describe if analytical methods were the object of a careful
On-line analysers □ □ investigation, mainly looking at the scope versus the principle:
Manual sampling □ □ Excellent Moderate Poor
1 2 3 4 5
10. Please describe if the following future plant sampling systems and
□ □ □ □ □
methodology will be used for sampling feed, tail and concentrates:
Yes No 18. Please describe if the inevitable amount of ore misclassification was
Cross-belt samplers □ □ roughly quantified as a function of the precision/reproducibility of
Stream discharge cross-stream samplers □ □ selected future sampling and analytical protocols:
Manual sampling □ □
1 2 3 4 5
11. Please describe if the following future plant sampling systems and
□ □ □ □ □
methodology will be used for process control sampling:
Yes No 19. Please describe the current practice of geologists, engineers and
On-line analysers □ □ metallurgists in the following areas:
Manual sampling □ □
Yes No
Has high grade been cut from the database? □ □
12. Please describe the overall confidence in the quality control of the
Are outliers believed to be bad data? □ □
implementation of all sampling, subsampling and assaying protocols:
Is sample recovery an issue well addressed in drilling contracts? □ □
1 2 3 4 5 20. Peer review: as a team, how do you feel about this project? For example,
□ □ □ □ □ is it worth the investment of many hundreds of millions of dollars? Would
the company involved be able to fully optimise the economic recovery of
13. Please describe the number of duplicate samples performed for the the predicted natural resources?
following areas:
21. In the event of a poor reconciliation event between mine grade and plant
5% 2% <1% None grade, describe the detailed procedure that will be implemented on this
Blastholes □ □ □ □
site to resolve the issue:
½ core □ □ □ □
¼ core □ □ □ □
Reverse circulation □ □ □ □
Underground channels □ □ □ □
Underground drawpoints □ □ □ □
Subsamples at sample preparation facility □ □ □ □
Assays □ □ □ □
Pilot plants □ □ □ □
Metallurgical tests □ □ □ □
Future plant sampling □ □ □ □

Contents
Sampling, corporate governance and risk analysis

G J Lyman1 and F S Bourgeois2
ABSTRACT
As prices of metals fall, financial control of mineral processing operations becomes
more and more important. Control of a mineral processing operation implies
measurement of grades, recoveries and losses, with every effort being made to recover
more product from the same feed. As a result, shareholders become more concerned
with the transparency of the operations and the corporate governance.
It is very important to have robust sampling and to make good use of the results
of that sampling. Sampling systems that provide only ‘rough’ figures are not good
enough. Similarly, a competitive position in the marketplace demands close control
of product quality. Maximisation of product output demands that the product meets
but does not exceed the specification so that product yield is maximised. Decisions
based on precise sampling results carry less risk that those based on imprecise results.
This paper reports some advances in sampling theory that permit full quantification
of sampling distributions, not just the sampling variance as in the past. These advances
can be used together with simulation methods to tie down sampling precisions and
develop calculations that quantify the risk when using the sampling results to make
decisions. Several examples illustrate the use of these advances and a final example
indicates the development of risk analysis curves in the coupling of sampling precision
to sale of a commodity.
INTRODUCTION
The control of mineral processing operations is becoming an increasingly complex task
as the grades of ores decrease, the scale of operation increases and financial control of
the operation becomes more critical. The role of sampling is gaining in importance,
as it is the measurement technology that empowers the control of operations. It is
no longer adequate to have sampling technology that carries financially significant
uncertainty in metal content; the sampling must have a precision that matches the
demands of shareholders for high quality information concerning production in terms
of objectives and the value of the resource. These are elements of corporate governance
and the data feeds directly into the assessment of risk of ongoing operations. Further,
control of product quality is critical in the marketplace and ensuring that the product
attains maximum value within the contract specification. Indeed, the ability to specify
the quality of a consignment of a product within tight tolerances, supported by
documentation, ensures that the seller and the buyer realise the value of the product.
Close control of a product quality also permits the supply of a product that just meets
the agreed specification, maximising the value of the resource.
This paper examines recent developments in sampling theory and shows how these
developments can be employed to provide full risk analysis associated with either the
purchase or sale of commodities and, in particular, precious metal ores.
SAMPLING THEORY AND COMPUTATIONAL DEVELOPMENTS

Beginning in the 1950s, Pierre Gy (1953, 1967, 1982) began the development of a
comprehensive theory of sampling. There had been some notable contributions
made to the science and practice of sampling as far back as 1895 (Brunton, 1895, 1909)
with the development of useful sampling equipment that was able to take unbiased
samples. However, the adoption of good sampling practice was slow in penetration of
the mineral industry and slower in the sampling of commodities such as grain. Even
with the publication in 1945 of Taggart’s Handbook of Mineral Dressing (Taggart, 1945),
there is support for devices and methods of sampling that are ‘mechanically incorrect’
meaning that there is a substantial risk, indeed an expectation, that the devices will
1. Principal, Materials Sampling & Consulting. collect biased samples. The theory of sampling espoused in this Handbook is also
Email: geoff.lyman@gmail.com complex, limited in application and difficult to apply. By 1972, when Gy presented his
2. Université de Toulouse, Laboratoire de Génie Doctoral Thesis, he had worked out the basis for a comprehensive theory and practice
Chimique, Toulouse, France. of sampling. Gy’s first book in English appeared in 1979, but is perhaps best known
Email: florent.bourgeois@inp-toulouse.fr for the second edition of that book that appeared in 1982.
349
The theory has two parts. The first part deals with the Finally, it was possible to make a complete estimate of the
heterogeneity of a particulate material such as a broken ore variance of sampling due both to the variation of grade in
and provides a basis for the calculation of the variance of the process stream and the particulate nature of the material.
sampling that applies to a sample of a given mass with the The two components are additive and, when one adds the
material in a given state of comminution. The second part variance of analysis of the sample to these, the total sampling
deals with the method of calculating the variance of sampling variance is available. The theory describes the variance when
from a process stream in which the grade of the target analyte all stages of sampling are mechanically correct.
is fluctuating with time. Many authors had dealt in various In a statistical problem, the analysis is not complete until
manners with the sampling variance due to the particulate it is possible to determine the entire probability distribution
nature of a mineral mixture, but none had been able to deal for the quantity under scrutiny. The variance of a statistical
with the problem of determining the variance of sampling due quantity measures the spread of the distribution, but gives no
to fluctuations in grade when sampling by taking increments information beyond that as to the full nature of the distribution.
of solids periodically from the process flow.
The authors have applied a mathematical method to the
Gy solved this problem by picking up the nascent theory particulate sampling problem that permits the calculation
of geostatistics as it was being developed by Georges of the full distribution of sampling (Lyman, 2014). This
Matheron at Ecole des Mines de Paris at Fontainebleau. In development provides the full distribution of one of the three
fact, the mathematics of time series analysis was already components of uncertainty of sampling.
available to enable the variance calculation, but this area of
Next, due to computational developments in the mid-1990s,
mathematics was essentially unknown to mineral processors
it is possible to generate accurate and very large realisations
and miners when Matheron began to construct the discipline
of random functions in any number of dimensions. Twenty
of geostatistics. Adding this capability to sampling theory years later, these new computational methods, known as
was Gy’s stroke of genius. Gy introduced the concept that circulant embedding, are beginning to be incorporated into
the time-variation of grade in a process stream could be the geostatistical packages offered by software specialists.
modelled by a stationary random function in the same way Materials Sampling & Consulting has its own software
that the geostatisticians model the variation of grade within implementing these methods.
a domain of a mineral deposit. A stationary random function
is one whose statistical properties do not change with time. Having captured the nature of the time-variation of grade in
However, Gy avoided introducing a lot of the classical time the process stream using a variogram and the distribution of
series analysis tools that existed. the grade about the mean, which may or may not be normal
(Gaussian), it is necessary to determine the distribution of the
Figure 1 shows such a random function after removal of uncertainty which derives from the sampling of the process
long-term movement of the mean. stream according to some protocol (systematic or stratified
The nature of the random function is characterised by a random sampling). When a process stream is sampled by
covariance function or a variogram, as used by geostatisticians. periodically extracting increments from the process stream,
There is nothing new in modelling process variations as there is a difference between the true average grade of the
random functions (stochastic processes); the methodology process stream over the sampling period and the grade of
has been in use in chemical engineering and general control the sample formed from the collection of the increments. It is
engineering since the 1960s. But it was certainly new to the very difficult to determine the distribution of this uncertainty
theory of sampling when introduced by Gy. theoretically, but the time-variation of the grade in the process
The variograms corresponding to the first 3000 and the last flow can be simulated on a very fine time scale using circulant
3000 points in the data set shown are given in Figure 2. The embedding methods and the sampling likewise simulated
two variograms are essentially identical and show a diurnal by extracting values from the process grade simulation. By
component with a period of 24 hours as well as a mixing generating many possible realisations of the grade variation and
component (an exponential covariance). The last component simulating the sampling for each realisation, the distribution of
is a measurement variance which is uncorrelated between the uncertainty due to sampling can be captured.
one reading and the next. This latter component shows up as Figure 3 shows a simulation of the grade of the feed to a
a non-zero intercept when the variogram is extrapolated to plant treating a mixture of low (2 g/t) and high-grade (10 g/t)
zero-time lag. gold ore in a circumstance where the two ore types cannot be
FIG 1 – Example of a random function. Variation in grade of a process stream after trend removal.

FIG 2 – Variograms corresponding to the first (red) and last (green) 3000 points in the data set of Figure 1. An approximate
95 per cent confidence interval on the red variogram is indicated by light blue lines. The black line is a fitted variogram (covariance)
model made up from an uncorrelated component, an exponential correlation and a damped periodic correlation.
effectively blended prior to feeding. For a 200 mm top size CORPORATE GOVERNANCE AND RISK ANALYSIS
of the ore and a feed rate of 1000 t/h, the primary increment
Good corporate governance demands that reports to
mass is 277 kg. The simulation was carried out on a 277 kg
shareholders and stock market regulators are accurate.
basis (one second of feed) for an eight hour period, with
Just how accurate the reporting of resources and their
32 samples taken systematically. The simulation required
reconciliation with actual production must be is a matter for
realisations of grade containing 28 800 points. An accurate
regulatory bodies or company directors. The determination
simulation of this magnitude is not possible without the
of production is ultimately finalised when the product is
circulant embedding method.
sold and is dependent on precise, unbiased sampling of the
The sampling simulation was carried out 5000 times to product. Prior to product sale, the mass of metal produced
obtain a good estimate of the distribution of the sampling is linked to sampling of the products of the beneficiation
uncertainty. Figure 4 shows the distribution. Surprisingly, processes employed and material balances around the
while the distribution is symmetric, indicating unbiased plant. Determination of losses of metal values is also based
sampling as it should, the distribution is not Gaussian. Instead, on sampling and material balances around processes.
it is a Laplace distribution (double-sided exponential), which Reconciliation of mine output with plant feed may be based
has heavier tails than the normal distribution. solely on sampling of plant feed or may involve the material
The last component of uncertainty is in the analysis, which balances on the plant as a whole. This reconciliation is critical
arises from the random errors that come into play when diluting to the assessment of the resource estimation and grade control
solutions, titrating or weighing. With instrumental methods procedures. The accuracy with which these balances can be
of analysis, the errors are usually due to solution aspiration, carried out is based solely on the accuracy of sampling and
instrument electronics, calibration and matrix effects. the distribution of sampling results.
The three components of uncertainty in a sampling result are Risk analysis in metal mining operations depends not
statistically independent and it is therefore possible to correctly only on the results of sampling but also on the distribution
combine their distributions using the method of characteristic of uncertainty attached to the results. To state that the
functions. This fact has not been appreciated until recently estimated recovery of metal is going to be 87 per cent over
proposed by the authors (Lyman, 2015). This methodology the next month is not meaningful until a confidence interval
makes it possible to accurately determine the total sampling has been placed on the estimated recovery. If the figure is
distribution due to all three components of uncertainty. known to ±10 per cent, such uncertainty may be completely
FIG 3 – A realisation of the feed grade to a gold plant showing the points in time at which primary sampling increments are taken.

FIG 4 – Distribution of sampling uncertainty for gold ore. Solid line is a fit to the simulation results (dashed).
unacceptable, while an interval of ±1 per cent may enable is a case in which comminution of the ore does not reduce
close planning of product shipments to consumers as well the sampling uncertainty until breakage of the gold particles
as close product quality control. The specification of a actually takes place.
confidence interval is based on knowledge of the distribution The successful marketing and sale of a commodity is heavily
of the figure involved. To simply assume that the distribution dependent on good sampling of the consignments delivered to
is normal introduces additional levels of risk in using the the client. When a product is to be sold at a specified content
result. First, the distribution may be symmetric, but it may not of valuable (or diluent in the case of coal), the maximum value
be normal. The distribution of Figure 4 is a perfect example of the product is realised by ensuring that the consignment just
of being misled by an unjustified assumption of normality. succeeds in meeting the specification both in terms of grade
The standard deviation of the difference in grade is 0.507 of the primary element and the allowable levels of deleterious
which would lead to a normal 95 per cent confidence interval elements, unless there is a premium paid for higher grade
of ±0.994 g/t. The actual 95 per cent confidence interval is material that operates from the specified grade upwards. To
±1.175, nearly 20 per cent wider. Second, the distribution this end, the seller should have in place sampling and blending
may be skewed, making the confidence intervals asymmetric; systems that permit very close control of product quality.
simply given a variance value, there is no way of knowing
Consider a producer of an ore with a specification of
whether there is asymmetry present.
>44 per cent contained metal which commands a price of
The sampling of ores containing coarse gold can lead to US$3.0/dmtu free on board, but no premium for supplying
unexpected problems if the size distribution of the gold higher grade. A tonne of ore at 44 per cent is then worth
is not well understood. Consider a case in which the ore is US$132. However, to ensure that the seller does not incur a
being treated by a third-party in their mill and the mill feed penalty for providing a product with less than 44 per cent
must be sampled to establish the value of the ore. The mill metal, they ship at 45 per cent metal. Had they shipped
is equipped with a sampling system through which all toll at 44 per cent, by blending lower grade material into the
ore is passed. The equipment in the system reduces the 45 per cent product, they would have shipped 2.27 per cent
ore to 95 per cent passing 6 mm delivers a 20 kg sample to more ore. On a 50 000 t shipment, this is worth about an extra
the laboratory for each eight hour period of operation. The US$0.15 M out of US$6.6 M for the 50 000 t at 45 per cent
sample is crushed to pass 2 mm and divided down to 2 kg metal. For an annual production of 3.5 Mt, the benefit is
prior to further pulverisation prior to fire assay. The sample about US$10.5 M.
grade is nominally 2 g/t. Figure 5 shows the assumed size The connection between sampling precision and ability to
distribution of the gold particles, assuming also that they are ‘sail closer to the wind’ is conveniently expressed by what
reasonably compact in shape. is usually called an operating characteristic (OC) curve.
While the sample protocol may seem to be reasonable, it However, it can simply be interpreted as a ‘risk’ curve. A true
is possible to determine the sampling distribution for 2 kg OC curve for a shipment of product would be based solely
samples of this ore, using the method developed by Lyman on the accuracy of sampling as a ship is loaded. Once the ore
(2014). The distribution is shown in Figure 6. It is wide, is on the belt on the way to the ship, it is usually too late to
skewed and has a relative standard deviation of 19.7 per cent. adjust the grade of the shipment unless the port is equipped
In fact, even the 20 kg sample from the sampling system will with a very fast turn-around robotic lab which can return an
have a standard deviation of 6.23 per cent. The skewness of assay for a sampling unit within minutes rather than hours.
the distribution will tend to produce sampling results slightly Assume that a contract negotiation begins with a discussion
below the true average value (the most probable value is of how the value of the product is to be decided. The seller
lower than the mean). For commercial purposes, one would explains that he has a very good sampling system and that
hope to achieve a more precise result. its performance has been documented in detail to show that
An appropriate procedure in this case is to retain a larger the sampling system is ‘mechanically correct’ and therefore
sample mass to represent the period of sampling, grind it unbiased. He has also documented the tests and calculations
fine while employing an accelerated leaching process to that have been carried out to prove the system and his
recover the gold and then recover the leached and ground analyses are precise; the seller has a ‘defensible’ sampling
residue for subsampling and fire assay. Sampling for gold system. The buyer does not have such a good system and no

FIG 5 – Size distribution of gold particles for sampling example.
FIG 6 – Sampling distribution for gold at 2 g/t and a 2 mm top size, with the gold particle size distribution of Figure 5.
FIG 7 – Risk curves for ore shipments. Solid curve: SD = 0.106 per cent. Dashed curve: SD = 0.0609 per cent.
solid documentation other than a statement that his system in the ore, he knows that the chemical analysis has a standard
complies with ISO specifications. He cannot prove the deviation of 0.11 per cent for a single analysis and he will
accuracy of his system. It is logical (and hopefully realistic) analyse 5 × 10 000 t sampling units in duplicate. The analytical
to conclude that settlement will be made on samples taken uncertainty is then:
by the seller’s sampling system, although final samples may
be exchanged for assay by the buyer’s third-party laboratory. (0.11) 2/10 = 0.0348%
The seller knows from the analysis of his sampling system so the total sampling and analysis standard deviation is:
that it will operate with a standard deviation of say 0.1 per cent
metal for a 50 000 t consignment and, for the payable element 0.1 2 + 0.0348 2 = 0.106%

On the assumption that his sampling uncertainty follows chemical or physical analysis. In this way, a complete picture
a normal distribution (which in truth may not be a valid of sampling uncertainties can be built up.
assumption), he can now draw a curve which will indicate his The complete picture can be used to optimise the sampling
probability of delivering a consignment with a grade below at any stage of a process and particularly at the sampling of
44 per cent, given that the loads ore at 44 + x%. If the sampling the final product. Adherence to an ISO Standard does not
system standard deviation can be reduced to 0.05 per cent guarantee the required precision.
for the total consignment, by sampling more frequently or
The ability to carry out risk analysis at any stage of sampling
carrying out more analyses per shipment, the total standard
is an important tool enabled by the recent advances and the
deviation will be 0.0609 per cent.
application of good quality control principles and statistics.
It is evident that if the sampling precision is improved and
the risk of an off-spec consignment is held at five per cent, the
grade offset can be reduced from 0.175 to 0.1, permitting more REFERENCES
product to be produced from the same resource. Brunton, D W, 1895. The theory and practice of ore-sampling,
Transactions of the American Institute of Mining Engineers, 25:826–
Of course, it is necessary to analyse, in conjunction, the 844.
precision with which a stockpile can be built to a specified
Brunton, D W, 1909. Modern practice of ore-sampling, Transactions of
grade. That task requires similar considerations to those
the American Institute of Mining Engineers, 40:567–596.
described here. Taken together with the sampling system
for loading the product, these risk curves can be used to Gy, P M, 1953. Error in taking a sample from a lot of ore (Érreur commise
dans la prélèvement d’un échantillon sur un lot de minérai), (Congrès
estimate how much expenditure on sampling can be justified
des Laveries des Mines Métalliques Francaises: Paris).
to provide a payback time of a few years. It is only with this
kind of analysis that the costs of excellent sampling systems Gy, P M, 1967. Sampling of Bulk Minerals, Volume 1, General Theory,
can be put into perspective. (L’échantillonnage des Minérais en Vrac, Tome 1, Théorie Générale), No
56, Éditions BRGM, 186 p (Société de l’Industrie Minérale: Paris).
Gy, P M, 1982. Sampling of Particulate Materials, Theory and Practice
CONCLUSIONS (Elsevier).
Recent advances in sampling theory, together with powerful Lyman, G J, 2014. Determination of the complete sampling
simulation methods, can be used to provide a full picture distribution for a particulate material, in Proceedings Sampling
of the distributions of sampling uncertainty. Previously, 2014, pp 17–24 (The Australasian Institute of Mining and
the Gy theory of sampling provided only an estimate of Metallurgy: Melbourne).
sampling variance. The theory could not be used to determine Lyman, G J, 2015. Complete sampling distribution for primary
whether sampling distributions were Gaussian (normal), or sampling, sample preparation and analysis, in Proceedings Seventh
skewed. The advances also permit the correct combining of World Conference on Sampling and Blending, Bordeaux, 10–12 June.
distributions of uncertainty due to the sampling from the
Taggart, A F, 1945. Handbook of Mineral Dressing – Ores and Industrial
process stream, the particulate nature of the material and the Minerals (Wiley).

Contents
Metallurgical management tools

for continuous improvement
W McCallum1 and R Dunne2
ABSTRACT
Continuous improvement might be defined as seeking incremental improvement over
time through enhanced equipment, processes, technology and the efforts of people.
Most people working in ore processing facilities embrace the general principles of
continuous improvement. However, given the industry-wide issues with skilled
personnel and the constantly changing environment in ore processing facilities, it is
often difficult to identify improvements in a facility over a given time period.
The driver for these outcomes is the use of basic principles and technology to achieve
best practice targets. The process involves four stages: develop standards of practice
to define best practice in specific areas; write procedures to guide the organisation
towards these best practices; implement the procedures and develop a review process
to provide a method of measuring performance against best practice.
By reviewing performance against best practice on a regular basis, the organisation
measures progress towards the best practice targets and the incremental improvement
between these regular reviews is a measure of continuous improvement.
This paper provides the background for developing standards of practice for metal
accounting, metallurgy and metallurgical management. The paper also describes how
standards of practice can be used to develop continuous improvement and the skill
set of the department’s leadership team.
INTRODUCTION
Most managers despair when attempting to quantify performance improvements
from one period to another. The apparent changes in results are generally small and
the variables sufficiently different to make it impossible to know whether performance
is static, has improved or even regressed. This paper considers the application
of performance drivers as the measure of performance. If the fundamentals of the
operation are properly applied (and aligned), then best performance should follow.
This involves a subtle change in measuring performance. The measure of
performance is no longer the outcomes, but the application of the drivers that generate
these outcomes. These drivers might be as fundamental as ensuring a weightometer is
calibrated or as complex as applying flotation models, however the more frequently
these drivers are applied, the better will be the metallurgical performance.
Moreover, by measuring the drivers of performance as distinct from the outcomes,
other areas can also be reviewed. One example is multiskilling of the workforce. The
impact of multiskilling is not directly reflected in recovery, concentrate grade or
throughput. However intuitively, a workforce where 50 per cent of the participants
can competently operate two or more areas of an ore processing facility will lead to
better outcomes than if only 20 per cent of workforce can competently operate two or
more areas of the facility.
It is measuring these fundamental drivers that allow for continuous improvement
and the development of people. This involves four steps:
1. develop standards of practice to formalise ‘best practice’
2. develop procedures and actions around the fundamental drivers that will lead
to the target best practice
3. implement these procedures and actions to ensure the business is moving
1. FAusIMM, Principal Metallurgist, towards best practice
Havilah Consulting PL, Tatura Vic 3616.
4. develop processes for measuring how well the procedures and actions are
Email: bill@havilah.com.au
applied and how the results they are generating reflect the best practice target.
2. MAusIMM, Metallurgical Manager Technical
Services, Newmont Mining Corporation, For example, the standard of practice might include as a best practice: ‘the
Englewood CO 80112, USA. weightometer measuring feed to the ore processing facility shall not vary more
Email: robert.dunne@newmont.com than one per cent from true accuracy ...’. In response, a site might write one (or
355
several) procedures for the plant feed weightometer that properly addressed. A standard should be a controlled
include calibration procedures, management of the data document and contain the following headings.
from calibrations, maintenance tasks and schedules, regular
inspections and house-keeping by operations people and Purpose
compilation of the records of all work completed. Describe the purpose and/or intent of the standard;
The site applies all the new procedures and actions, and for for example, ‘provide guidelines that will improve the
the next few months, weightometer accuracy is 1.5 per cent. accuracy of metal accounting within ore processing facilities.’
A review of this specific best practice might award the site
three (out of five) for this improvement. There are procedures Scope
and actions, there is evidence the site is applying these Detail where the standard applies; for example, ‘this standard
procedures and actions, however the target of one per cent applies to all mineral processing facilities in the XYZ Group.’
has not been met.
The site reviews its procedures and decides to reduce the Definitions
period between calibrations. Over the next year the accuracy Standardise the terms and their meaning. For example,
increases to 1.2 per cent. This result still does not meet the ‘availability is the amount of time (expressed as a percentage)
best practice target but is a significant improvement on the in any period the primary mill can operate,’ and ‘utilisation is
previous year and would now be worthy of perhaps a four the amount of time (expressed as a percentage) in any period
(out of five) after the review. The site reviews the whole the primary mill is operating.’
weightometer operation and finds the only way forward is
to completely redesign the weightometer installation using a Body
new weightometer. The site decides the costs far outweighs The specific best practices covered under this particular
the benefits and settle for their four-star rating in this area. standard of practice. ‘The reported value of (dry) ore milled is
Despite not meeting the target best practice, the site has: within 2 per cent of the true value,’ and/or ‘the sampled head
•• defined a best practice for one of the fundamental grade is within 2 per cent of the reconciled head grade.’
drivers for metal accounting
•• gained an understanding of this fundamental driver
Responsibility
by developing processes and actions that improve the Implementation of the standards of practice will require
performance of this fundamental action persons with the responsibility and authority to drive the
process. It is important these people are named and their
•• improved the performance of a fundamental driver in
responsibilities and authority clearly defined.
metal accounting over a two-year period.
References
BEST PRACTICE Provide references to documents that support the standards
A quick review of literature provides many similar definitions of practice.
for best practice with the differences between each definition Understanding the key fundamental drivers for a process
being driven by the industry providing the definition. and/or facility significantly reduces the difficulty in
A suitable definition for the mining industry could be: ‘a developing any standard of practice.
system of procedures and actions that lead to consistent
results above typical industry performance.’
Standard of practice for metallurgy
This definition allows for different organisations to focus on The best metallurgical results for a specific ore processing
specific areas that will make their business better. For example: facility will occur when the plant and equipment and operating
•• An organisation in the bulk commodities sector might conditions are configured to ‘best’ suit the properties of the
utilise blending of many materials to meet marketing ore. This is generally not too much of an issue when a plant
specifications. This business may require very high has been designed to process a specific ore. Older plants
weightometer accuracy. treating new ores may require significant ‘tuning’ and even
•• A business operating a copper heap leach might settle then the results might not be spectacular. However, if the
for a less accurate weightometer but requires the particle plant and equipment and operating conditions are correctly
size distribution of the material being stacked be within aligned with the ore characteristics, the results will be as good
a tight range. as can be achieved.
•• Another organisation might decide management of There are three important areas in developing this
reagent additions to the flotation circuit is a key driver understanding of ore properties, plant and equipment and
in their performance. operating conditions:
When considering the best practice targets for a standard 1. Metallurgical testing of the ore to determine such data as
of practice, it is important to understand the fundamental comminution properties, optimum grind size, residence
drivers that have the biggest impact on the business. Consider time, reagent consumption, nominal recovery, leach
the current understanding and performance of these drivers rates, concentrate grade, tailings properties etc.
in your business and that of the industry as a whole. Set best 2. Modelling and simulation of the plant to ensure
practice targets that will make your business an industry leader. equipment and operating conditions are optimised to
suit the ore. Typical of the technology available is the JK
Preparing a standard of practice SimMet software package.
It is unlikely one standard of practice will satisfactorily cover 3. Specialised instrumentation to monitor plant
a site or an organisation. Writing several standards that performance. Instruments can range from plant based
apply to separate functions allows each standard to focus machines such as ‘on stream analysers’ and ‘cyanide
on a specific area and also ensures organisational goals are analysers’ through to services provided by external

Metallurgical management tools for continuous improvement
providers such as mineral liberation analysers and Time- Samples

of-Flight Secondary Ion Mass Spectrometry (ToF SIMS).
The samples collected through the ore processing facility
A standard of practice for metallurgy should include the must be representative. Examples of best practice might be:
following best practices in the body of the document. ‘the mass of the sample collected is sufficient to represent
the material sampled’ and ‘the frequency of stream cutting is
Ore characterisation sufficient to represent the material sampled.’
Detail the minimum test work that should be conducted (and
the regularity of such work) that would make your business Assaying
a leader in this area. Examples of best practice might be:
Whether an internal or an external assay laboratory is used,
‘the comminution properties of the ore shall be evaluated
annually,’ or ‘the net acid generating potential of the plant the actual assay results must be accurate. Examples of best
tailings will be determined biannually.’ practice are: ‘the assay laboratory has a quality control
process’ and ‘the assay technique is suitable for the material
Process optimisation assayed.’
Process optimisation involves modelling and simulation of the
circuit. Examples of best practice might be: ‘the comminution Arithmetic
circuit will be modelled annually using suitable software,’ The arithmetic is what pulls the different bits of information
and ‘the performance of the comminution circuit will be together. The formulae must accurately calculate the variable.
evaluated annually.’ Examples of best practice are ‘the metallurgical accounting
spreadsheets and/or databases are audited by an independent
Metal loss analysis party annually’ and ‘there is a change management process
Analysis of tailings losses provides the basic data for that records changes made to data.’
(potentially) improving recovery. Examples of best practice
might include: ‘size-by-size recovery is analysed on monthly Reconciliation with production
tailings composites,’ and/or ‘extended leach tests are It is important to reconcile all metal accounting with the final
conducted daily.’
metal content of the product. An example of a best practice
might be: ‘the sampled head grade and the reconciled head
Reagent optimisation grade are within 2.5 per cent.’
Reagent optimisation provides an opportunity to reduce
operating costs and/or improve metallurgical performance. Data management
An example of a best practice might be: ‘regularly test reagent
additions and evaluate alternate reagents.’ It is important to maintain the files used and generated by
metal accounting. Example of best practice might be: ‘backups
Trade-offs of all metal accounting files are kept in a separate building’
The site should be aware of the impact changes to the current and ‘there is a change management procedure to manage
operating conditions will have on performance. ‘Evaluate the changes to the data.’
relationship between milling rate and recovery.’
Standard of practice for ore processing facility management
Standard of practice for metal accounting Metallurgical management is a broad area. It includes:
Metal accounting is the mechanism that records and reports •• meeting production targets in a safe and environmentally
the distribution of metal through an ore processing facility. sound manner
Perhaps the most important function of metal accounting is
•• the preparation of accurate production, operating cost
the Mine to Mill reconciliation which provides a check on the
and capital cost budgets
geological model of the resource. The geological model is the
basis of mineral reserves calculations and the mine model, •• training
which generates the ore for processing. •• reporting.
The information generated by metal accounting is also The organisational structures on mine sites vary, but
used in financial accounting (metal stocks), corporate generally safety, training, human resources and environment
reporting and determining metallurgical performance. are managed by departments independent of ore processing.
Whatever the use, it is important the data accurately reflects Irrespective of the organisational structure, the ore
the metallurgical operation.
processing facility should operate in conjunction with (and
Data generated by metal accounting results from a series support of) these groups. Any standard of practice should
of mainly independent processes including measuring, consider the relationships between ore processing and these
sampling, assaying, arithmetic and data management. If external functions.
the fundamentals of these largely independent processes
are completed correctly, the data generated by the metal The underlying driver for accurate production and
accounting process is likely to be accurate. operating cost budgets is knowledge of the ore to be processed
in the budget period. Admittedly this relies on the mining
Some areas that might be considered in a standard of
practice are as follows. department having their own realistic budget, but given such
a document exists, it is the responsibility of the management
The measurement of ore processed team to ensure grinding rates, reagent consumptions,
recovery, concentrate grades and tailings properties are
This is probably (arguably) the most important aspect of
metal accounting. An example of a best practice in this area understood and written into budgets.
might be: ‘the dry weight of mill feed in any period will be Components of the standard of practice for the management
within two per cent of the true value.’ of an ore processing facility should include the following.

Production schedule Site 1

The metallurgical characteristics of the ore scheduled for This site checks the calibration every six weeks (on the
processing during the next year provide the most accurate planned maintenance day), records the error and resets the
data for inclusion in the production schedule. An example of calibration. Moisture content is estimated as 4.5 per cent.
best practice might be: ‘the comminution properties of ores
schedule for processing in the next budget period are known.’ Site 2
This site also checks the calibration every six weeks, records
Operating costs the error and resets the calibration. The site also corrects the
The operating cost budget is developed from unit costs (labour, ore milled figures based on the calibration error. Moisture
content is determined by taking a 3 m sample off the mill feed
reagents, electricity) and consumption rates. Examples of
conveyor on the last day of the month and using the moisture
best practice might include: ‘the reagent consumption for
in this sample to correct throughput for the entire month.
ores scheduled for processing in the next budget period are
determined from metallurgical testing’ and ‘reagent purchase
Site 3
costs are agreed with the supplier.’
Site 3 checks the calibration on the middle and last day of
every month. They turn the feed to the mill off for ten minutes
Capital costs and check the calibration using static weights. They correct
Capital costs included in a budget should be scrutinised to the calibration and then correct the ore milled for the previous
some degree prior to inclusion to develop some confidence two weeks. Moisture samples are collected every day by
the project is viable, the cost estimate is reasonable and the stopping the feed belt and taking a sample.
timing appropriate. An example of best practice might be:
‘items requiring capital expenditure are supported by a Site 4
preliminary capital expenditure requisition including a basic This facility completed a lot of work on weightometer
project description and evaluation.’ calibrations a couple of years ago and found if they do a full
calibration check every 11 days, the calibration error is always
People less than one per cent. The site does a calibration check every
While there may be a separate human resources (HR) 11 days, resets the calibration and reworks the ore milled
department, many of the factors impacting the people for the past 11 days; keeping a record of exactly what they
within a processing department are less a function of the did. Moisture samples are taken every 37 minutes (based
HR department and more a function of the leadership and on statistical analysis) with the sample sized at 50 mm on a
management within the processing department. A very grizzly. The oversize is weighed and the undersize is used for
simple best practice might be: ‘processing department moisture analysis.
turnover is less than the average for the entire site.’ Perhaps the only site that meets the best practice is Site 4
and this site may score five stars for this area. The other
Safety sites, probably score one, two or three stars respectively.
With respect to continuous improvement, Site 1 decides to
Similarly, safety is significantly impacted by the leadership
improve metallurgical accounting by improving the accuracy
and management within the processing department. An
of the weightometer. Over the next few years the site moves
example of best practice might be: ‘the number of Job Safety
from the calibration management described for that site to
Analyses in the processing group is greater than two per day something like that described for Site 3. During these years
per employee.’ the five star review result for this best practice has increased
from one to three stars. This improvement in the application
Environmental incidents of a fundamental driver for metal accounting is effectively
Again, the attitude the workforce in the processing continuous improvement.
department has towards environmental management of the Over this period of time, Site 1 has reached a high level of
site is impacted by leadership and management within the weightometer calibration management. The site might decide
department. A best practice might be: ‘environmental incident the efforts required to further improve weightometer accuracy
reports by the processing department are less than 25 per cent will be better spent in another area and simply choose to
of the site wide incidents.’ maintain the current level. This is a very mature approach to
continuous improvement. Site 1 management have taken a
Five star reviews pragmatic view – bring all of the operation to a high standard.
An audit is essentially a ‘tick and flick’ process showing:
•• a procedure/document covers the subject in question PEOPLE DEVELOPMENT
•• this procedure/document is available for employees to There is no doubt people are the most important asset in
access any ore processing facility. The ever increasing numbers of
•• the procedure or document has been applied. mines and the changing expectations of the population in
general have created significant pressures on the skills of
Five star reviews are more intense and an excellent tool for the workforce. The industry has responded well at the grass
pulling continuous improvement and best practice together. roots level. Recruiting, training and multiskilling etc have
A five star review measures how well the standards of developed competent and flexible operators and maintainers.
practice are applied. Consider a best practice: ‘the dry weight The demand for technical people is perhaps greater than the
of mill feed in any period will be within 2 per cent of the true demand for skilled operators and maintainers. As a result, an
value’ and the following four sites, explained in detail, to be employer’s expectation of experience and knowledge base for
measured against this best practice. specific technical roles has fallen with time.

Metallurgical management tools for continuous improvement
Almost in compensation for the decreasing skills set, the FURTHER READING
availability of complex optimisation software packages and Adel, G, Smith, B, Kojovic, T, Thornton, D and Richardson, J M, 2005.
specialised instrumentation has increased significantly in the Mine-to-mill optimisation of aggregate production, in 2005 SME
past few years. The implementation of standards of practice Annual Meeting (Society of Mining Engineers).
and five star reviews will encourage technical personnel Langford, A and Hunter, D, 2003. PIMS – The road to efficient, cost-
to make use of the tools available and both develop their effective plant operation, in Proceedings Eighth Mill Operators
personnel skills set and improve the performance of the ore Conference, pp 197–202 (The Australasian Institute of Mining and
processing facility. Metallurgy: Melbourne).
Mackenzie, J and Lumsdaine, I, 2002. Characterising cobalt losses for
CONCLUSION improved recoveries, in Metallurgical Plant Design and Operating
Strategies, pp 373–382 (The Australasian Institute of Mining and
The change from analysing outcomes to measuring the Metallurgy: Melbourne).
application of the fundamental drivers that delivers the
outcomes is a subtle but significant shift in evaluating the Molone, M J, Peberdy, N J and Morrison, R D, 1991. The application
of simulation technology to enhance operator performance, in
performance of an ore processing facility. Understanding
Proceedings Fourth Mill Operators Conference, pp 147–151 (The
and applying improvements to these fundamental drivers Australasian Institute of Mining and Metallurgy: Melbourne).
will result in better performance and improve the skill sets
of the people managing these facilities. Adopting a five star Orlich, J, Dunne, R and Ching, A C, 2010. Use of AMIRA leach
and adsorption modelling and simulation at Newmont Carlin
review process, to measure the application of improvements
Mill 5, in Proceedings Precious Metals ‘10 (Minerals Engineering
to the fundamental drivers at regular intervals, will create International: Falmouth).
continuous improvement.
Quast, K B, 1994. Simple laboratory experiments to teach gold
metallurgy to students and operators, in Proceedings Fifth Mill
ACKNOWLEDGEMENTS Operators’ Conference, pp 291–292 (The Australasian Institute of
The authors appreciate and acknowledge the support provided Mining and Metallurgy: Melbourne).
by Newmont Mining Corporation in preparing this paper.

Contents
Measuring the influence of sample size on the

precision and accuracy of gravity gold estimation
A Giblett1 and T J Napier-Munn2
ABSTRACT
This paper presents the results of a test program evaluating the impact of sample
mass on the precision of Stage 1 gravity recoverable gold (GRG) test results. Four ore
types ranging from low to high GRG contents were the focus of testing. A method
to predict the influence of sample mass on the confidence limits of the Stage 1 GRG
test result is demonstrated. The results suggest that although small samples tend to
overestimate Stage 1 GRG in samples with low gravity gold values due to higher
mass pull to concentrate, such samples can give acceptable estimates of the Stage 1
GRG content with focused sample treatment protocols.
INTRODUCTION
Gravity gold presents a considerable sampling challenge to the process engineer striving
to develop a solid basis for plant design or optimisation. As would be expected when
an ore sample contains a distribution of low quantity, high-grade liberated particles,
the amount of sample analysed has a significant impact on the quality of grade related
measurements. Here ‘quality’ can be defined in terms of both accuracy (the extent to
which bias is present) and precision (the uncertainty in the estimation of grade).
The quantity of sample used for determining gravity gold contents has been
observed to vary significantly in metallurgical test programs performed around
the world. The accuracy of subsequent estimates of GRG may be expected to vary
significantly as a consequence. Further, the impacts of using inappropriate sample
quantities for metallurgical testing can significantly influence overall gold recovery
estimates when GRG is present. This is due in part to a lack of clear demonstration
of the impact of sample mass on measurement accuracy and precision, and the
inappropriate metallurgical testing procedures that can be a consequence of such a
limited understanding of fundamental sampling requirements.
To quantify and demonstrate the influence of test sample mass on the accuracy and
precision of gravity gold determinations, an internal research program was recently
performed by Newmont Metallurgical Services in Denver, Colorado. This program
studied the influence of sample mass on the accuracy and precision of Stage 1
Laplante GRG test results (Laplante, Woodcock and Huang, 2000) generated from
four ore samples ranging from fine to coarse GRG size distributions.
A procedure to predict the confidence limits of the GRG test result as a function of
sample mass and GRG size distribution was also evaluated during this program. This
paper presents the results of the test work and the determination of confidence limits
based on test sample mass. It is intended to contribute to a discussion on appropriate
sample size for the determination of GRG content.
PROCEDURE
Large mill feed samples were collected from four active gold mining operations,
selected to reflect a range of free gold concentrations from low to high. Each operation
selected has employed gravity concentration for gold recovery, and in all cases historical
GRG test data, as well as gravity plant performance data, was available for reference.
Fundamentally it is the demonstrated correlation between Stage 1 GRG content and
1. FAusIMM, Senior Technical Advisor – Mineral plant scale gravity recovery identified by Giblett (2011) that supported the focus of the
Processing, Newmont Mining Services, test program on only the Stage 1 GRG content, rather than the full three-stage Laplante
Subiaco WA 6018. GRG test value more commonly determined. As the first stage of the Laplante GRG test
Email: aidan.giblett@newmont.com
collects the coarsest gold particles, it therefore also presents the largest challenge in a
2. FAusIMM(CP), Emeritus Professor, Julius sampling context, and is the most appropriate focus for this investigation.
Kruttschnitt Mineral Research Centre,
Sustainable Minerals Institute, The University The test program sought to assess whether meaningful GRG values could be
of Queensland, Indooroopilly Qld 4068. obtained using small test masses. Laplante suggested that a typical minimum
Email: t.napier-munn@uq.edu.au sample mass of 40–70 kg was required for most ores, but may be as little as 25 kg for
361
high-grade ores, and as much as 140 kg for low-grade ores Table 2

(Laplante, Woodcock and Huang, 2000). The present program Stage 1 gravity recoverable gold (GRG) values for Comp 2,
evaluated sample masses as low as 10 kg. It also aimed to as a function of sample mass.
demonstrate the level of variability that could be expected at
smaller sample masses. The modification of the Gy sampling Sample mass 80 kg 25 kg 25 kg 15 kg
formula described by Barbery (1972) was applied to estimate
Stage 1 GRG content 25.40% 26.20% 30.20% 28.90%
the impact of sample mass on the fundamental sampling error,
and therefore the confidence limits on the GRG test results. Mass pull to concentrate 0.14% 0.31% 0.32% 0.52%
RESULTS were selected to facilitate this evaluation, Comp 3 (high Au)

Two bulk samples, designated Comp 1 (low Cu-Au) and and Comp 4 (low Au).
Comp 2 (moderate Au-Ag) were selected for the initial phase The Comp 3 bulk sample had previously been
of testing. An initial Laplante GRG test was conducted at independently tested as a part of a gravity circuit optimisation
a sample mass of 80 kg, consistent with the guidelines of study, and the testing had shown some inconsistency. A
Laplante, Woodcock and Huang (2000). This intended to single stage GRG test on a 25 kg sample yielded a calculated
establish a baseline GRG content that could be reasonably total head grade of 3.29 g/t, while a larger three-stage GRG
expected to represent the ‘true’ value as determined by test demonstrated a considerably higher head grade of
standard industry practice. 4.88 g/t. The uncharacteristically low Stage 1 GRG content
Comp 1 was then rotary split into two sets of duplicate of 48.3 per cent determined in the three-stage test was as a
charges for Stage 1 GRG testing at 15 kg sample mass and result considered questionable.
25 kg sample mass. The results are summarised in Table 1. It is immediately apparent from Table 3 that there is scatter
The results from Comp 1 were promising. The smaller in the results, reflected in both the calculated head grade
scale tests were prone to slightly exaggerate the Stage 1 GRG and the Stage 1 GRG content for the five 25 kg tests. A poor
content, which may be attributed to the higher mass pull to comparison is generally observed for both parameters against
concentrate compared to the 80 kg test. The increased mass the 100 kg test sample results, which were generated from
pull at smaller masses is due to the relatively constant volume an independent testing program at an external laboratory.
As previously noted a single stage GRG test performed at
of the GRG concentrate recovered during the test. Nonetheless
the same independent laboratory returned a calculated head
the accuracy of the subsequent Stage 1 GRG estimates, at
grade of 3.29 g/t, which is consistent with most of the Comp 3
worst ±2.5 per cent absolute at sample sizes down to 15 kg,
results. This demonstrated that the variability in results was
was considered encouraging.
not limited to the Newmont testing program.
Comp 2 was tested by a similar process although due to
As the GRG stage concentrate should recover the larger
sample quantity limitations only one 15 kg test was conducted,
free gold particles it is a practical assumption that analysis of
along with duplicate 25 kg tests. Again a tendency for the
the GRG concentrate grades should reflect a similar degree
smaller sample sizes to slightly overstate the GRG content
of variability, and be the main cause of the variability in
was observed; however the overall precision at +/-5 per cent
calculated head grades and GRG contents. It is evident from
absolute on a moderate to high GRG content ore was again
the data in Table 3 that for the five 25 kg tests the concentrate
within the target range. The results from Comp 2 test work
grades are reasonably consistent (630 g/t ± 11 per cent). The
are shown in Table 2.
relative variability in the tailings grades is much higher, and
Further analysis of the Comp 2 results was enlightening. the 1 g/t spike in the fourth 25 kg test result shown in Table 3
Despite the ‘noisier’ results at the 25 kg sample size is significantly affecting the calculated GRG content. This
(26.2 per cent GRG versus 30.2 per cent GRG) the concentrate provides some indication that the GRG test tailings sampling
grades were remarkably similar, and the primary difference and assaying protocols warrant closer attention. A similar
in the two results was the gold content measured in the GRG observation was made by Woodcock (1994) during his initial
tailings stream. This was of interest given that the concentrate work to establish GRG test protocols, and resulted in the use
would be expected to present the greater sampling challenge of larger tailings samples for analysis.
by far.
Further insight is provided by comparison of the GRG yield
This initial phase of testing on the Comp 1 and Comp 2 per unit sample mass for the large sample against the small
samples was considered successful in demonstrating that samples. This simple comparison is achieved by dividing
sample sizes down to 25 kg might be suitable for GRG testing the gold content of the test concentrate, in micrograms, by
for plant design purposes. It was however noted that all the test mass in kilograms. Across the six single stage tests in
smaller scale test results reported higher GRG values than Table 3 the average GRG yield rate was 2.17 with a standard
the original result based on the larger sample mass, though deviation of ±0.23 mg/kg. This provides confidence that the
the apparent bias was generally not excessive. A secondary test is consistent in terms of GRG recovery efficiency, and
testing program was devised to challenge this conclusion, and
to further test if the 15 kg sample mass was indeed suitable for Table 3
ores with low coarse gold content. Two further bulk samples Stage 1 gravity recoverable gold (GRG) content for Comp 3,
as a function of sample mass.
Table 1
Stage 1 gravity recoverable gold (GRG) values for Comp 1, Sample mass 100 kg 25 kg 25 kg 25 kg 25 kg 25 kg
Head grade (Au, g/t) 4.88 2.91 2.93 2.45 4.11 3.11
Sample mass 80 kg 25 kg 25 kg 15 kg 15 kg Stage 1 GRG content 48.3% 71.7% 76.4% 70.9% 57.3% 71.5%
Stage 1 GRG content 13.40% 14.50% 14.40% 15.90% 14.40% GRG con grade (g/t) 2578 617 662 509 687 673
Mass pull to concentrate 0.12% 0.39% 0.34% 0.59% 0.63% Tail grade (Au g/t) 2.52 0.83 0.69 0.72 1.76 0.89

Measuring the influence of sample size on the precision and accuracy of gravity gold estimation
that the test samples also provide a consistent GRG content

ftd 3m
at the smaller mass. Subsequently it was concluded from this i=
program that the GRG content can be predicted reasonably MP
consistently from 25 kg samples on an ore that contains a high where:
coarse gold content. Increased focus on the collection and
processing of test tailings samples is likely the best option to M is sample mass
improve test reproducibility. f is shape factor (0 <f <1)
The fourth composite chosen for testing was one that ρ is material density (g/cm3)
contained moderate (30–40 per cent) levels of fine GRG, as dm is mean size in range of interest (cm)
measured by the single-stage GRG or full three-stage Laplante
P is proportion of material in the range of interest
GRG test, and typically low levels (<10 per cent) of Stage 1
GRG. It was deemed an ideal sample to demonstrate if smaller θ is the relative standard deviation of P
test masses than 25 kg could reliably be used, on samples with Confidence limits can then be calculated as ±zq where z is
low coarse gold content, further exploring the encouraging the two-sided standard normal deviate (= 1.96 for 95 per cent
results achieved in the two 15 kg tests performed on Comp 1. confidence).
A baseline 50 kg test was performed on Comp 4, followed by In this calculation, the sample mass distribution (P) is
five 10 kg Stage 1 GRG tests. The results are shown in Table 4. replaced by the GRG content in the test feed, by size, back
The assay head and GRG results from Comp 4 are reasonably calculated from the GRG concentrate mass and grade
consistent at the smaller sample mass, except for one notably distribution by size.
higher GRG value, but show a clear bias towards higher GRG While the smaller scale tests exhibited a positive bias in all
values than the baseline 50 kg sample. In this instance, the instances compared with the large sample values, they typically
grades of the GRG concentrate are very low, and at almost fell within the prescribed confidence limits determined by the
1 per cent mass pull to concentrate it is apparent that the GRG application of Barbery’s equation, suggesting that this is a
estimates at such a small sample size are overstated. With low plausible method of estimating GRG confidence intervals as
free gold contents in the test feed, the gravity concentrate will a function of sample size (see Figure 1).
consist of a larger proportion of gold bearing coarse middlings The non-linear dotted lines in Figure 1 reflect the 95 per cent
particles than usual. These particles have little probability of confidence intervals which increase at smaller sample sizes,
being recovered in this form at the very low concentrate mass as expected. The linear dotted lines dictate the ±5 per cent
pulls typically employed in plant scale gravity concentration absolute tolerable absolute error band on the Stage 1 GRG
circuits. The quality of the results for tests of this size may test result, nominated as the required level of precision for
possibly be marginally improved by hand panning the the test. In all instances the bias at the smaller sample sizes
concentrate to reduce the concentrate mass pull. However, is observed. For the Comp 1, Comp 2 and Comp 4 tests the
the value of testing such small samples to determine GRG Barbery approach appears to appropriately estimate the
content is clearly questionable, particularly when the gravity confidence intervals.
gold contents and sample head grades are low. In retrospect,
For Comp 3 the analysis is compromised by the uncertainty
testing the smaller sample masses on material with higher
around the ‘true’ GRG value of this sample. The majority
GRG content would have been more informative, as despite
of the tests reflect a higher GRG content, and lower head
the higher scatter in the results that would be expected, such
grade, than that determined by the original large volume
samples would be less prone to such a dominant bias because
test. A Stage 1 GRG content of ~70 per cent is certainly more
of higher mass pulls.
consistent with historical values and actual plant gravity
gold recoveries. Where the Comp 3 tests did not appear to
PREDICTING THE IMPACT OF SAMPLE SIZE ON STAGE 1 be significantly influenced by variable tailings assays, the
gravity recoverable gold VALUE ACCURACY GRG results were consistent (71.7 per cent, 76.4 per cent,
An approach to estimating the confidence interval on the 70.9 per cent and 71.5 per cent). As a result it is feasible that
GRG result as a function of sample mass was adapted from if the tailings sampling and assay protocols can be improved,
that described by Barbery (1972) based on Gy’s sampling then smaller sample masses, in the 25–50 kg range, may be
theory. Barbery’s method is applied to estimate the influence suitable for such materials.
of sample mass on the precision of a measurement particle
size distribution, as a function of particle size. As the CONCLUSIONS
GRG test methodology is also an analysis of particle size The results of the research program performed by Newmont
distribution, specifically the distribution of low frequency Mining Services have demonstrated that:
high-grade liberated gold particles, this method should
•• The equation reported by Barbery (1972) can be used
give an approximation of the uncertainties involved in the
to provide practical indications of the influence of the
estimate of GRG:
sample mass on GRG test result precision, should a
reasonable estimate of the GRG size distribution be
Table 4 available.
Stage 1 gravity recoverable gold (GRG) content for Comp 4,
•• For samples with low to moderate Stage 1 GRG contents
(10–30 per cent), the use of smaller GRG test sample
masses (15–25 kg) is appropriate and can be expected to
Sample mass 50 kg 10 kg 10 kg 10 kg 10 kg 10 kg
generate Stage 1 GRG values within ±5 per cent of the
Head grade (Au, g/t) 2.41 2 2.11 2.15 2.14 2.33 true value.
Stage 1 GRG content 14.4% 19.4% 26.8% 20.3% 21.7% 2270.0% •• Smaller samples are prone to a high bias on the Stage 1
GRG con grade (g/t) 205 28 67 52 55 64 GRG test result as a function of increased mass pull
to concentrate. Secondary treatment of the GRG
Tail grade (Au g/t) - 1.63 1.56 1.73 1.69 1.81 concentrate to reduce the mass pull can be considered in

FIG 1 – 95% confidence intervals on Stage 1 gravity recoverable gold (GRG) content as a function of test sample mass.
such instances, particularly for samples with low GRG ACKNOWLEDGEMENTS

concentrate grades.
The authors acknowledge the permission of Newmont Mining
•• Significant error can be introduced into the GRG result to publish this paper.
and the calculated head grade based on sampling of
the GRG test tailings. The collection of larger tailings
samples and the use of screen fire assaying at 75 µm is
REFERENCES
being evaluated to mitigate this issue. Barbery, G, 1972. Derivation of a formula to estimate the mass of a
sample for size analysis, Transactions of the Institutions of Mining
•• This test program has been less conclusive than expected and Metallurgy, Mineral Processing and Extractive Metallurgy,
in determining the appropriate sample mass for high 81(784):C49–C51.
GRG content samples, based on the observed issues Giblett, A L, 2011. Gravity Gold Concentration at Newmont Mining,
with tailings grade reproducibility. Nonetheless there World Gold Conference 2011, in Proceedings 50th Conference of
is some evidence that, given careful consideration and Metallurgists, pp 217–228 (Met Soc: Montreal).
implementation of suitable test protocols, reasonable Laplante, A R, Woodcock, F and Huang, L, 2000. Laboratory
results can be generated on smaller sample masses, procedure to characterize gravity-recoverable gold, Society for
down to a lower limit of 25 kg. This has a significant Mining, Metallurgy and Exploration Inc, Transactions, 308:53–59.
implication for the costs of collecting and treating
Woodcock, F C, 1994. Use of a Knelson unit to quantify gravity
samples for GRG content. recoverable gold in an ore, Masters thesis, Department of Mining
and Metallurgical Engineering, McGill University, Montreal.

Contents
Defining practical metallurgical accounting

discrepancy limits for gold operations
A Giblett1, R Dunne2 and K McCaffery3
ABSTRACT
Maintaining a metallurgical accounting system in a functional state, in addition to
performing routine metal balances, demands a significant commitment of time and
effort from site metallurgists. Additionally, the discrepancy between indicated and
reconciled gold production for a given month will often result in time-consuming
debate and investigation that is based on prevailing expectations around acceptable
discrepancy limits within the site or company.
In order to avoid unwarranted and potentially unproductive investigations into
perceived accounting discrepancies that may well be nothing more than natural
variations within a given system, it is necessary to ensure some rigour is applied to
establish the discrepancy limits that define system performance. In so doing one can
ensure that an operations technical resources can focus on optimisation of the business.
Through the diversity of its metal recovery operations, Newmont has collected a
detailed database of metallurgical accounting discrepancy data over several years
of operation. The operations range from low to high levels of compliance with best
practice metallurgical accounting standards. The Newmont data shows that the
level of compliance with best practice is not always consistent with the greatest
metallurgical accountability.
The plant operator, and site management, must be cognisant of the value generated
through improvements to the metal accounting process given the often considerable
expense and ongoing effort that can be required to support these changes. Change
should only be made where value, either direct or indirect, can be added.
INTRODUCTION
The release of the Metallurgical Accounting Code of Practice (AMIRA International,
2007) through AMIRA and the Metallurgical Accounting Monograph (Morrison,
2008) by the Julius Kruttschnitt Mineral Research Centre (JKMRC), have provided
useful guidance in the design and execution of metallurgical accounting standards
for mineral processing, hydrometallurgical and pyrometallurgical metal recovery
facilities. Such guidelines can facilitate a thorough gap analysis of existing systems
against a metallurgical accounting ‘utopia’ where all specifications of the Metallurgical
Accounting Code of Practice are fully satisfied, and all measurement biases removed.
In some instances the gap analysis will identify remedial actions associated with
significant capital expenditure. An example is the need to retro-fit a large footprint,
multistage auto sampling station where no allowance was made for such an installation
in the as-built plant layout. There will also be instances where increased resources
are required to execute improved systems such as expanded QA/QC programs in
both the laboratory and the field. This can directly lead to increased operating costs
through additional labour resource requirements, increased sample processing costs
or in some cases the diversion of already limited resources away from more focused
1. FAusIMM, Senior Technical Advisor – Mineral process optimisation activities. Subsequently the business value of any recommended
Processing, Newmont Mining Corporation, improvements to the metallurgical processes at a given site must be well defined. This
Englewood CO 80112, USA. is particularly challenging where the recommended improvement may not deliver a
Email: aidan.giblett@newmont.com direct increase in metal production or cost efficiency, but may in some instances allow
2. MAusIMM, Metallurgical Manager Technical an improved ability to quantify the impact of improvement or change.
Services, Newmont Mining Corporation,
With this in mind, Newmont Mining Corporation has compiled a comprehensive
Englewood CO 80112, USA.
Email: robert.dunne@newmont.com
internal standard and guideline, based on the AMIRA Metallurgical Accounting
Code of Practice, to guide and govern the management of metallurgical accounting
3. FAusIMM(CP), Principal Adviser
Process Operations (Business Excellence),
processes at existing operations and for design and implementation at new operations.
Newmont Mining Corporation, This recognises that while the fundamentals of sound metallurgical accounting
Greenwood Village CO 80111, USA. are common, variability of ores, plant and processes at our operations means that
Email: karen.mccaffery@newmont.com acceptable discrepancy must be identified for each site. Metallurgical accounting
365
systems must be fit for purpose. A gap analysis of sites and analysis of leach feed, cyclone overflow or flotation feed,
against this standard allows identification of compliance with as the case may be, on a shift-by-shift basis. The cumulative
both minimum mandatory expectations and best practice. indicated grade is built up during the course of a one month
While sites are encouraged to work towards best practice, period and compared to final reconciled gold production based
implementing a capital or operating solution to close a gap is on product shipments (bullion, concentrates), changes in circuit
only supported only where a valid business case is identified inventory and measured tailings gold contents.
or, where a mandatory regulatory requirement dictates. The accounting discrepancy is then calculated as the
This paper presents actual metallurgical accounting difference between the indicated and reconciled gold
performance data from a variety of gold operations and contained in mill feed, expressed as a percentage of the
relates that performance to current metal accounting contained reconciled gold. In some cases, the same calculation
practices at those sites. This comparison allows a sobering is performed only by comparing indicated and reconciled gold
assessment of the likely value of a non-discretionary overhaul production in total ounces, and the methods are considered to
of metallurgical accounting systems to meet a single, all- be equivalent for the purpose of this evaluation.
encompassing standard. In the context of Figure 1, a positive result represents an
It has long been recognised that the nugget sampling effects understated indicated feed grade, such that final reconciled
of gold ores set them aside from base metal ores in the precision gold production exceeds the indicated value. While this
of sampling and grade determination that can be practically typically proves a stress-free result for the owner of the
achieved. The data in this paper attempts to demonstrate that accounting system at month-end, a positive discrepancy is
the variability of the nugget effect between very different an equivalent reflection on the reliability of the system as a
orebodies requires the practitioner to consider that a standard negative discrepancy. If a positive bias develops it is a serious
set of plus/minus tolerance limits for accounting discrepancy cause for concern as the cause of the bias is typically unknown.
in gold ore processing to be applied across the board in all Risky assumptions can be made such as applying an upgrade
instances is not a pragmatic approach. Certain ores are factor to indicated production when forecasting month-end or
naturally more prone to larger sampling errors and in such year-end production. Such practices are a recipe for disaster
cases more variability in the accounting precision should be when the cause of the bias is not identified and, thereby, no
expected. In the case of complex metal recovery flow sheets guarantees exist that the bias will be consistent or even occur
with coarse gold effects, this issue is even more relevant. at all in a given accounting period.
Conversely, a negative discrepancy, where reconciled
GOLD PLANT METALLURGICAL ACCOUNTING DATA production is less than indicated production, often leads to
The 12 month cumulative gold metal accounting discrepancy a month-end witch-hunt and time-consuming fault finding
for ten Newmont gold and copper-gold operations is shown exercises, depending on the magnitude or consistency of the
in Figure 1, noting that all discrepancy data presented in this negative discrepancy.
paper relates to gold and no data is presented for copper. Clearly there is a need to have well defined and well
This data has been compiled over four consecutive years considered expectations with respect to acceptable ranges
with the four bubbles for each site representing annual metal of metal accounting discrepancies. It is in this challenging
accounting performance. area where references and standards become much less
The discrepancy calculation represents the accuracy of the prescriptive. Subsequently, anecdotal expectations tend
original indicated gold head grade values, determined by the to rule the day, with associated variability in expectations
measurement of mass, gravity gold production and the sampling depending on the experiences of those involved.
25%
20% Site 5
15%
Site 2
10%
Site 7
5% Site 8 Site 9
Site 3 Site 6
Discrepancy
0%
0 Site 4
Site 1
-5% Site 10
-10%
-15%
-20%
-25%
FIG 1 – Cumulative annual accounting discrepancy by site.

Defining practical metallurgical accounting discrepancy limits for gold operations
SITE DATA – LONG RANGE ACCOUNTING DISCREPANCIES is subject to frequent heavy rainfall events and high clay oxide
content in ore resulting in highly variable ore feed moisture.
Figure 1 shows a shaded region of ±5 per cent as a nominal
Using flowmeters with density measurement has reduced
target range of discrepancies for the 12 month cumulative
month-end metallurgical accounting discrepancy by three to
metal accounting discrepancy for these operations. This is
five per cent compared to use of weightometers combined with
a commonly applied target range for monthly gold metal
moisture measurement. The Newmont standard recommends
balances. Generally the sites fall within this band, with the
weightometers as a more precise choice for wet weight
notable exceptions of sites 2 and 5, where a consistent positive
determination, but recognises that under certain circumstances
bias is evident and the size of the bubbles demonstrate a high
alternative methods such as flowmeter and densitometer
degree of variability in the short-term accounting discrepancy
measurement may yield an overall lower error for dry mass
levels. In some cases the cumulative discrepancy is consistently
determination. The dry mass determination method must
quite low (sites 1, 3, 6, 7 and 8), with cumulative discrepancies
therefore be fit for purpose and the particular site’s application.
often closer to ±2 per cent to ±3 per cent. It is instructive then
to consider what we know, at a high level, about these sites In one case, the site uses pipe samplers for feed grade
with respect to expected major influences on sampling and determination and is non-compliant with the prescribed
measurement precision, as presented in Table 1. sampling practices for primary metal accounting. In another,
a cross stream primary sampler is used. However, it is not best
The data presented in Table 1 allows a number of useful
practice design due to the vertical alignment of the cutter. In
high level observations to be made with respect to metal
these cases the margin for improvement in metal accounting
accounting performance and systems. Green shaded cells
systems by enforcing compliance with best practice is more
reflect substantial compliance with best practice metallurgical
than likely minimal. Consequently, there is less likelihood
accounting standards, red shading denotes non-compliance
of creating demonstrable business value by changing the
and the orange shading partial compliance. Excluding sites 2
existing systems. In these particular instances the fine grain
and 5 which clearly have some issues and will be discussed
size of gold particles in the mill feed is a dominant influence on
later; the remaining sites consistently achieve a 12 month
overall metal accountability, substantially reducing sampling
cumulative discrepancy below five per cent. Of this data,
and analytical errors at these sites. The careful and sustained
with the noted exclusions, 85 per cent of points fall within a
custodianship of the accounting systems at these sites is also
discrepancy range of ±4 per cent, 71 per cent within ±3 per cent
an important influence.
and 64 per cent within ±2 per cent.
When considering ores with fine grained gold occurrences
we cannot ignore the performance of site 2 that has all the
INFLUENCES OF GOLD GRAIN SIZE advantages of reduced sampling errors associated with the
Sites 3 and 4 represent medium to fine grained ores and absence of any real nugget effect, yet counter intuitively
are largely compliant with metal accounting and systems high levels of discrepancy are observed. The extent of
requirements. These sites also demonstrate relatively metal accounting discrepancy at this site is affected by
consistent and moderate to low annualised discrepancies. poor sampling practice, where compliant auto sampling
Instances of higher than ±5 per cent discrepancy do however stations were perceived to be unsustainable due to high
occur for shorter accounting periods and are most likely maintenance requirements, removed from operation and
related to maintenance and sustainment of existing systems replaced with ineffective manual sampling of slurries from
and practice. tanks. This inaccuracy is compounded by a high portion of
Sites 6, 7 and 8, all represent very fine grained gold ores gold contained in leach feed solution, resulting from high
and show quite low and relatively consistent annualised levels of cyanide in process water used for milling. Errors
discrepancies. This is despite in all cases being non-compliant associated with sampling and determinations of pulp
with recommended best practice mass measurement standards, density for head grade calculation are likely to be significant
using volumetric flow and densitometer measurements for at this operation. This site is a clear case of where improved
mill feed mass determination rather than high precision sampling practices would be expected to have a positive
weightometers and measured ore moisture levels. Site 6 ore impact on metal accountability.
TABLE 1
Overview metal accounting performance and systems compliance.
Site Discrepancy Gold Grain Mass Gravity Feed Tails

Year 1 Year 2 Year 3 Year 4 Sizea Measure Recovery Sample Sample
Site 1 -1% -1% 0% 3% C Manual Manual
Site 2 5% 11% 7% 8% F N/A Manual Manual
Site 3 1% 1% -4% 1% M
Site 4 -1% -5% -2% -3% M/F
Site 5 12% 15% 13% 10% C
Site 6 0% -3% -1% 0% F N/A
Site 7 3% 0% 0% 1% F N/A
Site 8 1% 0% 1% 0% F N/A
Site 9 1% -2% -3% -4% M/F
Site 10 -6% -2% 0% 0% M/F
a. Gold grain size references are coarse (C), medium (M) and fine (F).

Consider the example of site 1 where gold grain size is very inadequate where coarse gold particles can make it through
coarse and manual sampling could be expected to have a high the plant to tailings, through a combination of inefficient
associated degree of error. This site consistently demonstrates classification and reduced solubility due to passivation effects
low levels of discrepancy which is not what might be (iron smearing or surface coatings). This is demonstrated in
expected from infrequent manual sampling. The gravity detail later in this paper.
circuit recovers a large proportion of gold from plant feed,
however; high levels of coarse gold are also present in the Coarse gold influence on grade determination – a case study
recirculating load. Consequently there is a high probability In Figure 2 the gold size distribution as a function of particle
that coarser gold particles will periodically purge into the size in the cyclone overflow is represented for two ores. Ore
cyclone overflow. The current sampling frequency is possibly A represents a coarse grained ore, where coarse gold recovery
too widely spaced to represent this type of event. Despite the from the mill recirculating load is applied. Ore B represents a
apparent high metallurgical accountability demonstrated by fine grained ore, not requiring gravity concentration ahead of
this site, it is considered that more rigorous sampling through downstream processing of the cyclone overflow.
high frequency auto sampling may indicate a higher degree of
The gold size distributions are comparable, suggesting that
leach feed variability and potentially an increased quantity of
the gravity circuit in Ore A is doing a good job of removing
recoverable gold losses to tailings than is currently recognised.
coarser gold and that the classification system is performing
In this example there is potentially a strong business case well at holding the coarse gold in the grinding circuit. This
for improving sampling practices at this site by enforcing demonstrates the importance of a well-operated gravity circuit
compliance with the company’s metal accounting standard and grinding circuit classification system to the effectiveness
requirement to implement statistically correct sampling of the metal accounting process. Calculating the fundamental
and optimising sampling frequency via nomographic and sampling error by Equation 1 (after Pitard, 1993) for these two
variographic experimentation (as described by Pitard, 1993). size distributions gives values of 3.6 per cent and 2.2 per cent
This previous example of site 1 relates in many ways to respectively, which is much closer than expected given the
the example of site 5 which demonstrates a large degree of large difference in gravity recoverable gold F80 in the mill feed
accounting variability and a clear systemic accounting bias. (~500 µm versus <200 µm).
Site 5 is similar to site 1 in that the gold grain size distribution
3
is coarse, although site 5 suffers from more variability in this s 2FSE = 0.8.d Au (1)
regard between coarse and very coarse gold grain size. Site 5 (Ms.aL)
also demonstrates a high degree of compliance with sampling
and weighing equipment requirements of the company’s Both of these distributions were determined during steady
accounting standard, yet the discrepancy levels are the state conditions as is typically the case of mill surveys. However,
highest observed for this operational data set. One might these operations have substantially different levels of coarse
expect that the cause of such an accounting bias could be gold accumulation in the mill recirculating load. Consequently
readily identified and corrected through thorough analysis. more variability in the gold size distribution in the cyclone
However, it is likely that the degree of variability in a process overflow is expected for Ore A, as a function of higher coarse
of this nature with such inconsistent coarse gold content gold recirculation in the milling circuit. It is expected that the
will always be large, unless significant improvements can be difference in the fundamental sampling error for these two ores
made in sampling and sample analysis processes. This would would be larger in practice than that estimated from the steady
specifically require collecting and assaying much larger state data in Figure 2, and only a focused sampling effort can
samples than is typical practice. reveal the full extent of this difference.
This issue directly relates to the recurring issue of ‘spurious’ The influence of gold grain size and gold accumulation
gold in tailings assays and the correct representation of coarse on grade estimation for process samples is further
gold occurrences in plant tailings samples. Current standard demonstrated by closer analysis of the gold by size data for
practices of duplicate fire assays of 50 g charges are clearly Ore B, remembering that this is the ore with very low levels
FIG 2 – Gold size distributions in cyclone overflow for two mills.

of coarse gold in the mill feed and is generally considered it is observed that the influence of imprecise inventory
as having a fine gold size distribution. A 10 kg composite measurements on shorter cycles is too great for weekly
sample of cyclone overflow was collected over one hour reconciliations to be accepted as conclusive. The month-end
during a recent AMIRA P420 project survey, processed metal balancing or reconciliation process is thereby the most
through a Knelson concentrator with gold by size analysis common short-term assessment of the performance of the
performed on the Knelson concentrate and tailings. The metallurgical accounting system as a whole.
results are summarised in Table 2.
The data presented in Figure 1 and Figure 2 demonstrates that,
The high-grade result for the +300 µm concentrate fraction is for the most part, the annual reconciliation discrepancy can be
an immediate observation from this data. There are a number expected to fall within one standard deviation of 2.5 per cent,
of possible reasons why this fraction may have returned particularly if the accounting process is relatively free from
what is an anomalously high-grade; however, it is feasible
systemic errors and that random error discrepancy converges
that a single grain of coarse gold such as a 400 × 400 µm
over larger time periods. Based on the principle of random
flake could account for most of the gold in this size fraction.
error convergence we can expect longer term discrepancy to be
This is unexpected given the size distribution of the cyclone
a function of shorter term discrepancy based on the following
overflow, the observed cyclone classification efficiency at the
site and the fact that this ore contains very little coarse gold. equation (McCaffery and Birnbaum, 2011):
However, this example does indicate the potential for coarse
gold to report to the cyclone overflow, even under close to v MONTH = 12.v 2YEAR (2)
ideal conditions, with severe implications for the precision of
subsequent grade determinations. Solving this equation for σYEAR = 2.5 per cent results in
σMONTH = 8.7 per cent. A monthly discrepancy of ±8.7 per cent
In this instance, due to the large mass of sample used for the
is somewhat higher than common anecdotal targets, such as
analysis, the impact of this rogue particle on the head grade
±5 per cent as used to frame Figure 1 of this paper, or even
calculation is not substantial, although at approximately
values of ±2 per cent or ±3 per cent as have been suggested
2 per cent of the total feed grade it is not insignificant either.
The significance is clearer when it is considered that during elsewhere. It is, however, a useful point of reference for
the course of this hour, a much smaller sample mass would discussion. Actual monthly data for Newmont operations is
normally be collected for routine head grade determination summarised in Figure 3 for comparison.
and much smaller masses of subsample submitted for assay. This monthly data for seven operations over a four-year
The influence of a rogue particle on a conventional head period totals 335 data points. The mean of the data set tends
grade determination using duplicate 50 g fire assay charges to zero (0.2 per cent) and the spread of the data is defined by
is well beyond the ability of the accounting system to manage a standard deviation of ±5.5 per cent. Subsequently, based on
appropriately. Where the presence of coarse gold particles observed long-term performance of these gold recovery metal
is likely, only the collection and assaying of large sample accounting applications, two in every three-monthly results
masses can be expected to give a reliable measurement of can be expected to fall within the band of 1σ or ±5.5 per cent,
stream grades. with 11 out of 12 results expected to fall within a band of 2σ
or ±11 per cent. This tends to support the long held view in
SITE DATA – SHORT RANGE ACCOUNTING DISCREPANCIES operational circles that an appropriate target for monthly gold
Typically, metallurgical accounting balances are finalised metal balance accountability is ±5 per cent, with ±10 per cent
on a monthly basis, representing a practical and convenient representing an upper tolerance limit for short-term variations.
short-term interval to reliably measure and report actual gold Maintaining the balance within these statistical margins can
production. Preliminary balances are frequently constructed be expected to deliver a cumulative annual accountability of
on shorter intervals, commonly on a weekly basis; however, ±3 per cent or better based on the error convergence principle.
TABLE 2
Gold by size data for 10 kg cyclone overflow composite.
Size µm Knelson concentrate Knelson tailings Totals

Wt% Au g/t Wt% Au g/t Wt% Au g/t
850 0.03% 0.00 0.03% 0.00 0.03% 0.00
600 0.85% 1.64 0.04% 1.41 0.04% 1.44
425 1.77% 3.15 0.19% 1.41 0.20% 1.52
300 6.99% 134.80 1.30% 1.41 1.34% 6.41
212 10.65% 15.00 3.47% 0.76 3.52% 1.07
150 15.44% 11.10 8.01% 2.88 8.06% 2.99
106 14.02% 26.20 9.38% 0.66 9.41% 0.94
75 17.79% 52.90 17.19% 1.16 17.19% 1.54
53 12.17% 83.90 8.35% 1.12 8.38% 1.98
38 8.97% 290.00 9.04% 2.33 9.04% 4.38
25 2.85% 1124.00 3.35% 2.48 3.34% 9.34
-25 8.46% 1739.00 39.67% 1.17 39.44% 3.84
Total 100.00% 241.20 100.00% 1.40 100.00% 3.11

FIG 3 – Distribution of monthly accounting discrepancies.
15%
10%
5%
Discrepancy
0%
0 10 20 30 40 50
-5%
-10%
-15%
Month
FIG 4 – Monthly metal accounting discrepancy (site example).
SUSTAINING METALLURGICAL ACCOUNTING SYSTEMS results in ineffective communication of the basis for many
metal accounting systems, ensuring their inevitable failure.
A major ongoing concern with any site based metallurgical
accounting system is maintaining the measurement and An example of metal accounting system deterioration is
calibration processes to the desired standard over long shown in Figure 4, with a negative metal accounting bias
periods of time. Frequently corporate or external resources attributable to system deterioration resulting from changes
are called in to address deterioration in metallurgical in sampling and measurement practices over time. This was
accounting performance over time. While in some instances associated with turnover of key personnel in combination
these external audits demonstrate the absence of processes with ineffective communication of the basis for the original
critical to the metal accounting effort, they often reveal that practices. Note the developing negative bias from around
such processes were in place at one time but either have been month 25 which then deteriorated over time. In this instance,
neglected or rendered ineffective over time due to changes in plant management was initially concerned with the sustained
practices or personnel. This is clearly not ideal in cases where negative bias despite the discrepancy remaining within
the operation has demonstrated high metal accountability the ±5 per cent range, and subsequently initiated several
over a sustained period of time, only to descend into a form of investigations into the cause of the bias. This concern was
semi-organised chaos. Attempts are then made to determine proven justified by the deterioration in accountability observed
‘where the gold is’ following the deterioration of good in the last two months shown on the trend. A full system
management practices leading to a sustained negative metal review subsequently demonstrated a number of examples of
accounting bias. process deterioration contributing to the discrepancy.
Changes in key personnel due to turn over and role
transitions, in addition to the diversion of key resources SHORT-TERM VARIATIONS – BEST CASE EXAMPLES
towards ‘business optimisation’ activities, are common reasons By analysing monthly discrepancy data from Newmont
for deterioration of metal accounting systems and, indeed, operations, it is clear that some periods are better than
reduced focus on process optimisation activities that are core others, and no doubt the level of system maintenance has
to sound metallurgical plant practice. Insufficient resource some part to play in this. Influences such as loss of key
availability, or simply effort, in knowledge management personnel, transitioning responsibilities to new staff, loss

12.0%
10.0%
8.0%
6.0%
4.0%
Month Discrepancy
2.0%
0.0%
0 2 4 6 8 10 12 14
-2.0%
-4.0%
-6.0%
-8.0%
-10.0%
-12.0%
Site 6 Site 7 Site 8
FIG 5 – Monthly accounting discrepancy – best-case examples.
of knowledge, suboptimal staff competency levels and of gold in the tailings stream, is a prerequisite for effective
changes in business priorities can influence functionality of plant optimisation. Ensuring tailings sampling and analytical
the metallurgical accounting process. Ongoing education to practices are consistent with best practice in metallurgical
ensure adequate depth and competence levels of personnel accounting provides a sound platform for a detailed
in metallurgical accounting practice is therefore essential. It understanding of the nature of gold losses to further feed metal
is also instructive to look for periods of best performance recovery optimisation efforts. The representative sampling of
and strive to operate at that level for sustained periods by plant tailings is one component of metallurgical accounting
implementing appropriate practice. standards that is also a non-negotiable requirement of
Examples of best performance in the area of metal accounting effective plant optimisation.
are expected to generally be associated with ores of low The gold by size distribution in Table 2 demonstrates the
variability, specifically ores with low coarse gold content potential for coarse gold to be present in the cyclone overflow
and some degree of consistency in other characteristics feeding downstream metal recovery processes. A case study
(mineralisation, specific gravity, moisture content) that on Kalgoorlie Consolidated Gold Mine (KCGM) presented
influence measurements made for primary accounting by Hillier (2003) further demonstrates the presence of coarse
purposes. Sites 6, 7 and 8 used in this study demonstrate (>100 µm), liberated gold grains in the carbon-in-leach tailings,
periods of very low annualised discrepancy (<±0.2 per cent) having survived both flotation and subsequent cyanidation
compared to the observed database average of ±3 per cent. of the flotation tailings. The concentration of such particles
Again, this is seemingly achieved in violation of some of is generally so low as to be rarely detected by conventional
the major principles of the metallurgical accounting code, as sampling and assaying practices. Consequently, such gold
highlighted in Table 1, and illustrates the forgiving nature losses may go undetected for long periods of time, hiding an
of well disseminated and fine metal values from a grade opportunity to optimise plant performance and exposing the
determination perspective. metallurgical balance to the potential for bias when reconciled
From Figure 5 it is seen that site 7 shows the lowest degree of against resource models.
variability, with 100 per cent of values for this period within
Using the gold distribution in Table 2 as the basis for a
±5 per cent. In fact, 11 of 12 results are within ±4 per cent,
discussion of appropriate sampling and assaying procedures
and eight of 12 results are within ±3 per cent. The other two
we can make the following observations:
examples realise effectively the same annualised discrepancy
(±0.1–0.2 per cent), however, show slightly more scatter. Site 6 •• A large component of the gravity concentrate gold
has eight of 12 results within ±5 per cent while site 8 has ten content of the +300 µm fraction could conceptually be
of 12 results within ±5 per cent. Subsequently, based on best- associated with a large 400 × 400 × 50 µm gold flake.
case performance, a monthly discrepancy limit of ±5 per cent •• Based on the approach to concentrate gravity gold from
remains a defendable criterion. a 10 kg sample, assay the concentrate to extinction then
assay subsamples of the gravity tailings, the true grade
SAMPLING AND ANALYSIS OF PLANT TAILINGS of the sample can be accepted to be 3.11 g/t.
The reliable sampling of plant tailings streams is a critical •• Had a screen fire or leach assay been conducted on a
component of both the metallurgical accounting system 1 kg subsample of the 10 kg bulk sample, the leach feed
and plant performance optimisation efforts. The reliable grade would have been determined as 3.17 g/t in the
reconciliation of plant feed grade for mine production presence of the gold flake, and 3.06 g/t in the absence of
reconciliation relies heavily on accurate determination of the gold flake. This gives a theoretical grade precision of
gold losses to tailings through representative sampling and ±2 per cent.
accurate assaying, in combination with accurate measurement •• Had two 50 g assays been conducted on subsamples of
of the mill feed tonnage. the 10 kg bulk composite, the leach feed grade would
Accurate sampling for tailings stream grade determination, have been determined as 3.06 g/t in the absence of the
combined with a complete understanding of the deportment gold flake (though with larger confidence limits as a

function of subsampling to a smaller mass), and 4.14 g/t analysis. Incremental improvements to the metallurgical
in the presence of the gold flake, an error of 33 per cent. accounting precision are in such cases less likely to impact on
This demonstrates the unreliable nature of small assay the cost-effectiveness of the mining operation.
charges in the presence of coarse gold particles. When Opportunity will often exist to improve management of
coarse particles make it through to the plant tailings, as in quality control and management operating systems designed
the example of KCGM, the commonly applied practice of to maintain the metallurgical accounting system at the
assaying 50 g charges is grossly inadequate for both plant desired levels of precision and accuracy. The focus in such
optimisation purposes and metallurgical accounting practice. cases is generally to maintain the functionality of the metal
Where coarse gold loss to tailings is possible a more focused accounting system over the long-term rather than to improve
effort of assaying larger tailings samples or concentrating the short-term precision of the balance.
high specific gravity component of the tailings for assay will Specifically, tailings sampling is an area where there may
reliably quantify the issue. A good example of where this has be opportunity to piggy-back off metal accounting initiatives
been practiced to the benefit of both metal recovery and metal to improve accounting for gold losses, and thereby identify
accounting is provided by KCGM in the paper by Giblett et al opportunities to increase metal production. Representative
(2012). The quantified improvement in plant metal recovery sampling, assaying and thorough analysis of the plant tailings
alone in this particular case was of the order of 0.5 per cent allows the metallurgist to identify opportunities to improve
absolute. metal recovery through better understanding of the nature
and quantity of gold lost to tailings.
CONCLUSIONS
Monthly gold reconciliation data from a range of Newmont ACKNOWLEDGEMENTS
operations has been presented and demonstrates that, in The support of Newmont Mining Corporation to allow the
most instances, the monthly metal reconciliation falls within use of site reconciliation data and publication of this work
±5 per cent of the indicated value. On an annual basis, the is acknowledged. The efforts of site personnel to provide
rolling discrepancy can be expected to be within a ±3 per cent data are appreciated; specifically George Rautenbach, Luis
error band. In certain instances it is feasible that wider ranges Saldivar, Indah Wardhani, George Amonoo, Radu Gheorghiu
must be applied to account for greater process variability and Tyron Rees are thanked for their contributions. Peter
or variations in ore characteristics. Where operations Birnbaum is also thanked for his expert review of the paper.
demonstrate performance outside of the target discrepancy
range the presence of bias within the metallurgical accounting
processes is likely and corrective action is required. REFERENCES
AMIRA International, 2007. P754: Metal accounting code of practice
Integration of improved metal accounting practice by and guidelines, Release 3, February.
developing company standards that adhere to the spirit of
the Metallurgical Accounting Code of Practice is supported. Giblett, A, Hillier, D, Parker, K and Ramsell, V, 2012. The impact of
gravity gold recovery at Kalgoorlie Consolidated Gold Mines,
It is recommended, however, that such standards focus on
in Proceedings 11th AusIMM Mill Operators’ Conference, pp 199–
ensuring that systems are fit for purpose, define realistic site 206 (The Australasian Institute of Mining and Metallurgy:
specific targets for discrepancy, provide for sustainment by Melbourne).
consistently educating metallurgical accounting practitioners
Hillier, D J, 2003. Gravity gold recovery at KCGM’s SuperPit,
and system owners and that investment for system presented at Symposium: Optimisation of Gold Processing Plants,
improvement is only made where a sound or regulatory AJ Parker Cooperative Research Centre for Hydrometallurgy,
business case to do so exists. Launceston.
It is possible that if given the investment of resources, the McCaffery, K and Birnbaum, P, 2011. Newmont Operations: global
precision and accuracy of accounting systems can be improved standards and guidelines to optimise metallurgical accounting
by implementing best practice equipment, along with best procedures, internal Newmont standard and guideline document.
practice operating, quality management and maintenance Morrison, R D (ed), 2008. An Introduction to Metal Accounting and
practices. However, it is difficult to demonstrate a clear business Reconciliation (University of Queensland).
value for such an effort where existing metal accountability Pitard, F F, 1993. Pierre Gy’s Sampling Theory and Sampling Practice,
is sound and within the boundaries recommended in this (CRC Press Inc: Boca Raton).

Contents
Manual control, process automation or operational

performance excellence – what is the difference?
P Thwaites1
ABSTRACT
The mining industry uses many types of mineral and metallurgical plants to produce
saleable product from mined ore. Plant design history has left current operations with
a mixture of manual operations and various forms of automated process controls.
Consequently, we typically see high variability in the continuous operations together
with a shortfall in the attainment of full capacity, or higher utilisation of consumables.
At a level of best practice, ‘operational performance excellence’ focuses on process
control, using automation and control systems to deliver process optimisation. This
more sophisticated delivery is a great deal more difficult than the first stage of
equipment selection/installation. It includes the appropriate selection of the right
instrumentation, control system, key process knowledge, individuals with a solid
control engineering background/experience, and the essential backing/support of
the operations management team together leading to higher value delivery. Robust
solutions can be realised, considerably minimising process variation, thus leading to
process optimisation. This approach results in an easier, efficient and safer process
while providing considerable returns for the plant owners.
How variable are your processes, and do you maintain optimised process
performance with dedicated resources, modern instrumentation, ‘best practice’
control systems and performance monitoring tools? In this paper, these questions and
their possible answers are discussed.
We can do much more and be ‘more efficient’ in our efforts to remain competitive.
This area remains a major opportunity for improved efficiency, minimising tailings
losses and other milling costs.
INTRODUCTION
In introducing this topic, it is worth reminding readers of these (global) historical
comments, as these key champions are purposely referenced in the paper:
•• Konigsmann (1992) states:
Process control is now an essential part of any (Canadian) concentrator operation. It
provides a proven vehicle for improving operation economics by increasing revenues and
reducing costs. ... The question one must ask is whether we are tapping the full potential
of this technology? The answer: is ‘No!’ [MITEC Mineral Processing Technical
Advisory Committee (TAC) acknowledged two problem areas: 1) small plant
and 2) those plants that use some control while not achieving maximum benefit.]
•• McKee (1999) states:
Process control is a broad term which often means different things to different people.
The technology required to obtain information in real time on process behaviour
and then use that information to manipulate process variables with the objective of
improving the (metallurgical) performance of the plant.
•• McKee (1999) also states:
There is a need to determine, and if necessary correct, the condition of the plant as a
prerequisite to control development. Correcting plant limitations should be seen as a
first step in the approach.
•• Blevins et al (2003) states:
Proper controller tuning is the largest, quickest and least expensive improvement one
can make in the basic control system to decrease process variability.
•• Fiske (2005) states:
1. Process Control Consultant, XPS Expert
Process Solutions, Falconbridge, Ontario, ARC (Advisory Group) believes that those companies which are able to methodically
Canada. Email: phil.thwaites@xps.ca implement, use, and maintain APC (Advanced Process Control) applications, at the
373
P Thwaites
lowest cost and generate the highest value, over its life cycle PROCESS AUTOMATION AND CONTROL ENGINEERING
will have a distinct competitive advantage. Process automation is understood, in a wider sense, to
•• Lynch (personal communication, October, 2012) states: include/cover the following:
Control is not a high priority here (Australia) even though •• measurement and instrumentation
the effect of inadequate control on the loss of mineral tailings •• process modelling and simulation
is obvious.
•• process monitoring and data reconciliation
As such, the opportunities to consider in operating plants are:
•• data mining and multivariate statistics
•• Is your feed stable?
•• fault diagnosis and fault tolerant control
•• Are your instruments calibrated and performing?
•• process control, monitoring of product quality and
•• Are you aware of wireless instruments (including control performance
vibration)?
•• offline and online process optimisation
•• Is your control system up to date and stable?
•• AI (artificial intelligence) methods: expert systems,
•• Are you in manual or auto control? neural networks, fuzzy control
•• Are your operators acting on alarms or are they •• sequencing and automation.
nuisance?
This is a very large, fascinating field, with major
•• Do you understand and accept your process variability? opportunities in mining, mineral processing and ideal for
•• Are you operating within the design targets and process control engineers! It goes far beyond what plant metallurgists
constraints? (pumps, cyclones, supplies, thickeners, and instrumentation support personnel can (typically) handle
roasters, furnaces etc) and often requires many specialised training courses (eg the
•• Are you using your surge capacity … or running tight ABB / DeltaV / Foxboro I/A / Yokogawa / Siemens PCS7 etc
level control? control system, PROFIBUS, OPC and APC etc) on top of
control engineering theory. Undergraduate courses very
•• Are you at optimum and are the controls robust? rarely teach the basics let alone have any ‘hands on’ courses/
•• Are you benefiting from asset management systems? labs to illustrate control theory, or instrumentation practices,
•• Are failure/fault detection systems implemented? and the consequences of different tuning parameters, or
control philosophies.
•• Can you make the same product for less energy, or
material consumption? Before we discuss process control, consider Figure 1 and a
key comment by McKee in his very important introduction
•• Do you operate with or without a Control Engineer?
to the AMIRA P9L project, The Optimisation of Mineral
•• Does your plant/process or processes run as well as Processes by Modelling and Simulation 1996 to 1999:
your car?
There is a need to determine, and if necessary correct,
•• How do you accept your new plant or newly expanded the condition of the plant as a prerequisite to control
process? development. A good example is the importance of classifier
Each of these points should be carefully considered, based operation and its effect on comminution circuit performance.
upon personal observations with the Process Control Group Techniques exist (plant sampling, modelling and simulation)
from XPS, external clients, Glencore, Xstrata, Noranda, to audit the actual plant operation. Correcting plant
Falconbridge, CDC (Canadian Development Corporation) limitations should be seen as a first step in the approach.
and Texasgulf, it is this author’s belief that we can do ‘much (McKee, 1999)
more’ and be ‘much more Efficient’ using the technologies, Indeed, as we often see high variability in (grinding
systems and expertise which we have available to us. classification) cyclone feed pressure, or when part of a circuit is
FIG 1 – Process control will not correct inherent design / flow sheet problems.

Manual control, process automation or operational performance excellence – What is the difference?
shutdown, for example one ball mill in a one semi-autogenous at a 95 per cent confidence level. These tools are essential
grinding (SAG) × two ball mill grinding circuit. In the author’s in checking and correcting plant performance, or new plant
own experience, it relates back to the copper cementation and design, ensuring minimal wastage of precious paying mineral
cobalt removal process of the (former) Kidd Zinc Plant and resources, to tailings or ensuring best possible concentrate
the work in the early 1990s to improve cobalt purification grades for upstream processing at lowest treatment penalties.
and reduce zinc dust (the key precipitation reagent). Under
Figure 2 illustrates a milling operation which introduces a
K Kangas, as the Project Metallurgist, a cyclone recirculation
sump discharge into a cyclone overflow stream and shows
system was recommended, tested, engineered and introduced
the effect on cyclone overflow density (COFD). Note, COFD
after the first (of three) cobalt purification tanks. Accurate
is the product from grinding – to the customer – flotation.
loss-in-weight (Brabender/Control and Metering) zinc dust
metering systems were introduced; as was a (Massbal – now Flotation always wants a consistent particle size, feed flow
Honeywell) mass balancing control model, and an Advanced and feed density; together with the liberated pay mineral. As
Control Supervisory program (developed in-house) – to the sump kicks on, it dramatically disturbs the COFD. When
accurately control the zinc dust feed rate set points to all three off, there is good control around the target set point. Clearly
purification tanks. The consequences of the process change – this is not a control problem, but a process problem. It has
introducing the cyclone recycling, was huge but together with run for many years, too many years and the consequences of
the modelling, instrumentation, control improvements, far the ongoing disturbance are not that visible to the operators,
greater and in total 4.4 t/d of zinc dust was saved, resulting therefore relatively painless. It certainly would be more
in an additional sale of $1.8 M/a in (zinc product) sales! The visible if there was an online particle size measurement on the
exciting hidden benefit though, was a substantial increase in cyclone overflow! Collection, (of the sump – originating from
online (processing) time from 88 to 95 per cent – itself worth a dust collection system in the crusher plant) into a holding
a considerable sum! tank, with a smaller, more controlled bleed into the process
Figure 1 also shows (Fragomeni et al, 2006) mineralogical stream is a better process solution – but easier said than done.
assessment equipment which XPS has (in the Centre at Alternatively, a smaller, more continuous flow from the sump
Falconbridge) for completely analysing existing plant stream is one which would be better for the process, but this requires
samples (offline), or drill core samples, as well as MPP (mini addition of extra water – unnecessarily diluting pulp density.
pilot plant) equipment to diagnose, develop and confirm Once the plant is operational, it can become very difficult to
milling processes and complete flotation circuits. With resolve simple, yet highly disturbing process issues, such as
correct sampling and statistics, surveys, results are obtained this one.
FIG 2 – Cyclone overflow density (COFD) with and without sump pumping.

P Thwaites
Figure 3 illustrates the Process Control Group’s classic reduction is a recommended precursor to optimisation. It
symbol, as we have championed process control for at least may have a small value and be an easier to operate, safer
three and a half decades now: process; however, the true value to the business (and plant
The figure is a general diagram to illustrate the overall manager) is the ability to push to the constraint and robustly
process control objective for any key variable: feed, density, hold the process at a different operating point. The value of
flows, pressures, temperatures, levels etc. Measure and this, when it can be consistently sustained, is huge and often
understand the initial variability, stabilise and then optimise very significant to the operation. It is our ‘carrot’.
to the constraint of the process. As a control engineer, we Figure 4 illustrates poor (no control) to optimised control
need to understand: as distributions, common to Six Sigma and Statistical Process
•• the metallurgical constraints of the process Control. Note in the second step, under control, the figure
•• the equipment constraints. shows a tighter distribution and in the third step, ‘optimise’ the
Typically, throughput benefits will come from an increase process mean has been shifted providing a key measurement
in the set point, and consumable savings will come from a to quantify the result. Holding the process at a particular
decrease in the set point. Once these have been achieved it is constraint can take as much (development) work as was
important to ‘maintain the gain’ by ensuring the changes taken to reduce the process variability. Note an offset (in the
are robust, thus benefiting the plant. (Thwaites, 2007) mean) may also arise from poor tuning parameter selection,
Classic examples which the audience can relate to are for example flotation level control ‘Integral’ parameter in the
‘cruise (speed) control’ in vehicles and temperature control level proportional integral derivative (PID) control, as also
in houses and rooms. Note, other logos which we have seen shown in Figure 4’s plot of ‘bias elimination’. Reducing the
may only show the variability reduction (with control) and variability requires a good foundation and knowledge of
not the optimisation step – often requiring APC. Variability control engineering and instrumentation. Filtering, retuning
FIG 3 – Measure, control, optimise.
FIG 4 – Poor to optimised control.

may be the solution, but also other control and instrumentation basic controls are working in auto and are contributing to
solutions may be, and often are, required. reduced variability of all processes.
The key process control objectives, leading to operational While process and area optimisation can have considerable
performance excellence, can therefore be simply summarised financial returns, it is well recognised (and reported), by those
as the management of tools and resources to consistently: practising process control, that this can only be achieved by
•• minimise process variation having a robust and solid regulatory lower level.
•• optimise process performance Svedala (Metso Minerals) advocate a systematic approach
•• minimise (fixed and variable) costs to process control:
•• control product quality •• start by defining what is important through an analysis of
•• maximise safety. the process (an audit) to determine what is required and the
level of control
Figure 5 illustrates the classical layering of process control
(Thwaites, 2007; Jones, 1994; Jones and Morrell, 1996). •• understand the process and ensure the basics for good
Like Figure 3, this key diagram has been referenced for control are in place (eg instrumentation)
over three and a half decades, where process control has •• match the installed system to the capability of the people
been championed throughout Texasgulf, CDC (Canadian available
Development Corporation), Falconbridge, Noranda, Xstrata
•• recognise the ongoing commitment required for success
and now Glencore. Jones and Morrell (1996) discuss a figure
showing levels of control (as used by one of the authors •• with regard to the actual control, consider three levels as
in a 1969 article), whereby Level 0 represents the process follows:
I/O (Input/Output), Level 1 the regulatory level, Level 2 •• basic PID – via DCS or equivalent
Supervisory Control, Level 3 Unit Optimising Control, etc.
•• supervisory (cascade control, dead-time compensation
As shown in Figure 5, shows the processes at the bottom
and so on – via DCS or equivalent)
level, followed by the Instrumentation Inputs/Outputs,
ie measurements. Without any control, this then is shown as •• optimising (eg expert system, adaptive) – via a package
manual control. Loop control, the ‘regulatory level’ is on the such as Gensym’s G2 or implemented within the DCS.
next level and can be done both in the field, or in a central (McKee, 1999)
location, by panel based controllers (typically single loop), Figure 6 (Ruel, 2012) is a nice illustration showing that most
or distributed control systems (DCS), or programmable logic plants have the bulk of their control investment in the basic
controllers (PLC). Note PLCs, in their history, were never regulatory DCS. Additional, (approximately +33 per cent)
designed to operate plant-wide control supporting many PID investment is required for ‘coordinated control’ and
control loops. APC is layered above the regulatory control ‘advanced automation’ – not nearly so common. However,
operating on the regulatory control loop set points. It is not
in terms of return-on-investment (ROI), ‘Coordinated
best practise for APC to directly control end elements – valves,
Control’, ‘Advanced Automation’, and ‘Optimisation’ can
variable speed drives, etc. Above the Process Optimising
have a much, much bigger ROI – directly impacting business
Control, is the Plant Optimisation. Moving up the triangle the
complexity increases, but also, so does the economic returns. profitability. Exactly what the plant management wants to
Any problems in the lower levels, or functions, will of course know, hear and report about.
often prevent the optimisation. Furthermore, expecting The control technology options available start with ‘manual
any advanced control to be successful when the regulatory control’ where we need manpower to adjust valves etc to
control has not been closed or tuned is naïve and unrealistic, adjust flow rates (like water, chemicals, flotation pulp levels,
especially in mineral processing plants. airflow etc). The bigger the plant, the higher the throughput
Operations are recommended to do a regulatory control then the number of manual valves, feed rate controls etc
review, audit prior to moving to APC – ensuring that these increase. A couple of important points are:
FIG 5 – Layering of process control.

P Thwaites
does not attempt to continuously improve operations in an

economically optimal manner. Fiske (2005).
A summary of our control options, are as follows:
•• manual control … usually acting on valve position
(ie percentage opening, or motor drive speed)
•• PID control … process variable (PV) and set points (SP)
in engineering units – ie automatic regulatory control
based upon the PID control algorithm:
•• feedback only (single input, single output; SISO)
•• feed forward and feedback (multiple input, multiple
output; MIMO)
•• ratio control and cascade control
•• split range control
•• smith predictor … nicely dealing with significant
time delay.
FIG 6 – Leveraging control technology (Ruel, 2012). •• MPC or model based control … based on true process
dynamics, and using recent history and a model based
trajectory
1. our operators, in such a plant, will know valve positions
•• rule or fuzzy control … based on human knowledge/
(two turns open etc) and not necessarily the engineering
experience
unit of flow rate, such as m/min of water flow, cc/min
of chemical and so on •• automation control of specific known actions or
sequences; batch processes and so on
2. each operator will have their own way of operating –
readjusting from shift to shift. •• advanced (optimising) process control.
It is far better to have our operators work in engineering More detailed descriptions of control technology options can
units and set the set points of the control loops, and let the be found in Meech (2006), Ruel (2006, 2014), Boudreau and
controller adjust the control valve – typically on a 1 s basis, McMillan (2007), Hodouin (2011), Flintoff and Mular (1992).
to keep the flows at the targeted set point. These ‘regulatory’ Let’s take a closer look at the standard (PI and PID)
loops then only require a periodic update of the set point, controller. Figure 7 shows two basic controllers, the first used
and can vastly reduce the operator input, saving valuable for room temperature control (cruise ship) and the second for
manpower costs – keeping the process at an acceptable dual zone vehicle inside temperature control (1997 vintage).
operating point. Regulatory control is predominantly done For the former a value is not shown, neither is any indication
by proportional integral (PI), PID control loops operating in where the output is (ie 50 per cent). Also the set point value is
the control system – the DCS (or in some cases, and usually not shown and the controller indicates the comfortable target
more difficult, the PLC). PI and PID control can often be set point, allowing the freedom to move this either up or down,
(but is rarely) enhanced, for negligible cost, by adding a improving comfort. Simple and effective automatic control,
feed forward signal/term allowing the controller the ability but not allowing: 1) manual control, or 2) any automatic
to make a final element valve change when a (measured) optimisation of the temperature during the day. The latter is
disturbance comes to the process. Disturbance rejection is a dual loop controller in ‘auto’ with manual control switching
(should be) a key deliverable of regulatory control. Final and allowing a different target set point. Again, basic control
adjustments are done by the feedback control on the error loops allowing the (driver and passenger) to manually adjust
between the measurement and the set point. It is important their set points.
to point out that PI and PID control cannot predict where the In our mineral processing plants, it is normal (usually) to
process is going in the future and if it (the controller) is going have more complex control systems, as shown simplistically in
to violate a constraint. However, this is possible, for the few Figure 8, as often there are several hundred regulatory control
loops which require it, with Multivariable/Model Predictive loops, as a minimum, together with different instrument input/
Control (MPC) (Sandoz, 2006; Boudreau and McMillan, 2007;
output (I/O) networks etc. Instrumentation can be on (classic)
Jonas, 2008; Yutronic and Toro, 2012). Mineral processing
4–20 mA networks, H1 (Foundation Fieldbus), DeviceNet,
plants could benefit with more MPC controllers, particularly
those from the control system vendors – as embedded tools.
PI and PID controllers can easily be enhanced further
by configuring them as cascade or ratio controllers; for
example, water feed: tonnage feed, and flotation chemical
additions: flotation feed metal units.
Many of our flotation cells operate with dual dart valves,
and as such, require a ‘split range’ controller. These are rarely
set-up properly – in new plants, and often are not operating to
their full potential, requiring constant operator intervention or
poor performance of the flotation cell (through cell level being
outside the process constraint), as illustrated in Figure 25.
The primary role of regulatory control is to ensure stable,
safe, and reliable operations while maintaining process units
at a desired or specified condition’ – reducing variability. ‘It FIG 7 – Simplest of controllers … we all know!

Profibus networks etc. The latter (most buses) provide both the while noting that for all tuning the importance of good
process measurements as well as other key digital information measurements and final control element design goes without
of the device on the network. Furthermore, individual wiring question. Figure 10 illustrates further work to improve
of the devices is not necessary, reducing installation costs. tuning of the flotation cell airflow control loop. The process
Above the controllers, and I/O modules, you will see the remained the same, as did the airflow measurement and
operator workstations, as well as a (OSIsoft’s PI) plant historian
control valve. Shown is the difference in controller tuning
database, accessible from the plant local area network (LAN).
parameters. They are significant and very important as
The authors always find it very surprising to learn of key one moves from regulatory control to optimised control.
processing plants and operations that do not have a plant
Before the changes, a change in air set point would take
historian to support the operations.
approximately 15 minutes (to return to the new set point) and
Figure 9 illustrates flotation cells in a mineral processing this is no issue if they are only changed once, or twice a shift.
plant. Prior to upgrading, there was one air supply for two
As under regulatory control, this is far better than a manual
flotation cells under manual control. Cell performance was
valve being periodically adjusted by an operator. However,
therefore impacted by air supply changes, blockages and
varying, uneven pulp levels, thus requiring constant operator if airflow set points are to be optimised – as they should be,
attention. Note operators would ‘know’ key valve positions or could be (under metallurgical control), then retuning can
(number of turns) but not individual airflow rates. After result in a much faster, more responsive loop – stabilising
upgrading, individual airflow control was provided to each almost immediately, as required by a froth velocity control
flotation cell with the measurement and control in engineering loop, cascading new airflow set points on a one minute
units, as these became classic regulatory PI, PID control loops. control interval, for example. As you might appreciate,
Next look at one of our ‘typical’ controllers in the mineral the consequence to the next level of control is huge, not to
processing plant, under automatic, regulatory control, mention allowing a completely different upper loop control
FIG 8 – Modern data acquisition and control system – Emerson DeltaV distributed control systems (DCS).
FIG 9 – Flotation rougher control and move from manual control to (individual) cell air control … and preparing for optimisation.

P Thwaites
FIG 10 – Importance of regulatory loop tuning – flotation airflow loop.
interval, and one more appropriate to the dynamics of the Figure 12 (Bouchard et al, 2010; Émond and Vynogradova,
flotation process. 2010) illustrates similar results on a grinding circuit in our
Flotation air systems are often complicated by fan and air platinum processing plant in South Africa. Through control
duct designs ahead of the flotation cells. Multiple PID loops to improvements on mill feed, reducing variability by 72 per cent
(ie standard deviation of 50 to 14 t/h); and better flotation feed
the different cells on the duct work very often fight each other,
density control (56 per cent reduction in variability) utilising
and the engineering design does not always take into account
an installed nuclear density metre rather than a steady state
that the duty for a rougher is different than a scavenger.
water addition model etc, primary grind improvements were
Challenges with flotation control requires the process design
measured to improve from a decrease in grind size from
covers controllability in the operating envelope that handles
42–52 per cent -75 µm (improving mineral liberation) and an
the variability from changing head grades and feed tonnages increase in primary rougher recovery from 60–73 per cent.
– whether automatically or manually controlled. Overall plant recovery was increased by 3 per cent and final
Systematically taking this approach through the flotation product grade by 10 g/t on a feed grade of 2.5 g/t platinum
circuit takes one to the cleaner circuit, where column flotation grade elements (PGE). These are very significant results
may be used for cleaner cell upgrading. Figure 11 illustrates a coming from the variability reduction efforts and subsequent
feed variability reduction of some 56 per cent which was then control improvements using the same processing equipment.
correlated to concentrate grade variability reduction: some Emerson (2007) have stated (in their DeltaV embedded
79 per cent reduction in the nickel (impurity) in the copper advanced control presentation) that, as an industry average
concentrate grade (Jin et al, 2013). These kinds of improvements 75 per cent of assets are under process control, but over
have an overall benefit on the mill’s performance, impacting 60 per cent of all control loops are underperforming, and this
the bottom line and pleasing the operation’s manager. poor performance results in:
FIG 11 – Column flotation – feed control … regulatory control.

FIG 12 – Eland Mill (2010) overall recovery improvements from primary grind and primary rougher improvements.
•• high product quality variability different tuning parameters, known as gain scheduling, for
•• reduced throughput acceptable performance.
•• increased downtime, lower availability The XPS Process Control Group recommends (Figure 13) for
new plant, processes:
•• poor financial performance.
Commissioning of the PCS and PLC systems will not
For the mining industry, poor loop performance is often (should not) be considered to be complete unless:
due to:
•• all systems are connected, trouble and error free
•• poor tuning knowledge/practise
•• all designed and configured regulatory loops have been
•• feed variability changing the tuning parameters – or a acceptably tuned and are operational in automatic
lack of ‘gain scheduling’
•• as well as specified start-up and shutdown sequences
•• measurement problems (ie scale on pH probes, flotation configured/tested; and documented
level measurement floats sticking or cracked floats etc) •• alarm/group alarming implemented.
•• problems with final control elements/actuators This then will allow effective, safe, centralised control.
•• poorly designed control loops – not identified in design It is also a precursor for optimisation and APC.
or commissioning.
At a site visit six months after commissioning – Figure 13,
Where are you at in your plants, or newly commissioned on one of our new mineral processing plants, soon running at,
processes? Yes, regardless of initial and greenfield, or exceeding, design of 70 kt/d, the following was observed of
brownfield investment, your new plant may in fact be worse the installed control system status:
than this (60 per cent of your control loops underperforming), •• commissioned status (DCS) = 94 per cent
if commissioning was not completed properly, if there are
•• regulatory loops on manual = 37 per cent
measurement/final control element problems, if regulatory
control loops were incorrectly configured, poorly designed •• regulatory loops on auto = 63 per cent
and (mostly) not tuned properly or optimally, following •• 16 per cent of manual status had field instrumentation
commissioning. Additionally, feed variability may require or logic issues
FIG 13 – Control system commissioning status after six months – new plant.

P Thwaites
•• OSIsoft PI, as process management information system the (power generation) process is relatively simple. Recently,
(PMIS) plant historian – just introduced, and not Pyrogenesis (Holcroft and Carabin, 2012) reported on their
available for start-up, commissioning. Plasma Arc Waste Destruction System for the US navy carrier,
The plant design had approximately 145 regulatory control marine vessels and mobile units – a fully automated process,
loops. Advanced control was shortly implemented on SAG required to be ‘idiot proof’ for the operator.
mill feed and was observed to be very beneficial (as compared Some automation examples that we commonly see in the
to manual control) with impressive initial results; however, minerals industry are:
the installation proceeded in the absence of building on a •• motor starts and stops eg crushing plants; and sump
strong, well-tuned regulatory layer, a necessary building pump controls
block and fall-back position, for any operation. •• remote starts, start-up and shutdown sequences, which
Thornton, Prinsloo and Tewu (2009) state: can be quite complex
The authors have found in practice that one of the •• mill lube systems
fundamental barriers to effective SAG mill control is simply •• conveyor belt material handling systems
poor tuning of feedback control loops.
•• mine drilling equipment and (mine) autonomous
Please now think how you accept a new automobile, and haulage:
the state of the engine controls etc! All sensors, controls have
•• ‘Vale puts Railveyor to test’ – Sudbury Mining Solutions
been tested, are fully functional, and the ‘check engine’ light
Journal, Sept 2012
will not be activated. This should be the standard for our new
plant, process commissioning. •• ‘A key pillar of Rio’s mine automation initiative is
autonomous (vehicle) haulage’ – Technology Case
As Blevins et al (2003) stated: ‘Proper controller tuning is the
Study: Rio Tinto, Mining, People and the Environment,
largest, quickest and least expensive improvement one can
Jan 2012
make in the basic control system to decrease process variability.’
Further discussion can be found in the book Advanced Control •• pressure filtration and batch processes
Unleashed. The book was inspired by DeltaV Advanced Control •• water purification (Outotec etc)
Products, Figure 14. Note also the course manual Fundamentals •• automated metal (plate) stripping, as used for nickel,
of Process Control (Ruel, 2006) from Top Control, now BBA. copper and zinc stripping and so on.
Figure 15 illustrates a particular area of speciality for Glencore
AUTOMATION AND ADVANCED CONTROL and Glencore Technology (formerly XT, or Xstrata Technology)
Dictionary.com (2014) defines automation as: – automated copper stripping machines. Prior to 2002, the
1. the technique, method or system of operating or controlling stripping of refined copper cathode from stainless blanks was
a process by highly automatic means, reducing human accomplished using custom electrohydraulic machinery:
intervention to a minimum The Kidd Creek Copper Refinery, one of the most automated
and mechanised refineries in the world, has been in
2. a mechanical device, operated electronically, that functions
operation since 1981. ... With the invention of the KIDD
automatically, without continuous input from an operator
PROCESS® stainless steel cathode blank (Canadian Patent
3. act or process of automating #1263627) it became necessary to design and build a
4. the state of being automated. material handling system which included an automatic
Note, automation is not process control in itself, however stripping system. Advantage was taken of the opportunity
process control may and should, use automation. So where to include special design features providing for a more
do we see it and how do we apply it? Both simple and more efficient cathode washing system, cleaning of the hanger
complex examples are presented here. bar contact-surface and better packaging of the end product.
(Laezza, Box and Scott, 1990)
Automation can be done on various platforms (PLC or DCS),
on specific process units or areas of the plant. While the author This was a great example of automation in our industry,
is not aware of a fully automated mineral processing plant, he however both the Kidd Copper Refinery and the Kidd Zinc
has recently visited an ‘Unmanned (hydro) generation station Tankhouse are no longer in existence.
(water turbine/generator)’ – at Sandy Falls (Timmins), where In particular, the Kidd Refinery stripping process was
extremely successful, but the systems were quite complex,
fairly maintenance intensive and were challenged to recover
quickly when off-spec product entered the system:
In some cases this resulted in less than acceptable machine
up-time and it was determined (in 2003) that the bulk of the
system downtime occurred, not during the actual product
processing, but during product handling.
In 2003 the system was completely redesigned by Ionic
Engineering (Sudbury, Ontario, Canada) in cooperation with
Xstrata Technology – now Glencore Technology (the Group
responsible for the Falconbridge Kidd Process®). The new
design featured the world’s first use of high capacity, high
reliability industrial robotics – Fanuc Robotics – in a copper
refinery application. The first two systems were installed
at the LS Nikko Copper refinery in Ulsan South Korea in
2005/2006. The next system was installed at the Jinlong
refinery in Tongling China in 2008, with subsequent further
FIG 14 – Control tuning summary – don’t underestimate the impact. installations at other facilities in China. Some key metrics, in

FIG 15 – Automation – first installations – robotic cathode stripping machine.
the technology improvements (by moving to robotics) that control closer to operating constraints and more favourable
were implemented are: economic operating conditions.
•• machine uptime – in excess of 95 per cent, far better than ARC (Advisory Group) believes that those companies which
all prior designs are able to methodically implement, use and maintain APC
•• power consumption – over 50 per cent reduction, and applications, at the lowest cost and generate the highest
390 kW down to 175 kW value, over its life cycle will have a distinct competitive
•• number of sensors – 48 per cent reduction, 140 down advantage. (Figure 16)
from 270 Thwaites (1993) states (Figure 17):
•• number of actuators – 53 per cent reduction – 22 down Up until the mid 1980s Supervisory Process Control had
from 47. only been applied in one of the Kidd Creek Metallurgical
In time, we will see much more robotic applications, and Plants – the Concentrator. The on-stream analysis system
it is the author’s belief that there will be far more automated and central control design were key to the development of a
sequences in our control systems, however presently this supervisory reagent system running on economic efficiencies
is a key area where much more development and novel calculated from smelter contracts. Optimum metallurgy and
applications are required. Don’t forget an automated sequence 25% lower reagent costs were the direct economic benefits
is one which is driven the same way, every time and not one from this development effort. In 1986 a decision was taken
open to operator interpretation – each operator’s interpretation. by Kidd Management to apply Supervisory Process Control
An automated sequence is also one which will highlight any throughout other areas of the Metallurgical Site. The Kidd
problem or failure in the system completing the sequence. Creek Zinc Plant was given a priority to apply Supervisory
Meech further discusses intelligent robotic systems Process Control at this time.
and robot applications in the Mineral Process Modelling,
Simulation and Control (MPMSC) (proceedings) from the June
2006 International Conference and made several key points
including:
An important aspect of the continuous operation of a robotic
system is its ability to respond swiftly and modify its actual
physical behaviour as new pieces of data are perceived.
(Meech, 2006)
Let’s move onto APC, with some quotes by Fiske (2005),
before presenting a couple of examples:
In terms of automation functionality, regulatory control
and APC serve much different purposes.
The primary role of regulatory control is to ensure stable,
safe, and reliable operations while maintaining process units
at a desired or specified condition – reducing variability.
It does not attempt to continuously improve operations in
an economically optimal manner.
APC, on the other hand, is a supervisory control application
that coordinates a large number of parameters to maintain FIG 16 – Advanced process control (APC).

P Thwaites
FIG 17 – Advanced process control (APC) – hydro management and site peak power control.
The example presented (Figure 17) was selected to illustrate is that at the end of the 24-hour period, and week, no zinc
the considerable economic potential that was realised with production was sacrificed in the control of the site peak
the application of (then) current computer and software and the rectifiers (66 MW total) were the best load to ‘trim’
technology to the Supervisory Process Control of Kidd Creek or ‘take advantage’ of any load changes in any of the site’s
Metallurgical processes in the Zinc Plant. processes – automatically on a one minute basis. A simple
Online steady state modelling, dynamic modelling but very effective advanced control strategy – programed
and multivariable predictive control, and conventional in FORTRAN, with a simple operator interface, using
programming have all been applied to improve the automation (for the automatic switching of the peak controller
competitiveness of the Kidd Creek Zinc Plant. and the automatic ramping up of each zinc rectifier circuit for
‘off-peak’ periods). The set-up also allowed the site to quickly
In 1989, Ontario Hydro introduced time of use (TOU) power
(within 15 minutes) and dramatically, reduce power demand,
demand and energy consumption rates in their monthly
for example to a contracted 89 MW (from approximately
billing, for both winter and summer (six months each) months
133, 143 MW etc), when Ontario Hydro (the Power Utility)
and defined on/off-peak periods. These were for all Ontario’s
had supply problems. This ability further saved substantial,
mining operations. Following close scrutiny of the monthly
additional penalties on several occasions. The cumulative
billing, and installing key metering equipment, XPS was able
(five year) total hydro savings to Kidd (from 1989 to 1993) – as
to monitor the entire Kidd Metallurgical site’s power demand
presented to Falconbridge President, F Pickard, was $17.6 M,
(approximately 133 MW) and consumption (approximately
as compared to no control and the known Ontario Province
1 089 851 MWh in 1989) on a one minute average time basis,
average mining rate. Clearly, Kidd being an integrated
extremely accurately. Using the MODCOMP (main frame)
metallurgical site was best able to implement such an effective
computer, as also used by NASA at the time, the Technical
site wide strategy – but this is rarely addressed in the design
Computing Group, under Mr J Carriere, soon was able
of new plant, let alone a fully integrated site. To be effective,
to duplicate the entire bill for both the whole site and also
such a strategy requires a shedable load, and (importantly)
each of the major plants: mill, zinc plant and copper plants.
over capacity, as well as real-time pricing and consumption
The (second by second) consumption data was also made
monitoring/control on a second by second basis plus the
available to a single loop (panel based) controller (as there
appropriate control technology. ‘APC can have considerable
was no DCS in the Kidd Zinc Cellhouse) which controlled
returns … to help the business.’ Referring back to Figure 5,
three panel based ratio controllers – each in control of a zinc
this example represents control at the very top of the triangle
electroplating circuit consuming, under maximum power
– optimised plant/site operation.
some 22 MW. For the three circuits, 66 MW soon proved
ample to control the site’s 133 MW power demand and peak The teams at Kidd Creek (including process control and
power demand on a one minute basis. The supervisory control Technical Computing, supported by the Operations and Site
computer switched the site peak control on automatically for Management) showed a consistent, systematic approach to
the start of each day shift (6:40 am), returning it to manual at improving process control and applying APC. The approach
the end of the day (11:00 pm) and each weekend and statutory was presented to Noranda (Thwaites, 1993) at one of their
holiday. Furthermore, at the start of these ‘off-peak’ periods, technology conferences.
the supervisory control computer would also ramp rectifier Figure 18, DeltaV, by Emerson also (later) show a nice
production up to maximum – thus allowing production to systematic approach in applying APC from within their DCS.
‘catch up’ at cheaper ‘off-peak’ periods. Important to note This is generally uncommon (though recently also available

from ABB and from the Siemens PCS7, among others) and (PID) controllers and flotation pulp level (PID controllers);
yet a huge area of need in our minerals industry. Even, to then be duplicated for the larger flotation circuit. Importantly
this day, the bulk of advanced control has been done with all the key control blocks are running in the control system
external systems (such as the MODCOMP supervisory architecture, where historically the DCS has not had the
computer systems at Kidd) or control packages – such as the computation ability to perform MPC. Figure 19, at Collahuasi
Honeywell’s ‘ProfitSAG’. Figure 19, Gensym G2 (Suiches (Yutronic and Toro, 2010) presented this MPC application
et al, 1999) – as applied at Noranda’s Brunswick Mill and of Honeywell’s ProfitSAG as a better solution than expert
Invensys’ Connoisseur – the MPC package (originally written systems (see also Jonas, 2008), for the control of the three
by Dr D Sandoz, and available to us since 1989), which we lines (SAG and Ball Mills) controlling then, approximately
were more familiar with at Falconbridge. More recently we 140 kt/d.
see offerings from Metso, FLSmidth Automation (PxP) and Figure 19, Honeywell’s ProfitSAG is an MPC solution for
others. SAG Mills:
Emerson’s DeltaV DCS (Figure 18) allows (if you purchase •• objective function designed to accomplish the goals
the license) advanced control using fuzzy logic, neural (maximise fresh feed rate), by manipulating (fresh) feed
networks and MPC. One of the benefits is it is like an rate, mill speed and solids content
extension to the existing control blocks and can be utilised •• fault-tolerant policies (anti-windup integration with
on small and large applications. For example a ‘model based’ regulatory control level)
flotation froth velocity controller could easily cascade set •• fully integrated with measured disturbances, such as
points down to (well-tuned regulatory) flotation froth airflow returned pebbles and particle size.
FIG 18 – DeltaV (Emerson, 2007) – systematic approach to apply advanced process control in the distributed control systems.
FIG 19 – The Collahuasi Solution: Multivariable Predictive Control ProfitSAG replaces Expert Systems (Rules).

P Thwaites
The Collahuasi operation has run many years on ‘expert

systems’ and now (2010) prefers the MPC approach provided
by the Honeywell ‘Profit’ controller. Incidentally, this was all
on top of ABB regulatory control system (DCS) and regulatory
controls. The revised Collahuasi approach came from much
learning at the Codelco plants and Kairos (Zamora et al,
2010), itself the result of the AMIRA P893 benchmarking
study (Mujica et al, 2005) – Review and Benchmark of Process
Control and Automation (in South America). For P893 review
Falconbridge allowed Collahuasi and Noranda allowed
Antamina operations to participate in the benchmarking and
comparison. It was an important and very rare benchmarking
exercise looking into the level of process control and
automation at key (South American) concentrators operated
by different mining companies.
Jonas (2008) states:
Use of MPC for mineral and mining industries offers
improved benefits and more sustained benefits over the long- FIG 20 – Elements necessary for successful process control in mineral plants.
term. This is achieved by the use of technology that delivers
better control and more optimisation, while at the same time that has been much more difficult, or impossible to get – in the
offers a system that is easier to use, modify and maintain. past, is the CiDRA SONARTrac flow measurements.
The sheer volume of applications in mining and other As shown in Figure 21, SONARTrac (Maron, O’Keefe and
industries are a tribute to the success of this technology. Rothman, 2010) is a new class of measurement with distinct
And goes on to say: application advantages – ie ability to primarily clamping onto
The mining and mineral processing industries can improve most installed piping negating the requirement to cut and
business results by adopting an automation strategy more flange piping. As Maron, O’Keefe and Rothman (2010) state:
aligned to the other industrial sectors. This can be achieved Sonar array-based flow measurement technology was
by embracing more mature and better advanced control introduced into the mineral processing industry five years
solutions, and reducing expenditures in other technologies ago, and has since demonstrated significant usefulness
that deliver fewer benefits. (Jonas, 2008) and value in many difficult and critical flow monitoring
Ruel (2014) describes how to make advanced control choices applications. This robust non-invasive technology has become
when difficult processes need improvement. He concludes: the standard for many companies in certain applications.
‘Advanced control is powerful but should be used after an Presented here is a summary of application experience,
evaluation of needs and expected performance’. lessons learned and best practices from installations
worldwide. Highlighted applications include: cyclone
ELEMENTS NECESSARY FOR SUCCESSFUL PROCESS CONTROL feed flow measurement, measuring aerated flows for mass
balance correction, stratification and sanding detection in
‘Elements necessary for successful process control in
horizontal slurry lines, slurry pipeline flow monitoring and
metallurgical plants’ has been discussed and presented in
leak detection. It will be shown how the basic volumetric flow
the past (Thwaites, 2007) and so is summarised again here in rate, combined with the unique additional measurement of
Figure 20. They are very important and must include: entrained air volume and the proper positioning of multiple
•• tools – instruments, systems, technology etc metres can enable novel solutions to monitoring and control
•• people – control engineers/process knowledge, problems that are not possible with other flow technologies.
instrumentation technicians/specialists, IT Figure 21 also shows a hydrocyclone feed installation at our
•• actions – support, management, company, technology Collahuasi Mill in Chile. Glencore have similar installations at
transfer Antapaccay and Antamina (both in Peru) and also installed at
•• successes – results and examples for further leverage, MMG’s Las Bambas Mill (formerly the second of Xstrata Cu’s
application. standard concentrator design – sold by Glencore) in Peru.
Success demands a team approach, and operational Figure 22 compares volumetric flow readings from an
performance excellence can rarely be achieved without electromagnetic metre to a CiDRA volumetric flowmeter at a
these key elements, so please take note. Furthermore, as Glencore grinding circuit (Eland Mill) in South Africa. Clearly,
discussed earlier, solutions need to be robust and layered more accurate measurements allow improved control, but
on good regulatory control. They also need to be discussed also start to become very valuable in metal accounting.
The South African installation, presented at Platinum 2012
and presented.
(O’Keefe et al, 2012), also includes flow measurements on
concentrate products, an area that has been particularly
INSTRUMENTATION – A SOLUTION TO BE AWARE OF difficult in the past, and is not common. Note mass flow will
Instrumentation, as if we cannot measure, or actuate, we still require a density metre and installation guidelines must
cannot control! Instrumentation technology has progressed be strictly adhered to obtain the best accuracies, for both
significantly over the last three and a half decades, and flow and density. As you can appreciate from this example,
several examples can readily be discussed, from particle size the best, most accurate instrument measurement will help a
analysis, cross belt assay analyses, stream chemical analysis great deal in achieving ‘operational performance excellence’.
to pH innovations, basic level measurements and flow rate Less variability in measurement allows a higher degree of
measurements. One new technology example worthy of note control, or even basic regulatory control where previously
and which is providing the industry with a key measurement no control was possible.

FIG 21 – Instrumentation technology – flow measurement – CiDRA SONARTrac.
FIG 22 – Instrumentation technology – flow measurement comparison mag versus CiDRA SONARTrac.
Have we endorsed this ‘new’ measurement technology? operational performance excellence, with a focus on four
Figure 23 summarises that approximately 140 SONARtrac major areas’ with one being Control Excellence (Figure 24):
volumetric flowmeters have been installed at approximately
16 Glencore mineral processing, mining facilities in the last Closed loop production and business control (being able
decade, so yes we have. to use real-time feedback to better run the operation)
will continue to be an important goal for clients.
OPERATIONAL PERFORMANCE EXCELLENCE (AMR Research/Gartner, 2010)
How do we define operational performance excellence? How does this relate to mineral processing plants? It is
‘Business optimisation, by definition, requires a focus on important to note that there are many key areas such as:

P Thwaites
FIG 23 – Instrumentation technology – 140 SONARTracs installed at 16 Glencore facilities.
grinding, filtering, drying, metal accounting etc; however, this a very important part in many operations. Optimum flotation
paper will focus on three key areas because these generally performance is reliant on good level control, air sparging and
present many opportunities: flow control, as well as precise chemical additions. Significant
•• flotation level control opportunities are found in flotation operations by attention
to these fundamental controls. Flotation control has, should
•• reagent control have and will continue to have, a lot of attention from plant
•• thickener control. metallurgists and control engineers (Thwaites, 1983a, 1983b,
1986; Flintoff and Mular, 1992; Gillis, 2000; Shean and Cilliers,
Flotation level control 2011; Hodouin, 2011). ‘It is the process area of significant
Flotation is a complex subject and has many facets to its upgrading of commodity minerals/metals and avoiding
optimal control, making it very interesting to both the mill smelter penalties, maximising net smelter returns is a key
metallurgist and our control engineers: objective of all mineral processing operations’.
In an ideal situation, collectors should absorb selectively on the In the operation of a flotation cell, Figure 25, it is necessary to:
valuable minerals, and depressants should absorb selectively •• always prevent flooding … or empty banks
on the unwanted gangue minerals. Unfortunately, this is •• maintain an accurate, responsive pulp level, on a
almost never the case and the right combination of collectors one second basis, thereby allowing control, then
and depressants require the dedicated chemicals; and a lot optimisation, of the set point.
of fine-tuning in the flotation process. (Akzo Nobel, 2014) Note, pulp level is the variable known to the operator and is
Flotation is so fundamental (Thwaites, 2007) to the usually highly correlated to mass pull and mineral recovery.
separation of copper, zinc, nickel and lead sulfides and plays Recovery can be calculated on typically a 5–15 minutes basis
FIG 24 – The need for operational excellence, control excellence (AAMR Research/Gartner, 2010).

for metallurgical control (Thwaites, 1983a, 1983b, 1986) and

mass pull can be measured on a one minute basis, from optical
cameras (Jin et al, 2013).
As discussed earlier, and shown in Figure 26, tuning can
play a big role, and this figure shows the effect of retuning,
resulting in a 56 per cent reduction in variability and the
elimination of bias – by applying ‘integral action’ to the
control loop. These results were obtained in a large flotation
plant. Without it, (please note) the operators were required to
memorise the bias and apply it consistently to their set points!
Several such loops were noted to be the same, and each were
resolved by retuning. We did not ask the operators how they
handled this, or even knew about it!
Figure 27 summarises our view of ‘operational performance
excellence’ as applied to flotation level control – often an area
for huge gains in flotation plants.
For flotation level control eight key steps are shown to
achieve best practices. Prior to tuning, the condition of
the measurement device, and the actuation, needs to be
assessed and addressed – as appropriate. Not all flotation
level (displacement) floats are the same, with many sticking
(in their guides), with unacceptable friction. Many utilise
ball floats (rounded bottoms) which are more susceptible FIG 26 – Control loop improvement – flotation level control,
to density variations and cracking at the top of the float – from retuning … regulatory control, bias removal.
letting water in and affecting the mass of the float. Our best
practices are noted in the XPSFloatTM, which is available both minute basis, as shown in Figure 29) to ‘infer’ mass pull
within and outside of Glencore; and also through Endress (related to recovery) in between updates from the online
and Hauser (E+H), or others. After the basic level device XRF (X-ray Fluorescence spectrometer). Froth velocity can
and retuning, flotation level performance can be further control regulatory PID controllers for both air and level – as
enhanced by utilising a feed forward term in the controller, best practised at Escondida, Chile; however detailed attention
to address upstream disturbance. While easily done in our must be paid to flotation chemical additions: pH modifiers,
control systems and applied by Mintek’s FloatStar (Schubert frother, activators, collectors and depressants.
et al, 1999), it is generally not widely practised and rarely
implemented in new plant commissioning. Schubert et al Reagent control
(1999) states: In Figure 30 and as observed in our industry, we see a multitude
Mintek has developed a level-control system, FloatStar, of different approaches to chemical addition and control,
for use in flotation circuits. Instead of controlling levels so like flotation pulp level control it is worth identifying
separately in each unit, the controller monitors levels some key aspects of flotation reagent control best practices.
throughout the circuit. It can thus take account of the global Chemical metering is an excellent application for coriolis flow
effects of any control actions. and density measurement – at the highest possible accuracy
Like the level measurement, an even tougher challenge ie 0.1 per cent measurement error, as practised at Antamina
is the engineering of the final control element (ie often dart and previously applied at Kidd. The author has recently
valves) to handle control in the operating envelope. Often been into one of XPS’s new mineral processing plants and
two dart valves are installed, therefore split range control observed that there were no measurements on the chemical
(shown in Figures 27 and 28) is a further enhancement as addition, there were minimal addition points (minimal staged
well as smart positioner feedback – allowing the controller addition), no automatic ratio to metal units, and finally a poor
to know that one, or both, actuators are saturated or closed. set-up for measuring/checking the actual chemical flow rates
Surge tank control and ultimately level set point optimisation going to the process. Consequently, and with also no flotation
will ultimately allow closed metallurgical control (Thwaites, airflows to adjust, the only variable left for the operator control
1983a, 1983b, 1986; Jin et al, 2013). was flotation pulp levels! It has been learnt over the years
In most flotation plants it makes sense to utilise froth (Thwaites, 1983a, 1983b, 1986) that air and levels generally
velocity cameras (typically providing measurement at a one move the flotation process outcome along the grade/recovery
FIG 25 – Flotation objective? Prevent flooding; maintain an accurate responsive level.

P Thwaites
FIG 27 – Flotation level control – best practices example illustrating operational excellence.
on the measurement. Often they do not and need the best

measurement possible, that is, coriolis!
Use of chemical metering pumps (Figure 31) allows very
fine and accurate – staged additions, but they must be
measured and automatically controlled for best practices;
and they should pump into (low) pressurised lines, allowing
instantaneous update/changes inflow at the discharge point
and instantaneous shutoff. The discharge point should be
open, after the back pressure injector valve, and above a
funnel, allowing easy visualisation and measurement for
the field operators – rarely practised. Best practices flotation
reagent control (Figure 31) can then be summarised as:
•• basic measurement of chemical – coriolis metre gives the
FIG 28 – Feed forward and split-range control for two flotation darts. best accuracy and measures the density
•• speciality collectors (such as AEROPHINE 3418A –
curve, while collector chemical changes move the process
Thwaites, 1983a, 1983b, 1986) allow much more precise
outcome onto different grade/recovery curves. control of flotation, along the grade/recovery to optimal
Chemicals are one of the (more) expensive consumables, net smelter returns
and as such should have the highest precision and accuracy •• use a local PID control loop for flow control
FIG 29 – Froth velocity control (camera) controlling pulp level and flotation air set points.

FIG 30 – Reagent dosing and measurements. Are you using coriolis?
FIG 31 – Flotation reagent control – best practices, operational performance excellence example.
•• cascade feed tonnage and automatically control the More accurate flow measurements will also provide much
reagent set point in g/t ratios better data for the exciting area of multivariate analyses – as
•• upgrade the cascade ratio to grams per metal unit discussed later in this paper.
(GMU) and automate this ratio addition – updating
every 5–15 min (while noting that this is still feed Thickener control
forward control) Thickener controls are not easy and are rarely set-up
properly following new process/plant commissioning and
•• enhance the metallurgical control with strategies for
can lead to overloading, failure (Villanueva and Garcia,
start-up and shutdowns and short stoppages (such as
2013) and unexpected, costly downtime. Furthermore,
rodding etc)
key measurements have been challenging, or not even
•• enhance metallurgical control with feedback trim based implemented. Cortés and Cerda (2010) reported:
upon actual mass pull, or recovery update, or froth
The Concentrator of División Codelco Norte, Chile has nine
quality (from froth flotation cameras)
tailings thickeners, which are used to thicken the tailings
•• best practices flotation reagent control will lead to and reclaim water from them. Several control strategies,
‘Optimum Metallurgy’ and can save up to 25 per cent in including expert control, have been implemented in the
reagent costs (Thwaites, 1983a, 1983b, 1986) thickener, but the results were not satisfactory, therefore the
Much more attention needs to be paid to the chemical thickening process continued to run in manual operation.
additions in our mineral processing plants (see also Shean Are your thickeners in automatic control? Do you add
and Cilliers, 2011; Hodouin, 2011). flocculant based upon mass flow into the thickener? Did you

P Thwaites
know that E+H coriolis metres on flocculant measurement can (2005) identified a key need for milling operations, following
also identify lower viscosity, identifying when you may be a concentrator benchmarking study:
pumping just water in your reagent stream? Do you operate The first need is to solve instrumentation problems by
the thickener on bed mass measurement and do you maintain means of predictive maintenance and fault detection. The
a basic, consistent bed level with minimal loss of mineral use of ‘hard technologies’ such as smart instrumentation
to the overflow? Do you measure the quality (turbidity) of
Fieldbuses and ‘soft’ technologies, such as multivariate
the overflow? Do you need to put a clarifier to capture pay
online analysis shows a great potential on this topic.
mineral loss on your concentrate thickeners?
E+H is but one supplier offering a comprehensive
Figure 32 illustrates the key measurements available for
instrument support and documentation system which could
thickeners. Bed pressure can be measured, allowing (real-
time) calculation of inventory when the right instrument is be more widely used (and not just for their instrumentation).
correctly located in the bottom cone. This should be used Figure 34 illustrates very important asset monitoring
to control the underflow pumps (speed control) and their capabilities in DeltaV (Emerson) and ABB control systems.
removal of material from the thickener. Bed level can usually Control loop diagnostic conditions monitored can include the
be measured (by a sonar-based instrument) on a stable following, as documented by the current ABB control system:
operation – one under control, and should be used to trim the •• final control element (FCE) stiction, size and leakage
chemical addition ratio; however, bed level is still difficult to
measure especially on an upset process! Resulting underflow •• excessive FCE action
density should also be measured, as well as underflow flow. •• tuning problem
Overflow clarity is a good indication/warning of process •• loop oscillations
problems, and mineral loss! Automatic rake arm lifting
•• external disturbances
should be controlled to torque, and often is.
•• loop nonlinearity
Cortés and Cerda (2010) concluded:
•• data reliability
The underflow solid average per cent has been reduced by
22.3% in variability but a similar mean value. … •• insufficient travel
Implementation of this control strategy in all the thickeners •• no response to signal change
of the concentrator will increase the underflow solid •• noisy or unstable output
average per cent by 0.5% in the thickener which also
•• sluggish response
represents 2.8% of water used by the concentrator for
processing per day for each thickener. •• valve body/seat wear.
Yet we see very little actual usage of these important
ASSET AND PERFORMANCE MONITORING capabilities, unlike the extensive process monitoring that
Asset and performance monitoring is crucial if we are going we (Glencore, Vale, Anglo, Barrick etc) typically do with the
to fully realise the optimum from the big investments which widespread and well-known OSIsoft PI system (Figures 35
are made in mining and mineral processing plants and and 36).
operations. At the instrument level this is far more possible Most readers can relate to the ‘Check Engine’ light found
now, than it was three decades ago. Figure 33 – Mujica et al on all modern vehicles! It is invaluable, though we would
FIG 32 – E+H: typical positioning of a thickener’s critical parameter measurement devices.

FIG 33 – Plant asset monitoring (at the instrument level) – E+H and AMIRA P893 Study.
Outside of the control system, there are few options and

we presently see the choice as either Honeywell’s Control
Performance Monitor (Honeywell, 2010) (Figure 37) or
ExperTune’s (now Metso) Triage or some homemade
alternative. While (significant) further investment is required
for any current product, there ultimately is a business case
to the operation in continuously monitoring the minute by
minute performance of ALL the plant’s critical control loops,
as we monitor our processes using OSIsoft’s PI systems.
For example: ‘CPM can enable a sustainable workflow that
enables tracking of maintenance efforts as well as archiving
maintenance history on each loop’. See Honeywell, Jin et al
(2013) and Ruel (2010) for further information and discussion.
Many years ago, Sandoz (2003) presented Perceptive
Engineering’s important view/vision of asset monitoring
(as seen in Figure 38) that integrated condition monitoring
(and failure/fault diagnostics) with APC. In Figure 38,
FIG 34 – Control system diagnostics, eg ABB, DeltaV, Yokagawa etc. multivariate statistical tools such as Principle Component
Analysis/Projection to Latent Structures, or Partial Least
all appreciate the correct fault message than just the ‘Check Squares (PCA/PLS) and their models, are run in real-time
Engine’ light. as early monitoring condition monitors, predicting key
Monitoring of the control assets, whether the plant has measurements, comparing them to the actual measurements
100, 600 or over 1000 control loops is extremely important, and squaring the ensuing errors to identify any break in
especially if we want to immediately deal with poor controller expected performance.
performance, or know as soon as a key control loop degrades – Perceptive Engineering share the opinion of Joseph Shunta,
requiring retuning, instrument maintenance or gain scheduling. that ‘Achieving manufacturing excellence requires the
FIG 35 – Modern data acquisition and control – Emerson’s DeltaV and OSIsoft’s PI system.

P Thwaites
joint use of Statistical Process Control, Advanced Process Multivariate analysis is the preferred approach as it takes
Control, Optimisation (and experienced engineering into account the interactions between multiple variables,
judgment). (Lovett, 2004) including input and process variables …
FIG 36 – Process and plant monitoring by OSIsoft PI (plant historian) systems.
FIG 37 – Control performance monitor (CPM) by Honeywell/Matrikon.
FIG 38 – Perceptive Engineering’s focus/vision (2003) – integrated condition monitoring and advanced process control (APC).

It allows for further learning about the process and the flotation circuit. This initial, early study ‘was successful in
various interactions, utilising two primary multivariate identifying three distinct clusters signifying different modes
methods: of process operation, or possibly, identifying different ores
•• Principle Components Analysis (PCA) – for fault being processed’.
detection: provides a concise overview of a data set. It is Figures 39 and 40 show the score plots for grinding and
powerful for recognising patterns in data: outliers, trends, flotation:
groups, relationships etc. PCA makes it possible to detect Score plots describe the relationship between all samples
abnormal operations resulting from both measures and in the data, while samples clustered together have similar
unmeasured disturbances. variable properties, ie each cluster represents specific
•• Projections to Latent Structures (PLS) – for quality combinations of variable values.
prediction: aiming to establish relationships between input In Figure 39:
and output variables thereby developing predictive models
Cluster 1 (green) represents samples between January
of a process. (Emerson Process Management, 2015; and September 2015; Cluster 2 (purple) represents
Blevins, Moro and Wojewodka, 2007) samples between October 2015 and March 2016; and
XPS have used such systems in its pyrometallurgical Cluster 3 (brown) represents samples between March
operations very robustly and very successfully (Thwaites and June 2016. Each of these clusters represents distinct
et al, 2004; Nelson et al, 2007) now for many years. They are process behaviours and future activities can be potentially
very robust and can cope well with instrument signal losses. classified into one of these.
Unfortunately XPS have seen fewer applications in its mineral In Figure 40:
processing plants, although interesting (offline) prior work
The score plot reveals three clusters representing the
was done by Hodouin et al (1993):
following: Cluster 1 (green) represents periods of operation
On the industrial data which were processed, PCA and PLS with mean total bulk flotation cell air flow at average values
were able to disclose regions of consistent operations and (approximately 370 m3/h) for the data range; Cluster 2
times at which changes occurred for both the grinding and (purple) represents the periods of operation with lower than
flotation circuit. It was much more difficult to pick out such mean total bulk cell airflow ie for low air flows; Cluster 3
regions and changes by inspecting the raw data. … (brown) represents the periods of operation with higher
Predictive models for both the grinding and flotation mean air flows. (Lawal and McEwan, 2016).
sections of the plant were developed using PLS. The observed In the present market, Perceptive Engineering (UK),
relationships were, for the most part, quite consistent with ProSensus (Canada), Emerson Process Management (USA),
prior knowledge about their behaviours, despite the high and Contac (Chile) all have software products to further
level of noise, a large number of missing input variables, and support interested customers in this fascinating area.
some raw data inconsistencies. These models could be used See Figures 41 and 42. In fact Sandoz (2003) and Lovett’s
to gain improved understanding or to infer infrequently (2004) visions (beautifully captured and presented in
measured variables. Figure 38) materialised (Zamora et al, 2010) – coming from the
Recently, through Perceptive Engineering, XPS have AMIRA P893 benchmarking study – a comprehensive control
explored the PCA technique on OSIsoft PI process (30 minutes) monitoring centre, on top of the Plant and local control rooms
data covering a year and a half for a grinding and primary but in the city. One of the earliest of these, initially focused on
FIG 39 – Score plot for grinding stage Principle Component Analysis (PCA), showing data with similar variable properties.

P Thwaites
FIG 40 – Principle Component Analysis (PCA) scores for flotation circuit showing clusters, with each cluster representing a unique class.
the Codelco plants was in Santiago, Chile. The original focus A new concept in process automation is proposed through
was automation deployment and support covering multiple the addition of a new control loop based on performance
sites and utilising key Honeywell resources – more available analysis methodology which is intended to feedback process
and automation operation information to system and process
in the city environment. Once established, the focus expanded
specialists, allowing for continuous manufacturing process
to also include asset monitoring:
improvement and sustainable productions benefits through
•• providing access/network to many service providers effectual control strategies.
for remote servicing/configuration and on demand A remote support centre can play a key role in the delivery of
analysis and diagnostics such services..., as an enabler for plant operational practices
change. (Zamora et al, 2010)
•• providing a Core Honeywell team ensuring success
In Australia (Figure 43) it was noted a similar approach
•• adding real-time expert monitoring for fixed equipment.
from Rio Tinto:
Considering that process automation is a key enabler to 13 March, 2014 Rio Tinto has unveiled the latest component
business performance improvement, sustaining effective of its Mine of the Future program – the Processing Centre
performance benefits over time becomes essential. of Excellence. Based in Brisbane, Rio says this ‘is a world
FIG 41 – Kairos Mining (Chile) – a comprehensive approach to sustainable benefits in automation.

FIG 42 – Honeywell remote support centre – Kairos Mining Santiago, Chile.
first (?), state-of-the-art facility that enhances monitoring the ‘process excellence centre’ opened yesterday had already
and operational performance by examining in real time more than paid itself back.
processing data from several Rio Tinto operations spread Unfortunately, this second example has been re-evaluated
across the globe’. Operated by a team in Brisbane, that will and a new direction has been set for the infrastructure.
provide processing solutions and initiatives to mine sites in
Mongolia (at Oyu Tolgoi), the US (at Kennecott), and across
Australia (at five different sites). A massive interactive
CONCLUSIONS AND RECOMMENDATIONS
screen will show, and analyse, technical data in real time, The mining industry uses many types of mineral and
‘allowing processing improvements to be immediately metallurgical plants to produce saleable product from ore
introduced and operational performance to be optimised,’ the mined. Plant design history has left current operations
miner said. Early trials have already led to improvements with a mixture of manual operation and various forms of
such as adjusting the flotation process for gold and copper automated process controls. Consequently, we typically
recovery at Oyu Tolgoi in Mongolia. Speaking at the launch see high variability in the continuous operations together
today, Rio CEO Sam Walsh said ‘the Processing Centre with a shortfall in the attainment of full capacity, or higher
of Excellence (PEC) marks another important step in the utilisation of consumables.
rollout of our exciting Mine of the Future program’. At a level of best practice, ‘operational performance
Shortly thereafter, (Chambers, 2014) reported: excellence’ focuses on process control, using automation and
control systems to deliver process optimisation. This more
Big data saving us $90M: Rio boss. sophisticated delivery is a great deal more difficult than the
MASSES of data generated by Rio Tinto’s plants, drill rigs first stage of equipment selection/installation. It includes
and automated trucks have become a key part of the mining the appropriate selection of the right instrumentation,
giant’s productivity drive, with a new Brisbane process control system, key process knowledge, individuals with a
centre already reducing costs by $US80 million ($90M). solid control engineering background/experience, and the
Rio chief Sam Walsh said analysis of so-called big data at essential backing/support of the operations management
FIG 43 – Control centre – processing centre of excellence (Rio Tinto).

P Thwaites
team together leading to higher value delivery. Robust REFERENCES

solutions can be realised, considerably minimising process Akzo Nobel, 2014. Flotation Depramin depressants website [online].
variation, thus leading to process optimisation. This approach Available from: <https://www.akzonobel.com/cs/products/
results in an easier, efficient and safer process while providing depramin/>.
considerable returns for the plant owners. AMR Research / Gartner, 2010. Survey.
For those working in mineral processing plants, those Blevins, T, McMillan, G, Wojsznis, W and Brown, M, 2003. Advanced
working on new process and plant design and those Control Unleashed (ISA Press: North Carolina).
responsible for applying technology, we can, we MUST do
Blevins, T, Moro, P and Wojewodka, B, 2007. Coupling process control
MUCH MORE and therefore become more efficient. systems and process analytics to improve batch operations,
The process automation area remains a major opportunity presentation at 2007 Emerson Global Users Exchange (Emerson,
for improved efficiency, improved milling, minimising 2007 Exchange).
tailings losses and other costs. Multivariate statistics and Bouchard, J, Émond, M, Monnapula, R and Molelekeng, N, 2010.
process control should be further applied to help us run to Asset optimisation through process control at Xstrata Alloys –
our process constraints, in effect maximising the process Eland Platinum – ‘optimal doesn’t always rhyme with capital’,
efficiency and process utilisation. However, we need more in Proceedings 42nd Canadian Mineral Processors, pp 419–438
(Canadian Institute of Mining, Metallurgy and Petroleum:
champions, control engineers in our plants and operations’
Littleton).
management support.
Boudreau, M and McMillan, G, 2007. New Directions in Bioprocess
Would you run a business without a financial accountant? Modelling and Control (The Instrumentation, Systems, and
Let’s see more control engineers and more process control, Automation Society: North Carolina), ISBN: 978-1-55617-905-1.
using automation and the products from our suppliers in our
Chambers, M, 2014. Big data saving us $90M: Rio Boss, The Australian,
minerals industry. March 14.
Cortés, G and Cerda, J, 2010. Thickeners multivariable predictive
The author’s recommendations control: Division Codelco Norte, in Proceedings Second International
•• Instrumentations and instrumentation solutions: Congress of Automation in the Mining Industry (Automining 2010),
•• make them work for you and use more of the great p 14 (Gecamin: Santiago).
instrumentation currently available. Dictionary.com, 2014. Automation [online]. Available from: <http://
dictionary.reference.com/browse/automation>.
•• Control systems:
Emerson, 2007. DeltaV embedded advanced control, Emerson Global
•• push your control system to keep delivering for you,
Users Exchange, September, Grapevine, Texas, USA.
reducing variability
Emerson Process Management, 2015. DeltaV Distributed Control
•• use more feed forward, cascade controls and ensure System, Batch Analytics, product data sheet, December 2015
controllers are tuned, and functioning at the regulatory [online]. Available from: <www.emersonprocess.com>.
level
Émond, M and Vynogradova, S, 2010. XPS GrindTM at work in Eland
•• seek input from your supplier and ensure you utilise Platinum South Africa, XPS Bulletin, (5):5.
what the system can do, with the correct, available Fiske, T, 2005. ARC Advisory Group – advanced process control: the
software right stuff, ARC Insights #2005–49MP, Nov.
•• automate tedious and repetitive, known tasks Flintoff, B C and Mular, A L (eds), 1992. A Practical Guide to Process
•• focus on optimisation and seek better business Controls in the Minerals Industry (University of British Columbia:
performance. Vancouver), ISBN 0-88865-517-7.
•• Employ one or more control engineers to complement Fragomeni, D, Hoffman, M, Kelly, A, Yu, S and Lotter, N, 2006.
Flotation mini pilot plant experience at Falconbridge Ltd, in
your operations and instrumentation teams
Proceedings Mineral Process Modelling, Simulation and Control
•• Use asset, performance monitoring tools to be more International Conference, pp 329–355 (Laurentian University:
proactive Ontario).
•• Analyse, act, use multivariate statistics and ensure data Gillis, J, 2000. Flotation control review, internal presentation to Mill
becomes knowledge for business ($$) actions – always. Staff, Kidd Creek Mill, Timmins, Ontario, 10 Nov, 2000.
For further insight, please follow up with the references Hodouin, D, 2011. Methods for automatic control, observation and
cited, especially the texts by Hollender (2010), Blevins et al optimization in mineral processing plants, Journal for Process
Control, 21:211–225.
(2003), Flintoff and Mular (1992) and Boudreau and McMillan
(2007). Hodouin, D, MacGregor, J, Hou, M and Franklin, M, 1993.
Multivariate statistical analysis of mineral processing plant data,
CIM Bulletin, Nov–Dec:23–34.
ACKNOWLEDGEMENTS Holcroft, G and Carabin, M, 2012. Game changing plasma based
The author would particularly like to acknowledge the solutions – APT Applications, presentation to XPS [online].
50 process control engineers whom he has, and had, Available from: <www.pryrogenesis.com>.
the privilege of working with. Each have had their own Hollender, M, 2010. Collaborative Process Automation Systems (The
interesting, unique backgrounds and extensive contributions, Instrumentation, Systems, and Automation Society: North
several of which have been discussed in this paper. In Carolina), ISBN: 978-1-936007-10-3.
addition to the author’s senior and core group members, Honeywell Process Solutions, 2010. Advanced Solutions Control
he would like to thank: A Hyde, P Nelson, R Bose, E Grant, Performance Monitor overview (presentation) Aug 27.
N Adnan and K Ahmed. The author also acknowledges our
Jin, X, Rioux, D, Graffi, D and MacPherson, G, 2013. From stabilization
consultants G Marrs, S Nino, L Ryan and M McEwan, and to optimization; process control in the flotation process at Xstrata
other colleagues including D Fragomeni (XPS), R Løkling Nickel’s Strathcona Mill, in Proceedings 45th Canadian Mineral
(Nikkelverk) and Glencore for permission to share this work Processors, pp 387–398 (Canadian Institute of Mining, Metallurgy
and publish this paper. and Petroleum: Littleton).

Jonas, R, 2008. Advanced control for mineral processing: better than Ruel, M, 2014. Advanced control decision tree, 46th Canadian Mineral
expert systems, in Proceedings 40th Canadian Mineral Processors, Processors.
pp 421–443 (Canadian Institute of Mining, Metallurgy and
Sandoz, D J, 2003. Multivariate statistical process analysis, monitoring
Petroleum: Littleton).
and control, slide 4, MonitorMV Course Introduction (Perceptive
Jones, R, 1994. What is process control? An introductory guided Engineering Ltd: UK).
tour, plant-wide control and real-time process databases (PMIS),
Sandoz, D J, 2006. The exploitation of model predictive control in
process control application notes – for the resource industries,
the mineral processing industry, Laurentian Advanced Control
numbers 19 and 23.
Forum, presentation to Mineral Process Modelling, Simulation
Jones, R and Morrell, R, 1996. Process Management Information and Control International Conference, Laurentian University,
System (PMIS) concepts and architecture, Noranda Technology Ontario.
Centre, (internal) process control application notes plant-wide
Schubert, J, Valenta, M, Henning, R and Gebbie, I, 1999. Improved
control and real-time process databases.
flotation performance at Karee Platinum Mine through better
Kidd Process®, n/d. stainless steel cathode blank, Canadian Patent level control, Journal of the SAIMM, Jan/Feb:35–39.
#1263627.
Shean, B and Cilliers, J, 2011. A review of froth flotation control,
Konigsmann, K, 1992. Foreword, in A Practical Guide to Process International Journal of Mineral Processing, 100:57–71.
Controls in the Minerals Industry, pp i–ii (University of British
Suiches, M, Leroux, D, Dechert, C and Trusiak, A, 1999. Model
Columbia: Vancouver). ISBN 0-88865-517-7.
predictive control of the flotation process, in Proceedings Canadian
Laezza, J, Box, R and Scott, J D, 1990. The Kidd Creek copper refinery, Mineral Processors, pp 425–441, (Canadian Institute of Mining,
in Proceedings International Symposium on Electrometallurgical Plant Metallurgy and Petroleum: Littleton).
Practice, pp 3–19.
Thornton, A J, Prinsloo, A and Tewu, R, 2009. Evolution of semi-
Lawal, N and McEwan, M, 2016. Perceptive audit – strategies for autogenous grinding mill control at Century Zinc, in Proceedings
advanced monitoring, control and optimisation, PEP5838_XPS_ Tenth Mill Operators’ Conference, pp 329–335 (The Australasian
Audit v2.0 internal report, 03 Nov 2016, Perceptive Engineering Institute of Mining and Metallurgy: Melbourne).
Ltd: Cheshire.
Thwaites, P, 1983a. Continued development of copper flotation
Lovett, D, 2004. Using MonitorMV to enhance Six Sigma initiatives control at the Kidd Creek concentrator, in Proceedings 15th
for continuous processes (unpublished), Perceptive Engineering Annual Meeting of the Canadian Mineral Processors Division of
Ltd: Cheshire. CIM, pp 287–316 (Canadian Institute of Mining, Metallurgy and
Petroleum: Littleton).
Maron, R, O’Keefe, C and Rothman, P, 2010. Sonar flow monitoring:
first five years of experience in the mineral processing industry Thwaites, P, 1983b. Continued development of copper flotation
and future directions, in Proceedings Second International Congress control at the Kidd Creek concentrator, CIM Bulletin, Dec:41–46.
of Automining, p 94 (Gecamin: Santiago).
Thwaites, P, 1986. Computer controlled reagent addition strategy and
McKee, D, 1999. P9LProcess control – the optimisation of mineral hardware for the control of a copper and zinc flotation process,
processes by modelling and simulation 1996 to 1999, final report, in Proceedings Third Meeting of the Canadian Industrial Computer
vol 1: process control, pp 1–43, (AMIRA). Society, pp 56-1–56-8 (CCICS: Ottawa).
Meech, J, 2006. The evolution of intelligent systems in the mining Thwaites, P, 1993. Economic benefits of supervisory process control
industry, in Proceedings Mineral Process Modelling, Simulation and at Kidd Creek, Noranda Technology Conference: Applying
Control International Conference, pp 1–33 (Laurentian University: Technology to Achieve a Competitive Advantage.
Ontario).
Thwaites, P, 2007. Process control in metallurgical plants – from
Mujica, L, Yacher, L, Coetzer, L and Cipriano, Z, 2005. A review an Xstrata perspective, Plenary (Industrial) Lecture, 12th IFAC
and benchmark of process control and automation at copper Symposium on Automation in Mining, Mineral and Metal Processing
concentrators, AMIRA Project P893. – IFACMMM’07, Annual Reviews in Control, 31(2007):221–239.
Nelson, P, Hyde, A, McEwan, M and Sandoz, D, 2007. Integrity Thwaites, P, Sandoz, D, McEwan, M, Kak, D, Gillis, J, Nelson, P
monitoring of Xstrata Copper’s Kidd metallurgical division and Norberg, P O, 2004. System and method for the furnace
Mitsubishi 3-line furnaces using multivariate methods, in monitoring and control, Canadian Patent #2469975, issued 2013-
Proceedings Sixth Copper Conference (Cu2007), vol VII, pp 229– 09-17, Filed 2004-06-04, Examined 2009-05-01, Opened for Public
239 (Canadian Institute of Mining, Metallurgy and Petroleum: Inspection 2005-12-04.
Littleton).
Villaneuva, J and Garcia, D, 2013. Antapaccay–Tintaya Expansion,
O’Keefe, C, Felix, J, Peacock, R, Huysamen, T and Thwaites, P, 2012. Tenth International Mineral Processing Conference (Procemin).
The impact of entrained air and enhanced flow measurements
Yutronic, I and Toro, R, 2012. The Collahuasi solution, in Canadian
at Eland platinum concentrator, in Proceedings The South African
Mineral Processors 2012 (Canadian Institute of Mining, Metallurgy
Institute of Mining and Metallurgy – Platinum 2012, pp 981–
and Petroleum: Littleton); and Second International Congress of
1001 (The South African Institute of Mining and Metallurgy:
Automation in the Mining Industry (Automining 2010) (Gecamin:
Marshalltown).
Santiago).
Ruel, M, 2006. Fundamentals of process control, course manual (Top
Zamora, C, Schifferli, G, Castelli, L and Bonomelli, A, 2010. Kairos
Control).
Mining: A comprehensive approach to sustained long-term
Ruel, M, 2010. Control system performance assessment – best benefits in copper processing plant automation in Canadian
practices, in Proceedings Second International Congress of Automation Mineral Processors 2010, pp 439–459 (Canadian Institute of Mining,
in the Mining Industry (Automining 2010), p 54 (Gecamin: Santiago). Metallurgy and Petroleum: Littleton).
Ruel, M, 2012. Leveraging control technology, BBA Presentation.

Contents
Unlocking processing potential by

empowering our operators
X Li1, M S Powell2 and W McKeague3
ABSTRACT
This paper explores the importance of operators in the overall efficiency of mineral
processing plants. With the potential of future control configurations becoming
more remote, it is time to review the implementation of control and utilisation of
operators. This paper is based on a preliminary review of operators and control room
operations and clearly shows that, in the minerals industry, operations are typically
run in a reactive mode (with a focus on ensuring equipment operation), rather than in
a proactive mode (with a focus on optimising the process performance). The primary
reason for this misguided approach is the short-sighted thinking applied to the
challenges of designing technology and organisational tools capable of addressing
the essential operational needs: an approach that inevitably results in a disconnect
between people and technology. Contrary to the popular approach of simply
replacing operators with expert systems, the thesis of this paper is that with the shift to
more complex circuits, and its higher demands on processing capability and control,
operations will be better served by upskilling their operators. Experience in mineral
processing and other industries indicates that expert systems cope well with small
changes but are typically unable to respond to major changes in feed material and this
is where the expert operator becomes crucial. Providing operators with ongoing high-
level training in process understanding, and restructuring how they interact with the
plant and interface with the control system is the approach being proposed here –
an approach surprisingly under-utilised and often absent in the minerals industry
today. An improved style of operator interface designed to empower the operator is
presented and discussed.
INTRODUCTION
Undoubtedly the future of the minerals industry will entail change, but in what form
and to what extent remains unknown. However, current experience tells us that:
•• it will be more complex than ever before
•• the rate of change will be faster
•• it will likely deviate from a simple linear trajectory, bringing unexpected and
potentially disruptive impacts.
One of the many benefits of future-casting your operations is being better prepared,
resulting in faster and more confident decision-making. This significant advantage
was demonstrated by Royal Dutch/Shell’s growth in the 1970s and again in the 1980s.
Within mineral processing, Munro and Tilyard’s (2009) ‘Back to the future’ paper
paints a rich picture of the mineral processing engineering practices of 40 years ago
and today, and flags some key areas of concern.
Attempting to look ahead to the future of site operations, there are so many aspects
that we would expect to be different; however, obtaining some useful level of
perspective is challenging. To provide some structure to what the future for mineral
1. Formerly Julius Kruttschnitt Mineral Research processing plants may hold, we propose to explore three primary segments:
Centre, Sustainable Minerals Institute, 1. people and organisational structure
The University of Queensland,
Indooroopilly Qld 4068. 2. processes and process knowledge
2. FAusIMM, Professorial Research Fellow, 3. automation and related technology.
Chair in Sustainable Comminution, Julius Let us ask, what has been the key development philosophy for the last 40 years
Kruttschnitt Mineral Research Centre, in each of these areas, and how well are we prepared for today’s challenges and
Sustainable Minerals Institute, The University
potentially for those of the next 20 to 40 years?
of Queensland, Indooroopilly Qld 4068.
Email: malcolm.powell@uq.edu.au This is a big question and too broad to address comprehensively in a single paper,
3. MAusIMM, Director, Adaptive Solutions Pty we have therefore chosen to focus on the one key link in this chain – the operator.
Ltd. Email: mckeague@me.com. Formerly Working on the front line of plant operation, operators are the whisperers of process
JKTech, Indooroopilly Qld 4068. knowledge across and around a 24/7 plant operation; in fact, they possibly have
401
more impact on day-to-day profit than any other position increased downtime. Although no statistical data is available
on‑site. However, this essential role tends to be overlooked in in the minerals industry, it is estimated that the global
the current organisational structure. As such, this paper will process industry loses five per cent of annual production
focus on the plant operators, making a few observations and due to unscheduled downtime and suboptimal performance.
highlighting some findings from control field studies by Li Almost 80 per cent of these losses were preventable and
et al (2011) and Li, Powell and Horberry (2012), to support approximately 40 per cent were primarily from so-called
the conclusion that expert operators with more process ‘human operator error’ (Woll, Hill and Polsonetti, 2002).
knowledge, and supported by – rather than replaced by – These findings lead to the conclusion that the inability
expert systems, are key to effective future plant operations. of control room operators to take up new technologies is a
key ‘barrier’ to maximising the potential of an optimal plant
Remote control/support centre – a solution for the future performance. To highlight and better understand this issue, it
It is not difficult to envisage some inevitable changes is essential that an evaluation of the current status of control
and challenges to the current state of mining operations room operations, with particular attention to identifying the
worldwide. The location of available mining deposits underlying ‘barriers’ that constrain operator control capacity
will become more remote. Mines will need to go deeper and the issues arising from the transition to remote operation,
underground, and orebodies will have a lower grade and be undertaken.
become more challenging to treat. At the same time, operations
will be constrained by energy consumption targets, the need PROJECT INTRODUCTION
to minimise their environmental impact and by the shortage
Three plant site studies were undertaken in Australia during
of a skilled workforce arising from the remoteness of these
the period 2009 to 2011: one in copper and two in gold
operations. Utilising small teams of experts in a centralised
processing. These three plants were selected following an
control/support centre to operate equipment and facilities
extensive prescreening exercise by the Julius Kruttschnitt
remotely is regarded as a step change in thinking to address
Mineral Research Centre and industry. The sites were
some of these challenges. Many mining companies are already
considered broadly representative of typical Australian
well advanced toward this type of remote operation.
minerals processing plants, having installed a range of process
In an attempt to navigate this transition, most solutions control systems (including expert control systems) from
immediately lean toward upgrading technologies; building mainstream suppliers, and were utilising various technologies
more complex circuits, more powerful equipment and in an effort to improve metallurgical performance.
installing more sensors, with more sophisticated software
to computerise the plant operation. However, any potential Method
solutions which deal with the organisational requirements or
To maximise the success of data collection, multiple methods
the management challenges of such remote operation – have
were used in these field studies: control room observations,
yet to be proposed.
interviews, questionnaires and document reviews. Previous
Clearly remote control or support centres will become central control room operation studies in other industries (Burns
to operating strategy. However, strategies for effectively and Hajdukiewicz, 2004; Burns et al, 2008; Jamieson, 2003;
supporting operators working in such new environments Li, 2008) indicated that the quality of human operator
have yet to be introduced into the discussion. Indeed, over the control is significantly influenced by how the operator
last 20 years, the minerals industry remains largely silent on perceives and interprets the surrounding information in their
the topic of control room operations. As a result, the essential decision-making. Therefore the study of the interactions and
role they play and the needs of control room operators remain communications between operators and information ‘in’ and
largely unexplored. ‘around’ the control room in this study was the primary focus.
Are humans the bottleneck? Observation and interview

At the other end of this spectrum, there is an increasing The primary methods were direct observations and a short
awareness that poor people performance may be the in situ interview based on the crucial decision method
bottleneck to effective control. A statement from Munro and developed by Klein, Calderwood and MacGregor (1989).
Tilyard (2009), summarises the process in the last 20 years, The key objective was to identify typical critical plant
and well reflects such a concern: disturbances and uncover operator decision processes in
The industry’s strength has been in finding technical or dealing with such events. These processes included how
hardware solutions, while its weakness has been at the the operators detected and interpreted available cues, each
people end of the business in maximising and consolidating of their decision points, and the plausible goals and typical
the gains from the technologies. operational strategies involved. In total, 24 operators were
Indeed, the industry’s continual investment in technology observed for approximately 130 hours, and most of them were
has not resulted in the sustainable performance improvement subsequently interviewed. The operators involved ranged
as expected. Some key performance metrics, such as from new trainee operators, who had worked in the plant
operating times and plant start-up performance, remain for less than eight months, through to experienced operators
static or have even declined (Munro and Tilyard, 2009). A with more than 30 years’ experience in this industry.
global process industry survey indicated that 60 per cent
of control loops installed worldwide are underperforming Questionnaires
(Thwaites, 2008). Mineral processing is not alone in this Based on recent industry alarm standards (EEMUA, 2007),
regard. A global grinding mill circuit survey in 2009 showed questionnaires were designed to capture the operators’
that only a few plant sites (eight per cent) were satisfied opinions regarding the usability of existing alarms and
with the performance of their installed control loops (Wei the human-machine interface (HMI) in their workplaces,
and Craig, 2009). Poor performance of these control loops including other ergonomic and organisational factors
inevitably results in unsatisfactory industry-wide production; affecting their jobs. Altogether, eight operators completed the
namely, high variability in recovery, reduced throughput and questionnaire on alarm performance at two plants. No alarm

Unlocking processing potential by empowering our operators
questionnaires were completed at the third plant because with other operational units is heavily dependent on radio
operators were only occasionally in the control room. and telephones. In the control area of the concentrator, two
dedicated control room operators monitor three plants in the
Interview with metallurgists operation, and the flotation operators have their workstations
Informal interviews with plant metallurgists were inside the plant. This control room is more like a modern
also conducted to provide an alternative perspective control centre and the addition of some improved ergonomic
to understanding operators’ performance in handling factors compared with other two control rooms (A and B); for
disturbances observed in these field studies. Metallurgists example, the height of the operator console can be adjusted
were able to comment on the impact of these disturbances on to allow operators to stand at work, and the six monitor
plant control objectives and outline the appropriate control screens on each operator console are designed to fit within the
strategies that should have been taken in those instances. In maximum horizontal zone of human perception (Hollifield
total, ten metallurgists at three plant sites were interviewed. et al, 2008).
Various process control systems and technologies are
Documentation review installed in all three of these control rooms, including
The information about plant operations was mainly obtained distributed control systems, expert control systems, online
through plant tours and a study of plant flow sheets and stream analysis tools, froth image analysis tools and CCTVs.
operator training materials. The documentation study was Information displayed on operator monitors comes from
supplemented by follow-up clarification questions with the above resources. The operators typically maintain their
operators and metallurgists. situational awareness of the plant status via:
Most information collected was in a descriptive and •• various overview displays
qualitative format, which was in line with the methodology •• an alarm browser
of field studies. Aggregate data analysis from the above
•• radio and telephone communications with mill
resources were used to gain a fundamental knowledge of
operations and metallurgists.
mineral processing control and a basic understanding of
operational tasks, particularly with respect to the cognitive Figure 2 shows some typical monitor displays used in the
demands required to successfully perform these tasks. This control rooms.
paper presents an analysis of issues and factors affecting the Although the automation capacity varies among these
quality of control room operation identified at these three three plants, the three control rooms represent the central
plant sites. locations used to coordinate plant operations and to handle
plant disturbances. Operators were observed to be extremely
Control room environment busy when the plant is not stable. The control room operator
Figure 1 shows the three control rooms involved in this study. often acts as the director in guiding field operations and
The layout of the operator consoles and task allocation in these technicians to troubleshoot the problem, so their decisions
control rooms varied. Control room A (shown on the left) has and communication style must be clear and immediate. The
no dedicated control room operator. When occasionally in the following comments from operators well illustrate such
control room, operators monitor their section of the process situations:
on individual workstations. The control room B (shown in the Too many actions are happening in the process at once.
middle) has a dedicated control room operator monitoring the When the pressure is on, then it is on, with too much to
grinding and flotation processes via more than ten monitor watch, too many distractions.
screens distributed on the operator console. Control rooms A
and B are traditional control rooms located within the plant. When something goes wrong, there are a lot of sequences,
Metallurgists, mill operators and maintenance workers often areas and pages to go through to ensure you don’t miss
come in, so these control rooms are noisy and busy due to the anything.
surrounding machine noise, the presence of visitors and the
frequency of audible alarms. REACTIVE OPERATOR CONTROL
Control room C, or more aptly named ‘control centre’ Although operators in these three control rooms are equipped
C, is a large operation where the control room integrates with different technologies and are assigned to different tasks,
the functions of mine and mineral processing control. This their control behaviour typically fall into the same category:
centre best demonstrates remote operations control, the largely that of a reactive control mode. Most of the time
room is setup to one side of the plant and communication human operators preferred not to make any adjustments to
FIG 1 – Control rooms involved in the field study.

FIG 2 – Typical displays on operator consoles.
the control systems and chose to let them run independently. discussion of technical issues can be found in Li et al (2011)
They tended to respond only when things were going wrong and Li, Powell and Horberry (2012).
(for example when the mill was going to trip-out). In fact, often
their actions were too late, occurring only when the demand Limitations in the operator interface
for human control intervention had become overwhelming. The observed shortcomings of the operator interface that
A typical example of such behaviour was the supervision of contribute to human reactive control are summarised in
the semi-autogenous grinding (SAG) mill – the most dynamic Figure 3. A simplified human supervisor model (Li et al, 2006)
and critical area of grinding control. It was observed that involving detection, analysis (understanding and projection),
operators only responded when either the mill was going determination of possible actions, evaluation of the impact
to overload above a set load limit, or when the mill was of specific actions, were used to categorise the interface
approaching a trip-out point on power overload. This was limitations, which might lead to human control failures.
usually signaled by a ‘pop-up’ alarm message in the alarm Within this model:
browser (see alarm browser in Figure 2), a colour change in •• detection – takes the form of a general observation/scan
digital readings, or flashing graphics in the SAG mill overview of visual displays that helps controllers to recognise and
display (see overview display in Figure 2). Often these signals identify a disturbance
or alarms were too late to prevent the mill from overloading, •• evaluation – is a more focused purposeful observation,
with the result that throughput had to be reduced for an geared to assessing whether stable production can be
extended period. In the worst-case observed, the mill tripped restored after an intervention.
out altogether due to overloading.
For each of the three stages in Figure 3, a brief analysis of
Moreover, the objective of control optimisation (managing sample findings is provided.
a stable and efficient production flow), was not translated
into operator behaviour, as operators were observed to be Detection
mainly engaged in incident prevention. Most of the time, Although it was not possible to quantify findings in this
operators were busy coordinating with the field operators or study, it was evident that late and inaccurate detection was
technicians to ensure that the equipment or instruments were the major cause of human control failures. Poor alarm support
working properly to keep production on target. Answers to was identified as the primary reason, since the alarm was the
the questionnaires also indicated that operators interpreted primary tool directing operator attention. The common alarm
their tasks or control priorities to be ‘equipment-focused’ problems reported by operators included: too many alarms,
only, ensuring that various pieces of large equipment excessive nuisance alarms, inadequate prioritisation of alarms
were prioritised and kept running for production. Process and incorrect alarm settings.
optimisation and stabilisation was not their focus, indeed, It was noted that operators were seriously disrupted by
process optimisation was generally considered the job of the alarm overload and excessive nuisance alarm triggers in the
process control system itself. three plants. During an interview, one operator outlined the
latest 50 alarm messages he received, with only one alarm
Where human control is lost being critical (a pump temperature). If the operator failed
Given the above reality, it becomes important to ask where and to respond to this alarm it would trip the mill. However,
how the human control is ‘lost’ in such an environment. The this critical alarm and the other nuisance alarms were all
findings from the field study indicated that the deficiencies displayed on the same page with little differentiation.
in the operator interface (HMI/alarms) and organisational Operators admitted that often too many alarms appeared on
issues were the key barriers constraining human control the screen and they did not have time to read and understand
ability. For the scope of this paper, only summary results are all of them. As a consequence of this alarm overload, missing
reported. A full analysis of the findings and a more detailed or failing to respond quickly to critical alarms became the

FIG 3 – Operator interface (human–machine interface/alarms) limitations.
most common ‘operator error’, which could ultimately lead Evaluation

to plant trip-outs. Moreover, and of even greater concern, was Usually, operators watch the same overview or the trend
the fact that alarms became a major source of distraction and displays used at the analysis stage, to assess whether the
disruption to operators’ work. In one plant, operators ignored process has been stabilised. The common challenges affecting
alarms most of the time simply because they distrusted the their judgement was either that critical information was
benefit and necessity of these alarms to their work needs. missing or that it had not been effectively integrated to
support a timely and accurate assessment and prediction.
Analysis For example, a close monitoring of the trends of power
It was found that operators relied heavily on the information consumption, bearing back pressure and feed rate in the SAG
displayed on various overview and trend displays to mill is critical. However, when these three process trends are
understand and analyse the production status. However, either hidden somewhere behind the monitor screens, or are
information on the current operator console failed to be presented on a distant separate screen (see trend display in
integrated in a way that provided sufficient insight into critical Figure 2) from the principle SAG mill overview display (see
process dynamics (for example, suboptimal performance) and overview display in Figure 2), monitoring this information
its impact on control objectives. The SAG mill control example becomes difficult.
discussed previously is a typical example. Most operators Therefore it can be concluded that human control needs
had no idea of the dynamic behaviour of the SAG mill and have not been considered in the interface design of process
its impact on mill production; for example why power control technologies. Such interfaces are unable to provide
and bearing back pressure (an indicator of SAG mill load) operators with an adequate awareness of the control goals,
fluctuated in the trends display and how this phenomenon process status, operating constraints and options. As a result,
affected the SAG mill production stability and quality. human control on these plants inevitably falls into the reactive
Without a complete and accurate understanding of process mode, rather than into the desired proactive/optimisation
states and interconnectivity, effective decision-making and mode.
control of these processes is easily lost.
Alternatively, large operational sites may install expert Lack of an operational-focused culture
control systems to stabilise and optimise mill control. In addition to the above limitations in the operator interface,
However, many of these systems were not performing as organisational inadequacies were the other contributors
expected. In reality, many were switched off due to lack of to this human reactive control mode. The primary reason
efficient online tuning. In a plant site where the expert system was that the industry had focused too much on improving
was online to control the SAG mill, the control room operator metallurgical performance (for example maximising
admitted that the rules and logic in this expert system were throughput and recovery), while little thought had been given
unknown to him, being viewed as a ‘black box’ of outputs. It to developing and sustaining an operational-focused culture.
provides only a table for the operator to change the set points; As a result, inadequacies in the organisation and management
indeed, the operator’s input into the control system is often of the system put human operators at a distinct disadvantage.
rejected without any explanation. As such, operators have The following section provides a brief summary of identified
no understanding of what the expert system has done, was organisational issues, a more detailed discussion is provided
doing and was expected to do. The irony is that if operators in Li et al (2011) and Li, Powell and Horberry (2012).
(or metallurgists) are not meant to be in charge of the expert
system, who exactly is expected to take on this role? Perhaps No structured authority and responsibility in plant control
the belief was that a so-called expert system should be ‘smart’ Unlike other process industries, it is common in some (small
enough to optimise itself. The feedback collected to date and medium scale) mineral processing plants to have no
suggests that this particular expert system was good at dealing dedicated operators working in the control room. Instead,
with small variations, but failed in handling large variations, field operators (typically grinding and flotation operators)
the point at which operator engagement becomes essential. occasionally use part of the information from the control

system to monitor/control the production within their control room handled a lot of unexpected maintenance
section. Often the control room is empty while a critical requests. However, when they ring the mill operation
alarm is triggered. For example, in one instance the SAG field staff, often no one was available. It was also observed
mill suddenly stopped while all the operators were outside that control room operators spent a large portion of
and no one had any idea why it had stopped. As operators their time on tasks not relevant to the control room. For
are not tasked with supervising the control systems, it is not example, operators often dealt with various external
surprising to find that most operators lack ownership of the information enquiries (such as chasing up a particular
plant control. alarm message) and some operators commented that
maintaining the running of the plant while at the same
Insufficient training time answering phones and responding to the radio was
Although most plant sites provided a formal process for very stressful.
operator training, it was typically presented as a set of •• Communication with metallurgists – it was also found
operating procedures, rather than a systematic education that a disconnection between metallurgists and control
designed to developing process knowledge. Indeed, the room operators is common in the industry. In a perfect
majority of operators reported that their most valuable world, controlling a mineral processing plant should
control system knowledge was obtained piecemeal from be more like conducting a continuous experiment on a
hands-on training from a more experienced operator. The dynamic variable process with the metallurgist acting
lack of systematic training and ongoing assessment is likely as the experiment designer who makes decisions
to be a key barrier to operator efficacy. to ensure the best experimental outcome and who
The observed associated consequences due to insufficient gives the experimenter guidelines on how to conduct
training include: the experiment (‘feed forward’ communication). In
•• Inadequate knowledge of the control system and this scenario, the control room operator becomes
technology – The interviews revealed that the majority the experimenter who measures the outcome of
of operators do not have an adequate knowledge of the experiment from time-to-time and then reports
the control systems and technologies used in their those finding to the experiment designer (‘feedback’
workplace. Among the 20 operators observed or communication). Good communication between both
interviewed in the first two plants, only six operators parties is critical. However evidence of either feed
reported or were observed to use the trend displays forward communication (from metallurgist to control
to monitor a critical process performance over time. room) or the feedback (from control room to metallurgist)
Five operators were observed to use camera images was seldom observed in the field study. As a result,
to check the online condition of the flotation cells some metallurgists admitted in the interview that they
(others preferred to physically go to the site) and only had no idea how to make sure the daily operation target
two operators claimed that they knew the definition and limits were met, while most control room operators
of bubble stability (considered by experts to be a key were not even aware of the control strategies, such as
process performance indicator). However, no operator recovery – throughput drivers related to variations in
stated that he or she knew how to use the flotation image the ore feed grade.
analysis tool in their work.
•• Variation in shift operation – although no quantitative DISCUSSION
data was available, it is clear from the study that operator
performance variability has a real impact on the stability Challenges in future control room operation
of the production process. A typical scenario found that
As available orebodies decrease in quality and there is an
the start of the new shift was often problematic, with
increasingly carbon/energy-constrained future, innovations
the operator in the new shift busy resolving various
in complex circuit design and equipment selection will
problems remaining from the prior shift, or changing
the control settings to their preferred mode of operation inevitably require control room operators to handle much
– often resulting in an unstable fluctuating performance more complex processing circuits and more sophisticated
for some hours. In one case, the day shift operator set the automation systems. Certainly, they will require a much
tank level control (pump box) to manual to do a task. The higher level of training and process understanding than is
night operator was not aware of this and did not change currently the norm in most mineral processing plants.
the pump box back to the remote mode, almost resulting With increasingly remote operations, the operators’
in a major plant spillage incident the next morning. This traditional expertise in ‘seeing and feeling’ equipment on-
human performance variation is to some degree related sites coupled with their local engagement with the mill
to operator training. operation unit (site operators and technicians) will diminish.
•• Poor coordination within the operational team – the Instead, their monitoring and control will become totally
plant operation relies heavily on effective coordination reliant on various HMI and alarm systems around them.
among all the operational units from metallurgists to Therefore enhancing their ‘mental models’ of the production
the control room and mill operators. However, clear and process and developing effective coordination/collaboration
consistent communication patterns were seldom found between the control centre and the remote sites for various
in this field study. plant critical operational situations will become essential.
•• Communication with mill (site) operation – when the Due to an increasing shortage of experienced control
control room was located outside the plant, all the room operators, there is increasing expectations that future
operators reported that their communication with mill operations will rely on a single operator in the control
operations became worse or more difficult than when centre controlling several distant mineral processing plants
they were based in the plant. The typical example cited remotely. In such a scenario, issues such as how to extend
was that sometimes they saw from the screen that the and elevate operator control ability through effective human-
mill had stopped but had no idea why. Additionally the system interaction design become even more crucial.

Control room operator of the future – ‘dumb’ or ‘expert’ controller? The key to unlocking production potential
Returning to the question raised at the beginning, it is – empowering operator control
important to ask what the role of the control room operator From the argument developed in this paper, it should be
should be in such a challenging context? Should this role clear that empowering control room operators to elevate
remain that of a mere reactive responder to the automated their contribution from a reactive, equipment-focused control
controls surrounding them, or will the control room operator to a more proactive and knowledge-based one will be a key
become an expert controller in their own right seeking to factor in moving toward more remote operations. To do this,
proactively manage (with the aid of automation) the many taking a human-system integration approach rather than the
and varied interconnected plant processes? technology-alone approach would likely prove most effective.
Remote control in other industries Li et al (2011) and Li, Powell and Horberry (2012)
provides some initial thought in this regard. The discussion
A good clue to answering the above questions stems from largely focuses on how to make systems or technologies
reviewing the practices of other industries that have relied on more usable and useful to operators. The proposed work
remote control/operations for many years. includes optimising HMI to meet supervisory control needs,
Although the power industry has a much higher level of developing early warning tools to handle feed variation,
automation than other process industries due to its real-time uniform HMI/alarm styles from multiple systems installed
safety-critical nature, the role of the human operator still on-site, and improving operator training to fill the skill gap.
remains pivotal. In the power generation section, the operator’s Unfortunately, the potential solution of how to enable the
job responsibility has largely shifted from monitoring organisational environment to improve operation efficiency
equipment/processes to meet dispatched production and resilience has not been much discussed yet due to
targets (indeed, this level of control is mainly executed by limited information access of this research. Hopefully, the
automation), to a more holistic and strategic level of control organisational inadequacies identified in the field study
or coordination of energy resources under a deregulated would be taken up by management and motivate their
market environment. Operators are actively involved in thinking in future improvement.
developing strategies to maximise profits or reduce loss/risks
As operator engagement and support is a most crucial part
in response to various operational contingencies. Similarly, in
in the human-system integration approach, the best human
power dispatch/distribution control, the task list of operators
factors practice from other process industries (Hollifield et al,
has not been constrained in supervisory control of the power
2008) has indicated that the optimisation of process overview
distribution network status; indeed, operators are working
displays often is regarded as the most powerful and easiest
in planning and scheduling of routine/non-routine network
way to involve operators and change their thinking. Following
maintenance/test tasks.
this approach, some initial efforts are illustrated in Figure 4,
Another example is in air traffic control domain. Driven which shows an initial prototype of a control overview display
by the demand for increasing air traffic control capacity, for a typical grinding circuit. Unlike the displays shown in
this industry is undertaking the transformation of the entire Figure 2, which are similar to engineering flow sheets and
aviation system – from an ageing ground-based system to piping and instrumentation diagrams, this new display
a satellite-based system, the so-called ‘next generation’ air focuses on supporting operators with a complete and accurate
traffic control system. The role of human operators will move mental representation of the circuit process and performance.
from direct control of individual flights to supervisory control Operational needs such as information integration, handling
of the shared air space. Human automation interaction is a multiple tasks, competing priorities, early warning triggers on
critical investigative area in designing the next generation of process abnormality and timely trouble-shooting (reasoning,
air traffic control systems. Much work has been conducted diagnosis etc) have all been considered.
and is still ongoing in designing the organisational functions
The design rationale and features of this new display
and the organisational management of the new systems,
concept are outlined below.
including the authority and task allocations between human
and automation resources, operator training and changes in Rationale:
the ‘safety culture’ etc. •• clarify data outputs
The above examples from other industries indicate that the •• provide the ‘right data in the right form at the right time’
greater centralised or remote control will inevitably require to support human perception and cognition
integration of more sophisticated control tasks (such as •• reduce operator stress or workload in handling plant
strategic control) with human operators, which in turn leads upsets
to a much higher expertise or training for such human control
and emergent needs of restructuring human interaction with •• promote more consistent performance between
technology and improving organisational behaviours. operators
In particular, our observations from field studies indicate •• provide short-cuts from screen to critical operator
that automating everything is not an effective solution. Expert actions
systems deal well with limited variability but cope poorly •• the information is limited to the outputs that dictate
with major changes in input. In the petro-chemical industry process performance
experts test the feed stock and reset the system for each new •• colours are subdued so as to not overload the senses
tanker-load of crude oil, yet the mineral processing industry with a flood of primary colours and to ensure that critical
expects their expert system to retune its response to massively information is instantly recognisable as such.
larger changes in feed. Controlling a mineral processing
•• the front layer provides context-sensitive link to
plant is more like conducting a continuous experiment on a
appropriately grouped deeper information as required
dynamic variable process rather than applying routine control
to a well-defined system. The inputs from human operators •• simple visual metrics are used – bold trends with just a
are therefore crucial to optimise or coordinate these systems few lines and simple bar diagrams
across various unexpected plant disturbances. •• numeric data is minimised

FIG 4 – Prototype of a primary control screen for grinding circuit control.
FIG 5 – Analogue bar indicators.
•• the entire process is summarised on one screen: there •• target plots, such as for the ball mill, show the distance
will undoubtedly be additional monitors to provide from the desired value of the key circuit parameters,
access to the wider range of equipment data across the such as mill power, cyclone pressure and cyclone feed
plant, but they will not be for the purposes of guiding flow rate, with the target rings highlighting the worst
operational performance. point for quick identification, as shown in Figure 6
Features: •• context-sensitive alarms to support early warnings on
•• simple graphic ranges are used to indicate set point critical processes
ranges, desired operating range and where the process •• an alarm summary section (Figure 7) highlights alarms
is in this range, as shown in Figure 5 by hierarchy and colours by criticality; clicking on any
•• trends dynamically reset ranges to maintain resolution one will display the full alarm message
to visually rapidly observe excessive rates of change; this •• only key process data is displayed on the total process
provides consistent gradients (that is, rates of change) layout
that the operators can become accustomed to •• sump and stockpile displays provide not only range
•• a pertinent time scale is chosen to suit the residence and current value, but also the trend over the relevant
time, control lag and response of the processes control time interval – a quick glance gradient display
•• a red button appears if operator attention is required; •• operating units in a cluster are shown by subdued
upon pressing this the display system switches to colouring
showing only the relevant information that is required •• the central circuit layout panel provides an overview
to interpret and diagnose the condition with the following features:

process knowledge and who can apply this knowledge to the

plant’s operation. This will require continuous upskilling of
operators to provide them with the processing knowledge
and tools required, and it will lead to a radical change in how
they interface with the plant and control systems. Re-thinking
and then re-designing how the operator and the control
system interface via the HMI is a key aspect to this evolution
to elevated proactive control that is aimed at maximising
productivity, that is, profit, rather than at just reactively
keeping machinery running. The expert control system then
becomes an effective process optimisation tool for the expert
operator. Essentially these expert controllers will have to be
empowered to take the mineral processing industry into the
advanced technological milieu of the 21st century.
poor good
operation operation REFERENCES
Burns, C M and Hajdukiewicz, J R, 2004. Ecological Interface Design
(Routledge).
FIG 6 – Polar performance plot.
Burns, C M, Skraaning, G, Jamieson, G, Lau, N, Kwok, J, Welch, R
and Andresen, G, 2008. Evaluation of ecological interface design
Unacknowledged alarms, for nuclear process control: situation awareness effects, Human
Factors, 50:663–698.
red= high priority, orange =
medium, blue = low, Engineering Equipment and Materials Users’ Association
(EEMUA), 2007. Alarm Systems: A Guide to Design, Management
Acknowledged alarms, and Procurement (The Engineering Equipment and Materials
colour turns pale. Users’ Association: London).
Hollifield, B, Oliver, D, Nimmo, I and Habibi, E, 2008. The High
FIG 7 – Alarm coding. Performance HMI Handbook: A Comprehensive Guide to Designing,
Implementing and Maintaining Effective HMIs for Industrial Plant
•• use variation in grey and thickness of lines to represent Operations, 207 p (PAS).
production flows and equipment status Jamieson, G A, 2003. Ecological interface design for petrochemical
•• simple circuit layout to minimise the complexity and process control: integrating task – and system-based approaches,
highlight the important control relationship PhD thesis, The University of Toronto.
•• integrate important process/control variables that Klein, G A, Calderwood, R and MacGregor, D, 1989. Critical decision
are hidden in separated screens to support human method for eliciting knowledge, IEEE Transactions on Systems,
decision-making: for example flow rate, water Man, and Cybernetics, 19:462–472.
level, splitter ratio, number of cyclones online, size Li, X, 2008. Evaluating ecologically-inspired displays for complex
distribution on output. systems: Hydropower system case study, PhD thesis, The
University of Queensland.
•• the coarse ore stockpile shows approximate distribution
between feeders, providing available feed time from Li, X, McKee, D, Horberry, T and Powell, M, 2011. The control room
each and warning when a feeder is running low, so as operator: the forgotten element in mineral process control,
Minerals Engineering, 24:894–902.
to prompt operator action of adjusting coarse-fine feed
split or feeder usage. Li, X, Powell, M and Horberry, T, 2012. Human factors in control
room operations in mineral processing: elevating control from
This new display has been reviewed on industrial sites and reactive to passive, Journal of Cognitive Engineering and Decision
the feedback has been overwhelmingly positive. Operators, Making, 6:88–111.
metallurgists and even plant managers involved in the
Li, X, Sanderson, P, Wong, W B L, Memisevic, R and Choudhury, S,
survey could immediately see the potential benefits of the
2006. Evaluating functional displays for hydropower system:
new display. The operators regarded it as a handy tool for the model-based guidance of scenario, Cognition, Technology, and
new operator and for the experienced operators, time saved Work, 8:269–282.
in switching display pages. The metallurgists thought this
Munro, P D and Tilyard, P A, 2009. Back to the future: why change
display would elevate the control focus from equipment to doesn’t necessarily mean progress, in Proceedings Tenth Mill
process efficiency, and the plant manager perceived it as an Operators’ Conference, pp 5–12 (The Australasian Institute of
enabling tool that would allow control room operators and Mining and Metallurgy: Melbourne).
metallurgists to work together to explore enhanced control
Thwaites, P, 2008. Process control in metallurgical plants: towards
strategies. The clear message from the above feedback is that ‘operational performance excellence’, presented at International
the interface design of control systems is not simply about the Congress on Automation in the Mining Industry – AUTOMINING
informational representation of the technology. It is also about 2008, Santiago.
designing for the way people work with the technologies; for Wei, D and Craig, I K, 2009. Grinding mill circuits: a survey of control
example, appreciating how their understanding is developed, and economic concerns, International Journal of Mineral Process,
how communication or coordination is developed, or even 90:56–66.
how the organisation itself functions.
Woll, D, Hill, D and Polsonetti, C, 2002. Collaborative process
automation drivers return on assets, ARC Advisory Group.
CONCLUSIONS
It is proposed in this paper that the industry needs expert
operators (not simply expert systems) who have more in- depth

Contents
Current developments in the operation

and control of autogenous and
semi-autogenous grinding mills in Australia
J Karageorgos1, Y Atasoy2 and D Baas3
ABSTRACT
With increased pressure for companies to cut costs during design of new mineral
processing plants, overall plant design has changed over the last decade with fewer
plants being designed for feed blending and less installed surge capacity. This has
changed the operability of new mineral processing plants, where disturbances
due to variable ore types have resulted in the evolution of new control techniques.
These techniques are necessary to maintain stability and optimise the performance
of these new plants.
This paper describes the benefit of applying these new control techniques in a new
mineral processing plant in Australia utilising Rockwell’s ControlLogix® system
architecture.
INTRODUCTION
In recent years, new mineral processing plants have been constructed and
commissioned throughout the world. These new plants have been constructed and
operated with tight budgets. Direct tipping of ore, significant ore type variation,
coarse ore feeder trips and the removal of surge bins from recycle crusher circuits has
required new control techniques to stabilise the grinding circuit in these processing
plants, specifically with semi-autogenous grinding (SAG) mills.
Manta Controls®, an Australian based company, has been engaged to optimise the
various processing units on-sites in Australia and Africa by improving the process
control strategies.
This paper describes the improvements in SAG mill control on one site in Australia
that has adopted the Manta Cube®, a new control technology developed by Manta
Controls®.
Overview of the Manta Cube®

The Manta Cube® process control technology utilises a variety of fundamental control
techniques, including the traditional expert system approach together with new
techniques specifically developed by Manta Controls®. The Manta Cube® ensures that
the dynamic behaviour of the circuit is understood and utilised in the design of the
control system. The Manta Cube® is primarily made up of four parts:
1. the Cube – this determines the operating mode of the unit process such as a
SAG mill
2. the Cube expert decision matrix – this describes what is required to get the SAG
mill back to the required operating band
3. the Cube engine – this is a fundamental control structure utilising multivariable
and decoupling techniques
4. the Cube optimisers – there are various optimisers developed to ensure that the
1. MAusIMM, Managing Director, Manta system is optimising the control objective such as maximum throughput.
Controls Pty Ltd, Grange SA 5022. Email: The primary advantages of the Manta Cube® Control system are:
john.karageorgos@mantacontrols.com.au
•• The system is configured locally on the site’s Distributed Control System (DCS)
2. MAusIMM, formerly Metallurgical
Superintendent, Gold Fields Ltd,
or Programmer Logic Controller (PLC) system using the inbuilt functionality
St Ives Mine, Kambalda WA 6442. available on these systems. This eliminates the need to maintain new systems,
Email: atasoyya@yahoo.com learn a new programming platform and removes additional hardware points of
3. MAusIMM, Senior Process Control Engineer, failure. Any system developed using the Manta Cube® maintains the robustness
PanAust Ltd, Fortitude Valley Qld 4006. and integrity of the original site control system as it is developed without
Email: dirk.baas@panaust.com.au reliance on communication to third party computer processors.
411
•• The system utilises the dynamics of the process in the St Ives – closed circuit semi-autogenous grinding mill
design of the control strategies, coupled with traditional Goldfields commissioned their St Ives gold plant located
expert system type approaches and new control near Kambalda in Western Australia in December 2004. The
techniques developed by Manta Controls®. The overall St Ives grinding circuit consists of three coarse ore feeders
system is very robust. and one soft ore feeder supplying a 10.97 m (36’) high aspect
•• The underlying architecture of the Cube engine is ratio single stage SAG mill in closed circuit with 508 mm
structured utilising a modular framework making (20’) Krebs hydrocyclones. The SAG mill is powered by an
maintenance and upgrades very easy. ABB variable speed 13 MW wraparound drive. The grinding
circuit feeds directly to a leach circuit without the aid of a
•• The business objectives are defined in the strategy
leach feed thickener. The grinding circuit also has a recycle
design of the Manta Cube® and the optimising logic
scats crusher without any surge capacity when the recycle
continuously drives to achieve these outcomes. crusher is bypassed. A schematic of the St Ives grinding
circuit is shown in Figure 2.
Integration of the Manta Cube® Due to the limited surge capacity it was important to
The SAG Mill Cube system graphic shown in Figure 1 establish very good control of the grinding circuit to minimise:
demonstrates that the system is fully integrated with the •• the effect of disturbances to the grinding circuit
Allen Bradley ControlLogix® System used at Goldfields St
•• grinding circuit disturbances sent to the leach feed.
Ives. The system is also integrated with the Yokogawa CS/
Prior to engaging Manta Controls®, the grinding circuit
CS3000, Foxboro IA, Emerson Delta V and other common
was being operated according to the original control strategy
control systems.
of the SAG mill and classification circuit as per the design
control philosophy.
Case study Controlling a SAG mill utilising older control techniques
The site that will be covered in this paper is Goldfields St Ives will give adequate control, however to provide control that
Lefroy Mill (St Ives). A brief description of the grinding circuit will continuously optimise the grinding circuit whilst coping
is given for St Ives with a description of the performance of with the above disturbances, new control techniques needed
the Manta Cube®. to be developed.
FIG 1 – Manta Cube® integrated on the Allen Bradley ControlLogix® System at Goldfields St Ives.

Current developments in the operation and control of autogenous and semi-autogenous grinding mills in Australia
FIG 2 – Schematic of the Goldfields St Ives grinding circuit.
The Manta Cube® technology utilises new control When the feeders were in auto mode, a feeder trip would
techniques that can be implemented on DCS and PLC not be corrected until the loss of feed was noticed on the
systems that apply the IEC61131–6 control standard. There weightometer. This is a common problem with large dead
is no need to have separate operating systems linked to the time systems and can be corrected by utilising the principles
plant control system. The Manta Cube® can operate on both of dead time compensation.
DCS and PLC systems, therefore the system integrity is the
Figure 3 shows the effect on the feed rate system on the SAG
same as the plant control system.
mill prior to the Manta Cube® feed control being implemented.
With increasing pressure to lower operating costs, a practice A reduction in the feed is noticeable when one of the feeders
that is also becoming more common is to direct tip feed to the tripped due to a fault condition. This reduction in the feed rate
primary crusher rather than creating blend fingers. can cause problems with the mill weight decreasing for short
The Manta Cube® is capable of detecting ore type changes periods of time. This has the potential to cause mechanical
online and compensating for this variation via manipulated damage to the grate and lifters of the SAG mill.
variables such as mill speed, mill feed rate, feed density, feed
With the installation of the Manta Cube® feed controller, the
size and other manipulated variables that are provided by the
reduction in the feed rate is virtually eliminated when a feeder
grinding circuit design.
trips. This is shown in Figure 4. The feed controller is capable
Manta Controls® redesigned the entire control strategy for: of distributing the lost feed of the tripped feeder to all other
•• the coarse ore feed system operational feeders in the space of one scan thus eliminating
•• large variability in ore competency in the mill feed a reduction in the overall feed rate. This provides a very
robust feed system that is capable of minimising mechanical
•• the classification circuit
damage of the mill grate and lifters due to a reduction in the
•• the SAG mill. mill weight when associated to a partial loss of feed to the
SAG mill.
The coarse ore feed system
The Manta Cube® feed controller accounts for feeder trips and
The initial control of the coarse ore feed system used an
provides ‘bumpless’ transfer for all feeders. ie a change from
operator-selected feed rate set point which manipulates the
one control mode to another no longer creates any disturbance.
feeder outputs via a configurable feeder ratio. This required
the feeder speed control to be in auto mode. When a feeder
was changed from auto to manual, the feeder speed was The effect of variable ore type
changed to the raw ratio value, eg if the auto feeder ratio was St Ives treats a variable orebody from an open pit and direct
20 per cent and the feeder speed was operating at 5 per cent, tipping is utilised as a strategy to reduce handling costs. Direct
when the feeder was placed in manual mode the feeder speed tipping introduces another disturbance where the different
would step up to 20 per cent. This caused significant feed ore type conditions will have to be accounted for to maintain
disturbances and made operator control difficult. stable SAG mill operation.

FIG 3 – The effect of a feeder tripping on feed rate – before the Manta Cube®.
FIG 4 – The effect of a feeder tripping on feed rate – with the Manta Cube®.
The Manta Cube® has been designed to detect such ore •• cyclone pressure control utilising automatic on/off
type variations utilising fundamental control techniques. The control of the cyclone feed valves
Manta Cube® speed system compensates by automatically •• grinding circuit particle size control utilising cyclone
changing the overall operating speed band of the SAG mill feed density and cyclone pressure control.
for the duration of the disturbance. This is a very robust and
The improvements made in the classification area were
elegant system with the result being stable SAG mill weight.
remarkable with stable control of the cyclone feed flow rate,
Improvements in stability of the SAG mill weight can be seen
cyclone feed density and cyclone operating pressure between
when comparing Figure 5 and Figure 6.
the allowed 15 kPa pressure band.
The classification circuit An additional benefit was reduced cyclone feed pump wear/
mechanical failure. The rebuild of the cyclone feed pumps
The St Ives SAG mill operates as a single stage SAG mill
was reduced considerably due to the consistent operation of
in closed circuit, and any disturbances passed onto the
the cyclone feed pressure as a result of good surge control of
classification circuit then return to the SAG mill. Therefore it
the cyclone feed hopper.
is critical to dampen out disturbances that may be fed to the
hydrocyclones.
The St Ives semi-autogenous grinding mill
The entire classification control system was redesigned Operational data was collected for a month prior to the
utilising the standard Manta Controls® classification strategy. installation of the Manta Cube®. The four sets of SAG
Areas that were improved were: operational data are graphed as histograms and are shown in
•• improved SAG mill feed density control utilising Figure 5. The wide variation of mill weight is noticed which is
cyclone underflow density and other water sources becoming a signature of operator controlled SAG mills.
reporting to the feed The Manta Cube® was configured on the St Ives SAG mill
•• cyclone feed flow rate stabilisation utilising surge and performance data was again analysed for a one-month
control of the cyclone feed hopper period, shown in Figure 6. During this time, the Manta Cube®

Current developments in the operation and control of autogenous and semi-autogenous grinding mills in Australia
FIG 5 – The performance of the Goldfields St Ives SAG mill – before the Manta Cube® was implemented.
FIG 6 – The performance of the Goldfields St Ives SAG mill – after the Manta Cube® was implemented.
was the only modification made to the SAG mill and as seen on The variability of the mill weight again reduced considerably,
the histogram on Figure 6, the mill weight is tightly controlled thus stabilising the grinding circuit. This additional benefit
around the required set point. The control system is working of stable mill performance is quite pronounced downstream
the mill harder by operating at a higher speed range. from the SAG mill. Additional improvements in the leach
Variations of feed type are also present at St Ives arising circuit due to the stable operation of the SAG mill were not
from natural variation of ore types from the mine and direct quantified in this study.
tipping at the primary crusher. As described previously, the With the improved control of SAG mill weight, the Manta
Manta Cube® self-compensates for these variations, with the Cube® was able to increase the overall feed rate to the SAG
result being a stable SAG mill weight. mill as shown in the feed rate histogram of Figure 6.

The improved performance of the SAG mill after only one By using the Manta Cube®, St Ives has achieved greater
month of operation with the Manta Cube® was calculated as a grinding circuit stability, tight mill weight control and higher
6.1 per cent increase in mill throughput. throughput. The SAG mill is working harder, continually
pushing for higher feed rates and consistently utilising the
SUMMARY available power draw. The site recorded an immediate and
sustained increase in throughput of 6.1 per cent.
Goldfield St Ives provides an excellent example of plant
design in the late 2000s. The emerging challenges presented The design and construction of new mineral processing
by streamlined plants demand higher control performance. plants presents increasing pressure for reduced capital and
The Manta Cube® successfully overcomes these issues using operating costs by eliminating such things as feed blending
a unique control algorithm implemented locally on the user’s and recycle surge capacity.
plant control system. High performance control systems such as the Manta Cube®
The Manta Cube® has proven itself to handle the common have become more than a nice-to-have, but rather a necessity
issues facing SAG mill designs. It provides additional benefits for optimal operation of minerals processing plants.
in operability by working directly on the local plant control
system. A platform-independent control algorithm, the Manta ACKNOWLEDGEMENTS
Cube® removes the cost and complications associated with The authors wish to thank the management of St Ives Gold
traditional, standalone expert systems. Local implementation Mining Company for permission to publish this paper.
provides operators and technicians alike with reliable access
The efforts of all the Lefroy processing teams are also
to detailed information on their SAG mills.
acknowledged for their valuable input during commissioning
of the Manta Cube® that has made the projects a success.

Environmental
management and
sustainability
Contents
Mine waste risk minimisation by integrated

waste management and process optimisation
D Brett1
ABSTRACT
Mine waste storages including tailings dams and waste rock dumps represent
arguably the biggest risk on a mine site from both physical and geochemical
viewpoints, related firstly to structural stability and the consequences of failure and
secondly from potential ongoing seepage or run-off of acid and metalliferous drainage
(AMD). The risks are most likely to increase during operations and continue through
the closure and post-closure phases, with the waste storages likely to be the most
difficult features to resolve in the mine closure process. Contaminated water quality
from tailings dams and waste rock dumps can require treatment ‘in perpetuity’ at a
cost of $100s of millions. The problem is increasing with ever larger mine operations
and waste generation volumes.
This paper identifies leading practice methodologies in planning and operating
mines and process plants to minimise the risks from tailings storages and how this
might also affect waste rock management and even the design of mine excavations.
These methodologies require a ‘big-picture’ approach to develop an understanding
of the overall issues and identify synergies between the different operational areas on
a mine site that can reduce overall risk. With a conservative approach and realistic
allowance for future costs it is possible to facilitate decisions at the mine planning
stage that can reduce risk and save cost over the long-term. This is in contrast to the
common approach where each operational area is looking to minimise its own costs
with a short-term view. Incorrect mine methodologies can be developed by allowing
net present value (NPV) accounting methods to trivialise future closure issues.
Methodologies discussed include:
•• Integrated waste management – significant structural improvements to tailings
dams can be achieved if the structures can be incorporated into the waste rock
dumps. Risk of AMD can also be reduced by utilising the tailings and waste rock
effectively. This could go even further to consider the placement of wastes to
limit mine void environmental issues.
•• Process optimisation – water quality and the nature of potential tailings dam
structures can be highly influenced by the geo-chemistry of the tailings. This can
be modified by processing, where simple changes such as maintaining separate
streams of geo-chemically different tailings rather than recombining them or by
initiating further processing to remove problematic minerals such as sulfides.
•• Enhancing geophysical properties – the geochemical performance of waste
repositories can be greatly influenced by the geophysical properties of the
stored materials, particularly the air-water characteristics, which are influenced
by particle size, density and permeability. The geotechnical properties can be
enhanced by strategic paddock dumping, blending and compaction.
These methodologies are demonstrated by case histories.
INTRODUCTION
Some of the major issues facing mining today involve preventing environmental
impact during operations but, possibly even more difficult, is achieving sustainable
closure of mine waste dumps, including tailings and waste rock storages post-mining.
Issues during operations can include, hydrocarbon spills, dust, noise, erosion, damage
to vegetation and many others. These can normally be addressed by engineering
solutions and operational controls. However, maintaining stable landforms and
preventing pollution from AMD in seepage and run-off from mine voids and waste
dumps are likely to be the most insidious and difficult problems to address.
1. Principal Engineer, Mine Waste and Water Robertson (2011) described the current increase in the volume of mine wastes from
Management, GHD, Hobart Tas 7000. large mines by a factor of ten each generation (1/3 century) and the consequent
Email: david.brett@ghd.com increase in risk from failure. With analogy to Murphy’s Law that ‘if something can go
419
D Brett
wrong then it will’, Robertson described Murphy’s ‘A-Team’ can be related if the structural failure leads to the development
as the Perpetual Forces of ‘erosion, weathering, desiccation, of AMD by exposure of sulfide waste. The consequences of
fires, biotic and anthropogenic activity’. Geochemical activity structural failure are likely to be very much reduced if the
should be added to this list. For engineers these forces materials released are not prone to AMD generation. AMD
are possibly more difficult to design for than Robertson’s can be caused by any exposure of sulfide material with the
‘B-Team’ of Extreme Events. AMD, in particular, is difficult to propensity to oxidise. The sources commonly involved include:
stop once it has started and can lead to a requirement to treat
contaminated seepage and/or run-off from the site for decades •• Waste rock dumps – these usually involve dumping of rock
or, more likely, considerably longer, to prevent unacceptable from a mine operation in a relatively loose condition,
environmental impact. Investments of hundreds of millions often tipped from a high ‘tip-head’ leading to significant
of dollars per site may be required to pay for collection and segregation and subsequently to highly permeable
treatment of contaminated water ‘in perpetuity’. Uncontrolled zones allowing entry of both air and water. This can
erosion of thin cover systems on tailings dams can expose create the perfect conditions for AMD generation. As
sulfide tailings and lead to rapid unravelling of an engineered oxidation commences, temperature within the dumps
waste dump. can rise leading to rise of heated air by convection
The international mining industry is coming to understand through the dump surface and suction of fresh air into
the concept of sustainable development and the need to follow the dump base. The extent of mineral content within a
‘leading practice’, which is simply ‘the best way of doing dump will depend on the ‘cut-off’ grade of the mineral
things’ and can be site specific. As such, leading practice is processing operation and it is usual for ‘waste’ rock to
difficult to define and it is important that it is flexible and be significantly mineralised and therefore a potential
innovative to allow the development of solutions that match source of toxic metal pollution.
site specific requirements. Leading practice is as much about
•• Tailings storage facilities – sulfidic tailings can represent a
approach and attitude as it is about a fixed set of practices or
significant source of AMD due to their fine particle size
a particular technology.
and consequent large surface area. Tailings transported
AMD pollution can be prevented with upfront planning as slurry are typically deposited initially in a saturated
and investment in appropriate methodologies, including state and oxygen penetration is limited. However, over
training of mine operators and provision of the resources to
a period of time the tailings can become unsaturated,
manage the wastes appropriately. The additional cost of these
allowing oxygen penetration, leading to a potentially
methodologies can be minimal or at worst, relatively minor in
long-term AMD source. Seepage from tailings storage
relation to the total project expenses, but could lead to huge
cost savings in the long-term. This paper identifies a range facilities can contaminate groundwater and surface
of methodologies that could prevent or minimise the risk water streams.
of AMD particularly from tailings dams and to some extent
waste rock dumps, ranging from simple waste classification UNDERSTANDING THE ISSUE
and strategic placement to downstream processing and waste The good news is that, if the potential for AMD is understood,
modification. The paper explores the difference in approach
then there are methodologies to plan and implement mining
from creating a hazard and trying to protect it with dubious
and milling operations to prevent or at least limit the
engineering controls to modifying the hazard to make it
generation of AMD to the extent that long-term treatment
benign enough so that it is robust and immune from localised
breakdown of the surface treatment. can be avoided. It is particularly important that the risk of
AMD is recognised before mining commences if adequate
The chemistry of AMD is well known, involving oxidation
methodologies are to be put in-place. It is usually very
of sulfide minerals that are exposed to both oxygen (from air)
difficult and expensive to re-engineer a mining operation to
and water. Acid water can then react with other minerals. To
prevent AMD once it has started.
some extent, the acid becomes neutralised by the minerals it
reacts with but normally at the expense of increased metal or The key to understanding the issue is to carry out appropriate
salt concentrations. Occasionally, the acid generated can be investigations right from the initial exploration phase of a
completely neutralised with precipitation of metals such as project. It is important during the premining phases that
aluminium, copper and lead. However, at near-neutral pH, the project team, including geologists, mine planners,
concentrations of toxic components such as zinc, arsenic, environmental scientists and AMD experts, ensures that an
nickel and cadmium can remain elevated. adequate geological and geochemical database is compiled to
The control of AMD essentially revolves around control of clarify baseline conditions and the risk of AMD. Knowledge of
the factors that cause AMD, including cutting off the supply the likely wastes that will be generated, materials exposed and
of either oxygen or water, or in changing the chemistry the constraints this will place on the mining operation is vital.
within the dump. Research work (Aubertin, 2005) has
In addition, a detailed and realistic closure plan needs to
shown that oxygen ingress is controlled by the degree of
be developed and costed during the feasibility phase. This
saturation and that if saturation levels are maintained above
must remain a ‘living document’, as the mine proceeds, with
approximately 85 per cent then oxidation of sulfides can be
effectively eliminated, or at least reduced to levels that can be regular reviews and updates based on new technologies,
assimilated into the environment as shown in Figure 1. This stakeholder inputs, changing mine conditions and community
relationship is a significant factor behind the technologies expectations.
used for prevention of AMD generation within dumps and
for design of cover systems. INDUSTRY LEADING PRACTICE FOR PREVENTION OF AMD
Assuming it is known that there is likely to be AMD pollution
SOURCES OF RISK FROM MINE WASTE STORAGE FACILITIES with ‘business as usual’ then what can be done to prevent
The source of long-term risk from mine waste storage facilities this? The following are various methods that are currently
can include risk of structural failure and risk of AMD. These considered industry leading practice.

Mine waste risk minimisation by integrated waste management and process optimisation
FIG 1 – Coefficient of diffusion versus degree of saturation for saturated porous media.
Waste rock dumps Recommended leading practice methods include the

following.
It is not good enough to build a waste rock dump from sulfide
bearing waste with normal uncontrolled high tip methods Strategic Placement of Materials (Zoning) – this requires
and rely on a relatively thin ‘engineered’ cover system to classification of waste materials and designing/operating
prevent AMD, particularly if it is constructed many years waste rock dumps to place each type in an appropriate location
after initial placement of the dump. Instead, there is a range in a dump. A common classification system could designate
of practices that can be implemented as appropriate for the waste into three or four categories based on geochemical
individual situation. properties as follows:
It should be noted that ‘blending’ is not one of these. The 1. potentially acid forming– high risk (PAF-HR)
theory of ‘blending’ is that by mixing potentially acid forming 2. potentially acid forming– low risk (PAF-LR)
(PAF) rock with acid neutralising rock that AMD can be 3. non-acid forming– (NAF)
prevented. The method has been generally unsuccessful. This
4. acid neutralising.
is due to several factors as follows:
Then there could be further designation based on
•• even if neutralisation does occur, the resulting run-off or geotechnical properties such as:
seepage water quality is still likely to be metalliferous,
or at best saline •• hard rock (erosion protection)
•• reaction of acidic water and alkaline material within a •• clay (waterproofing – encapsulation)
dump is likely to result in insoluble precipitates that •• sand (filters – drainage).
‘coat’ the alkaline materials and prevent further access The zoning design is likely to provide for PAF materials to
•• preferential flow pathways are likely to develop within a be placed deep in a dump, where they can be encapsulated
dump, which means that water flow will not necessarily with clay, with NAF materials forming an outer protective
contact the neutralising materials. zone. Ideally the dump design may also allow for placement
of PAF materials in a way that they can be maintained in a
Similarly, ‘capping’ is not necessarily leading practice.
permanently saturated condition, either returned to mine
Capping is the normal fall-back plan for closure of waste
voids below the long-term water table, within the saturated
rock dumps and tailings storages. The theory of capping is to
zone of a tailings dam or with the waste rock dump itself being
reduce penetration of oxygen and/or water and thus reduce
designed to develop and maintain a saturated condition.
the resulting AMD release. However, cover systems are
unlikely to be effective due to these reasons: Paddock dumping and compaction – significantly better
geochemical and geotechnical performance can be achieved
•• if oxidation has already occurred there is likely to be a
by dumping waste in piles then spreading in relatively thin
built-up store of oxidation products that will continue
layers of 1–2 m. This results in minimal segregation and, with
to be flushed out as AMD for potentially many decades traffic by mine trucks, particle size breakdown and compaction
•• a reduction of infiltration to a dump will logically lead to to an increased density. For even better compaction, heavy
a reduction in outflow, due to water balance, but typically vibrating rollers can be used. Impact rollers can be successful
this results simply in an increase in the concentration of in achieving high density with relatively thick layers. The
the resulting seepage, with maintenance of the ‘mass improvement in density leads to increased geotechnical
balance’ of pollutants strength but more importantly, reduced porosity and
•• the effective life of a relatively thin cover system is permeability. This means that pore water is more likely to be
likely to be limited due to a range of events during retained and the degree of saturation increased, potentially
the expected post-closure lifetime. These events can be limiting oxygen ingress.
erosion, vegetation effects such as disruption by root In addition, the potential for encouraging oxygen entry
systems or settlement causing ponding of water. due to convection flow will be reduced and preferential flow

D Brett
paths for water will be disrupted. Compaction is likely to be scope for appropriate placement of the waste streams
effective for softer rock types that can be broken down to a separately rather than combined.
soil consistency by compaction. If this form of separation is not possible then a more active
Co-disposal – depending on the nature of the waste rock processing phase may need to be added, most likely in the
material, it is likely that, particularly for hard rock, relatively large form of sulfide flotation to split the tailings into low sulfide
voids can remain after placement and even after compaction. and high sulfide streams. Hesketh, Broadhurst and Harrison
This situation allows potential for ‘co-disposal’ where tailings (2010) describes a case study from copper sulfide tailings
are recombined with waste rock to produce a mixed, well- where desulfurisation by flotation was able to process an
graded material with strength properties dominated by the acid forming tailings containing 3.7 per cent sulfur to a low
rock material and hydraulic properties controlled by the sulfur tails comprising 91 per cent of the raw tails stream
tailings. This material will have low permeability, good water and a sulfide tails including the chalcopyrite fraction with
retaining properties and therefore good resistance to oxidation 36 per cent sulfur but only comprising nine per cent of the
and generation of AMD. Co‑disposal is becoming common in original. Bruckard and McCallum (2007) showed how a range
the coal industry, where coarse and fine washery reject can be of separation processes could be used to remove sulfides
combined and pumped as a mixture. It is, so far, less common from tailings and produce low sulfide sand suitable for
in hard rock mining due to practicality of combining and construction use, plus a range of potentially valuable metallic
transporting the mixed material, but methods for mixing in by-products. Processes included simple screening, gravity
cells have been considered. methods, magnetic separation and froth flotation. Bois
et al (2005) describe the desulfurisation of a tailings sample
Tailings containing typically 20 per cent sulfide to a low sulfide NAF
tailings and a high sulfide tailing. Cement stabilisation tests
As previously described, the best method of prevention of
showed no detrimental effect in using the high sulfide tailings
AMD is to maintain tailings at a high level of saturation with
for underground backfill.
the best, proven method being subaqueous disposal into
a permanent water body. There are many examples of this This later example highlights the potential scheme for
in literature. However, a drawback of subaqueous disposal environmentally sustainable disposal of PAF tailings
is the resulting density of tailings being significantly lower by desulfurisation to produce NAF tailings that can be
than other methods with consequentially larger tailings economically disposed of in several ways with very low risk
storages being required. In addition, tailings storage dams for while the high sulfide residual can be safely used for mine
subaqueous disposal need to be designed and built as water backfilling or used for commercial production of acid. This is
storage dams with implications for long-term management. shown diagrammatically in Figure 2, modified from Hesketh,
Broadhurst and Harrison (2010).
The density problem can be reduced if the rate of oxidation
is low enough to allow short-term beaching when density can Saturation of sulfide tails – arguably the best method for
be increased by settlement, drainage, evaporation of excess preventing oxidation of sulfide tailings is to keep the materials
water and shrinkage consolidation. The key to successful saturated or at a high level of saturation. This can be achieved
in a variety of ways including:
subaerial beaching is to spread thin layers of tailings by
well-managed spigotting, with frequent re-covering so that •• re-placing underground where water table will flood
relatively high levels of saturation are maintained. them
A potential problem with spigotting of conventional tailings •• storing subaqueously in a purpose designed dam either
is the segregation that occurs on the beach. This can lead to with a water cover or a ‘dry’ cover designed to maintain
coarser sand particles settling out near the spigot points with saturated subsurface conditions
finer ‘slimes’ being transferred further into the storage. The •• in-pit storage, where water table will saturate the tailings
sands can be relatively permeable and consequently prone •• compacting tailings to a level where the hydraulic
to drainage, which allows oxygen ingress and generation of characteristics of the tailings mass retains water and a
AMD. On the other hand, this segregation and drainage is an high level of saturation can be maintained by infiltration
important factor when considering ‘upstream construction’ of rainfall.
where tailings storage dams are raised by construction over
This is demonstrated by review of the probable density of
previous tailings beaches. These structures then rely on
a tailings deposited under various conditions as shown in
drainage to lower the phreatic surface within the outer face
Figure 3, where the permeability of a typical tailings can vary
of the dam to maintain structural stability. The outer, drained
by nearly four orders of magnitude depending on relative
zone can then become a potential source for AMD pollution.
density. Tailings discharged subaqueously would typically
The further dewatering of tailings to form ‘paste’ and be loose, after drying on a beach could densify to medium
filtering to allow ‘dry stacking’ may provide improved compact but would require mechanical compaction to achieve
performance of some options for AMD management for compact to very compact status.
PAF tailings but once again, only if relatively high levels of Paste and dry stacking – paste and dry stacking technologies
saturation can be maintained. have potential to be leading practice as part of other options
Some leading practices for AMD management from tailings as described above. Paste has an advantage in allowing higher
are detailed below. density to be achieved without needing segregation as part of
Segregation of sulfide tails – this can be as simple as reviewing beach development. This means that the high permeability
the mill process to evaluate the various sources of tailings and zones associated with subaerial discharge can be avoided
consider if they are better dealt with separately or combined. while still potentially allowing upstream construction.
Quite often a low sulfide tails will be produced by a gravity Both paste and dry filtered tailings technology also allow
circuit or a rougher tails circuit. This can be a substantial increased scope for co-disposal of tailings and waste rock,
proportion of the total tailings stream and may not have the which can lead to the production of a low permeability but
same level of AMD potential as the remaining high sulfide geotechnically stable material that can retain a high level of
component. If this is the case then there is possibly significant saturation in appropriate climates.

FIG 2 – Conceptual approach to sulfide tailings management.
FIG 3 – Permeability versus density.
CASE HISTORIES AND EXAMPLES OF LEADING by reverse flotation as a separate tailings stream comprising
3 per cent of the total tailings volume. Due to the low
PRACTICE FOR AMD PREVENTION neutralising capacity of the NAF tailings, recombining the
tailings posed a risk of AMD from the total. Accordingly,
Renison Bell Mine, Tasmania the sulfide stream was kept separate and stored in a lined,
The first discovery of tin in the Renison Bell area of western water-covered facility. This allowed the main tailings stream
Tasmania was made in 1890, with mining having been almost to be dealt with as super-thickened tailings in a stable storage
continuous from that time. Mining of the sulfide ore was faced with 27 m wide zone of NAF waste rock. Using the
limited by poor recoveries until the development of the sulfide waste rock allowed construction of a conservative structure
flotation process in the 1930s. Mining remained a marginal with very low risk. Modelling showed that the tailings mass
proposition until a major four-year development program in would become unsaturated and risk generation of AMD if the
the late 1950s developed one of the largest underground tin separation had not occurred. The scale of the different tailings
mines in the world. storages is shown in Figure 6 and the simple cross-section in
The Renison process is a two-stage flotation process similar Figure 7.
to that which is shown in Figure 4.
Brukunga
The tailings streams were originally mixed but separated
in 1996 when the potential benefit of the low sulfide tailings Brett et al (2011) describes an example of co-disposal at the
Brukunga Mine. The Brukunga Mine in the Adelaide Hills,
was recognised. The sulfide component of the tailings is high
40 km east of the city, was a source of pyrite for fertiliser
in highly reactive pyrrhotite, which resulted in rapid onset
manufacture driving the development of South Australian
of oxidation and generation of AMD. By providing a second
agriculture during the mid-twentieth century. The legacy is
tailings line for low sulfide tin-float tails, as shown in Figure 5,
an ongoing source of AMD requiring continuous collection,
a discharge system was developed to allow progressive
treatment with lime and disposal of sludge. The sources of
covering of sulfide tails with tin-float tails to prevent exposure
AMD include a tailings storage facility, waste rock dumps
of the sulfides. Trials of revegetation of the tin-float tails have and the exposed mine highwall. The site is being managed by
been successful, allowing development of a closure plan that the South Australian Government, which initiated a study to
does not require capping of the Renison tailings dams with review options for remediation of the site using a Technical
imported soils. Advisory Group (TAG) of specialist consultants. The TAG’s
recommended solution comprised remixing the wastes and
Unnamed magnetite mine constructing a new disposal structure allowing continual
The ore processing for a proposed magnetite mine resulted saturation of the co-disposed material. This was considered to
in sulfides being floated with the concentrate then removed offer the lowest risk, most cost-effective and most achievable

D Brett
FIG 4 – Tin process flow diagram.
the waste materials, to determine the density achieved and to

assess the impact of saturation on the rate of oxidation.
The geochemical testing indicated that crushed limestone
addition of 1.0 per cent by dry mass would provide sufficient
neutralisation capacity to deal with current and presaturation
sulfuric salts within the waste rock/tailings mixture, with a
stoichiometric safety factor of approximately 4 SF.
The particle size distribution (PSD) curves for the various
mixes are summarised in Figure 8, in which the ‘mixture’
comprised of waste rock and tailings, in the dry mass ratio
63:37, plus one per cent crushed limestone. This ratio is based
on the conservatively estimated volumes of the existing waste
FIG 5 – Twin tailings pipelines at Renison Mine. rock and tailings. As can be seen, the mixture is somewhat better
graded than the sampled waste rock, and much better graded
strategy for a ‘walk-away’ remediation solution, obviating the than the narrowly- graded tailings, which facilitates achieving
need for ongoing active water treatment. Studies, including a high compacted dry density. Table 1 presents the calculated
field trials of mixed tailings and waste rock to determine saturated hydraulic conductivities. This shows a weighted
geotechnical properties, demonstrated that this can be increase in density of the co-disposed waste of 15 per cent and
achieved despite the relatively dry climate of the site with an a halving of the permeability. Results would be expected to be
average rainfall of 600 mm but a significant water deficit over more significant if coarser waste rock was involved.
most of the year.
Two field trials, each measuring about 15 m long by about
The final concept required that the majority of waste rock 4 m wide by about 0.5 m high, were carried out on 3 June, 2009
and tailings to be excavated from the current storage areas on the Brukunga Mine bench, comprising the compaction of
and ‘co-disposed’ by mixing and compacting them in areas mixed waste using a 10 t, smooth-drum, self-propelled roller.
where local streamflow can be diverted to maintain saturation. The co-disposal trial, indicated that a compacted dry density
Crushed limestone would be added to the co-disposed wastes of up to about 1.85 t/m3 (up to 97 per cent of laboratory
to neutralise stored oxidation products. The co-disposed standard maximum dry density (MDD)) is achievable at a
waste would be placed behind containments comprising moisture content of 16.5 per cent, which is 3.2 per cent wet of
embankments of clay and rock with appropriate foundation laboratory standard optimum moisture content (OMC). This
preparation to ensure minimisation of seepage from the corresponds to a calculated degree of saturation of 90 per cent
saturated wastes placed upstream. The co-disposed wastes (assuming a specific gravity of 2.8).
would be covered with a horizontal layer of permeable crushed The materials used in the field co-disposal trials (reworked
rock overlain by a soil cover suitable for vegetation growth, with waste rock, surficial tailings excavated from the centre of the
recreated ‘creeks’ over the final surface able to feed water into tailings storage facility (TSF), and crushed limestone) are
the permeable rock layer to distribute water to the underlying shown in Figure 9.
wastes and to maintain them in a saturated condition. The
The compacted waste rock/tailings mixture achieved near-
resulting landform would be landscaped and revegetated, with
saturated conditions. To fully saturate the compacted waste
the landscape including wetlands and natural meandering
rock/tailings mixture would require only a 3.3 per cent
creek lines, with a range of native vegetation.
addition of water on a volume basis.
The key to the success of the proposal is the maintenance of
Due to the higher dry densities achieved in the field,
saturated conditions within the waste under the most extreme
permeability testing gave saturated hydraulic conductivities
environmental conditions, facilitated by the co-disposal of the
an order of magnitude lower than was achieved in the
various wastes to produce a better grading of material able to
laboratory (about 1.0 × 10–8 m/s in the field).
be compacted to a significantly higher dry density than the
component parts and with improved engineering parameters, Oxidation tests on the compacted materials showed that:
most importantly lower permeability. A field trial was •• the low hydraulic conductivity, reasonable air-entry
undertaken to assess the feasibility of mixing and compacting value (AEV), high water storage capacity and relatively

FIG 6 – Plan of non-acid forming and potentially acid forming tailings storage facilities.
FIG 7 – Cross-section of non-acid forming tailings storage facilities.
small drying/wetting hysteresis of the compacted waste •• design for avoidance – investigate and understand the
rock/tailings mixture suggest that it would have a geochemical and geotechnical properties of the waste
high capability to hold water and remain saturated (or materials
tension saturated) between rainfall events •• don’t rely on thin cover systems – these are ‘band-aid’
•• the oxidation rate of a compacted co-disposed waste solutions and carry significant risk of failure in the long-
varies approximately linearly with the moisture content term
of the mixture over a range of moisture contents from •• ‘business as usual’ is not good enough where there is a
five per cent (representing desiccated conditions with a risk of AMD
degree of saturation of only 24 per cent) to 19 per cent
•• separating sulfides should be the first priority with
(near-saturated at a degree of saturation of 93 per cent),
sulfides stored permanently underwater – this can
from 1.54 per cent/annum to zero
make disposal of the bulk of the tailings economical and
•• the higher dry density and hence higher degree of safe; this allows a flexible approach to the construction
saturation achieved in the field for a compacted waste and closure of storages for the remaining NAF waste
rock/tailings mixture would be expected to lead materials including relatively free dumping of waste
to lower oxidation rates than those achieved in the rock and economic options for tailings dams including
laboratory using scalped waste rock, and an oxidation upstream construction, cyclone sand and paste dry
rate of close to zero would be achievable in the field stacking which might not be feasible otherwise
following compaction
•• the only other reliable option is to permanently maximise
•• full saturation of the compacted waste rock/tailings saturation levels; this may be assisted by:
mixture in the field would ensure an oxidation rate close
•• minimising segregation (paste?)
to zero.
•• maximising density – dewatering, compaction
CONCLUSIONS •• thinking ‘co-disposal’.
Leading practice for mine waste management most likely Other thoughts are:
to become more appropriate in the near future can be •• do it properly and a viable mine closure plan might be
summarised as follows: possible after all

D Brett
FIG 8 – Measured particle size distribution curves and estimated mixture for co-disposal of trial waste rock, tailings and crushed limestone.
A B C
FIG 9 – Materials used in field co-disposal trials: (A) reworked waste rock, (B) surficial tailings (C) crushed limestone.
Table 1 Bois, D, Poirier, P, Benzazoua, M, Bussière, B and Kongolo, M, 2005.

A Feasibility Study on the Use of Desulphurized Tailings to Control
Properties of individual wastes and co-disposed wastes.
Acid Drainage, Bulletin 98, (1087):74 (Canadian Mining and
Metallurgical).
Waste Moisture Dry Saturated hydraulic
content density conductivity (m/s) Brett, D, O’Kane, M, Scott, P, Taylor, J, McCleary, M and Williams,
D, 2011. A water-covered waste dump in an arid climate? A
(%) (t/m3) remediation concept for Brukunga Mine, in Proceedings Mine
Scalped (<4.75 mm) trial 10.3 1.62 4.8 × 10-7 Closure 2011, pp 623–633 (Australian Centre for Geomechanics:
waste rock Perth).
Trial tailings 12.8 1.43 8.2 × 10-7 Bruckard, W J and McCallum, D A, 2007. Treatment of sulfide tailings
from base metal and gold operations – a source of saleable by-
Co-disposed WR:T = 63:37 14.0 1.78 3.0 × 10 -7
products and sustainable waste management, in Proceedings
World Gold Conference, pp 85–92 (The Australasian Institute of
•• beware of NPV accounting that will put off decisions on
Hesketh, A, Broadhurst, J L and Harrison, S T L, 2010. Mitigating the
closure until it’s too late.
generation of acid mine drainage form copper sulphide tailings
impoundments in perpetuity: a case study for an integrated
REFERENCES management strategy, Minerals Engineering, 23:225–229 (Elsevier).
Aubertin, M, 2010. Industrial NSERC Ecole Polytechnique – UQAT Robertson, A M, 2011. Top 10 things that go wrong with plans for
Chair, 2005, unpublished presentation referenced in GARD mine closure, keynote address to Mine Closure 2011, Alberta,
Guide, 2010, International Network for Acid Prevention (INAP). September.
Available from: <http://www.gardguide.com>.

Contents
Sustainability, made easy – a business

improvement case study
G Corder1, N Currey2 and G Becker3
ABSTRACT
The Sustainable Operations (SUSOP) framework is a technique that helps incorporate
sustainable development principles and objectives into the design and operation of
businesses. Developed through the Co-operative Research Centre for Sustainable
Resource Processing (CSRP), it brings a design methodology to sustainability, in a
similar way to what hazard and risk techniques such as the hazard and operability
study (HAZOP) have done for safety.
In 2012, Metallica Minerals Limited (‘Metallica’) adopted the SUSOP framework
to help identify opportunities and risks that could directly improve or affect its
scandium-cobalt-nickel (SCONI) project. This project has the potential to become
one of the world’s largest producers of scandium oxide (‘scandia’), which is used to
stabilise the electrolytes in solid oxide fuel cells, but more importantly strengthens
(and allows superior weldability of) aluminium alloys. Located north-west of
Townsville in Queensland, the proposed plant site will be located adjacent to the
previously mined and partially rehabilitated Greenvale Nickel Mine.
By applying the SUSOP framework through a series of multidisciplinary workshops
and offline analysis, Metallica was able to identify numerous sustainability
opportunities and risks. Opportunities were clustered together to form concepts to
develop a more holistic understanding of the project’s overall potential contribution
to sustainable development as well as determine where key risks could be mitigated.
Four key concepts were identified:
1. Waste and Water
2. Workforce
3. Community and Enterprise Development
4. Energy and Greenhouse Gases (GHG).
Metallica included a range of initiatives that could deliver enhanced environmental,
social and community outcomes into its mine development plan. These initiatives
included measures to reduce reliance on water, reduce the volume of waste, provide
local opportunities for training and skilled employment and create support for local
enterprise development. These measures also resulted in a significant reduction in the
capital cost estimate of the project.
This paper provides an overview of the SUSOP framework, details its application
to the SCONI project and discusses the outcomes that will help establish the project’s
strong sustainable development credentials and improved business outcomes.
INTRODUCTION
This paper presents the application of the SUSOP framework to Metallica’s SCONI
project. The SUSOP framework is a project risk and opportunity framework that is
applied in the early phases (concept, prefeasibility and feasibility) of industrial projects
(mining, chemical, energy, oil and gas). It uses corporate sustainability principles as
the basis for identifying innovative opportunities that deliver sound environmental,
community and social outcomes as well as identifying risks that have the potential
1. FAusIMM(CP), Principal Research Fellow, to impact on the environment, local communities and broader society. The approach
Sustainable Minerals Institute, The University utilises a structured process of multidisciplinary workshops and supporting analysis.
of Queensland, Brisbane Qld 4072. Early applications of SUSOP were presented at the MetPlant 2011 conference by
Email: g.corder@smi.uq.edu.au Corder and Green (2011) and have been reported elsewhere (Corder, McLellan and
2. MAusIMM, Director, Sustainable Mining Green, 2012; Corder et al, 2012; Corder, McLellan and Green, 2010).
Strategies Pty Ltd, Cleveland Qld 4161. Email:
ncurrey@sustainableminingstrategies.com.au During the study, Metallica was committed to understanding how its activities,
products and operations align with global efforts to move towards a more sustainable
3. FAusIMM(CP), Principal, Gavin Becker and
Associates, Ormiston Qld 4160.
use of resources. The foundation principles behind its activities include:
Email: gavin@gbecker.com.au •• working closely with local communities in which the company operates
427
•• preventing where possible, or minimising, adverse of the CSRP thus resulted in the development of the SUSOP
environmental impacts framework (Corder and Green, 2012).
•• reducing the project’s environmental footprint by
continually improving efficiency Features of the Sustainable Operations (SUSOP) framework
•• adopting a safe and environmentally conscious lifestyle This section provides a brief overview of the framework
both at work and at home highlighting its features, application and typical outputs
and outcomes. Other publications provide further detail and
•• endeavouring to protect and to restore biodiversity background on the SUSOP framework (Corder et al, 2012;
through land stewardship and rehabilitation. Corder, McLellan and Green, 2012; Corder and Green, 2011;
To help Metallica meet these aims for SCONI, it applied AACE, 2005).
the SUSOP framework to help identify the opportunities and SUSOP is a project risks and opportunities framework
risks that could directly improve or affect the overall project to facilitate a proper contribution to sustainability by the
performance. industrial facilities being studied, designed, built or operated.
The SUSOP framework has three major elements:
PROJECT OVERVIEW 1. Sustainability opportunities and risks identification
At the time of this SUSOP study, SCONI was Metallica’s (SUSID), which includes a significant characteristic,
flagship project. SCONI had the potential to become one of ‘new ideas’ generation, and is made up of four steps:
the world’s largest producers of scandium oxide. Metallica •• step 1 – familiarisation with sustainability concepts
recently announced a farm-in agreement by Australian Mines and project context
Limited (Metallica Minerals Limited, 2016). SCONI is located
•• step 2 – goal scoping and opportunities and risks
in North Queensland, approximately 250 km from Townsville
identification
and consists of several nickel-cobalt rich and scandium-rich
lateritic deposits. The three main deposits in the southern •• step 3 – analysis of sustainability opportunities and
area of SCONI are Greenvale, Lucknow and Kokomo. The risks
Greenvale deposit is on the previously mined and partially •• step 4 – prioritisation of sustainability opportunities
rehabilitated Greenvale Nickel Mine. and risks.
The project proposed to mine the remaining nickel, 2. Preparation of action plans for conducting a detailed
cobalt and scandium laterite resources at Greenvale with evaluation of the shortlisted or high-priority opportunities
supplementary ore feed from the Lucknow and Kokomo and risks.
satellite deposits. At a cut-off grade of 1.0 per cent Ni eq, 3. Decision support for assisting with decision-making at
Greenvale and Kokomo contain approximately 23.3 Mt of project toll gates.
0.72 per cent nickel, 0.09 per cent cobalt and 63 g/t scandium All elements are supported by the SUSOP knowledge base
(167 kt nickel and 21 kt cobalt), while Lucknow comprises which includes information on sustainability frameworks
10.1 Mt of 145 g/t scandium (2189 t scandium oxide) with and principles, details on case studies, relevant sustainability
minor nickel and cobalt. tools and databases to assist in the evaluation and assessment
Mining would require only minor drill and blast, and then stages, resources for workshops and relevant public domain
truck and shovel operations. The project plan was to construct information and data. A schematic summarising the SUSOP
a 750 kt/a high pressure acid leaching (HPAL) plant on the site framework is presented in Figure 1.
to process the ore for more than 20 years. The residue was to be The main outputs from a SUSOP study are presented in the
neutralised and stored in an engineered residue storage facility. Sustainability Register. It works in a similar manner to the
A scandium focused phase, treating just 200 kt/a of conventional risk register and includes:
Lucknow ore, was proposed as the initial development •• opportunities for improving the contribution to societal
project (Metallica Minerals Limited, 2013). An on-site sulfuric sustainability and business performance of the project
acid plant can generate power for use in the processing plant •• supporting Sustainable Development Balance Sheets
and associated infrastructure, as well as supply steam for the for top ranking opportunities to schematically show the
HPAL plant and sulfuric acid lixiviant. positive and negative impacts across the ‘five capitals
model’ of sustainability, which comprises natural,
SUSTAINABILITY FRAMEWORK OVERVIEW human, social, manufactured and financial capitals
•• sustainability risks that could potentially impact on the
Background project’s viability
The SUSOP framework was a research outcome from the •• action plans for each item (opportunities and risks) in
CSRP. The CSRP, which was formed in 2003, brought the register before proceeding through the next project
together industry, government and researcher stakeholders to toll gates.
investigate ways of improving the sustainable development
performance of the Australian and international minerals SUSOP workshop participants
industry. While its initial focus was on several areas of a The mix and nature of the participants will vary depending
technical nature, such as less energy and water consumption on the stage that the project is at and the type of project being
and less waste production and emissions, there was a considered. Typically, core project staff (eg process engineers
developing parallel stream of work concentrating on and plant designers) plus environmental and community
investigating various ‘sustainability frameworks’. The focus of experts are part of the SUSOP study team. A trained SUSOP
this work was on measuring the effectiveness of sustainability practitioner directs the overall study and a trained SUSOP
initiatives and driving better behaviours, in a similar manner facilitator runs workshops. The team should include people
to how safety analysis and statistics started the drive towards of differing backgrounds, experience and attitudes, and
safer workplaces more than 30 years ago. This expanded focus expansive, unconstrained, lateral thinkers should be included

Sustainability, made easy – a business improvement case study
(SD frameworks and principles, public domain information and data,

Familiarisation with Sustainability
Concepts and Project Context
details on other SUSOP® case studies, relevant SD tools)

Opportunities and Risks
SUSID™
Identification
Analysis of
Knowledge Base
Prioritisation of
Preparation of Action Plans
Sustainability SD Balance
Decision Support
Register™ Sheets™
FIG 1 – Sustainable Operations framework.
and encouraged. This greatly enhances the potential for comprehensive understanding of community issues and
generating innovative and alternative opportunities as well concerns. As a result, they provided very beneficial context
as identifying possible sustainability risks. on the local area which was important in understanding the
potential risks and the possible opportunities.
APPLYING SUSOP TO THE SCONI PROJECT This initial workshop created a common understanding
amongst the participants of the project’s key characteristics
Given that the SCONI project was in the enhanced scoping
and features and core sustainability knowledge, such
study phase, it was most appropriate to apply the SUSID
as SUSOP’s default sustainability (five capitals model)
part of SUSOP. The other two steps of SUSOP are typically
framework (Forum for the Future, 2005), the International
applied at later project development phases when more data
Council on Mining and Metals ten sustainable development
and details are available.
principles (ICMM, 2003) and Metallica’s Environment Policy
The process and outputs from applying SUSID to this and Indigenous and Community relations policy.
project are presented in the subsections below. Using this knowledge of the project and sustainability
concepts, the participants established the linkages between
SUSID step 1 – familiarisation workshop the project and key business, community and environmental
Prior to the first workshop a briefing document discussing the impacts.
project context and the SUSOP approach was prepared and The aim of this exercise was to identify the potential positive
issued to the workshop participants. impacts and potential negative impacts of the project across
The SUSID step 1 workshop was held over half a day and, the five capitals model. The five capitals model, developed by
excluding the SUSOP facilitator, there were ten workshop the not-for profit organisation, Forum for the Future, utilises
participants comprising nine Metallica personnel and one the concept of capitals, briefly described below, to organise
person from Metallica’s consulting engineering company. The sustainability thinking (Forum for the Future, 2005):
positions of the workshop participants are presented in Table 1. •• Natural capital is the basis not only of production but of
While there was no community relations person in the life itself and is any stock or flow of energy and material
workshop, the CEO, site manager and EIS manager had a that produces goods and services. It includes:
•• resources (renewable and non-renewable materials)
Table 1 •• sinks (that absorb, neutralise or recycle wastes)
Job positions of first workshop participants. •• processes (such as climate regulation).
Corporate development manager •• Human capital consists of people’s health, knowledge,
skills and motivation, which are needed for productive
CEO
work.
Consulting process manager •• Social capital concerns the institutions that help maintain
EIS manager and develop human capital in partnership with others;
Engineering manager eg families, communities, businesses, trade unions,
schools and voluntary organisations.
Engineering project manager
•• Manufactured capital comprises material goods or fixed
Exploration and geology manager assets which contribute to the production process rather
Managing director than being the output itself – eg tools, machines and
buildings.
Site manager
•• Financial capital plays an important role in our economy,
Tenement manager enabling the other types of capital to be owned and

traded, and is representative of natural, human, social Each participant selected up to ten opportunities they
or manufactured capital; eg shares, bonds or banknotes. considered would have a high ‘achievability’ – that is, there
Table 2 presents examples of some of the impacts from the is good potential to implement the opportunity now or in
linkage exercise. the future. Achievability is a broad term and is related to a
series of factors including cost, resources, timing, logistics etc.
By considering these possible positive and negative impacts,
For example, a cost-effective commercially proven, broadly
the participants began to scope out the range of issues that
accepted technology with minimal environment impact could
could affect the project. This then formed a sound basis
be considered to have a high ‘achievability’ rating.
for the next step in SUSID, ie goal scoping and identifying
opportunities and risks. The results of this opportunities rating exercise are presented
on the impact-achievability plot of the opportunities shown in
SUSID step 2 – goal scoping and opportunities and risk workshop Figure 2. The lines on the graph divide it into four quadrants
– high and low in terms of positive impacts, and high and
The same participants from the SUSID step 1 workshop
low in terms of achievability. The positions of the lines are
participated in the SUSID step 2 workshop with the addition
determined by the number of opportunities and the number
of Metallica’s Process Manager who was not available for the
of participants.
first workshop. This workshop ran over a full day and was
held two days after the first workshop. Eight opportunities lay in the large positive impact and
high achievability quadrant in the top right-hand corner these
The main aim of this workshop was to employ the SUSID
being identified in Table 3.
approach to identify opportunities and risks that could
directly improve or affect the SCONI project.
Rating of risks
The key steps comprised:
A similar process was then conducted with the risks. Each
1. establishing sustainability goals that would provide the participant selected up to ten risks that they considered
basis for the identification of opportunities and risks would have a high ‘negative impact’ and up to ten risks that
2. identifying opportunities that could achieve one or they considered would have a high likelihood of affecting
more goals and risks that could prevent one or more the project’s outcome. A risk was considered to have a high
goals being achieved ‘negative impact’ if a large negative consequence occurred as
3. conducting initial ratings of the opportunities and risks. a result of the risk. Likewise, a risk was considered to have
high ‘likelihood’ if it is probable that the risk would occur.
The first two steps above were conducted in groups using a
structured prompting process. This produced nearly 50 goals The results of this exercise are presented in the negative
across the five capitals, which then led to the identification impact-likelihood plot of the risks shown in Figure 3. Similar
of nearly 50 individual opportunities and over ten individual to Figure 2, the lines on the graph divide it into four quadrants
risks. – high and low in terms of negative impacts, and high and
low in terms of likelihood. Similarly, the positions of the lines
Rating of opportunities are determined by the number of risks and the number of
participants.
Each participant selected up to ten opportunities they
considered would have a high ‘positive impact’ – that is the Five risks lay in the large negative impact and high
scale of the benefit of the opportunity, should it be realised, is consequence quadrant as listed in Table 4.
large (recognising that it might not be able to be implemented At the end of the SUSID step 1 and step 2 workshops, the
now). This benefit might be in one or more capitals but sustainability register for the SCONI project was created and
generally would not benefit one capital at the expense of all the outputs from both workshops were entered into the
another capital. register.
Table 2
Possible impacts across five capitals.
Capital Positive impacts Negative impacts

Natural Water source for wildlife Short term environmental impacts; eg noise, dust etc; generation of CO2
Reuse of water Impact on environmental values
Nurseries
Nature flora
Possible aquaculture
Human Employment Population growth (drive-in/drive-out, fly-in/fly-out)
Increased skills Changes in ‘life style’
Cultural heritage knowledge base
Social Population increase will bring in more skills; eg teachers Change in demographics
Increase in services and jobs; eg supermarkets, emergency services Rents and land values go up
Improved communications Competition with farms to get labour
Increased traffic
Manufactured Increases in housing No market for magnesium sulfate
Increase in water/communications/power infrastructure
Possible limestone works
Financial Profitability Scandium oxide uncertain price
Injection of funds to region with flow on benefits to state and national governments Ni and Co not seen as ‘green’ metal
230 persons employed directly in operation

Opportunities
12 Achievability vs Positive Impact
O31
10 Hi - Hi
Lo - Hi
8
Achievability Votes
O22 O32 O20 O33 O34

O23 O21
O19 O35
6
O26
O01 O03 O13 O02 O06 O15

4 O12 O30 O48
O42 O04 O27
O07 O40 O05 O28 O24 Hi - Lo

2 O46 Lo - Lo
O29
O08 O09 O36 O43
O44
O10 O11 O16 O45 O14
0 O25 O47
0 2 4 6 8 10 12 14
O17 O18 O37 Positive Impact Votes
O38 O39 O41
FIG 2 – Graph of positive impact versus achievability for opportunities.
Table 3
Large positive impact and high achievability.
ID Title
O20 Establish high-technical facilities, such as improved communications
O21 Create best practice rehabilitation program; eg develop a plan to rehabilitate the site during operations
O26 Demonstrate a Scandium fuel cell (pilot or large scale) on site with a learning centre / tourist attraction (‘walk the talk’)
O31 Conduct ore beneficiation studies
O32 Create training schemes for Traditional Landowners
O33 Reuse waste dump as supplementary feed
O34 Reduce residue storage area by extracting moisture, improving dewatering
O35 Produce saleable by-products; eg develop a marketable magnesium product and maximise water recycling and efficiency
SUSID step 3 – analysis of opportunities and risks 4. Community and Enterprise Development concept –
In SUSID step 3, the opportunities and risks that were which envisages opportunities to enhance the quality
identified in the SUSID step 2 workshop were initially of life and services, as well as developing business
reviewed to identify any inconsistencies. enterprises in the Greenvale region.
To develop a more holistic understanding of the overall Importantly, several key risks could be reduced or possibly
potential contribution to sustainable development, mitigated from some of these concepts:
opportunities were grouped into clusters. These clusters •• lack of water (high risk – see Table 4) – utilising water
represent potential dependencies between opportunities, opportunities and dry stacking of residue from Waste
such as new technologies supporting new enterprises and and Water concept could create a substantial reduction
enhancement of local skills, and are called concepts. Key risks on the reliance of water from the natural environment.
were then reviewed to determine if they could be mitigated or This potential reduction is estimated to be from 3.5 GL/a
their impact reduced by any of the concepts. to 1.1 GL/a
By applying the procedure in SUSID step 3, four main •• lack of appropriately experienced staff (high risk – see
concepts were generated for the SCONI project: Table 4) – the cluster of opportunities in the Workforce
1. Waste and Water concept – which focuses on reducing concept should help reduce significant issues related to
the footprint of the residue storage facility and recycling lack of skilled workers
and re-using water. •• change in landform and no net benefit to environment – the
2. Workforce concept – which envisages a combination of cluster of opportunities in the Waste and Water concept
opportunities to enhance workforce skills and reduce would allow for superior rehabilitation practices that
staff turnover. should minimise the chances of these risks eventuating
3. Energy and GHG concept – which focuses on reducing •• labour shortage for local farmers – specific opportunities
energy usage and greenhouse gas generation. in the Workforce concept should help facilitate

Risks
Likelihood vs Negative Impact
16
14 R1
Hi - Hi
12 Lo - Hi
R4 R2
10
Likelihood Votes
R3
8
R5
6
4 R6 R7
Lo - Lo Hi - Lo
2 R8
R12 R9
R11
0 R10
0 2 4 6 8 10 12 14 16 18 20
Negative Impact Votes
FIG 3 – Graph of negative impact versus likelihood for risks.
Table 4 Using this approach has proved very useful to illustrate

Large negative impact and high consequence risks. the interdependencies between opportunities and risks
and demonstrate schematically a holistic picture of the
ID Description sustainability benefits of a concept. In some cases, several low
R01 No joint venture funding in time due to high Australian dollar rating opportunities can be clustered together to produce a
concept which has significant sustainability benefits. The
R02 Lack of water due to a dry season and need to source water from individual opportunities in isolation can have limited value
Burdekin River but when matched with other similar or complementary
R03 Loss of Scandium intellectual property or competing intellectual property opportunities a much stronger case for implementation is
R04 Lack of appropriately experienced staff presented.
R07 Cost of water recycling SUSID step 4 – prioritisation of opportunities and risks
The final step in this study was conducting SUSID step 4
employment opportunities on the SCONI site as well as where the outcomes from SUSID step 3 were presented to
for local farmers and other businesses Metallica in a wrap-up half-day workshop.
•• competition between businesses – specific opportunities The participants included a subgroup of the Step 1 and
in the Community and Enterprise Development Step 2 workshop participants and comprised:
concept should help businesses adapt to the changing
demographics and business conditions as the Greenvale •• an EIS manager
region grows. •• a consulting process manager
To illustrate the process of applying the cluster approach •• an exploration and geology manager
and demonstrate its advantages, the Waste and Water concept •• a tenement manager
is discussed in more detail here.
•• a CEO
Given that some of the high rating opportunities were related
to employing best practices for rehabilitation, minimising the •• an engineering manager.
residue storage footprint and maximising water recycling, The aim of this step was to agree on the prioritised list of
clusters were generated using opportunities related to the concepts. The key criteria for prioritising the concepts were
residue area and reducing water usage to produce the waste based on the estimated benefits of the opportunities and the
and water concept, schematically shown in Figure 4. The lines potential reduction or mitigation in high rating risks. On this
represent dependencies between the opportunities. Further basis, the ranking of the concepts were:
inspection of the clusters identified that a key risk, ‘lack of 1. Waste and Water concept – five high-high opportunities,
water’, and other risks such as ‘changes to landform’ and one high risk and two other risks.
‘no net benefits to the environment’, could be significantly
mitigated by this concept. 2. Workforce concept – two high-high opportunities, one
high risk and two other risks.
The key benefits of this concept are a reduced reliance
on local water sources by recycling water from the reside 3. Community and Enterprise Development concept –
facility which will result in a reduced volume and footprint three high-high opportunities and one other risk.
for the residue storage facility and will allow for progressive 4. Energy and GHG concept – no high-high opportunities
rehabilitation during the life of the operation. and no risks.

FIG 4 – Waste and Water concept (note: opportunities are in oval shapes – lightly shaded red with black text represents high-high
opportunities; darkly shaded blue with white text represents on-site opportunities; and white background with black text represents
off-site opportunities; risks are in rectangles – shaded represents high-high risks and white represents all other risks.)
It is important that all opportunities and risks are not each of the five capitals – natural, human, social, manufactured
discarded from the Sustainability Register at this early stage of and financial capitals. The overall results are presented in the
the project even if they have been assigned low ratings. As the preliminary SD balance sheet in Figure 6. Given that the project
project progresses and more data and information becomes is in an early phase of development these ratings are indicative
available, and as any changes in the project plan occur, these and can be refined as the project progresses.
opportunities and risks might create more benefits than were As a reference, estimates of the impacts across the five
initially considered possible. capitals for this project were made if a ‘business-as-usual’
One of the valuable aspects of the prioritisation workshop approach had been adopted. This is shown in the left-hand
was the contribution from participants on the additional diagram of the preliminary SD balance sheet in Figure 6.
potential benefits related to each of the four concepts. The key It illustrates that there would be negative impact on the
benefits are summarised in Table 5. natural capital, minimal impact in terms of human and
social capitals and positive impact for manufactured and
Sustainable Development (SD) Balance Sheets financial capitals.
The impact of the risks and opportunities was estimated across If the opportunities with high positive impact and high
each of the five capitals using Figure 5. That is, each risk and achievability were implemented (middle diagram in
opportunity was given a rating for its estimated impact on Figure 6), there would be an improvement across each capital
Table 5
Key benefits of concepts raised in prioritisation workshop.
Concepts Benefits
Waste and Water Capital costs could be approximately 50% lower compared with a wet residue storage facility.
Recycled water from the residue could reduce the natural water draw from the Burdekin River from 3.5 GL/y to 1.1 GL/y.
Residue storage facility could have a smaller footprint, as well as fewer and smaller ponds.
Residue storage facility should be easier to rehabilitate during operation resulting in faster closure.
By not drawing water from the Burdekin River, pumps and piping capital costs could be offset by $80M.
Magnesium sulfate would be removed and might be sold as a by-product in the agriculture industry.
Workforce Training hub could help deliver skilled workers both for the SCONI project and the broader community; eg provide skilled labour for farmers.
Training hub could be based in Townsville to capture the wider population.
Roster to allow farmers to work at the site and continue working on their properties, thus demonstrating that miners and farmers can work together.
Source local workers and up-skill them from similar style operations in Townsville or in Papua New Guinea.
Energy and Greenhouse A larger sulfur plant could generate excess electricity and feed into the grid, which could help stabilise the system.
Gases Excess acid thus produced could be used on a nickel laterite heap leach to enhance production and revenues.
Waste oil could be reused in the lime kiln, if constructed.
Community and Mt Isa and Townsville Economic Zone (MITEZ) have a 50-year vision for the region and Metallica should work with MITEZ.
Enterprise Development Local business opportunities could include establishing:
• accommodation camps and messes near the mine
• a light vehicle maintenance workshop and other maintenance spin off businesses to service the mine and surrounding region.

Many ‘good’ projects lose …and fail to capitalise on

support here… gaining support here
Community/Regulator Reaction
Controversial Little Reaction Likely strong support
Unacceptable Poor <------------------Business as Usual-----------------> Innovative Breakthrough
Impacts Negative Neutral Positive
Extreme High Significant Moderate Minor Negligible Minor Moderate Significant High Extreme
Score
-5 -4 -3 -2 -1 0 1 2 3 4 5
FIG 5 – Impact ratings.
Financial Financial Financial
Manufactured Manufactured Manufactured
Social Social Social
Human Human Human
No delta
+ve delta Natural

Natural Natural
-ve delta
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
Impact Impact Impact
Standard outcomes from Improved outcomes from
implementation of opportunities Consequences of Risks Occurring
business-as-usual approaches
FIG 6 – Preliminary Sustainable Development balance sheet.
compared with its corresponding capital in the ‘business- helped develop a more holistic understanding of the project’s
as-usual’ approach diagram. This indicates that there would overall potential contribution to sustainable development as
be strong sustainability benefits for implementing the high well as determining where key risks could be mitigated. Four
positive impact and high achievability opportunities. key concepts were identified and ranked in terms of priority
The risks that were rated as having a high negative impact as follows:
and high likelihood are plotted on the right-hand diagram 1. Waste and Water
in Figure 6. This shows the consequence of these risks (if
they occurred) would result in negative financial, social and 2. Workforce
environmental outcomes. However, it is important to note 3. Community and Enterprise Development
that these risks, as discussed in the section ‘SUSID step 3 4. Energy and Greenhouse Gases (GHG).
– analysis of opportunities and risks’, can be mitigated to
various degrees by implementing the relevant identified Each concept has the potential to deliver sustainability
opportunities. benefits for the project. In the Waste and Water concept,
recycling and re-using water could result in a smaller residue
By presenting the outcomes of the SUSOP process in the
storage facility, producing a significant capital cost reduction
form of an SD Balance Sheet, this helps illustrate the areas of
of up to 50 per cent compared with a wet residue storage
improvement and potential areas for concern in a concise and
schematic format. facility, and would also allow progressive rehabilitation over
the lifetime of the operation.
CONCLUSIONS
By applying the SUSOP framework through a structured
ACKNOWLEDGEMENTS
process of multidisciplinary workshops and supporting The authors would like to acknowledge and thank Metallica
analysis, Metallica could identify several sustainability Minerals Limited for granting their permission to publish the
opportunities and risks for the SCONI project in North original paper in the MetPlant 2013 Conference and this paper
Queensland. The clustering of opportunities to form concepts in the Landmark Papers publication.

REFERENCES Corder, G D, McLellan, B C, Green, S R, Bangerter, P B and Van Beers,

D, 2012. Engineering-in sustainability through the application of
AACE, 2005. Cost estimate classification system – as applied in
SUSOP®, Chem Eng Res Des, 90:98–109.
engineering, procurement and construction for the process
industries, in AACE International Recommended Practice, Forum For The Future, 2005. The Five Capitals [online], Forum for the
(Association for the Advancement of Cost Engineering). Future. Available from: <http://www.forumforthefuture.org/
project/five-capitals/overview> [Accessed: 12 March 2012].
Corder, G D and Green, S R, 2011. Integrating sustainability principles
into mineral processing plant design, in Proceedings MetPlant ICMM, 2003. International Council on Mining and Metals 10
2011 (The Australasian Institute of Mining and Metallurgy: Principles [online]. Available from: <http://www.icmm.
Melbourne). com/our-work/sustainable-development-framework/10-
principles> [Accessed: 3 June 2013].
Corder, G D and Green, S R, 2012. Using a sustainability assessment
framework to achieve enhanced legacy outcomes, in Proceedings Metallica Minerals Limited, 2013. ASX Release 28 March 2013
Life-of-Mine 2012, pp 311–317 (The Australasian Institute of SCONI Scandium Project – Positive Pre-feasibility Study (PFS)
Mining and Metallurgy: Melbourne). [online]. Available from: <http://metallicaminerals.com.au/
sites/default/files/SCONI%20Scandium%20PFS%20positive%20
Corder, G D, McLellan, B C and Green S, 2010. Incorporating
pre-feasibility%20study%2028.03.13.pdf> [Accessed: 6 May 2013].
sustainable development principles into minerals processing
design and operation: SUSOP®, Minerals Engineering, 23:175–181. Metallica Minerals Limited, 2016. ASX Release 10 October 2016
Metallica – Australian Mines Joint Venture at SCONI project
Corder, G D, McLellan, B C and Green, S R, 2012. Delivering solutions
[online]. Available from: <http://www.metallicaminerals.com.
for resource conservation and recycling into project management
au/wp-content/uploads/2016/09/161010-ASX-Release-SCONI-
systems through SUSOP®, Minerals Engineering, 29:47–57.
JV.pdf> [Accessed: 22 November 2016].

Contents
Continuous improvement – how changes in

metallurgy can have a substantive impact on
environmental impact and closure: lessons
learnt from environmental auditing
P Mulvey1 and G McMillan2
ABSTRACT
Geochemical characterisation is done based upon the geology and metallurgical batch
testing prior to commencement of the operation of the plant. Tailings repository
design is based upon geochemical assessment of the waste material, as well as the
geomechanical issues associated with storing the waste. The resultant conceptual
and detailed design is incorporated into the environmental impact statement (EIS),
upon which mine approval is based. The entire environmental monitoring and harm
mitigation program, which is usually a condition of the mine lease approval, is based
upon the applicability of the original characterisation.
Tacit to this process is that there is no change in the geochemistry of the tailings,
ie no change in the implications for acid and metalliferous or saline/neutral mine
drainage (AMD/SNMD) generation. But any changes in milling and processing can
have a substantive impact on the geochemical behaviour of the tailings metallurgy,
and potentially render an operation non-compliant within the conditions of the
mine license.
Environmental and statutory approval implications for continuous improvements
of metallurgy need to be understood because unintended consequences could cost
more than the benefits of the improvement, and in some cases cost more than all
the profit made from the mine. Examples uncovered during audit and consultant
practice will be presented together with a process to ensure implications for adverse
environmental harm and license conditions are fully considered.
INTRODUCTION
The role of auditing the environmental performance, and more so the regulatory
compliance of mining operations, has been a feature of the Australian resources
sector since the early 1990s. These audits have generally followed the development of
the ‘systems’ audit approach, whereby an organisation and/or operation is audited
against regulatory requirements and commitments made, and also to demonstrate
continuous improvement and competency of responsible agents in the organisation
(or operation).
While being able to demonstrate compliance with the organisation’s environmental
management systems (EMS) and regulatory commitments, ie the operation can
show that it has been undertaking the monitoring program outlined in its licence
and approvals, this auditing process may not necessarily address changes in the
operations since the original approvals were granted.
Many mines are subject to changes over their operating life, such as pit designs,
metallurgical process improvements, and external stakeholder requirements
(ie changes in water allocations, encroaching residential areas). These changes
are generally noted in annual mine management plans, including environmental
1. MAusIMM, Founder and Senior Principal management plans, however; the impact of these changes may not be felt immediately
Scientist, Environmental Earth Sciences, and as such, the mechanism for mitigating any adverse impact may not be addressed
North Sydney NSW 2059. through a traditional ‘systems’ audit approach.
Email: pmulvey@eesigroup.com
2. MAusIMM, Senior Hydrogeologist, CDM This paper will outline how changes in a major base metal mine’s metallurgical
Smith, Richmond Vic 3121. processes were highlighted through a statutory environmental audit process and its
Email: geordie_mcmillan@hotmail.com implications for the environmental performance of the operation.
437
STATUTORY ENVIRONMENTAL AUDITING OF MINES •• reviewing environmental audits, assessments,

management systems and monitoring undertaken by
Regular and periodic independent environmental auditing
the regulator
of mines is a regulatory requirement in several states and
territories in Australia, and is also encountered through state •• presenting the audit findings to the local community.
or provincial jurisdictions in North America. The mechanism
for directing these audits, particularly in Australia, is often Risk assessments and gap analysis
a statement in approval conditions to the effect that ‘an The risk assessment and gap analysis methodology described
environmental audit must be undertaken every X years by a below was implemented in the early stages of the audit scoping.
suitably qualified independent person or organisation…’.
The audit process generally follows the ISO 14001:2015 Risk assessment process – metallurgy and the environment
Environmental Management System approach, where the The objectives of the risk assessment are to identify
organisation or operation is assessed against its capacity to environmental risks that are considered significant in the
demonstrate compliance; resources, roles, responsibilities and opinion of the audit team, and to evaluate whether the
persons of authority; competence, training and awareness; environmental monitoring and management practices are
communication; and documentation. adequate and appropriate to mitigate the risks of potential
The audit system of the operation being considered herein, environmental impacts.
one of the largest base metal operations in the world, differed The risk register considered the following aspects that
to those described above. The statutory requirement for an required interpretation of the monitoring and not just
independent environmental audit was stipulated in a mine- compliance:
specific amendment to the state’s Mining Management Act,
•• potential duration of impact
which was necessitated due to significant external scrutiny of
the operations’ change in layout, mining and metallurgy. It •• location of impact
should also be noted that the state regulator was also to be •• causes
subject to the independent environmental auditing process, •• existing controls, monitoring or assessment undertaken.
which initially was for an ongoing five year program.
Risk evaluation was conducted on the basis of residual
To independently evaluate the environmental performance risk with known controls in place. The risk associated
of the operation, the annual audit program developed in with each potential impact was determined using a
conjunction with the audit team, the mine operator and the matrix of likelihood and potential consequence whereby:
regulator included: risk = consequence + likelihood (Table 1).
•• review of the mine’s monitoring data, management ‘Consequence’ was determined to be the reasonable
systems, and assessments undertaken during the year maximum impact there may be on the environment if existing
via a statutory compliance assessment, a technical review monitoring and assessment controls were inadequate or
of data and procedures, a site inspection and interviews inappropriate. This consequence was considered with regard
with mine personnel (particularly the environment, to both the location and duration of the impact (Table 2). The
metallurgy and mining managers) reasonable consequence and likelihood of occurrence was
•• updating the audit team’s risk assessment and gap considered for each impact in terms of the scales provided
analysis for the operation in the risk matrix and the results of the risk assessment are
Table 1
Risk matrix.
Likelihood (regardless of potential time latency)

1 2 3 4 5
Certain Likely Possible Unlikely Improbable
1 Catastrophic 2 3 4 5 6
2 Major 3 4 5 6 7
3 Moderate 4 5 6 7 8
4 Minor 5 6 7 8 9
5 Insignificant 6 7 8 9 10
Table 2
Risk rating explanation.
Risk matrix result Risk rating Description

2 to 3 E Extreme – immediate intervention required to eliminate or reduce risk at a senior management/government level.
4 to 5 H High risk – it is essential to eliminate or reduce risk to a lower level by the introduction of monitoring and assessment measures
implemented by senior management.
6 to 7 M Moderate – corrective action required, and monitoring and assessment responsibilities must be delegated.
8 to 10 L Low risk – corrective action should be implemented where practicable, and risk should be managed by routine monitoring and
assessment procedures.

Continuous improvement – lessons learnt from environmental auditing
recorded in the risk register (Tables 3 and 4) along with the Table 3
risk matrix. Key to risk register.
Mine risks have to further be interpreted against the area
and duration of impact. Thus scale and duration become very Location of impact
important. When these factors are considered as well, the single RI Regional impact (>2 km radius outside mining lease)
item with the most extreme risk or the likelihood of moving to
extreme risk is the tailings dam and the waste rock enclosures. OM Impact outside mine lease area (<2 km radius)
Chemical and physical failure of tailings dams can occur WM Wide impact within mining lease boundaries
decades after placement and waste rock hundreds of years after L Localised area within mining lease boundaries
placement. The operation of the tailings dam typically comes
under management of the mill/plant superintendent/manager. P Small point source within mining lease boundary
But once placed the assessment of the dams performance often Potential duration of impact
comes under the environmental department particularly for
monitoring, rehabilitation and closure. This joint responsibility G Geological long term (>100 years)
can often create communication and data gaps, conflict and in L Long term (30–100 years)
the extreme render the operation potentially non-compliant
with initial commitments in the EIS. M Medium term (5–30 years)
S Short term (1–5 years)
Gap analysis – metallurgy and the environment
E Ephemeral/seasonal impact
A gap is defined as ‘a discrepancy between the monitoring
program that is taking place, and the monitoring program that Risk rating number and letter colour coding
should be taking place if the mine’s environmental performance
Black Risk rating has remained the same since the last IM audit
is to be maintained at industry best practice standards’.
Each identified gap was evaluated in accordance with the Red Risk has increased in consequence and/or likelihood since last IM audit
gap analysis process flow chart – developed by Environmental Green Risk has decreased in consequence and/or likelihood since last IM audit
Earth Sciences and included in Figure 1 – and used to
Grey This risk item has been added since the last IM audit
categorise identified gaps as described in Table 5.
All gap categories are considered to have equal weighting;
for example, not undertaking appropriate assessment of Table 4
monitoring data or not undertaking appropriate mitigation Risk register – tailings dam geochemistry uncertainty issue.
measures, a Category 3 gap, may have the same adverse
impact as not monitoring at all (a Category 1 gap). The Risk issue–potential Acid/metals leaching from tailings storage facility
gap analysis register for the tailings dam geochemistry is hazard/loss scenario (TSF) into creek
presented in Table 6.
Potential duration L
METALLURGICAL CHANGES AND THE ENVIRONMENT of impact
– A SHARED Responsibility Location of impact RI
The audit process described above highlighted the significance Causes Known conduit to creek, capacity of tailings to go acid
of metallurgical process on the environmental performance of has been confirmed, but no quantification or timing
the mine through two key functions: has been determined through investigation.
1. field inspections of the tailings dam
Existing controls / Seepage recovery bores
2. interviews with the metallurgy, mining and Shallow Cut-off barrier (ineffective)
monitoring and
environment managers. Monitoring of surface water, groundwater and
assessment
undertaken incoming tailings.
Audit field inspections of the tailings dam Completion of clay cap of cell 1.
During the first two years of the audit process, the audit Geophysical analysis undertaken.
team essentially re-evaluated the conceptual model of the
mine and its interaction with the environment. This involved Consequence 2
disseminating the data and the assumptions made in the Likelihood 1
principal environmental approvals documents and correlating
those with the observations in the field and measurements Matrix result 3
in the years immediately following the change in the mine’s
Risk rating E
operation (that were based on said approvals).
A key aspect of this was the evaluation and assumptions of Additional controls, Ascertain velocity of groundwater (and acid and
the geochemistry of the historic and proposed mine tailings. monitoring, dissolved metals).
Assumptions of the tailings geochemistry were made before assessment or Establish long-term oxidation rate of tailings.
the mine’s expansion and alteration from an underground to actions required Respond to monitoring results of current tailings.
an open cut operation. Geochemistry of tailings is yet to be understood.
Consider acid production within the mine closure plan.
The mine operated from 1995 to 2005 as an underground
Establish likely metal and acid concentrations in creek.
operation. In the EIS the tailings were predicted during
Continue to work towards the eventual recovery and
the feasibility test work to remain alkaline, even though
reprocessing of tailings in the TSF Cell.
sulfides were significant, as buffering capacity was said to
Gain a more comprehensive understanding of the
be substantial. In 2006, the tailings from the underground
geology under the TSF.
operation had an average sulfur content of ~15 per cent and

It should be observed that the assessment described above

1. Is monitoring
undertaken in in 2006 noted that:
accordance with
associated potential
‘..the tailings NAPP value had been increasing since 1998, a
risk? trend that can be attributed to a higher sulfide content lower
No
ANC (acid neutralising capacity) for ore reporting to the
process plant, as the underground mining has progressed
2. through different parts of the orebody’.
Category 1 No other interpretation of the daily record of the mill tailings
Is monitoring sufficient
in design (frequency, discharge pH was provided. What was not discussed was that
No type, location etc.) to the mill after 2006 was intended to become more efficient in
address and mitigate
potential risk? separating concentrate from pyrite.
Iron minerals have colours that are strongly associated with
Category 2 Gap
pH. A yellow-orange colour occurs as iron sulfate minerals
such as jarosite or schwertmannite precipitate at pH around
Yes 3.8. As the minerals further oxidise and the pH is elevated
above 4.5 (often by rainwater), goethite is formed (which is
brown). Abundant protons associated with low pH causes
3.
suspended matter to coagulate and clarify the water. Thus
Is monitoring data/output acid water is very clear.
information assessed,
interpreted and managed to
The audit in 2008–2012 involved site inspections with the
No environmental team and the relevant managers of the facility.
track risk alteration and
evaluate the need for The audit team and the mine inspections team breasted the
improved risk mitigation?
tailings dam wall and viewed orange iron minerals in the
Category 3 Gap tailings and clarified water as shown in Figures 2 and 3. The
geochemist in the audit team informed the environmental
manager that the tailings were acidic. The environmental
Yes manager denied this was the case. It however transpired
during discussions with the mill/plant superintendent that
the tailings leaving the plant had been acid for some time, and
No Gap identified that they had been sending the results to be processed with
the other monitoring data.
FIG 1 – Gap analysis flow chart. Figures 2 and 3 demonstrate the reactivity of the near

surface (ie most recently deposited) tailings in 2009. Despite
net acid producing potential (NAPP) of ~390 kg H2SO4/t assumptions in 2006 and earlier that the tailings decant and
(ie strongly positive potential to produce acidic leachate upon leachate water would be neutral (pH ~7), the ponded water
oxidation), but still were considered to be ‘non-acid forming’ exhibited acidic conditions (pH ~3.8) along with evidence of
(NAF) as the static test work and ‘end-of-pipe’ analysis sulfide oxidation by-products such as jarosite, goethite and
from the mill exhibited neutral to alkaline pH conditions sulfo-salt crystallisation. The team responsible for reporting
(>pH 7.5–8). Thus the test work on the synthetic Stage II compliance were unaware this occurred, ie this constituted a
operation material was ignored. Category 3 gap (Figure 1).
Table 5
Gap evaluation categories.
Gap category Description

Category 1 Monitoring to mitigate potential associated environmental risk is not undertaken.
Category 2 Monitoring is undertaken, but is not sufficient in design – ie frequency, location, type and so on, are insufficient to identify or quantify
potential environmental risks.
Category 3 Monitoring is undertaken and is appropriate in design; however, data/output information is not adequately assessed, interpreted or
managed to appropriately mitigate potential environmental risks.
Table 6
Gap analysis register for tailings dam geochemistry.
# Monitoring area Monitoring gap Gap category Recommendations/comments

1 2 3
Mine site
2 Tailings geochemistry Acid/base accounting × Results be reviewed annually in terms of initial projections of tailings
geochemistry, acid production and long-term weathering effects.
3 Tailings geochemistry Inadequate follow-up × × The seepage prediction investigation work was not appropriately
targeted to ensure a valid assessment of the time prediction for when
the system will go acid, and travel time to creek.

Continuous improvement – lessons learnt from environmental auditing
minus buffering capacity. The data presented in Figure 4

represents samples taken at various depths of the tailings dam
and analysed for NAGP subject to the tailings’ mineralogy
(determined by X-ray diffraction). There are three phases
in the evolution of the geochemistry and acid producing
potential of the tailings:
1. the initial phase; below 7 m depth (ie earliest tailing
deposition), the NAGP is generally ≤-9 kg H2SO4/t and
is considered to be not acid forming (NAF)
2. between 7 and 3 m depth (ie mid-stage of tailings
deposition), there is a general increase in the NAGP
to warrant classification of the tailings as potentially
acid forming (PAF), associated with a slight increase in
sulfides
3. from 3 m depth to surface (final stage of tailings
deposition, which was also final tailings height before
FIG 2 – Sulfide oxidation by-products jarosite, goethite and placement of temporary cover), significant increase in
NAGP, associated with a significant increase in sulfides,
sulfo-salts expressed at surface of tailings dam.
indicating tailings were highly likely to produce acid.
Further examination of the original data, and the follow-
up studies instigated by the audit team as a result of the
observations and findings, indicated that while the tailings
may have had a high ANC content (average ~200 kg H2SO4/T),
this buffering capacity is sparingly reactive as the carbonate
content was more associated with dolomite rather than more
reactive carbonate minerals such as limestone.
Coupled with the increasing sulfide content (as per the
2006 assessment), the low reactivity of the acid neutralising
(carbonaceous) materials in the tailings indicates that the
tailings were increasingly likely to produce acidic leachate
upon oxidation at some point. As the sulfide minerals
include pyrite (FeS2) but also less reactive galena (PbS) and
FIG 3 – Sulfidic tailing by-products from localised oxidation of tailings. sphalerite (ZnS), the uncertainty in the desktop geochemical
evaluation lay in determining the reaction times and rates to
This tailings dam was already leaking and impacting the produce potential acidic leachate upon oxidation. Galena and
local waterholes. A detailed investigation of the tailing data sphalerite do not produce acid when oxidised, whereas pyrite
was requested and undertaken. Tailings were sampled at does. Knowing the mineralogy is essential in predicting what
varying depths in the dam. This data was reinterpreted by the will happen as the tailings weather.
audit team and plotted with depth and is presented in Figure 4. The data review and field observations made were
Net acid generation potential (NAGP) is total oxidisable conducted in the presence of, and in collaboration with,
sulfur minus buffering capacity and differs from net acid the three critical leaders of the operation associated with its
producing potential (NAPP), which is total sulfur (as sulfides) metallurgical processes and the environment:
FIG 4 – Net acid generation potential (NAGP) plotted against depth.

1. the metallurgy manager – responsible for the mill The escalation of this risk was exacerbated by the pre-
(including run-of-mine (ROM) pad) and the tailings existing major seepage issues from the tailings dam wall that
dam, including civil works and water management had resulted in the expression of sulfo-salts and increased
2. the environment manager – responsible for the salinity in an adjacent creek. The seepage through and
environmental performance of the mine, mill and beneath the tailings dam wall occurred within two years
tailings dam, including surface water and groundwater of commencement of tailings deposition, which was again
(seepage) monitoring exacerbated by its unlined construction on an alluvial
floodplain with extensive paleochannels and macro-porosity
3. the mine manager – responsible for the operations of the
in the underlying fractured dolomite bedrock (and aquifer).
open cut mine, disused underground operations.
The improvements in the process had created an
The sporadic oxidation of the tailings at various depths
unappreciated potentially significant regional environmental
and locations within the tailings dam was discussed between
problem, given the pre-existing seepage. Investigations into
the audit team and the metallurgy, environment and
whether acidic leachate from the acid upper tailings will
mine managers. Fortunately, the managers had significant
migrate through the alkaline lower part of the tailings are
experience at the operation and could call upon their own
ongoing.
and colleagues’ recollections of past practices. It was found
that pyrite had not increased in the orebody to justify this
response. Another solution was sought. CONCLUSIONS
The environmental auditing of mines has traditionally
Integrated mine environmental management followed the ISO 14001:2004 and more recently ISO 14001:2015
and metallurgy and auditing Environmental Management System approach, where the
organisation or operation is assessed against its capacity to
As part of the technical aspect of the audit, it was considered
demonstrate compliance, procedural documentation and
imperative to ascertain how and why the observed conditions
continuous improvement. While suitable in some situations,
and data differed from the conceptual model presented
a procedural or compliance audit such as this may not reveal
in the approvals process only three years previously.
operational and technical issues that are adversely impacting
Metallurgically, the issue was to ascertain why the tailings
the overall environmental performance of the mine, and in
were oxidising and exhibiting acidic leachate conditions
turn reducing its capacity to successfully meet its regulatory
when none were expected.
and corporate obligations for compliance, rehabilitation and
Observations of the roles and responsibilities of the various ultimate closure.
mine departments, which is also a function of a compliance or
The adoption of risk assessment and gap analysis techniques,
system audit (ie ISO 14001) revealed the following:
reviewed on a periodic basis, along with a technical review of
•• The mine manager indicated that the latter stages of the environmental monitoring and management (ie not just
underground mining and commencement of open cut reviewed against conditions but an evaluation of trends),
operations had encountered higher sulfide content can provide prioritisation of actions to improve the mine’s
in both the ore and waste rock along the ore zone environmental performance.
boundaries. As stated, this alone could not justify the
To achieve this improvement requires all departments
results in the tailings dam.
of a mine to ‘buy-in’ and understand that all aspects of the
•• The metallurgy manager indicated that over the life of operation contribute to its performance. Although compliance
the operation (~13 years), the mill had undergone two with regulatory conditions and commitments is crucial for
significant upgrades that had improved the efficiency mines, strict ‘end-of-pipe’ monitoring does not account for
of the flotation circuit, which increased the concentrate changes to mining and milling operations that invariably
purity but also the sulfide content of the tailings occur due to improvements in processing technologies,
reporting to the dam. Furthermore, there had been mining fleet management, fuel consumption, water reuse and
several modifications to the operation of the tailings orebody size and quality.
dam, including spigotting and water management that
Where changes occur in mining and metallurgical
resulted in varying rates of consolidation in the dam
techniques at an operation, it is critical to re-evaluate the
cell, and hence development of preferential pathways,
conceptual model of the operation and its assumed interaction
leading to localised oxidation and leachate generation.
with the environment. Communication of these changes
•• The environment manager indicated that their role between the mining, metallurgy and environment functions
(and department) was almost over-committed to are imperative to enable the refinement and adjustment of
managing environmental compliance monitoring and the conceptual model(s), which in turn provides cause and
management, including periodic tailings geochemistry reason to adjust monitoring and management conditions,
testing, dust monitoring, seepage etc. The environment commitments and techniques.
department, at that stage, was not consulted about
changes to the metallurgy and tailings dam management
such as process improvements, dam sizing, water REFERENCE
recycling etc. Department of Industry Tourism and Resources, 2007. Tailings
Management – Leading Practice Sustainable Development Program
As such, the lack of internal communication and for the Mining Industry, (DITR). Update available online from:
integrated planning, particularly between the metallurgy <https://industry.gov.au/resource/Documents/LPSDP/
and environment departments, resulted in the escalation LPSDP-TailingsHandbook.pdf>, 128 p.
of an environmental risk from a medium-to-longer-term
consideration to a near-term-imminent event.

Process improvements
and case studies
Contents
Improving fines recovery by grinding finer

J D Pease1, M F Young2 and N W Johnson3
ABSTRACT
Despite overwhelming evidence to the contrary, many practitioners still believe
that ‘fines don’t float’. This misperception arises from the typical ‘hill shaped’ size-
recovery curves, with high recovery in the mid sizes and low recovery at the fine and
coarse ends. These graphs tell us that fines don’t float well in many plants. But this
doesn’t mean that fines can’t float well. Base metals operations like McArthur River
and Century relied entirely on ultra-fine flotation in conventional cells. With very
complex ores they achieved over 80 per cent recovery into clean concentrates at an
average particle size of 2.5 µm.
Practitioners often attribute losses in ultra-fines to ‘overgrinding’. But paradoxically,
sometimes fines don’t float because we don’t grind fine enough!
The paradox is explained by the different flotation needs of fine and coarse particles.
This was largely ignored in the search for ‘simpler’ circuits with larger equipment.
Most plants now treat all particles together in a ‘soup’ of widely varying particle sizes.
Reagent conditions are set for the dominant coarser particles, which starves fines of
collector. To make it worse, weakly-hydrophobic coarse minerals (eg composites or
iron sulfides) must be rejected from concentrate, but the conditions to reject them –
low collector and high depressants – also hinder recovery of fine liberated particles.
The old concept of having separate ‘sand and slimes’ circuits addressed this, but
is rarely actually practised. A more practical approach is to apply staged grinding,
regrinding and flotation to progressively liberate and recover in narrower size ranges.
While the flow sheet looks larger and more complex, in fact it is smaller and simpler to
operate because it addresses the needs of each particle class and virtually eliminates
circulating loads.
An excellent case study is the use of staged primary grinding and regrinding in
the Mount Isa lead zinc concentrator. The final regrind product was P80 of 7 µm.
Previously, most plant losses had been in the <10 µm fraction. Grinding to produce
more of this fraction increased shifted grade/recovery performance by ten per cent
lead and 16 per cent zinc. Circulating loads dropped, reagent additions dropped
despite the higher particle surface area, and the plant became much more operable
and responsive.
Yes, fines do float, and they float very well indeed.
The Conventional View

Figure 1 is the conventional view of the flotation size-recovery curve. Indeed, almost
every flotation plant reports a similar curve. Fines do float poorly in most plants, and
so good operators carefully avoid ‘overgrinding’ and ‘sliming’ of feed.
This conventional view is supported by plant results. Batterham (2008) reported
Figure 2 for large-scale copper concentrators (note the size scale is linear, not
logarithmic, making the downturn at the fine sizes look more compressed).
THE NEW REALITY

While this conventional view reflects common performance, it does not mean that
fines don’t float. This misperception has been thoroughly disproven by modern
1. FAusIMM, Senior Principal Consulting practice. McArthur River (MRM) and Century mines made their entire production
Engineer, Mineralis Consultants Pty Ltd,
at a P80 of 7 µm and recovery over 80 per cent using conventional equipment. Mount
Brisbane Qld 4066.
Email: jpease@mineralis.com.au Isa Mines, who pioneered the use of ultra-fine grinding in base metals, produced lead
and zinc concentrates at P80 of 15 µm (and components at 7 µm) and over 80 per cent
2. Principal Consultant, Mineralis Consultants
Pty Ltd, Brisbane Qld 4066. recovery. The world’s largest zinc producer, Red Dog, regrinds to 13 µm in its final
Email: myoung@mineralis.com.au stage for zinc production. Platinum Group Metals’ (PGM) operations in South
3. Principal Consultant, Mineralis Consultants Africa use fine regrinding to improve concentrate grade (Rule and de Waal, 2011).
Pty Ltd, Brisbane Qld 4066. An increasing number of copper producers regrind cleaner feed to P80 of 20 µm to
Email: b.johnson@mineralis.com.au improve concentrate quality.
445
is no choice. Figure 4 shows photomicrographs of Broken Hill

ore and McArthur River (MRM) ore at the same magnification.
Broken Hill never had a liberation problem – its metallurgists
could focus on simple grinding and flotation circuits and
rightly avoid ‘overgrinding’.
But for MRM and Mount Isa, no amount of fiddling with
circuit design or reagent testing will help (trust us; we wasted
years of our lives trying). For these ores, there is only one
question that matters: ‘Will this circuit change or new reagent
increase mineral liberation?’ If not, the best thing to do is to
do nothing.
Some Definitions and Perspectives

Communication between different operating plants is often
FIG 1 – Conventional view of flotation recovery by size.
confused because people use different definitions of ‘coarse’,
‘intermediate’ and ‘fine’ particles. An operator of a ‘coarse’
grained orebody may call -37 µm ‘slimes’. To an operator
at Mount Isa or McArthur River this is ‘gravel’. To them,
anything above 20 µm is coarse, between 10 and 20 µm is
intermediate, and less than 10 µm is fine or ultra-fine.
When people say that ‘fine’ particles don’t float well, they
usually mean either -37 µm or cyclosizer C5 (around 10 µm).
But even the finest definition doesn’t explain why so many
operations now routinely produce millions of tonnes a year
of concentrates at an average sizing less than 10 µm at high
recovery from complex ores, in conventional flotation cells
with simple reagent system.
Particle surface area per tonne increases rapidly as size gets
finer. One tonne of 7 µm particles has five times the surface
area of a tonne of 37 µm particles. For 2.5 µm particles, the
surface area triples again. This explains why grinding energy
increases exponentially as grind size decreases and why
finer media with higher surface area is needed for grinding
below about 30 µm, since grinding energy is proportional to
new surface area created. Fines also require more collector
to satisfy the same adsorption density of collector per unit
surface area as coarser particles.
By convention a grind size of 7 µm refers to the 80 per cent
FIG 2 – Recovery by size for major copper concentrators (note linear scale).
passing size (P80). For a regrind size of 7 µm (MRM, Century),
Prominent Hill and Phu Kham and Somincor, are excellent 80 per cent of particles in flotation concentrate are finer than
case studies. Figure 3a (Bennett, Crnkovic and Walker, 2012) 7 µm and 50 per cent are finer than 2.5 µm. Converting this
shows that recovery at Phu Kam keeps increasing as grind to particle volume, 50 per cent of particles by weight means
size gets finer. Figure 3b explains why: liberated fines have that 96 per cent of the individual particles recovered are less
than 2.5 µm. Of the successful bubble-particle interactions,
very high recovery to at least 3 µm but composites have lower
96 per cent occur for 0 to 2.5 µm ultra-fines. Popular
recovery in all size fractions. Finer regrinding produces more
perception is wrong. Fine particles float extremely well in
particles in the liberated, fine, high recovery fraction.
conventional cells.
The grain size of ores being treated is decreasing. There is
increasing pressure to remove fine-grained contaminants such
as pyrite and arsenopyrite from copper concentrates. In the DESIGN of Ultra-FINE Grinding TO SUIT FLOTATION
1990s, lead zinc operations had to adopt ultra-fine grinding to As described above, achieving the right liberation is
survive. In recent years, and into the 2020s, copper operations fundamental. Then fines flotation is straightforward – but it
will become the dominant users. is more sensitive to chemistry conditions than intermediate
(20–75 µm) particles. When you only have a tiny surface area
The industry is fortunate that fines do float so well.
per particle, you can’t afford to lose much to hydrophilic
‘patches’.
Resolving the Paradox Therefore, the control of grinding and chemical conditions
Why do many plants fear fines production, yet an increasing is more important for ultra-fine grinding. Three areas need
amount rely on it? The answer lies in three areas: attention and are explained in more detail:
1. mineralogy and liberation 1. power efficiency
2. operating constraints and strategy 2. classification within the grinding circuit
3. flotation rates and surface area. 3. control of surface chemistry.
Mineralogy is the fundamental driver. Fine grinding is Power efficiency – even with modern high intensity stirred
power-intensive and expensive, so should only be used when mills, ultra-fine grinding is power-intensive. Firstly,
the minerals require it. But for fine-grained mineralogy there you can’t afford to waste energy. Secondly, high energy

FIG 3 – Increasing recovery at Phu Kam by producing more ultra-fines.
FIG 4 – Different grain size of (A) Broken Hill and (B) McArthur River ores (shaded square in each is 40 µm).
means increased pulp temperature which affects surface 1993) and Table 1. The same sized media must be used in the
chemistry and depositions (for example, gypsum exhibits final plant as used in laboratory test work – if the laboratory
reverse solubility with higher precipitation on surfaces test evaluates power using 6 mm media, then using 12 mm
at elevated temperatures). So, it is essential to maximise media in the plant would significantly impair grinding rates.
energy efficiency – below 25 µm this almost certainly means Good classification – this is important to the efficiency of any
using a high intensity stirred mill with fine grinding media; grinding, and is vital for ultra-fine grinding. The target is to
typically 2 mm diameter (or less). Conventional mills using, maximise mineral liberation at the lowest possible power
for example, 25 mm diameter would require uneconomic and draw, which means the sharpest possible size distribution.
impractical grinding energy (Figure 5, from Gao and Weller, This requires either small diameter high pressure cyclones

Pease et al, 2006), with a disproportionate effect on fine

particles which can ill afford to lose hydrophobic surface
area. A successful response to surface contamination after
fine grinding with steel was at Hellyer, where High Intensity
Conditioning was used for attrition cleaning of surfaces
(Holder, 1994). The advent of high intensity stirred mills
superseded this, since these mills are more energy efficient
(fine media), use inert media, and their grinding mechanism
provides clean surfaces by attrition. There are still challenges
– high pulp temperature and low Eh – but the clean surfaces
give the flotation metallurgist a ‘clean sheet’ to work with.
Flotation Circuit Design and Operating Strategy

The view that ‘fine’ particles don’t float is caused by circuit
design and the constraints of operating strategy. Simply,
flotation works best when applied to narrow size distributions.
A 5 µm particle has ten times the surface area of a 50 µm
particle. It has fundamentally different hydrodynamics. Yet
many circuit designs assume they will behave the same and
treat them together in flotation. Texts as old as Taggart (1927)
described the benefits of ‘sand/slimes’ splits into separate
FIG 5 – Breakage rates in different grinding devices and different media size. circuits. This simple concept has been largely ignored in the
push for circuit simplification and larger flotation cells.
Table 1 We are not advocating complicated flotation circuits.
Comparison of various grinding technologies for ultrafine grinding. However, for fine-grained minerals the circuit must be
designed to suit the needs of fine particles, not coarse particles.
Feature IsaMill Tower mills Vertical pin mills The Mount Isa circuit developed into an excellent balance of
Grinding intensity (kW/L) 0.54 0.005 0.15–0.18 the needs of different minerals, relying on several stages of
grinding and flotation. The design principals are:
Residence time to 15 µm (min) 0.6 154 7–9
•• Don’t grind anything finer than needed – fine grinding is
Power usage to 15 µm (kWh/t) 17.4 59.6 37.5–39.0 expensive; technically the best solution for Mount Isa
Media material Various Steel Steel would be to grind everything to 12 µm, but this was not
economic. The solution was to stage grind and float to
Media size (mm) 0.8–1.6 9–12 6–8
suit the mineral behaviour – at Mount Isa this meant a
37 µm grind before roughing. Some mineral is liberated
or sharp internal classification within the grinding mill (as at this size and can go to concentrate. Other minerals in
the IsaMill achieves with the centrifugal separator at mill rougher concentrate need to be ground to 12 µm. Some
discharge). Circuit designers should pay particular attention to of these are rejected in cleaning and need to be reground
classification and size distribution – some laboratory ultra-fine further to 7 µm.
grinding tests are done in batch mode which will produce a •• Float minerals in narrow size distributions – this happens
sharper size distribution than the continuous plant installation. automatically with the staged grinding approach
Managing the impact of grinding on surface chemistry – this described above, and was assisted by the inherent sharp
is crucial for ultra-fine grinding before flotation. Usually size distribution produced by the internal classifier in
this necessitates the use of inert (usually ceramic) media. IsaMills.
Consuming so much power in a steel environment means •• Minimise circulating loads, and open-circuit as much as
high retention time and high steel contamination of the pulp. possible – this is an added benefit of staged grinding.
The resultant low pulp potential changes flotation behaviour Once a stream is reground it should then be floated
and reduces selectivity. Iron oxidation products are known separately with similar sized particles, and the tailings
to form hydrophilic ‘patches’ on particles surfaces (Figure 6; moved onto the next stage – either a stage of even finer
regrind and flotation, or to tailings. It is pointless to
recirculate a composite particle unless it is going to be
reground finer.
These principles can be seen in the simplified Mount Isa
flotation circuit in Figure 7. Though the circuit may appear
complicated, it is better than the alternatives of either:
•• grinding everything to 12 µm and floating together (too
expensive)
•• recirculating regrind products and trying to float them
with coarser minerals (causing poor performance of
the reground minerals, high circulating loads and low
recoveries).
Contrary to appearance, these developments at Mount Isa
greatly simplified circuit operations. Lead recovery increased
by five per cent and lead concentrate grade by five per cent,
FIG 6 – The impact of steel media on hydrophilic surface coatings. zinc recovery by ten per cent and concentrate grade by

FIG 7 – Mount Isa lead zinc stage grind and flotation circuit.
two per cent. More surprisingly, reagent needs dropped, the liberated 10 µm particle which has less sphalerite
circulating loads dropped and the circuit became far more surface area than the composite we need to reject.
stable. Flotation suddenly became as easy and predictable as Some further problems arise when floating coarse and fine
the textbooks say it is! particles together:
•• There is no point depressing the 37 µm composite particle
Competing Flotation Rates of Different Particle Sizes unless it can be subsequently liberated. While plants
often ‘send it to regrind’, this is often a conventional ball
The profound impact of a narrow size distribution to flotation
mill or Tower Mill that has very low breakage rates on
feed is explained by mineralogy and operating constraints.
-30 µm particles. This just causes high circulating loads
In a system with just pure liberated sphalerite and quartz, of composite particles.
flotation could achieve good recovery in all size ranges, even
•• The high circulating loads then take up volume and
though the ‘fines’ have slower flotation rates. But in real
reduce residence time in roughing and cleaning. Since
circuits there are two crucial constraints: fine particles are slower floating, this reduction in
1. Other contaminant minerals such as pyrite and residence time further hurts their recovery.
pyrrhotite also exhibit some flotability – a ‘coarse’ pyrite •• If high pH is used for depression and lime is used to get
particle may have the same flotation rate as a ‘fine’ high pH, this introduces a potential surface chemistry
sphalerite particle. problem. Circuit water can become saturated in calcium
2. Composite particles – in Figure 8, the simple 37 µm ions. This leads to reaction with sulfate ions, which
sphalerite-pyrite binary must be rejected to achieve causes gypsum to precipitate on the nearest surface –
usually a mineral particle. Scanning electron microscope
target concentrate grade. To reject it needs low collector
work at Mount Isa before IsaMills were installed showed
and high depressant. But these conditions will also reject
that up to 80 per cent of sphalerite surface was masked
by gypsum. This has a more serious effect on -20 µm
particles that can’t afford to lose precious hydrophobic
surface area.
Reconciling the reality with the perception

Common perception is that ‘fines’ don’t float. In reality they
float very well in the right conditions. Figure 9 explains this
conceptually. In most plants, -20 µm particles do perform
poorly because they are mixed with coarser particles with
very different needs. If these particles were floated in a
narrow size distribution, flotation conditions can be tailored
to them and they perform well. This explains the performance
of the Mount Isa staged grind and float circuit.
Case Study – Mount Isa Lead Zinc CONCENTRATOR

These principles were successfully applied to the Mount Isa
FIG 8 – How can I reject this particle without rejecting this one? lead zinc circuit (Young, Pease and Fisher, 2000; Pease et al,

The project predicted five per cent higher zinc recovery

(and no extra concentrate grade) from the extra liberation.
This was achieved instantly. The surprise was the ‘second
wave’ of a further five per cent zinc recovery increase and the
two per cent increase in zinc concentrate grade.
This was because fines flotation improved after grinding finer
– though it took us about six months to discover this because
we had to ‘unlearn’ what we thought we knew about fines.
Our three biggest mistakes were:
1. expecting to increase reagents after IsaMilling in
proportion to the to match due to the huge new surface
FIG 9 – Resolving the paradox. area created
2. not taking the depressant (lime to pH 11) off the zinc
2006). The project added six IsaMills to regrind lead rougher
cleaners
concentrate to P80 of 12 µm, most zinc rougher concentrate
3. ‘pulling’ flotation banks harder because we assumed the
to 12 µm, and a final zinc regrind to P80 of 7 µm (Figure 7).
fines would float slowly.
Previously most of the losses were in the ultra-fines fraction.
Installing more ultra-fine grinding improved lead recovery These ‘logical’ assumptions underestimated four
by five per cent and increased lead concentrate grade by fundamental effects:
five per cent (equivalent to ten per cent upward shift of the 1. After inert attrition grinding we had fresh surfaces that
lead grade-recovery curve). Zinc recovery increased by didn’t need high collector additions to compete with
ten per cent in two steps and zinc concentrate grade increased surface coatings.
by two per cent (equivalent to 16 per cent upward shift of 2. The virtual elimination of composites meant it was easy
the zinc grade-recovery curve). The story of zinc metallurgy to eliminate contaminants without high depressants –
is told in Figures 10, 11 and 12. minerals behaved like minerals, not composites.
FIG 10 – Initial zinc recovery increase after finer grinding.
FIG 11 – Improving fines recovery by grinding finer.

FIG 12 – Slimes? What slimes? Zinc recovery by size from rougher concentrate.
3. The narrow size distributions in each regrind-float stage sphalerite increased from about 78 per cent to about 88 per cent
meant minerals behaved predictably according to their after IsaMilling. Very high recoveries were achieved even in
surfaces, not their size. the 0–4 µm fraction – the fraction many operators call ‘slimes’
4. Eliminating circulating loads (previously up to (Figure 12).
300 per cent) between stages meant we didn’t
mix different particle classes and sizes. The lower The big picture – mineralogy, liberation,
recirculation meant higher residence time and lower
pulp density (which assisted cleaning). The lower circuit
chemistry and recovery explained
volumes reduced also stopped spillage; previously a Figure 13 is a graphic summary of the impacts of ore type,
significant disruption. liberation, steel grinding, inert grinding and chemistry
We learnt to challenge everything we knew about plant changes unfolding over 20 years at the Mount Isa operation.
behaviour. Despite adding 6 MW of grinding power to the Each month Mount Isa collected inventory samples of all feed
circuit, unit costs returned to their pre-regrind levels since and exit streams. Each exit stream was sized and cyclosized
minerals behaved like the text books said, with low reagents and size fractions submitted for quantitative mineralogy.
and no circulating loads. This generated a huge amount of data on a compelling story
Figure 11 shows that not only did fine grinding put more of sphalerite liberation and recovery. ‘Recalculated feed’ is
material into the high recovery fines, it also improved created by mathematically combining all the plant exit streams
performance within the fines – recovery of 8 µm liberated according to their relative tonnage. It captures the impact of all
FIG 13 – The Big Picture – Mount Isa Sphalerite Liberation and Zinc Recovery.

Stage 1 – The sickening decline. Stage 3 – A pleasant surprise.

the grinding and regrinding stages of the circuit. The Mount Isa we made two steps forward from liberation and one step
journey unfolded in four stages, as described herein. backward from surface chemistry. Liberation increased by
18 per cent, while zinc recovery increased by seven per cent.
Stage 1 – the sickening decline The old rule ‘recovery is liberation plus ten per cent’ had
During the 1980s ore sources became increasingly fine-grained. broken down, but this was the inevitable price of grinding
Liberation declined, and recovery inexorably declined with fine with steel media.
it. Like most operators under profit pressure we increased
feed tonnage. This increased revenue but further reduced Stage 3 – a pleasant surprise
liberation. We tried dozens of reagents, dozens of circuit The IsaMill technology was developed in the Mount Isa
changes and dozens of metallurgists. Finally, we accepted we concentrator with the first installation to regrind lead
were just ‘shuffling the deck chairs on the Titanic’: reagents cleaner feed. We wanted to liberate some galena-sphalerite
don’t grind. A fundamental rule seemed to guide plant binaries to reject zinc from lead concentrate. This was
performance despite our metallurgical efforts – recovery achieved, though the total effect on sphalerite liberation was
equalled liberation plus ten per cent. A small amount of small. However, zinc recovery to zinc concentrate increased
composite particles could be accepted into concentrate, until by four per cent. This was because the clean surfaces after
the quality constraint on impurities is reached. IsaMilling improved flotation chemistry – the fine clean
liberated sphalerite particles ‘behaved properly’ and were
Stage 2 – use a bigger hammer easy to depress from lead concentrate. We had achieved
We had to increase liberation to increase recovery. The only what no amount of chemistry or circuit changes could
available technology was conventional ball milling, and the achieve in the past – increasing recovery increase without
significantly changing liberation.
then emerging Tower Milling. We installed 6 MW of ball and
Tower Mills, effectively doubling the grinding capacity. We
knew this would increase liberation but also that there would Stage 4 – cooking with gas
be a downside – creating finer particles with so much steel Breaking the quandary between increasing liberation and
media would create surface chemistry problems. As expected, harming surface chemistry was a revelation. With steel milling
Stage 2 – Use a bigger hammer. Stage 4 – Cooking with gas.

we had reached to stage of one step forward from liberation, Then, contrary to common belief, very fines particles float
but two steps back from chemistry. Suddenly a new option extremely well.
was available – improve both liberation and chemistry in
the same step. We put 4 MW IsaMills into the zinc circuit. REFERENCES
Liberation increased four per cent, but recovery increased by
Batterham, R, 2008. Introduction to AMIRA P260E Mini Workshop,
ten per cent. Concentrate grade also increased by two per cent May 2008.
– we could have taken this as an extra six to eight per cent
recovery, but the higher grade was more profitable. Making Bennett, D, Crnkovic, I and Walker P, 2012. Recent developments at
the Phu Kham copper gold concentrator, in Proceedings 11th Mill
the higher grade had not been possible in the past without a
Operator’s Conference (The Australasian Institute of Mining and
punishing recovery loss. Metallurgy: Melbourne).
Recovery of the finest sphalerite from rougher concentrate Gao, M and Weller, K R, 1993. Review of alternative technologies for
is over 95 per cent after IsaMilling (Figure 13). However, fine grinding, AMIRA Project P336, Report P336/20, November.
total circuit sphalerite recovery is only 82 per cent, with
Holder, R K, 1994. Improvements in copper and silver flotation at
seven per cent lost to lead concentrate and 11 per cent Hellyer using high energy conditioning, in Proceedings Fifth Mill
reporting to rougher-scavenger tailing. The losses to scavenger Operators’ Conference (The Australasian Institute of Mining and
tailings are about half is in coarse composites and half in fine Metallurgy: Melbourne).
liberated particles.
Pease, J D, Curry, D C, Barns, K F, Young, M F and Rule, C, 2006.
If a particle is too coarse, or the surface is badly contaminated, Transforming flowsheets with inert grinding – The IsaMill,
then it doesn’t float in the roughers, and doesn’t get a chance Canadian Mineral Processing Conference, January 2006.
to get clean liberated surfaces in fine attrition regrinding. Rule, C and de Waal, H, 2011. IsaMill design improvements and
operational performance at Anglo Platinum, in Proceedings
CONCLUSION MetPlant Conference (The Australasian Institute of Mining and
Overgrinding should always be avoided, it wastes energy. But
Taggart, A F, 1927. Handbook of Mineral Dressing, pp 12-92–12-97 (John
if composites are present then regrinding is necessary. This
Wiley and Sons: New York).
needs a grinding environment that produces clean surfaces
and sharp size distributions, followed by flotation in narrow Young, M F, Pease, J D and Fisher, K S, 2000. The installation of the
size distributions with minimal circulating loads. When George Fisher flowsheet in the Mount Isa Lead/Zinc Concentrator,
in Proceedings Seventh Mill Operators Conference (The Australasian
these simple principles are followed, grinding significantly
improves fines behaviour.

Contents
Cunning solutions to process improvement

T J Napier-Munn1
ABSTRACT
This paper proceeds from the premise that process improvement is a key part of the
mineral engineer’s role, and one in which he/she can make a significant contribution
to shareholder value. It presents four tools to help with process improvement projects.
Two are designed for process monitoring, to determine as soon as possible whether
there has been a true change in performance: the cumulative sum (cusum) control chart
plot with V-mask, and the control chart using a process model. The other two are used
for process optimisation: simplex self-directing evolutionary operation (SSDEVOP) and
a response surface model developed using a central composite rotatable experimental
design (CCRD). These methods are robust, have a rigorous theoretical basis and are
easy to implement. They also promote project continuity, which increases the chances
of success, particularly in fly-in/fly-out (FIFO) operations. Variants are used in other
industries, but they are rarely if ever applied in mineral processing today for reasons
which are unclear. The paper gives examples of the application of each method in
concentrators which demonstrate that they have the potential to be useful additions to
the mineral engineer’s professional toolbox.
INTRODUCTION – THE VALUE PROPOSITION

Mineral engineers are in the business of process improvement. Enhancing shareholder
value by increasing plant grades and recoveries and reducing costs is a large part
of the job description. The methods by which mineral engineers go about this task
should therefore be of more than passing interest.
Mineral processes are complex and rarely yield to a superficial or simplistic analysis.
This is one reason why the 20-year site veteran of yesteryear was so valuable; he/
she had learned on the job, understood the process to a considerable degree, and
had developed effective ways of approaching the task of process optimisation. The
consequence of the terminal decline of such experience in recent years has been
eloquently and passionately addressed in an earlier Mill Operators’ Conference
(Munro and Tilyard, 2009).
However, all is not lost. By taking a systematic approach to the business of
improving mineral processes the objectives can still be attained. The solution is to use
formal procedures which are designed specifically to bring rigour to the monitoring
of process performance and to the search for the optimum. The purpose of this
paper is to show examples of some techniques which can be used for this purpose.
These techniques have been known for a long time and are well established in other
industries, but for reasons which remain a mystery have not yet found general favour
in mineral processing. It is time that this omission was rectified.
The paper outlines four methods that mineral engineers should consider adding to
their professional toolbox, with examples of their application in concentrators (this is
not a complete list – there are others). The first two allow effective process monitoring
and analysis so that a real decline in performance can be quickly identified and
rectified. The second two are formalised methods of finding the optimum process
performance. All can be handled in Excel. The methods are particularly useful in
FIFO situations in which project continuity is often a challenge, because they provide
a systematic structure which can easily be followed by successive rosters. The prize
is a more efficient and cost-effective approach to process improvement with a greater
chance of success than the ad hoc and often unsuccessful campaigns that sometimes
destroy rather than enhance shareholder value.
CUMULATIVE SUM PLOT WITH V-MASK

The cumulative sum (cusum) plot is a very simple way of expressing any data
1. FAusIMM(CP), Emeritus Professor, Julius collected in a meaningful sequence, such as a time series of process data. The formula
for the cusum is:
of Queensland, Indooroopilly Qld 4068. t
Email: t.napier-munn@uq.edu.au Ct = / (yi – T) (1)
i=1
455
T J Napier-Munn
where: plot, thus converting it into a true control chart. Control

Ct cusum at time t charts are designed to detect ‘significant’ changes in process
yt original observation at time t mean, where ‘significant’ is defined both by the magnitude
T a constant (usually a target value or the mean of of the change and by the usual statistical risks of making a
historical data) false conclusion. The mask is constructed with these values in
The cusum is then plotted in time sequence. Its key feature is mind, as shown in Figure 3.
that changes in cusum gradient indicate changes in the process The dimensions d, h and k are chosen to detect a change of
mean, and this shows up changes in the prevailing mean mean of a given magnitude with a given risk of a false alarm
much more clearly than the original data. A positive gradient (a ‘Type 1 error’), based on the prevailing standard deviation
indicates that the process is running at a mean greater than
of process noise. The wider the arms of the mask the smaller
T, a negative gradient that is it running below T, and if the
cusum plot is horizontal then the mean is running at a value the change that can be identified as significant, and/or the
around T. T can be chosen to accentuate these effects. lower the risk of a Type 1 error.
Figure 1 shows a simple time series plot of daily concentrate The origin of the mask is placed on the latest point in the
grade in a copper concentrator over about six months. series (or any other historical point of interest) and if any prior
Figure 2 shows the cusum plot of the same data using T = points lie outside the arms then the process can be considered
35.5 per cent Cu, the mean of days 1–60 which was regarded as to have moved out of control, ie a significant change in process
a satisfactory operating period and therefore an appropriate mean has occurred. If the points lie outside the upper arm the
performance target. process mean has decreased, and if they lie outside the lower
The cusum plot of this initial period is indeed horizontal, arm the mean has increased. Figure 4 shows the application
confirming that the mean grade was relatively constant of the V-mask using the data of Figures 1 and 2. This mask
around the target value. There is then a change to a shallow is designed to detect a change of one standard deviation
negative gradient at about day 72 and then another more (2.4 per cent Cu in this case) with a 3σ Type 1 error risk as
significant shift at day 116, and another change for the worse
often used in conventional control charts.
around day 158. In each case the gradient is getting more
negative and so the mean is declining. The key on the chart can In this case the process is deemed to have moved out of
be used to estimate the grade associated with each gradient. control (an unacceptable drop in grade) at day 119, somewhat
By the end of the period the concentrate grade has declined later than visually apparent in Figure 2. The chart can be made
from 35.5 per cent Cu to about 30 per cent, a significant drop. more or less sensitive by tuning the V-mask dimensions.
This interpretation is purely visual. However, it is possible The charts shown in Figures 1, 2 and 4 were drawn in Excel
to make it more rigorous by imposing a V-mask on the cusum (including the V-mask of Figure 4), but statistical software
such as Minitab® Version 16 (Minitab Inc) offers specific
50 tools for generating cusum charts with decision criteria such
as the V-mask.
Cu con.grade (%)
45
40
35
Origin
30
d
25
20 h
k
0 20 40 60 80 100 120 140 160 180
Day
FIG 1 – Time series plot of Cu concentrate grade. FIG 3 – Construction of a V-mask for a cumulative sum control chart.
50
Cusum con. grade (% Cu)
0 116
-50 72 158
-100 35.5
34.3
-150
33.0
-200 30.5
-250
28.0
-300
0 20 40 60 80 100 120 140 160 180
Day
FIG 2 – Cumulative sum (cusum) control chart plot of Cu concentrate grade. Key shows mean values corresponding to particular gradients.

20
Cusum con. grade (% Cu)

-20
-40
-60
-80
-100
0 20 40 60 80 100 120 140
Day
FIG 4 – Cumulative sum control chart with V-mask.
CONTROL CHARTS USING A PROCESS MODEL not well predicted by the model, so they flag a real and
The classical form of control chart (sometimes called a unexpected process change requiring investigation (such
Shewhart chart after the originator) is a simple plot of the time points are regarded as outliers in model development, so
series with control limits imposed to warn when the process the model should be obtained from data representing what
is moving ‘out of control’ and an ‘assignable cause’ must the mineral engineer would regard as normal operation, and
be found. This is the language of control charts for quality outliers should be removed).
control in manufacturing popularised by W Edwards Deming Figure 6 shows a control chart constructed using this
after World War II, which led to the concept of total quality approach to monitor recovery in a nickel concentrator. The
management (TQM). Figure 5 shows a Shewhart chart for the model included as predictors Ni and S feed grade, the
Cu concentrate grade data, with warning and control limits square of S feed grade, throughput (t/h), and the previous
set at the conventional values of 2σ and 3σ respectively, where day’s recovery to ensure uncorrelated model residuals. R2adj
σ is the standard deviation of the process when in control. = 87.1 per cent indicating a very good model. The points
There are several conventions about what constitutes a in Figure 6 are the real daily process performance. The
warning and what requires action. In Figure 5 the first point continuous lines are the predictions of the model ±3σ where
below the lower control limit (LCL) occurs at day 159, but here σ is the standard error of the model. If the process is
day 121 marks seven days in a row below the mean. This following the model predictions, ie if the process is in control,
can be interpreted as an action limit because its chance of then the chance of a point falling below the lower confidence
occurring, if the process is in control, is only 0.57 or one chance limit (lower control limit) is 0.135 per cent (1 in 741) assuming
in 125 (0.8 per cent). the data are normally distributed; limits can also be calculated
A simple Shewhart chart contains no information about the for other distributions that may apply such as the Weibull.
process other than a historical record of previous values of Most of the points lie well within the 3σ confidence limits
a performance criterion such as concentrate grade. A more and so are well predicted by the model. The circles identify
highly tuned tool can be obtained by developing a regression recoveries that fall below the LCL. The first is at day 34, with
model of the process in terms of predictors such as feed another at day 51, but from day 54 there are several alarm
grades, throughput etc and using the model to determine points which clearly require attention. We know these points
the control limits. This has the advantage of incorporating are special because the model does not predict them, so
known effects. Any points falling outside the limits are points clearly something has changed which is not accounted for by
50
121
45 UCL
Cu con.grade (%)
40
35
30
25 LCL
159
20
0 20 40 60 80 100 120 140 160 180
Day
FIG 5 – Shewhart control chart for Cu concentrate grade. Dotted line: warning limits (±2σ); continuous line: action limits (±3σ).

T J Napier-Munn
85
Alarm points
80
3 sigma upper confidence limit
75
Ni recovery (%) 70
34
65
60 51
54 69
55
3 sigma lower
confidence limit
50
0 50 100 150 200 250
Day
FIG 6 – Ni recovery control chart using control limits calculated from a process model.
the model. This could be operating error, equipment failure, two dimensions representing different levels of the two
or some new unexpected process effect such as an ore type variables being varied. There are simple rules governing the
change whose influence is not predicted by the consequent construction of the simplex which are designed to ensure
change in Ni or S content. that the process moves up the ‘hill’ towards the top (the
Using a model to determine action limits in a control chart optimum). Essentially each new simplex is ‘flipped’ or
provides a more informed environment for decision-making reflected on the condition of the previous worst performing
than simply using the standard deviation of ‘normal’ operation vertex. The principle is easier to appreciate in two dimensions
as in a Shewhart chart, as it can immediately exclude causes (as in Figure 7) but there is no theoretical limit to the number
that are incorporated in the model. of variables, which can be included.
Figure 8 shows the progress in maximising solids recovery
SIMPLEX SELF-DIRECTING EVOLUTIONARY in steps corresponding to the simplex numbers in Figure 7.
The progress is quite efficient, particularly in view of
OPERATION (SSDEVOP) OPTIMISATION the non-linear behaviour of the process contours shown in
A real-time optimisation procedure for an industrial process Figure 7 which of course would normally not be known to
needs to achieve three things: the experimenter (these came from a model developed in a
1. it should be efficient, that is, it will drive the process to pilot plant).
the optimum region by the shortest possible route SSDEVOP (and EVOP) steps are easy to calculate in
2. it should avoid large changes in process conditions a spreadsheet. It is also recommended by experienced
that might disrupt the process and cause a reduction practitioners that an SSDEVOP committee is established
in performance comprising those with an interest in the outcome of the
3. it should be easy to implement. optimisation exercise, including operations management,
Evolutionary operation (EVOP), developed for optimising domain experts (usually a senior site or head office
chemical processes in the 1950s, uses the principles of efficient metallurgist) and the operators who will have to implement
experimental design as a strategy for systematically improving
the performance of the process. It makes small changes to
process input variables, observes any change in performance,
and moves the process in small steps in the direction which
demonstrates most improvement. The changes are made
using simple factorial designs, which are an efficient way
of doing experiments to determine the effect of factors on a
response. Whilst EVOP mandates small changes, ensuring
no process disruption, the conclusions are made statistically
robust through replication.
Two disadvantages of EVOP are that it is not efficient
at chasing a moving target (a common problem in mineral
processing) and it can be slow. Simplex self-directing EVOP
(SSDEVOP) was developed to overcome these disadvantages.
Both techniques, which are forms of hill-climbing optimisation,
were championed for mineral processing by Mular (1972) and
Mular and Bull (1973) among others, but have not been widely
used for reasons that are not clear. They deserve re-visiting.
Figure 7 shows a simulated SSDEVOP program, in which
the spigot and vortex finder diameter of a small hydrocyclone FIG 7 – Optimisation path for a simulated two-factor
are varied to maximise solids recovery. The process is run simplex self-directing evolutionary operation (SSDEVOP)
at levels defined by a simplex – in this case a triangle in experiment. Contours are solids recovery (per cent).

80
previously been thought to be unsuitable for plant scale
78
optimisation because it is seen to be difficult to maintain
76
plant operating conditions at the required level for long
Solids recovery (%)
74 enough. However, three such experiments have recently been

72 successfully carried out in three different concentrators on
70 three continents and it is now clear that, given commitment
68 by the site staff, CCRDs can be effectively used under plant
66 conditions. The statistical rigour of the design and subsequent
64 analysis help mitigate the effects of process noise, and the
62 product is a model that can be directly used to optimise
60 performance.
0 2 4 6 8 10 12 14 16 Figure 9 shows response surface (3D) graphs of
Step (simplex)
concentrate solids mass flow (as a proxy for recovery) and
FIG 8 – Progress of hydrocyclone solids recovery. concentrate per cent solids (as a proxy for concentrate grade)
in a single flotation cell in the rougher bank of a platinum
any changes. This ensures well informed technical decision- group metal (PGM) concentrator, in terms of airflow rate and
making and ‘buy-in’ by all interested parties, which greatly cell level in arbitrary units (Venkatesan, Harris and Greyling,
increases the chance of success. 2014). These were drawn using equations of the form of
Equation 2 in which the coefficients were determined from
RESPONSE SURFACE OPTIMISATION USING A CENTRAL a CCRD experiment conducted over a few days in the plant.
Twelve experimental runs were conducted in random order,
COMPOSITE ROTATABLE EXPERIMENTAL DESIGN including four repeat runs at the centre point of the design to
A process response such as recovery or concentrate grade can evaluate experimental error.
be optimised if we can generate a model of the response in
The plots tend to be mirror images of each other, which reflects
terms of measurable process variables such as feed grades,
the expected inverse relationship between recovery (t/h) and
throughput, ore type etc. The optimum variable settings can be
concentrate grade (per cent solids). Maximum per cent solids
found mathematically from the model and then implemented
is realised at low air and level, and maximum t/h at high air
on the plant. The model needs to accommodate linear, non-
and level, although the exact relationships are complicated
linear and interaction effects (interaction occurs when the
by non-linear trends and interactions. A good example of
dependency of the response on the level of one variable itself
interaction occurs in the per cent solids response; there is very
depends on the level of another variable, not uncommon
little effect of cell level at low air rates but a strong negative
in mineral processing and undetectable without the correct
experimental design). Non-linear effects are represented in effect at high air rates.
the model by squared terms and interaction effects by cross Once a model has been obtained it can be used to find a process
product terms. Such a ‘response surface’ model in two process optimum subject to whatever constraints are appropriate.
variables takes the following form: Figure 10 shows an overlaid contour graph which defines
a ‘feasible region’ (white area) given prescribed process
y = a0 + a1x1 + a2x2 + a3x1 2 + a4x2 2 + a5 x1x2 (2)
where:
y is the process response to be optimised Feasible
x1 and x2 are the two process variables 22% solids min.
a0–a5 are coefficients to be determined from data
4.0 t/h min.
The model is in principle extensible to any number of
variables. Non-feasible
The CCRD is a form of factorial experiment that allows
such a model to be developed with maximum efficiency
in the minimum number of test runs. CCRDs mandate the
conditions under which the experiments should be run, once
the limits of interest of the variables have been defined. Whilst FIG 10 – Overlaid contour graph for platinum
widely used in laboratory and pilot plant test work they have group metal flotation cell performance.
FIG 9 – Response surface graphs of solids t/h and concentrate per cent solids in terms of air rate and cell level for a flotation process.

T J Napier-Munn
Optimum
factor levels
Set factor
levels
manually
Max. con.grade
(% solids)
Max. recovery
(tph)
FIG 11 – Minitab® Response Optimisation tool output.
constraints, in this case minimum acceptable performance active monitoring of process performance to identify failures
criteria of 4 t/h and 22 per cent solids. The feasible region quickly, and in the true optimisation of process performance,
is small, reflecting the strong trade-off between concentrate however that may be defined. These methods are robust, have
grade and recovery and the features of this particular process. a rigorous theoretical basis and are easy to implement. They
Figure 11 shows the output of Minitab®’s Response also promote project continuity, which is likely to significantly
Optimisation tool which allows several performance criteria increase the probability of success. They deserve to be more
to be optimised according to constraints defined by the user widely used.
(Excel-based options are also available). These include both The methods described in this paper are discussed in more
process limits and the relative importance of different criteria. detail in Napier-Munn (2014).
For the criteria used in this case the optimum level and air rate
are 89.5 and 39.4 respectively, delivering 22.2 per cent solids REFERENCES
(concentration grade) and 4.66 t/h (recovery), both close to
Mular, A L, 1972. Empirical modelling and optimisation of mineral
the lowest acceptable limits. This confirms the conclusions of processes, Minerals Sci Eng, July, 4(3):30–42.
the contour plot of Figure 10, that there is little room to move
when there is such a strong relationship between concentrate Mular, A L and Bull, W R, 1973. Mineral processes: their analysis,
optimisation and control, course notes, 1969–1973, University of
grade and recovery.
British Columbia and Colorado School of Mines.
Munro, P D and Tilyard, P A, 2009. Back to the future – why change
CONCLUSIONS doesn’t necessarily mean progress, in Proceedings Tenth Mill
Process improvement is a key part of the mineral engineer’s Operators’ Conference, pp 5–12 (The Australasian Institute of
role and one in which he/she can make a substantial Mining and Metallurgy: Melbourne).
contribution to shareholder value. However, too often the Napier-Munn, T J, 2014. Statistical Methods for Mineral Engineers:
path followed tends to be ad hoc, arbitrary and subject to How to Design Experiments and Analyse Data, 645 p (JKMRC, The
personal whim. Clearly professional experience must play an University of Queensland).
important part, and this paper has suggested some systematic Venkatesan, L, Harris, A and Greyling, M, 2014. Optimisation of air
ways of putting that experience to better use, both in the and froth depth in PGM flotation using a CCRD design, Minerals
Eng, 66–68:221–229.

Contents
How to prioritise process improvements

D J Hill1
ABSTRACT
Mining companies routinely invest in improvements to their processing operations,
but often these enhancements do not work to full effect.
Improvements can be ineffectual because underlying process problems are not
identified or prioritised for correction or repair.
This is frequently because ‘black box’ solutions deployed to implement process
control improvements typically rely on instrumentation to provide data for
algorithms to work effectively, and these solutions are often ineffective or disabled
at mineral processing sites.
There are three key reasons why these improvements can fail to deliver:
1. vital instrumentation is not calibrated or functioning as required
2. basic process control and loop-tuning is not implemented correctly or requires
tuning
3. there is an inability to support the ‘black box’ solutions on-site.
Many of these problems can be overcome by implementing operational analytics
that identify where problems are occurring and prioritise where improvements
need to be made.
Incorporating operational analytics into process improvement is not a new
methodology. It is more a back-to-basics approach to make more efficient use of
capital and systems already deployed.
Data collection systems, such as the OSIsoft PI System, provide real-time and
historical process information for process control experts to analyse and identify areas
for improvement. However, most systems are not used effectively. With some simple
enabling, data collection systems can provide evidence for prioritising investments in
improvements and determining an investment’s effectiveness.
MIPAC advocates that the initial step, with analytical evidence, is that clients review
the control strategies and implement small influential instrument repairs/calibrations
and operator training. This can significantly boost process productivity without the
need to spend large sums of money on new capital or employing expensive, difficult
to maintain ‘black box’ solutions.
A simple Stage-Gate®-style approach ensures that small investments in evidence-
based process improvements deliver significant benefits to mining operations.
INTRODUCTION
All plant managers must make crucial decisions about what process improvements
to invest in and which to address first. This paper determines how to prioritise asset
productivity improvements in the context of a fully commissioned, operational plant
from a control and instrumentation perspective.
The challenging market conditions and relatively high costs in Australia continue to
drive a focus on capital discipline and improving productivity and cost competitiveness
in the minerals industry (MCA, 2014). In the past decade, mining productivity has
declined in Australia as it has in other mature resource-producing nations such as the
United States and Canada (Figure 1). PwC (2013) reports that productivity indexes
indicate that since 2002, mining is achieving 56 per cent less output per hour for labour
employed and 44 per cent less output in terms of capital employed, despite dramatic
advances in technology across this period. Over the last ten years, both labour and
capital productivity in mining has been declining (ABS, 2013) (see Figure 1).
For Australia to remain a world leader in the resources and energy sector,
productivity and competitiveness must increase and investments in asset productivity
improvements are required.
1. Business Development Manager, The Honourable Ian Macfarlane, the then Minister for Industry, stated that the
MIPAC, Brisbane Qld 4000. mining sector has responded with a heightened interest in sustainable productivity
Email: darren.hill@mipac.com.au initiatives and process improvements that boost output (Macfarlane, 2014). As an
461
D J Hill
FIG 1 – Ten-year productivity decline (source: ABS (5260), Australian Bureau of Statistics, 2013).
example, BHP Billiton has placed extreme focus on issues of There are many differing ways of achieving APC that are
productivity and capital discipline (Mackenzie, 2013). The supported by a plethora of white papers that illustrate APC
sector is interested in what initiatives to implement, how systems and process improvement strategies. APC theory
and when to modify or enhance processes, which processes is widely accepted in the process control engineer world,
are optimal for intervention and how to enable technical and and evidence demonstrates that APC can deliver process
people-based execution. improvements. However, such improvements are invariably
Interest aside, the lack of clear evidence-based analysis to not maintained when a change is made at the plant, either
support decision-making around new control strategies, knowingly or unwittingly. For example one simple field
new technologies, expert systems and/or training of plant instrumentation change or a simple control-loop issue can
operators means that decision-makers are less able to reduce the effectiveness of APC. This downturn in process
performance can be puzzling for operators, who find it easier
quantify and compare return on investment (ROI) value
to operate the plant in manual mode by switching off or
and therefore find it difficult to prioritise these options. As
bypassing the APC.
a consequence, beneficial gains to productivity are often not
realised, sometimes due to simple basic correction and repairs Typically, these systems are expensive, costing tens if not
being overlooked. hundreds of thousands of dollars to purchase, and often even
more to deploy. While some process plants benefit from APC
McKee (1999) defines process control as the technology
and expert systems, these appear to be in the minority.
required to obtain information in real-time on process
behaviour and then using that information to manipulate
process variables with the objective of improving the REALITY VERSUS THEORY
metallurgical performance of the plant. A plant requires solid regulatory performance and process
It is the recognition of the influencing factors of process optimisation to operate effectively. This can only occur if the
control fundamentals that should form part of prioritising plant’s control system and its strategies use reliable data from
process improvements. Otero (2008) states that in many cases, field instrumentation. This requires a proactive maintenance
simple procedures based on the fundamentals of process regime that ensures the reliability of assets and the
controls can affect the performance of not only the control functionality of the control system. Thwaites (2013) states that
system, but also the bottom line of the plant. the mining industry has an average of 75 per cent of physical
assets under process control and that more than 60 per cent
of all control-loops underperform. In specific regards to
CHASING PROCESS OPTIMISATION proportional-integral-derivative controllers (Figure 3) used
As a first step, many processing plant managers deploy for feedback loop control in mineral processing plants,
advanced process control (APC) or ‘black box’ expert only 25 per cent of them contribute to reducing variability.
systems (Figure 2) to instigate improvements to stability, According to Bouchard and Émond (2009), the main reason
yield and throughput. for this situation is poor tuning.
Numerous vendors provide APC and optimisation tools Poor performance of field instrumentation and control-
backed by successful case studies. Some vendors also have loops results in:
support mechanisms for APC or expert systems that generate •• high product quality variability
reports to suggest improvements.
•• reduced throughput
All APC and expert process optimisation systems rely on •• increased downtime and lower availability
basic field instruments to provide inputs into their logic or
algorithms. These are relied on for calculations and rules •• poor financial performance.
(modelling) to determine plant performance measurements MIPAC is a process control and optimisation company
and to identify improvement opportunities. Field that has successfully delivered more than 250 engineering
instrumentation status and changes are usually collected projects globally. Client engagements usually involve control
by the control system as ‘time-stamped data’ that is then system upgrades and control process improvement projects
stored in a plant-wide historian (essentially a large pool of at mineral processing plants.
information). These systems work in real-time (as it happens), Most engagements require a simple site audit and historical
analysing current events while reviewing past events by data analysis to check a plant’s health before implementing
interrogating a plant-wide historian database. These are any changes or improvements. MIPAC’s engineers’ experience
usually collated into a report with supporting evidence for with smelters, concentrators, grinding, flotation and
process tuning, adjustments or changes. electrowinning/refining has led to some significant findings:

FIG 2 – Chasing optimisation at supervisory level.
FIG 3 – Proportional–integral–derivative controller diagram.
•• poorly maintained instrumentation These benchmarks should be derived from the validated
•• inappropriately tuned loops plant historian data that are interpreted by process experts
and plant operators. The implementation of improvements
•• degradation of interacting control-loops
should be conducted in a staged approach. The key reason for
•• operational differences between crews/shifts this suggested method is that each measured success should
•• systems in manual mode (not automatic) be validated before proceeding onto the next improvement.
•• disabled expert systems This means that one can control investments tightly to ensure
that the benefits are attained according to the desired ROI.
•• alarm flooding
Typically, this methodology involves the tried and tested
•• unvalidated historical data. Stage-Gate®-style approach (Figure 4).
These findings indicate that fundamental control issues It is recommended that optimal process performance is
are limiting the performance of plants to achieve a stabilised achieved in three main phases: fundamental fixes, proactive
state and prevent step change productivity gains. Some of maintenance and upgrades of systems and, finally, APC and
these findings are not a surprise to the plant management and expert systems (Figure 5).
operators, however, they are not aware of the impact that they The first phase is to review the plant status and then
have on plant performance. As a result, most choose the safe determine all the parameters and controls that are not
theory option of deploying an APC or an expert system rather functioning as they should or as designed. These factors
than fixing the fundamentals that tend to be overlooked or should be fixed to get the plant to a stable and steady state.
considered unimportant. The impact of these factors should be reviewed in the context
of historical data evidence.
STAGES OF IMPROVEMENTS The next phase is to ensure that process health monitoring
To achieve optimal process improvements, one must first and an informed proactive maintenance regime is in place.
establish an initial performance benchmark on which to This is perhaps the most difficult step to implement and is a
improve from and also set a benchmark goal to achieve. common failure in many plants due to the factors mentioned

D J Hill
FIG 4 – Stage-Gate® methodology diagram.
FIG 5 – Stages of process improvements.
earlier and continual changes to the plant process. Again, a necessitates scopes of work to be developed that deliver
review and analysis of historical data can support and aid strategic process improvement projects. These individual
prioritisation of works. projects can then be assessed against the overall business
Once the process plant is in a maintained state, but perhaps ‘transformation strategy’ so that each project dovetails
into one another and can be sequenced (prioritised) for
not optimal in performance, it is the right time to move to
implementation.
the third phase and introduce APC and expert systems if
required. This is because there will be a level of certainty and As suggested by Gershon (2010), there are numerous
confidence that the fundamental instrumentation is recording improvement process methodologies, such as Sixth Sigma,
and correctly providing the real-time data on which these lean manufacturing, just-in-time, Kaizen and Total Quality
systems rely, and that the control strategy and loops are Management. Many share features such as the define,
measure, analyse, improve, control process (Figure 6). The
working to the flow sheet design of the process.
methodology’s success relies on two factors: statistical
information (truth) and plant operators’ process knowledge
IMPROVEMENT METHODOLOGIES (faith in applying).
Currently, plant management operates, budgets and In essence, there are multiple choices as to the methodologies
appropriates the monies for maintenance, upgrades and to be employed for process improvements, with most
improvements. There is little evidence that the mining used throughout other industry sectors. Unfortunately,
industry applies the process of strategic project prioritisation due to skill (naivety), insufficient time, a lack of detail or
and cost benefit analysis to ensure that the business focuses an inability to prioritise, these are not often applied to the
its resources on areas that will add the most value. This decision-making process.

PRIORITISING IMPROVEMENTS BASED ON EVIDENCE

In today’s process environment, where the control specialist
and the information technology specialist converge, decision-
makers are able to overcome many improvement failures
by using operational analytics derived from collected plant
data. Deloitte (2013) suggests that the use of analytics to
uncover true cost drivers is a key strategy to be considered for
identifying and realising enormous savings or productivity
gains. This statistical information needs to be reconciled,
validated data from key operational inputs by experts with
the appropriate domain knowledge.
In ‘Envisioning the future of mining’ (IBM, 2009), IBM
proposes a new vision for information in the mining industry
that interconnects all entities in the environment, including
instrumented assets and equipment, transportation,
people, supplies and plans, into integrated views. Named
‘visualisation’, this integration of information provides
production and maintenance operations, analysts, field crew
and other decision-makers with a real-time view of their
entire operation via consolidated, synoptic interfaces. Alerts, FIG 6 – Define, measure, analyse, improve, control diagram.
alarms and triggers enable the mine to be hyper-responsive
to change and challenges. Advanced analytics is enabled criteria form part of the ROI business case for each proposed
through plant-wide enterprise access to a comprehensive set improvement in the process. The Australian Standards Guide
of facts (‘single version of the truth’) that assist miners to (SA, 2012) suggests prioritisation process and criteria for:
predict and plan for the future, not just react. •• preliminary assessment; quality and capability
This is achievable today as MIPAC has found that most •• proposal evaluation; net benefit and assessment measures
process plants already have managerial software systems, •• costing; resource requirements.
such as OSIsoft PI System, that store and generate data
Once the criteria (Figure 9) has been agreed to and applied
capable from the ‘single version of truth’, with visualisation
across the portfolio of (proposed) process improvements,
enabling better decision-making (Figures 7 and 8).
a review should be conducted, with help from experts, to
Assets and people can be instrumented with location-aware determine the sequence that will deliver the maximum
technology that provides real-time metrics on performance benefit to the process.
and status, which reflects the current business reality,
delivered in real-time or right-time. When this reporting and
CASE STUDY
analysis of operations is visualised, mining companies can
spend more time ‘looking forward’ and making informed As many readers of this paper will be aware, the competitive
commodity market causes mining companies to be very
decisions instead of reporting on the past.
conservative with their overall plant performance information.
Decisions based on this reliable data should then be costed This makes it very difficult for ‘white papers’ (like this paper) to
on a ROI (net benefit) basis, reviewed as part of a portfolio of include specific details of named plant process improvements.
process improvements and prioritised. Nevertheless, process improvements prioritised using the
Criteria for judgement and prioritisation will differ with each methodology referred to in earlier sections is conveyed in this
company (each has differing drivers for their business). These section without compromising confidentiality.
FIG 7 – Visualisation example of concentrator.

D J Hill
FIG 8 – Visualisation example of power.
improvements, a comparison was made using operational

analytics in the comminution circuit and overall plant
performance. Data collected from the PI historian was
analysed for a period prior to the first site visit and after the
second visit (a ten-month period). It was demonstrated that
there was an increased throughput of the grinding circuit, and
the overall effectiveness of the flotation circuit was improved
for both gold and silver.
These improvements cost the mining company ~$1.2 M
(<$200 000 in process control engineering) and led to an
overall improvement in recoveries for both gold and silver
(Figure 10) worth more than $15 M/a. This is an ROI of 0.08,
which is payback of less than a month.
CONCLUSIONS
When auditing deployed control systems, instrumentation
and data, MIPAC has found that most:
•• process plants’ control systems are capable of delivering
better performance
•• managerial software systems are capable of enabling
better performance
•• operators and maintenance crews can be educated to
maximise plant efficiency.
However, plant managers may not know their plant’s true
FIG 9 – Australian Standards Guide judgement criteria.
state because of poor quality data or knowledge of where
inefficiencies exist.
Experienced process control engineers undertook several
site visits to a gold-silver plant site in 2013. After an audit The first step to achieve asset productivity improvements
of the plant and review of the analytical evidence from is to audit the plant and address the fundamental control
their plant historian (OsiSoft PI system), process control and instrumentation issues before applying APC or expert
improvements were prioritised using costed ROI business systems. Each improvement project should be underpinned
cases and implemented using a modified Stage-Gate®-style with validated analytical evidence that is compiled into an
methodology. First stage improvements corrected some ROI business case. These multiple improvement projects
basic instrumentation, while second stage improvements can then be prioritised and sequenced within the three main
were made in the grinding and flotation circuits, principally phases using a suitable methodology and criteria.
achieved by performing basic loop-tuning and improvements If the mining industry is to increase productivity to achieve
to control strategies. These changes, together with other competitiveness into the future, it is crucial that decisions
recommended mechanical and process changes, improved about what process improvements to invest in and which
process stability and performance. to address first are made to deliver the maximum ROI. It is
To help the plant control manager justify the improvement essential that the industry adopts best practice improvement
expenditure and seek further budget for continued prioritised methodologies, that technology is at best maintained and that

FIG 10 – Recovery productivity gains graph.
it ensures that data for analytical purposes is validated from Macfarlane, I, 2014. Australian productivity relies on innovation,
the fundamental process inputs. Resourceful, March, pp 10–12 [online]. Available from: <https://
www.csiro.au/en/Research/MRF/Areas/Resourceful-
These factors, if addressed, will help to achieve higher magazine/Issue-06/Australian-productivity>.
productivity of processing and keep Australia as a world
Mackenzie, A, 2013. BHP to drive focus on costs in new era, The
leader in the resources and energy sector.
Sydney Morning Herald, 21 February.
McKee, M, 1999. The mineral processing project, AMIRA project P9L.
ACKNOWLEDGEMENTS
Minerals Council of Australia (MCA), 2014. 2014–15 pre-budget
Eddie Derivera, Tony Mathison and Bruce Northway at submission [online]. Available from: <http://www.minerals.
MIPAC. org.au/file_upload/files/submissions/MCA_Pre_Budget_
Submission_2015-16_final.pdf>.
REFERENCES Otero, F, 2008. Never forget the fundamentals of process control,
Australia Bureau of Statistics (ABS), 2013. Labour productivity index. InTech, June.
Bouchard, J and Émond, M, 2009. Who you gonna call? [online], CIM PwC, 2013. Productivity not austerity: productivity scorecard –
Magazine, August 2009. Available from: <http://magazine.cim. mining focus [online]. Available from: <https://www.pwc.kz/
org/en/August-2009/upfront/Who-you-gonna-call.aspx>. en/publications/new_publication_assets/productivity-not-
austerity-apr13.pdf>.
Deloitte, 2013. Tracking the trends 2014: the top 10 issues mining
companies will face in the coming year [online]. Available Standards Australia (SA), 2012. Standards Australia guide – project
from: <https://www2.deloitte.com/content/dam/Deloitte/ prioritisation process and criteria [online]. Available from:
ca/Documents/international-business/ca-en-ib-tracking-the- <http://www.standards.org.au/StandardsDevelopment/
trends-2013.pdf>. Developing_Standards/Documents/GU104%20-%20SA%20
Guide%20to%20Project%20Prioritisation%20Criteria%20and%20
Gershon, M, 2010. Choosing which process improvement Process.pdf>.
methodology to implement, Journal of Applied Business and
Economics, 10(5):61. Thwaites, P, 2013. Time to take control [online], CIM Magazine,
December 2012/January 2013. Available from: <https://issuu.
IBM, 2009. Envisioning the future of mining [online]. Available com/cim-icm_publications/docs/cim_dj12lr>.
from: <https://www-935.ibm.com/services/multimedia/wp_
envisioning_mining.pdf>.

Contents
Some practical problems in running statistically

valid plant trials and their solution
T J Napier-Munn1
ABSTRACT
Operating metallurgists spend a significant part of their time running plant trials to
compare two or more conditions in order to improve the process, such as a new reagent,
flow sheet or piece of equipment. The best way to conduct such a trial is as a paired
experiment when testing two conditions (eg ‘old’ and ‘new’) or a randomised block
design when testing more than two conditions. These experimental designs should
always be preferred over alternatives if their requirements can be met, because they
block the effect of uncontrolled variables (‘covariates’) and are the most statistically
efficient, ie they will lead to the required level of confidence in the outcome faster than
other methods.
Sometimes, however, it is not possible to switch condition easily, which is a prime
requirement of these methods. In such cases the only option may be to switch
condition once only, and compare the process performance before and after the
switch. Examples are testing grinding media, large equipment, or long residence time
processes such as leach trains or counter current decantation (CCD) thickeners. This
paper considers some data analysis methods to deal with this situation. Modelling
with intervention analysis (IA) uses regression models with a dummy variable to
indicate the presence or absence of the test condition. IA with a time series model
ensures that the residuals are uncorrelated. IA with a process model helps to deal
with the problem of covariates. The reference distribution is a method free of any
statistical assumptions but is less powerful than the other methods. Cusum charts
are a helpful visual aid to interpreting time trends. A combination of these methods
can help improve the confidence in the final decision. None are as good as a formal
experimental design such as a paired trial.
PROBLEM STATEMENT
Metallurgists spend a lot of their professional lives planning, conducting and
interpreting plant trials, and then making decisions based on the results. Plant trials
are usually carried out to determine whether there is a benefit in introducing some
change in process operating conditions, or to identify and quantify the effect of key
process variables on metallurgical performance. Examples are:
•• testing a new reagent in a flotation plant
•• testing an alternative piece of equipment, or equipment setting
•• testing a new circuit configuration
•• answering the question: does feed rate affect recovery?
•• answering the question: are two parallel circuits performing the same way?
This process often starts in the laboratory with experiments to test a new idea, or a
new reagent or piece of equipment recommended by a vendor. If these experiments
are promising, the next step is often a full plant trial to test whether the benefit can
be realised in practice. A plant trial is usually the only way of determining with
confidence whether a process benefit really exists. Improvements seen at the laboratory
or pilot scales sometimes evaporate when implemented at the full production scale.
However, plant trials are often expensive to conduct, and can interfere with normal
plant operation, so they are not undertaken lightly.
Therefore, it is important to conduct the trial in such a way that a firm conclusion
is reached (one way or the other) in the minimum time. The problem, as in all
experiments, is to understand and deal with the basic experimental error – what
electronics engineers call ‘noise’. In plant trials, as often as not, this is apparent mainly
1. FAusIMM(CP), Emeritus Professor, Julius
Kruttschnitt Mineral Research Centre, as a random variation of performance with time which masks the improvement
Sustainable Minerals Institute, The University being sought. It may also include systematic changes in performance due to factors
of Queensland, Indooroopilly Qld 4068. that have nothing to do with the subject of the trial, most often feed characteristics.
Email: t.napier-munn@uq.edu.au Figure 1 shows the daily variation in gold recovery in a copper-gold concentrator.
469
T J Napier-Munn
A one per cent recovery improvement may be worth several often lead to no clear result, or worse still, the wrong result, or
million dollars a year in a large operation, but how is one to take much longer than they should. The irony is that the right
detect it in data which varies over a 25 per cent range? The way to do things is usually easier and cheaper than the wrong
answer is: easily, if the right experiment is done, which allows way, and leads to the correct result more quickly. There is
the difference being sought to be reliably detected and its therefore no real excuse for not doing a good experiment.
magnitude estimated, against a noisy background. There are, however, situations in which these powerful
If the object is to compare the performance of two methods cannot be used for practical reasons. Examples are:
conditions, such as ‘old’ and ‘new’ or ‘on’ and ‘off’, then the •• Situations in which the experimental conditions cannot
best experimental design is the simplest, ie running pairs be easily changed, preventing the regular switching of
of ‘on’ and ‘off’ conditions in sequential time periods, or one condition to another. For example, grinding media
across parallel circuits, with the order or position within each cannot be repeatedly switched in short time periods.
pair being randomised. The duration of each ‘on’ and ‘off’ The new medium is generally made up over a period of
condition should be short relative to the natural frequency weeks or months, and grinding media trials have been
of the basic process variation (see Figure 1) but long enough known to last for years. Large items of equipment may
to allow plant conditions to stabilise after a change. A shift also be difficult to switch regularly (though with a little
or a day are typical durations, which also take advantage of thought and effort this is sometimes easier than it may
the standard plant reporting protocols to collect the necessary first appear).
data. The data (in the form of recovery for example) are then
•• Very long residence time processes, particularly
analysed by a paired t-test. If more than two conditions are to
be assessed then this idea can be extended to a randomised leaching. Here the problem is the time taken to stabilise
block design which is analysed by an Analysis of Variance to the new condition, meaning that pairs of conditions
(ANOVA) (Napier-Munn, 1995). may take too long to investigate, and there will always
be doubt as to whether the process has truly stabilised
Why is this the best approach? Because, assuming that certain under the new condition if it is being regularly switched.
statistical assumptions underlying these forms of hypothesis This situation is exacerbated in cases whether there are
test are not seriously violated [note: the main assumptions of complex recycle streams.
the paired t-test are that the paired differences are normally
distributed and uncorrelated. This is usually (though not In these cases, the metallurgist is often obliged to make a
universally) the case with mineral processing data. In any case single switch of process condition and analyse the resulting
the test is robust to departures from these assumptions, and data as a ‘before and after’ experiment, or at the very least
a modified test has been developed for data autocorrelated operators can make only a small number of switches with
in time (Napier-Munn and Meyer, 1999)], this is the most long operating periods in between. This paper describes
efficient and powerful way of reaching a conclusion at any some data analysis methods appropriate for such situations.
desired level of confidence (or risk), in the shortest possible The purpose, as always, is to show that the magnitude of
time. It also almost always ensures that any change which the observed performance change is (or is not) significantly
is observed is attributable to the factor being investigated greater than the background noise (error), at a desired level
(eg a new reagent) because the effect of other uncontrolled of confidence. The confidence level defines the risk of making
factors such as feed conditions, have been ‘blocked’ by the an incorrect conclusion. This is controlled by the magnitude
experimental design and thus removed from the comparison. of error (measured as a standard deviation) and the number
For these reasons, there must be extraordinarily good reasons of data points collected for each condition.
not to adopt either the paired t-test or randomised block
design to conduct the trial. TIME SERIES MODELLING WITH INTERVENTION ANALYSIS
It is therefore astonishing how often metallurgists do not Data collected in time sequence such as those shown in Figure 1
use these methods when there is no practical or theoretical can be modelled as a time series. These models exploit the
impediment to doing so. This author has seen countless fact that such data are often serially correlated – that is, each
situations in which plant trials were conducted without result is to some extent dependent on the previous results.
anything more than a primitive plan with no real idea about The degree of dependence is measured by ‘autocorrelation’,
how the results were to be analysed and interpreted. These defined similarly to the correlation coefficient, R. Serial
FIG 1 – Daily variation in recovery.

Some practical problems in running statistically valid plant trials and their solution
correlation violates the assumption of data independence on highly significant, demonstrated by the very low P-value
which the two-sample t-test is based. (5×10–5). The two-sided 95 per cent confidence limits for the
There is a family of time series models known as Box- effect are -3.3 to -1.2.
Jenkins models which take this approach. A simple time series The autocorrelation coefficient is 0.52 (also highly
model that has been found useful in mineral processing is the significant), which is normal for this kind of data. Thus, the
autoregressive model of order one (AR(1)), which simply time series model of Equation 3 is appropriate.
says that the current result is dependent to some extent on the As a linear least squares regression procedure has been
previous result (Napier-Munn and Meyer, 1999): followed, the usual assumptions need to be tested, eg normality
y t = n + z (y t - 1 - n) + e t (1) and independence of residuals (Draper and Smith, 1998;
Mendenhall and Sincich, 2003). In this case, some outliers were
where: identified (points poorly predicted by the model). Removing
yt performance (eg recovery) in current time period them improved the model fit (R2a increased to 46 per cent) but
(eg today) the magnitude of the media effect remained unchanged.
yt–1 performance in previous time period (eg yesterday).
ϕ autocorrelation coefficient (0 ≤ ϕ ≤ 1), which measures
PROCESS MODELLING WITH INTERVENTION ANALYSIS
how strongly yt depends on yt–1 Intervention analysis involves a time series model of the
process. This principle can be extended to a full process model
µ process mean, the value about which the process is
incorporating available variables such as feed grade, feed rate
randomly varying
etc, again including the intervention term in the model, eg:
et error for time t, a measure of the residual error after
the model is fitted Recovery = a0 + a1(feed grade) + a2 (feed rate) + a3T (5)
Equation 1 can be rearranged in the form of a linear
expression which can be fitted by linear least squares where T takes the value 0 or 1 depending on the experimental
regression: condition. The normal procedures and protocols are followed
in fitting the parameters in Equation 5. An advantage of
y t = n (1 - z) + z (y t - 1) (2) this model over Equation 3 is that it can account for those
interfering factors (‘covariates’) that can be measured that
When using a spreadsheet such as Excel, yt is the column of also affect recovery, such as feed grade.
daily data and yt–1 is this column displaced down by one cell.
Equation 6 shows a model developed for a trial in a base
(The first yt value and the last yt–1 value must be eliminated to
metal concentrator in which a new reagent was being tested
produce two columns with the same number of datapoints).
against the standard one, run (incorrectly) as a ‘before and
A term, T, can now be added which describes the absence after’ trial with 134 daily datapoints:
or presence of the new condition. It takes the value 0 for
the old condition and 1 for the new condition. This is called Recovery = 78.1 - 0.00011 (tonnes per day)
(6)
‘intervention analysis’ when used to analyse time series data - 0.19 P80 + 12.9 (Feed grade) + 1.84T
(Box, Hunter and Hunter, 1978). The new model is:
Table 2 shows part of the (edited) Excel output for the
y t = a0 + a1(y t - 1) + a2T (3)
regression.
where: Note that the purpose of this model is not to predict recovery
a0 µ(1-ϕ) for a given set of process conditions (that is a legitimate but
separate objective) but to decide whether the new reagent is
a1 ϕ
having a beneficial effect compared to the old, and if so, the
µ a0/(1–a1) magnitude of the effect. The former question is determined
a0, a1 and a2 are estimated by multilinear regression, for example from the P-value for the reagent coefficient, which is very small
using Excel’s Regression tool. a2 measures the magnitude and (~10–5), showing that the effect is statistically significant. The
significance of the change to the new condition. magnitude of the effect is given by the value of the reagent
Figure 2 shows a time series of average daily specific power coefficient, 1.84 per cent, indicating that the new reagent
before and after the make-up of a new grinding media in a mill. delivers, on average, a recovery 1.8 per cent higher than the old.
Specific power is expressed here as kWh/t of -75 µm product. The 95 per cent confidence limits of this estimate are also given:
The dividing line separating the old and new media regimes 1.05–2.63 per cent. An approximate one-sided lower confidence
represents a period of about ten months during which the new limit can be estimated as 1.84–1.64 s = 1.2 per cent, where s =
media was being gradually made up to replace the old. The standard error of the reagent coefficient (0.40). This can be
line therefore represents a notional ‘instant’ switch between the interpreted as a risk estimate in being 95 per cent confident that
old and new media. Clearly other things could have been, and the improvement in recovery with the new reagent is at least
probably were, going on during the make-up period. 1.2 per cent. So in a cost benefit analysis, if the additional cost
Equation 3 was fitted to the 328 observations, and Table 1 of the new reagent is no more than 1.2 per cent recovery then it
shows part of the (edited) Excel output: is 95 per cent sure that a net benefit will accrue.
The model therefore is: Again, as in any regression analysis, it is important to ensure
that the usual assumptions are met, eg normal residuals,
kWh/t - 75nm = 16.05 + 0.523 kWh/t - 75nm residuals independent of predictions and independent variable
(4)
(t - 1) –2.21 Media values, and no serial correlation, otherwise the calculation of
the P-values, and thus the determination of whether the effect
Table 1 shows that the effect of the media is -2.2 (that is, with being sought is significant, will be compromised. In the case
the new media the specific power is reduced on average by of serial correlation, a term such as that in Equation 3 (yt–1) can
2.2 kWh/t -75 µm, a desirable trend). This value is statistically be included in the model.

T J Napier-Munn
FIG 2 – Time series plot of daily mill kWh/t -75 µm.
THE REFERENCE DISTRIBUTION Suppose that the trial period gives ten results with a mean
One of the problems in using parametric tests such as the t-test recovery three per cent more than the mean of the last ten
or regression analysis is that these methods make assumptions results immediately before the change. The question, as
about the independence of the data and the normality of ever, is ‘is this difference significant (real) or has it arisen by
residuals that are not always fulfilled in mineral processing. chance?’. This can be re-stated rather more formally as ‘what
The non-parametric reference distribution method overcomes is the probability with which such a difference would arise,
this problem as it makes no such assumptions (Box, Hunter if no change had occurred?’. This is answered directly by
and Hunter, 1978; Cavender, 1993). Bruey of Cytec has interrogating the RD.
advocated the use of the reference distribution for reagent
Suppose that the RD comprises 100 sequential datapoints.
trials (Bruey and Briggs, 1997).
Now look at the 100 − 10 + 1 = 91 sequential sets of ten results
The method assumes that a body of consistent data is (1–10, 2–11, 3–12 etc) and see how many times one set of ten
available representing the condition of the process before (eg 1–10 or 2–11) is followed by another set of ten (eg 11–20 or
the trial starts (the ‘reference distribution’ or RD). The
12–21) with a mean recovery higher by three per cent or more.
characteristics of the trial data are then compared with those
If this occurs rarely in the RD (eg ≤5 per cent of times), then
of the RD to determine the probability with which they are
the trial result must be regarded as significant.
likely to have arisen, assuming there has been no change in
the process. If that probability is small (eg ≤0.05), then it is This approach was applied to the media change data.
concluded that a real change has occurred. So essentially In a situation like this, in which a period under standard
it is another significance test. However, in this case the conditions is followed effectively immediately by a period in
probabilities are measured directly from the historical data the trial condition:
rather than from a mathematical distribution such as the
t-distribution. P = R - 2T + 1 (7)
TABLE 1
Excel output for model fit of Equation 1.
Term Coeffs Standard error t Stat P-value Lower 95% Upper 95%
Intercept 16.046 1.619 9.9116 2.09E-20 12.861496 19.231439
Autocorrelation 0.5233 0.04724 11.078 2.07E-24 0.4303556 0.616216
Media -2.209 0.53747 -4.1105 5E-05 -3.266644 -1.151927
R2a = 39.0 per cent, which is not unusual for this kind of data.
TABLE 2
Excel output for model fit of Equation 6.
Coefficients Standard error t Stat P-value Lower 95% Upper 95%

Intercept 78.101518 1.8997005 41.11254 1.28E-75 74.342636 81.8604
Tonnes -0.0001077 4.799E-05 -2.245152 0.026474 -0.000203 -1.28E-05
P80 -0.1839094 0.0552288 -3.329954 0.001135 -0.293189 -0.07463
Feed grade 12.947722 1.3316255 9.723246 4.53E-17 10.312874 15.582571
Reagent 1.8424002 0.4002107 4.603575 9.88E-06 1.050515 2.6342855
R2a = 50.5 per cent.

Some practical problems in running statistically valid plant trials and their solution
where: RD is homogenous. Also, it cannot help with the nuisance

P number of pairings in the RD influence of covariates such as feed grade. However, it is free
of the statistical assumptions required by many of the other
R number of values in the RD (the standard condition)
methods.
T number of trial values
Normally R and T would be chosen with a view to CUMULATIVE SUM (CUSUM) CHARTS
maximising the power of the comparison. Here historical
The cusum chart is a simple but effective way of plotting time
data is being dealt with in which R = 199 and T = 130, so P is
series to show up changes in the process mean. It is a time
negative and therefore a smaller value of T must be chosen.
sequence plot of the cumulative sum of the current value
In this choice, we need to balance the need for sufficient RD
minus some mean value plus the previous cusum. In symbols:
pairings to give a reasonable estimate of confidence with a
sufficiently long period under the new condition to ensure Ct = Ct - 1 + Rt - n (8)
representative behaviour. If, for example, it is decided
100 pairings are needed from the RD, then T = 50. Excel can where:
then be used to calculate the 100 rolling pairs of differences Ct cusum at time t
between the means of 50 successive values, and Excel’s
Percentile function can be used to estimate any desired Ct–1 cusum at time t–1
confidence level from these values. R t value (eg recovery) at time t
In the present case, the difference in kWh/t -75 µm between µ some mean or target value (often the process mean
the first 50 values with the new media and the last 50 of the over the period)
old was -2.7 (compare with -2.2 from the time series model). Figure 4 is a cusum plot of the kWh/t -75 µm data from
The 100 RD pairings give a mean percentile P = 0.1 (ie one- Figure 2, with µ = 31.8 (the overall mean).
sided 90 per cent confidence limit) of -5.9.
In a cusum chart the absolute value of the cusum is not
This suggests that the experimental value of -2.2 is not important. It is the gradient of the line over a characteristic
significant. The comparison is shown graphically in Figure 3a. period that indicates the prevailing process mean over that
Clearly the experimental difference is no different to what period, and changes in gradient indicate changes in process
might be expected by chance based on previous history (the mean, in a much more striking way than the simple time
RD). series plot. If the gradient over a given period is negative,
Using ten trial values and 180 RD pairings gave an then the prevailing mean is less than µ, if positive then the
experimental difference of -4.5 and 90 per cent confidence prevailing mean is more than µ, and if flat then the prevailing
level of -5.9, still not quite significant (Figure 3b). mean is equal to µ.
Note that a re-run of the time series intervention analysis In Figure 4 there is a clear move to a negative gradient
model with just the first 50 new media datapoints returned a at the switch from the old to the new media, confirming
difference of -2.5 and a one-sided significance of 99.97 per cent; the conclusions of the time series modelling and reference
with ten new media datapoints the difference was -2.6 and a distribution discussed earlier that a real change did occur that
one-sided significance of 95.9 per cent. It seems therefore that coincided with the change in media (the cusum chart says
the RD approach is not as powerful as intervention analysis in nothing about the cause of course).
time series modelling. The cusum also clearly shows a major change in the process
Another problem with the RD method is the key assumption mean around observation (day) 135, which had nothing to do
that the RD is stable and represents the constant condition of with the change in media because it precedes it. This clearly
the process before the change. This is an assumption that is demonstrates that long periods of operating data are rarely
often not met – the process mean moves around systematically homogenous, as significant systematic changes occur from
and it is difficult to define what is the appropriate RD. The time to time for whatever reason. The cusum chart is therefore
cusum chart (see below) can help to determine whether the also useful in forensic analysis of historical data, and choosing
FIG 3a – Experimental sample of 50 results compared with reference distribution of 100 pairings.
FIG 3b – Experimental sample of ten results compared with reference distribution of 180 pairings.

T J Napier-Munn
450
Cusum of kWh/t -75micron

400
Old media New media
350
300
250
200
150
100
50
0
-50
0 50 100 150 200 250 300 350
Observation number
FIG 4 – Cusum chart of daily mill kWh/t -75 µm.
homogeneous data sets for use as a reference distribution or changes have occurred, and provide support for the more
for making comparisons between experimental conditions. In rigorous methods. Using a combination of these approaches
the present case, for example, one might want to compare the will help improve the confidence of the final decision. None
relatively homogeneous periods immediately before and after are as efficient as a formal experimental design.
the change (eg days 136–199 versus days 200–260). The methods described in this paper are discussed in more
detail in Napier-Munn (2014).
CONCLUSIONS
If it is desired to compare two process conditions (eg a new ACKNOWLEDGEMENT
versus standard reagent or a new piece of equipment) then Thanks to Dr Denny Meyer for many useful discussions over
a randomised paired trial, analysed by a paired t-test, is the many years.
appropriate experimental design. It will give the correct
decision at the desired level of confidence in the minimum
time. The t-test is robust to departures from the assumptions REFERENCES
underlying the test. In particular, the paired differences are Box, G E P, Hunter, W G and Hunter, J S, 1978. Statistics for
Experimenters (John Wiley and Sons).
usually not serially correlated in the way that the raw data
sometimes is. If three or more conditions need to be compared, Bruey, F and Briggs, D, 1997. The REFDIST method for design
then the paired concept can be extended to a full randomised and analysis of plant trials, in Proceedings Sixth Mill Operators’
block design. These designs have the added key advantage Conference, pp 205–207 (The Australasian Institute of Mining and
of excluding the effects of nuisance variables which may
otherwise influence the result, and should always be used if Cavender, B W, 1993. Review of statistical methods for the analysis of
practical (which is often). comparative experiments, SME Annual Meeting, Reno, February
(SME), Preprint, pp 93–182.
There are, however, situations in which the process
condition cannot be easily switched to conform to such Draper, N R and Smith, H, 1998. Applied Regression Analysis, third
edition (Wiley).
designs. Examples are grinding media trials, trials of large
pieces of equipment, or trials involving long residence Mendenhall, W and Sincich, T, 2003. A Second Course in Statistics:
time processes. In these cases modelling with intervention Regression Analysis, sixth edition (Pearson Education).
analysis is probably the best way of assessing the effect of Napier-Munn, T J, 1995. Detecting performance improvement in
the new condition. Intervention analysis can be used in trials with time-varying mineral processes – three case studies,
either time series modelling or full process modelling. The Minerals Eng, 8(8):843–858.
latter can account for the effect of covariates which may Napier-Munn, T J, 2014. Statistical Methods for Mineral Engineers:
also have changed between the ‘before’ and ‘after’ periods. How to ‘Design Experiments and Analyse Data’, 645 p (JKMRC, The
The reference distribution method can be used, but is likely University of Queensland).
to be less efficient. Cusum charts are a helpful visual aid Napier-Munn, T J and Meyer, D H, 1999. A modified paired t-test
in identifying points in time at which significant process for the analysis of plant trials with data autocorrelated in time,
Minerals Eng, 12(9):1093–1100.

Contents
The optimisation of semi-autogenous grinding

and ball mill based circuits for mineral
processing by means of versatile and efficient
high pressure grinding roll technology
S W Kirsch1 and M J Daniel2
ABSTRACT
Mineral resources all over the world are often heterogeneous and have variable
ore properties and ore grade. As a result, the ore processed in concentrating plants
varies over the life-of-mine. Consequently, the circuits often suffer from bottlenecks
typically at the stages of crushing, semi-autogenous grinding (SAG) and ball milling.
Bottlenecks may be remedied with equipment retro-fits, however; design engineers
should consider a phased approach to the circuit design that delays initial capital
expenditure over the life-of-mine.
This paper deals with the debottlenecking of existing mineral processing plants by
means of high pressure grinding roll (HPGR) applications. It also deals with the use
of innovative flow sheet designs using HPGRs for the design of greenfield plants by
engineering or mining companies. In the case of greenfield plant design an analysis
with respect to energy efficiency and operating cost (OPEX) is presented. The analysis
compares the performance of SAG based circuits against HPGR circuits that may be
influenced by variable ore properties.
Around the world (except China) some 750 HPGRs are installed in various
applications in the cement and minerals industries. In excess of 250 of these HPGRs
are being used since the mid 1980s for crushing and grinding of various ore types
including ‘hard rock’ applications. HPGR technology is mature enough to be
considered as an ideal debottlenecking device. HPGRs are versatile, they have a low
installation footprint and they are flexible in their machine settings (pressure, roll
speed), which can be adjusted to react to ore variations whilst maintaining a high
throughput. Examples for the benefits of using HPGR in terms of energy efficiency,
OPEX and reduced carbon footprint are presented in this paper.
‘Eco-efficient’ flow sheets and sustainable development initiatives are linked to
energy savings as an ‘energy cost’ and benefit to the environment. Energy savings can
also be analysed in terms of the ‘dollar cost’ (OPEX) savings. This paper demonstrates
that HPGR is more energy efficient and may reduce the overall grinding energy
requirements in the order of 16 per cent depending on the target grind size and ore
properties. The HPGR is more cost efficient and may reduce overall grinding costs by
as much as 26 per cent depending on the cost of energy, grinding media and the rate
of grinding media wear in the comminution circuit.
Introduction
Energy efficient mineral processing using HPGR technology has been an item of
great interest at major conferences (Patzelt et al, 2000, 2001, 2005; Klymowsky et al,
2002; Oestreicher and Spollen, 2006; Morley, 2006). Much of the published literature
focuses on alternative flow sheet configurations that should be considered with this
paper. The six alternative flow sheets presented here are similar, but discuss and
analyse the benefits of the HPGR in terms of measured energy efficiency, reduced
OPEX, reduced overall environmental impact and carbon footprint through the
consumption of less energy.
Often, the focus of attention lies in the presentation of alternative flow sheets that
1. MAusIMM, Director, HPGR Solutions, are claimed to have extraordinary energy benefits over one another. Rosario (2008)
Stirling WA 6021. used JKSimMet to compare SAG versus HPGR equivalent circuits and reported
Email: stephan.kirsch@hpgr-solutions.com comminution circuit energy savings of 25 per cent and 30 per cent, but when the
2. Director, CMD Consulting Pty Ltd. entire circuit was analysed the energy benefits were reduced to still 11.7 per cent and
Email: cmdconsulting@bigpond.com 18.4 per cent respectively.
475
Marsden (2008) evaluated SAG versus HPGR circuits Concentrators vary in size. A smallish concentrator may
and followed an approach that evaluated the total energy treat 17 000 t/d (6.2 Mt/a), a medium concentrator 48 000 t/d
required and saved in producing a pound of copper. The (16.2 Mt/a). Whereas some of the very large concentrators
analysis was presented with respect to changing ore grade, being considered these days might treat as much as
but did not evaluate the potential differences attributed to 260 000 t/d (95 Mt/a) with multiple processing trains. Two of
ore properties. Marsden concluded that energy reductions the largest HPGR concentrators in operation are Boddington
of 10.7–19.3 per cent with respect to decreasing grade can be (in operation since 2009) and Cerro Verde (in operation since
expected for processes that included solvent extraction and 2008 (C1) and 2015 (C2)), which treat some 130 000 t/d or
electrowinning and 7.318 per cent energy saving where the some 360 000 t/d respectively. Toromocho in Peru is a good
process includes the production of a concentrate, smelting
example which is looking at multiple 40 ft × 26 ft SAG mills
and refining. This aspect was not considered for the analysis
(28 MW) and the world’s largest ball mills (22 MW) in a
of the results of this paper.
conventional circuit designed to treat 120 000 t/d (44 Mt/a).
The analysis here is based on the interactions between the
ore properties, required throughput rates, cost of energy and
cost of grinding media and the resultant effect this may have HIGH PRESSURE GRINDING ROLL CIRCUIT
on energy savings per se and ‘dollar cost’ savings. TYPES BROWNFIELD AND GREENFIELD
Energy efficiency is often misquoted or misrepresented. Globally some 750 HPGRs are installed in various
In this paper, energy efficiency or ‘energy cost’ is defined applications in the cement and minerals industries. In excess
as the energy that can be saved by the entire processing of 250 of these HPGRs have been used since the mid-1980s for
circuit, which includes an HPGR and which is compared to a crushing and grinding of various ore types including ‘hard
conventional tumbling mill processing circuit. In addition to rock’ applications. HPGR technology is mature enough to be
this the operating cost or ‘dollar cost’ is compared as well as considered as an ideal debottlenecking device. HPGRs are
the environmental impact or carbon footprint resulting from versatile, they have a low installation footprint and they are
reduced energy requirements. flexible in its machine settings (pressure, roll speed), which
can be adjusted to react to ore variations whilst maintaining
‘ENERGY’ COST, ‘DOLLAR’ COST a high throughput. Examples for the benefits of using HPGR
The relatively new and important concept of reduced carbon in terms of energy efficiency, operating costs and reduced
and energy footprints that are linked to so-called greenhouse carbon footprint are presented here and analysed in the next
gas emissions and to climate change, have received much two sections.
attention in the wake of the global financial crisis. The ‘new- Combining the two new concepts, ‘eco-comminution’ and
emerged’ economy is set to feature various aspects of these
‘compact design’ has a significant impact on project net
new concepts as the world begins to better understand
present value and carbon footprint. In this regard Kirsch
sustainable growth within the context of ‘eco-efficiency’.
designed tailored process flow sheets for various projects in
Marginal differences (<5 to 15 per cent) in the capital cost of the minerals industry, integrating HPGR technology, which
SAG versus HPGR flow sheet designs are not studied in detail are the basis of this paper. Further Ausenco Minerals has
in this paper, however, there may well be larger differences in
embraced ‘e4’, a new systematic approach to engineering
the operating cost and in the ‘energy footprint’.
process design that includes a formal analysis of energy,
Innovative comminution flow sheets need not be complex economics, environment and efficiency.
but rather based on very simple open circuit concepts. This
As part of this paper, there are six different flow sheets
has the advantage of simplifying eco-comminution concepts
and the reality of having to deal with increased engineering (Table 1 and Figures 1–6) that are used to evaluate SAG
design cost and risks associated with these new ideas. versus HPGR based circuits. These consist of a conventional
Lane (2008) has demonstrated a path of engineering design SAG milling flow sheet (Baseline SAG), and variants of this,
that still includes conventional wisdom, but that compact where the HPGR is either introduced as a debottlenecking
design concepts can equally improve the environmental and device (Debottleneck options 1 and 2) or a replacement for the
economic impact. SAG mill (Baseline HPGR).
Much greater ‘eco-efficiency’ can also be realised by reducing Variants of the Baseline HPGR/ball mill circuit mill are
the consumption of mill liners and grinding media. Though presented as Variant HPGR (1) and Variant HPGR (2), which
the ‘dollar cost’ of comminution is normally accounted for are better suited to the initial size of the concentrator being
as a direct electricity expense in the process, and is rarely designed.
considered for its overall energy cost or ‘embodied energy’ of When the ‘debottleneck’ flow sheets are compared to the
manufacturing the steel, eg for grinding media, which amounts
Baseline SAG circuit option for the sake of improving the
to up to 4–6 kWh/t. The section SAG versus HPGR compares
overall circuit throughput, the energy saving is likely to be
and analyses the benefits of each type of circuit flow sheet.
based on a pro-rata basis of installed HPGR power. In addition,
Direct energy use in comminution processes was reviewed the overall circuit throughput will be increased significantly.
by Daniel (2007) and showed that 0.56 per cent (87 TWh) of the
global net electrical energy consumption of 15 500 TWh/a is Thus, the following analysis examines the benefits of
used to crush and grind non-ferrous ores. Of this, 33 per cent HPGR circuits against the performance of conventional SAG
and 53 per cent of the energy is required to process gold and mill grinding circuits. SAB refers to a circuit without pebble
copper ores respectively. The option to use HPGR technology crushing (SAG/ball), and SACB refers to pebble crushing
for these applications is a significant contributor to energy product stream returning to the ball mills. SAB, SACB are
savings, which suggests that the HPGR should be targeted variants of the SABC (SAG/ball/crusher) circuit shown in
at gold and copper mining operations in the future to be Figure 1. The benefits of HPGR are measured and analysed
effective in reducing operational costs, energy consumption in terms of energy efficiency, operational costs and carbon
and carbon emissions. footprint relative to the conventional SAG milling circuit.

The optimisation of semi-autogenous grinding and ball mill based circuits for mineral processing
TABLE 1
Flow sheet considered in the semi-autogenous grinding versus high pressure grinding roll circuit analysis.
Flow sheet ID Flow sheet description Figure

Baseline SAG Conventional SAG (SAB, SABC, SACB) Figure 1
De-bottleneck (1) Conventional SAG (SAB, SABC, SACB) + HPGR plant throughput expansion projects where Brownfield Figure 2
new feed is split between the existing SAG mill circuit and a new HPGR module
De-bottleneck (2) Conventional SAG (SAB, SABC, SACB) + HPGR plant throughput expansion projects where Brownfield Figure 3
the HPGR treats products from the SAG mill pebble crusher
Baseline HPGR Conventional HPGR (HPGR, BM) – Boddington, Cerro Verde <40 Mt/a Figure 4
Variant HPGR (1) Variant of circuit option C using multiple pass HPGRs typically configured as HPGR2 – BM Greenfield Figure 5
suitable for very large concentrators >40 Mt/a
Variant HPGR (2) Variant of circuit option C using a single HPGR with edge recycle, HPGRe – BM suitable for Greenfield Figure 6
small concentrators <10 Mt/a
FIG 1 – Conventional SAG circuit (SABC).
FIG 2 – Conventional SAG (SAB, SABC, SACB) + HPGR – suited to brownfield circuit capacity expansion options or debottlenecking option 1.
SEMI-AUTOGENOUS GRINDING MILLS typically at the stages of crushing, SAG and ball milling as
only few operating parameters can be varied to cope with
LIMITED BY ORE COMPETENCY different ore types (eg ball size, ball load, mill speed, trommel
Mineral resources all over the world are often heterogeneous screen aperture, pebble ports and pebble production). These
and have variable ore properties. As a result, the ore circuits are often termed as SAG limiting circuits.
processed in concentrating plants varies of the life-of-mine. The SAG mill becomes limiting when the coarse ore
Consequently, the circuits often suffer from bottlenecks becomes more competent. Ore competency or resistance

FIG 3 – Conventional semi-autogenous grinding (SAB, SABC, SACB) + high pressure grinding roll – suited to brownfield circuit capacity
expansion options or debottlenecking option 2.
FIG 4 – Conventional HPGR (HPGR, BM) – Boddington, Cerro Verde <40 Mt/a.
FIG 5 – Multiple pass HPGRs typically configured as (HPGR2 – BM) for very large concentrators >40 Mt/a.

Furthermore, the ore can be characterised in the JK drop

weight testing device (DWT) and classified in terms of a
new ore strength parameter called the Drop Weight Index
(DWi®). The DWi is measured in kWh/m3 with a higher value
indicating a higher resistance to impact breakage. The DWI is
related to the JK DWT A × b parameter (Morrell, 2004).
As a general rule, ore competency increases with depth
within the pit. Over time, the throughput of a given sized mill
will decrease as the SAG mill becomes limiting. The decreased
throughput can be alleviated to a small degree by either
increasing ball charge and mill speed, or through secondary
crushing, but typically at the expense of a coarser grind size.
For the purpose of the analysis of this paper, three ore
competency types at three mines X, Y and Z have been chosen.
These all have a density of 2.65 g/cm3 and have JK DWT A × b
values of 38 (hard), 46 (medium) and 64 (relatively soft). The
three mines were chosen on the basis that if used with three
commonly sized SAG mills, this will result in a throughput
of approximately 2000 t/h or 16.2 Mt/a as determined using
an empirical SAG mill throughput model (Table 2). This
was deemed necessary in order to demonstrate how ore
competency affects throughput. These results are also used to
establish a baseline from which to compare the HPGR circuit
and associated energy efficiency, operating cost comparison
and carbon footprint.
The SAG performance with respect to ore competency
(A × b) for these three mines is compared with the Newmont
FIG 6 – Single HPGR with edge recycle, HPGRe – mine database as published by (Veillette and Parker, 2005). In
BM suitable for small concentrators <10 Mt/a. Figure 7, the specific energy of the SAG mill for each of the three
mines is overlayed against the Newmont database in order to
to impact breakage within the SAG mill is measured using, support the accuracy of the SAG limiting model prediction.
eg the SAG Mill Comminution (SMC®) Test. The SMC® Test Maximum copper recovery on a revenue/cost basis is assumed
is a precision test with a high degree of control over particle for each process grind size. More resource based economics
would be required should the mine economics be optimised in
size and breakage energy (Morrell, 2004; JKTech, 2009).
Mine-to-Mill® type studies that integrate blasting practices and
SMC® Tests can be performed in order to profile an orebody costs with mill operations and costs.
for greenfield or in existing operations. Several tests can be
Table 2 assumes conventional blasting practices with a SAG
carried out quickly to get information about the comminution mill feed that is prepared by means of a gyratory crusher
circuit throughput. with a closed side setting of 150 mm. Detailed operating costs
TABLE 2
Performance of the semi-autogenous grinding mill for three different mine sites.
Mine X Mine Y Mine Z

Mine production capacity Mt/a 16.2 16.2 16.2
Predicted SAG F80 mm 117 102 81
Primary crush CSS mm 150 150 150
Ore property (JK DWT) A×b 38 46 64
Ore density g/cm3 2.65 2.65 2.65
Slurry % solids % 70 70 70
Mill diameter (Shell) ft 40 38 36
Mill length (EGL) ft 20 18 15
Liner thickness mm 100 100 100
Charge density t/m3 4.4 4.4 4.5
Charge density t/m3 3.9 4.0 4.2
Mill speed (Crit) % 75.0 75.0 75.0
Ball charge % 12.0 12.0 12.0
Mill filling (Jt) % 24.7 23.8 22.6
Rock charge % 12.7 11.8 10.6
Ball to rock ratio 0.94 1.02 1.13
SAG gross power kW 17 967 14 167 10 276
SAG specific energy kWh/t 8.8 6.9 5.0
SAG throughput t/h 2033 2036 2035

Mine X
Mine Y
Mine Z
Mine X
Mine Y
Mine Z
FIG 7 – SAG specific energy of the three mines with respect to ore competency (A × b) and the Newmont database (Veillette and Parker, 2005).
which include liner and grinding media consumption are Table 4 uses an estimated power cost of A$7 c/kWh and
included, however, grate replacement costs are assumed to A$1200/t of grinding media. Grinding media consumption
wear at the same rate as the liners. rates are 0.45 kg/t for SAG mills, 0.55 kg/t for SAG circuit ball
mills and 0.6 kg/t for HPGR circuit ball mills due to a coarser
SEMI-AUTOGENOUS GRINDING VERSUS HIGH PRESSURE transfer size experienced in the HPGR circuit.
Based on empirical data and for the purpose of this exercise
GRINDING ROLL TRADE-OFF STUDY ANALYSIS
the frequency of liner replacement for the SAG mills is two
Table 2 summarises the performance of SAG mills for three per annum and for the HPGRs 1.5/a. Replacement costs for
operating mines. This data establishes the throughput baseline liners are presented in Table 4 at A$2–2.5 M for each SAG
from which the circuit’s total specific energy and target grind and A$2.8 M for two HPGRs. The ball mill liner replacement
size may be calculated. The established baseline circuits are cost is not listed for the cost comparison of SAG versus HPGR
then evaluated against equivalent sized HPGR/ball mill based circuits as ball mills are equally installed for both type
circuits as defined by the flow sheet represented in Figure 3. of circuits. Hence this has no impact on the comparison.
Based on pilot testing, these results are later used to calculate
the HPGR circuit performance. This provides the basis for the Additional power required for ancillary equipment such
comparison for the SAG versus the HPGR trade-off. as screens and conveyors is not included and may offset the
energy saving by 35 per cent.
Table 3 presents the SAG versus HPGR energy related or
direct ‘energy cost’ comparison. For all three cases, mines X, The results show that a reduction of 25.3 to 26.0 per cent of
Y and Z, the expected energy saving is relatively constant the operating costs (dollar cost) can be realised by means of
and expected to be in the 15–17 per cent range. Although HPGR circuits. These savings are on the conservative side as
the absolute energy saving in terms of reduced installed 1.5 HPGR liner changes have been considered for all three
power is greater for mine X at 5.8 MW than for mines Y and mines (X, Y, Z). Typically, one HPGR liner change would
Z at 3.6 MW and 2.1 MW respectively. This analysis implies apply for mines X, Y and mine Z respectively.
that the energy saving benefits are greater with HPGR in The reduction in cost per tonne treated ranges from
absolute terms when the ore is more competent. The total A$0.54/t to A$0.73/t which is somewhat higher to what was
circuit specific energy was determined using empirical predicted for Cerro Verde at U$0.37/t at current exchange
process models (Table 2), which were used to generate rates (Vanderbeek et al, 2006). The main reason for this is
SAG mill specific energy and used to compare these with the reduced HPGR liner cost which dropped significantly in
the total circuit specific energy using Morrell’s method in recent years as the HPGR original equipment manufacturers
Table 3 (Morrell, 2009). The data show that the two methods lost their market monopoly for liners and other high-quality
produce very similar results. products are available for the industry.
There is a tendency to overlook benefits of using HPGR Variations are sensitive to energy costs, grinding media
over a conventional circuit ie the complete elimination of costs, consumption rates and the rate of currency exchange.
using grinding media when the HPGR replaces the SAG mill. Screening costs and conveyor costs could account for $0.04/t
Grinding media consumption rates for SAG mills and ball – $0.08/t, and crusher liner costs $0.04/t, as well as dust
mills are well documented. Table 4 provides an operating extraction at $0.02/t. The impact may reduce overall costs
cost summary that includes the ‘dollar cost’ (OPEX) benefits per tonne by $0.1/t with a range from $0.2/t to $0.39/t.
of using less energy and less grinding media in the HPGR Production ramp up time for HPGR circuits is faster, which
circuits. may positively affect capital loan costs that may offset

TABLE 3
Energy efficiency of semi-autogenous grinding versus high pressure grinding roll circuits.
SAG versus HPGR Units Mine X, SABC Mine X, HPGR Mine Y, SABC Mine Y, HPGR Mine Z, SABC Mine Z, HPGR
circuit analysis 40 ft × 20 ft 2 × 24/17 38 ft × 18 ft 2 × 24/17 36 ft × 15 ft 2 × 24/17
20 MW SAG 5 MW 18 MW SAG 5 MW 14 MW SAG 5 MW
Plant throughput Mt/a 16.6 16.6 16.6 16.6 16.6 16.6
Plant throughput t/d 45 400 45 400 45 500 45 500 45 400 45 400
Tonnes per hour t/h 2032 2032 2036 2036 2035 2035
Ore property
A*b 38.0 38.0 46.0 46.0 64.0 64.0
SG t/m3 2.65 2.65 2.65 2.65 2.65 2.65
BWI kWh/t 16.0 16.0 14.0 14.0 10.0 10.0
Circuit data
F80 mm 117 117 102 102 81 81
P80 µm 106 106 150 150 150 150
Metric t/h t/h 2033 2033 2036 2036 2035 2035
(SAG versus HPGR)
Secondary crushing kWh/t 0.6 0.6 0.6
HPGR kWh/t 3.4 3.0 2.9
Ball mill kWh/t 12.4 8.5 5.3
Pebble crushing kWh/t 0.1 0.1 0.1
SAG (Ecs) kWh/t 8.7 6.9 5.0
Ball mill (Ecs) kWh/t 10.5 6.9 4.8
Total circuit specific energy kWh/t 19.3 16.4 14.0 12.2 9.9 8.8
Total (Wt) Morrell's method SABC kWh/t 19.4 14.2 10.1
Energy reduction kWh/t 3.2 2.3 1.6
Circuit pinion power MW 39.2 33.4 28.4 24.8 20.1 17.9
Installed MW' saved MW 5.8 3.6 2.1
SABC versus HPGR energy saving % 16.5 16.2 15.8
the screening, crushing and conveying costs. Overall cost ACKNOWLEDGEMENTS

benefits require analysis on a case-by-case basis. Examples This is a joint contribution by Stephan Kirsch and Dr Mike
provided here are illustrative. Daniel who acknowledge HPGR Solutions and CMD
Consulting Pty Ltd for their support to publish.
CONCLUSIONS AND FUTURE DIRECTION
HPGRs are effective for brownfield retro-fits to increase REFERENCES
or maintain plant capacity as ore types become more Daniel, M J, 2007. Energy efficient mineral liberation using HPGR
competent. Greenfield eco-comminution circuit design with technology, PhD thesis, JKMRC, School of Engineering,
HPGRs as opposed to SAG mills are more energy efficient, University of Queensland.
hence cost-effective with a reduced carbon footprint. The JKTech, 2009. Comminution testing [online]. Available from: <http://
main conclusions of this paper are thus as follows: jktech.com.au/comminution-testing> [Accessed: 31 July 2017].
•• HPGR is more energy efficient and may reduce the overall Klymowsky, R, Patzelt, N, Knecht, J and Burchardt, E, 2002. Selection
and sizing of high pressure grinding rolls, in Proceedings Mineral
grinding energy requirements in the order of 16 per cent Processing Plant Design, vol 1, pp 636–668 (SME: Littleton).
depending on the target grind size and ore property.
Lane, G, 2008. Engineering design of concentrators in Australia, Asia
•• HPGR is more cost efficient and may reduce overall and Africa – what drives the capital cost?, in Proceedings Procemin
grinding costs by as much as 25.3 to 26.0 per cent 2008, V International Mineral Processing Seminar, 22–24 October
depending on the cost of energy, grinding media and 2008, Santiago, pp 29–39.
the rate of grinding media wear in the comminution Marsden, J, 2008. Energy efficiency and copper hydrometallurgy,
circuit. These savings are on the conservative side for Hydrometallurgy 2008, 6th International Symposium, Phoenix.
the three mines considered for this comparison. Morley, C, 2006. Flowsheets for HPGR, SAG 2006 conference
Vancouver.
In the future, brownfield and greenfield comminution circuits
should be evaluated using the above-mentioned concepts to Morrell, S, 2004. Predicting the specific energy of autogenous and
semi-autogenous mills from small diameter drill core samples,
re-enforce the decision-making process when choices between
Minerals Engineering, volume 17/3.
different comminution circuit concepts are proposed.
Morrell, S, 2009. Predicting the overall specific energy requirement of
The next step would be to equate the HPGR efficiency into crushing, high pressure grinding roll and tumbling mill circuits,
carbon offsets or carbon credits. Minerals Engineering, 22(6):544–549.

TABLE 4
Operating cost comparison between SAG and HPGR circuits
SABC versus HPGR Units Mine X, SABC Mine X, HPGR Mine Y, SABC Mine Y, HPGR Mine Z, SABC Mine, Z HPGR
operating cost summary and 40 ft × 20 ft 2 × 24/17 38 ft × 18 ft 2 × 24/17 36 ft × 15 ft 2 × 24/17
A$ 20 MW 5 MW 18 MW SAG 5 MW 14 MW SAG MW
Plant throughput Mt/a 16.6 16.6 16.6 16.6 16.6 16.6
Plant throughput t/d 45 400 45 400 45 500 45 500 45 400 45 400
Tonnes per hour t/h 2032 2032 2036 2036 2035 2035
Power cost c/kWh 7 7 7 7 7 7
Grind media cost $/kg 1.2 1.2 1.2 1.2 1.2 1.2
SAG Mill relining cost $M 2.5 2.3 2
Liner replacement cost for 2 HPGRs $M 2.8 2.8 2.8
(new/refurbished)
SAG grinding media consumption kg/t 0.45 0.00 0.45 0.00 0.45 0.00
rate
Ball mill media consumption rate kg/t 0.55 0.60 0.55 0.60 0.55 0.60
Total media consumption kg/t 1.00 0.60 1.00 0.60 1.00 0.60
Operating cost power $/t 1.35 1.15 0.98 0.85 0.69 0.62
Operating cost media $/t 1.20 0.72 1.20 0.72 1.20 0.72
Operating cost liners $/t 0.30 0.25 0.28 0.25 0.24 0.25
Comminution circ operating cost $/t 2.85 2.12 2.46 1.82 2.13 1.59
TOTAL cost saving $/t 0.73 0.64 0.54
HPGR cost saving power $M/a 3.3 2.1 1.2
HPGR cost saving for media + $M/a 8.8 8.5 7.8
liners
Cost saving $M/a 12.1 10.6 9.0
Cost saving % 25.6 26.0 25.3
Oestreicher, C and Spollen, C F, 2006. HPGR versus SAG mill Rosario, P, 2008. Analyses of the total required energy for
selection for the Los Bronces grinding circuit expansion, SAG comminution of hard ores in SAG mill and HPGR circuits,
2001 Conference. Procemin 2008, V International Minerals Processing Seminar, 22–
24 October 2008, Santiago, pp 129138.
Patzelt, N, Klymowsky, R, Burchardt, E and Knecht, J, 2001. High
pressure grinding rolls in AG/SAG mill circuits – the next step Vanderbeek, J L, Linde, T B, Brack, W S and Marsden, J O, 2006.
in the evolution of grinding plants for the new millennium, SAG HPGR Implementation at Cerro Verde, in Proceedings SAG
2001, Vancouver. Conference 2006, IV45.
Patzelt, N, Klymowsky, R I, Knecht, J and Buchardt, E, 2005. HPGRs Veillette, G and Parker, B, 2005. Boddington expansion project
for hard rock application, in Proceedings of the Randol Metallurgy comminution circuit features and testwork, in Proceedings Randol
Forum 2005, Perth. Gold Forum.
Patzelt, N, Knecht, N, Burchardt, E and Klymowsky, R I, 2000.
Challenges for high pressure grinding in the new millennium,
in Proceedings Seventh Mill Operators Conference, pp 44–55 (The

Contents
Optimisation opportunities for

high pressure grinding rolls circuits
M S Powell1, M M Hilden2, C M Evertsson3, G Asbjörnsson4, A H Benzer5,
A N Mainza6, L M Tavares7, B Davis8, N Plint9 and C Rule10
ABSTRACT
The Mogalakwena North Concentrator (MNC) of Anglo Platinum Operations was
built with the capability to allow for future projected expansion. Several research
groups, currently working with Anglo American on collaborative research over
a range of projects, were brought in to review the existing operation, with a view
to releasing some specific bottlenecks and providing specialist design input to the
potential plant expansion.
Based on a thorough review of the high pressure grinding rolls (HPGR) and
crushing circuit, a wide range of debottlenecking opportunities were identified. This
was followed by a series of surveys to provide performance measurements and to
model the circuit in detail. From this effort, a full simulation of the plant has been
developed, including updated models of the HPGR, crusher, bins and screens. These
models have been implemented in the researchers’ Model Development Kit (MDK)
1. FAusIMM, Professorial Research Fellow, version of JKSimMet, and in a prototype dynamic simulator based on the Matlab
Chair in Sustainable Comminution, Julius Simulink software.
The observations, measurements and simulations have provided guidance and
of Queensland, Indooroopilly Qld 4068. quantified the benefits of some proposed circuit modifications. Based on these findings
Email: malcolm.powell@uq.edu.au a number of upgrades have since been implemented, with the aim of ensuring full
2. Senior Research Officer, Julius Kruttschnitt
utilisation of the installed equipment. The circuit demonstrated a potential expansion
Mineral Research Centre, Sustainable of over 30 per cent through minor improvements, and a 70 per cent increase in capacity
Minerals Institute, The University of with the addition of some strategic equipment.
Queensland, Indooroopilly Qld 4068. The HPGR-crushing upgrade project demonstrates the potential benefits of
Email: marko.hilden@uq.edu.au nurturing disparate research groups into a strong inter-disciplinary technical team
3. Professor Product and Production capable of supporting operational needs while developing future technology.
Development, Chalmers Rock Processing
Research, Chalmers University of Technology,
Göteborg SE-41296, Sweden. INTRODUCTION
Email: magnus.evertsson@chalmers.se With the rise of HPGR processing plants over recent years, a new focus on process
4. Research Fellow, Chalmers Rock Processing design (utilising tools such as modelling and simulation), has been aimed at these
Research, Chalmers University of Technology, dry processing circuits. HPGR technology offers an array of well-publicised potential
Göteborg SE-41296, Sweden. advantages over conventional semi-autogenous grinding ball milling circuits, albeit
Email: gauti@chalmers.se
with a new set of design and operation challenges. These challenges were well
5. Departmental Head, Professor, Department expressed at the SAG11 conference in Vancouver by Hart et al (2011a, 2011b), Koski,
of Mining Engineering, Mineral Processing
Vanderbeek and Enriquez (2011), Villanueva et al (2011), Banini et al (2011) and Powell
Division, Hacettepe University, Beytepe-
Ankara 06800, Turkey. et al (2011).
Email: benzer@hacettepe.edu.tr Under-design of the crushing circuit is a common theme, along with the associated
6. Professor, Centre for Minerals Research, operating and maintenance issues. Overloading of an inadequate screening capacity
University of Cape Town, Rondebosch 7700, formed a constraint at a number of sites. This limits the circulating load about the
South Africa. Email: aubrey.mainza@uct.ac.za HPGR, and thus its utilisation. The HPGRs were noted to operate above design
7. Professor, Universidade Federal do specification, at a markedly lower specific energy, leaving about 40 per cent spare
Rio de Janeiro, COPPE-UFRJ, Department of motor power.
Metallurgical and Materials Engineering,
Rio de Janeiro, Brazil.
In choosing to take the relatively untried route of an HPGR circuit, for the extremely
Email: tavares@metalmat.ufrj.br competent ore at the MNC, the Anglo Platinum design team chose to include excess
capacity in the key capital equipment items, so as to cover uncertainty in operation
8. Formerly: Business Area Manager,
Mogalakwena Concentrator, Anglo Platinum, and to provide potential for expected future expansions. Once in full production it was
Mokopane, South Africa. found that the circuit did not perform as planned, with some unexpected constraints
9. Director, Sustainable Minerals Institute, The to the operation. A specialist technical team (comprised of research groups funded
University of Queensland, St Lucia Qld 4068. under a range of long-term collaborative research projects) was brought in to help
Email: n.plint@uq.edu.au. identify the remaining bottlenecks, and to then provide advice on the proposed plant
10. Formerly: Head Concentrator Technology, expansion. The team consists of the comminution groups from the Julius Kruttschnitt
Anglo Platinum, Johannesburg 2107, Mineral Research Centre, University of Cape Town, Hacettepe University, Chalmers
South Africa. University of Technology and Universidade Federal do Rio de Janeiro – all members
483
M S Powell et al
of the Global Comminution Collaborative (GCC). This paper Approach

highlights the approach used to identify and address circuit The project team followed a classic investigative approach:
bottlenecks, quantify circuit performance and provide design steadily building up relevant data and improved process
input to the circuit expansion. understanding, and then applying these to detailed modelling
and simulation – the latter being used as a tool to investigate
OBJECTIVE circuit optimisation and design scenarios.
As the MNC circuit could not consistently achieve its design A detailed list of issues and opportunities was presented
capacity of 1000 t/h through the dry circuit, the initial objective by Powell et al (2011) so has not been repeated here. Instead,
key points are noted as they arise through application of the
of the project team was to seek out the underlying causes of
systematic strategy. It was observed that many of the problem
the restrictions and to provide advice on how to address them.
areas were difficult to identify and address by an operating
While this work was progressing, how best to fully utilise the team unfamiliar with these new circuits; emphasising the
installed spare capacity became clear. This helped prompt the need for expert technical input and training when moving
proposed expansion, an option both achievable and attractive into new styles of operation.
to Anglo American Platinum, given this is their only open pit
operation. SCOPING REVIEW
The circuit is on a large-scale, spanning 1 km in length, as The team typically favours an initial quick review conducted
illustrated in Figure 1. The wet milling section, which lies over one to two days. This review is used to capture an image
beyond this part of the circuit, consists of two 16 MW ball of the circuit operation via several channels: harvesting
mills, five M10 000 Isa mills and three stages of flotation; invaluable operational input from the operators and plant
forming a sizable operating plant. metallurgists; assembling a list of operating issues and the
strategies used to date to address them; collecting some
The dry section flow sheet is illustrated in Figure 2, showing
operating data; and most importantly, walking the plant and
how numerous components link together into a series of
getting a feel for it. The team is renowned for taking hours to
circulating streams. These result in knock-on interactions, walk round a plant, all the while applying careful insight into
with time lags and fluctuations passing between equipment all aspects of the operation. With an experienced metallurgist
with different duties and capacities. The first step was to to provide relevant input, this approach can provide the most
develop a sound approach to addressing these challenges. benefit of all.
Size ranges in each stream are indicated and the red triangles Using this approach, issues such as uneven feed into
show the location of weightometers. the crusher chambers resulting in rapid cyclic loading of
FIG 1 – A view of the Mogalakwena North Concentrator crushing and high pressure grinding rolls circuit.
FIG 2 – Mogalakwena North Concentrator crushing and high pressure grinding rolls flow sheet.

Optimisation opportunities for high pressure grinding rolls circuits
crushers (Figure 3), segregation on belts and in feed bins •• excessive wear on step washers (oil film lubricated plain
leading to uneven utilisation of screens (Figure 4), HPGR bearing) carrying the axial crushing forces from the
product flake density, and the noting of worn parts which main shaft.
speak to operational issues, are just some of the challenges The rapid cycling of the pressure in the chamber is not
which were uncovered by the experienced team during this observed in the control room due to smoothing of data sent
thorough review. It was particularly noted in feedback from from the Sandvik ASRi control unit data logging system.
the operators of other crushing–HPGR circuits that materials However, it was clearly observed in the field at the local
handling issues are critical to the circuit performance and control boxes, with a period of 0.26 s (3.8 Hz) and variation
must be given utmost priority in the design phase. in pressure of 1 MPa, Figure 3. This cyclic stressing fatigues
the crusher and leads to increased maintenance and reduced
Crusher feeding and operation reliability of the equipment. Additionally, it can be noted that
As a consequence of poor feed presentation, the crushers the crusher is operating well below the set point maximum
received segregated feed and dramatically uneven chamber of 5.9 MPa, as shown by the blue line in Figure 3, due to
loading, illustrated in Figure 3. The feed can be seen to fall on the inbuilt protection system which keeps the fluctuating
the near-side (relative to the feeder) of the spider, with larger pressure maximum below the set point.
rocks cascading over the far side. When observing from the From the site review work alone, a number of
side there was about a 0.5 m height variation of the rock level recommendations can be provided without recourse to any
across the feed chamber. With the level detector directed at modelling work. Indeed, many of these improvements lie
the higher side, the low side was kept well below the chamber outside the scope of the simulation and calculation advice
entrance. The net effect was that the crusher was not choke fed. offered by remote consultants.
The consequences were: A couple of useful generic tools are noted here.
•• high stress on one side and low on the other
•• high filling on one side and low on the other Overall equipment effectiveness
•• coarse rocks on one side and fine on the other. To quantify the available capacity of the key equipment, the
application of a standard manufacturing assessment tool –
The consequences of these were:
overall equipment effectiveness (OEE) calculations – was
•• non-choke fed conditions and thus poor overall used.
reduction
•• uneven liner wear OEE = availability × performance × quality
•• rapidly fluctuating power and pressure causing undue Availability is the amount of time the equipment is available
stress on the main shaft and bearings and crusher fatigue for use and is not shutdown due to maintenance, breakdowns
(potential risk of cracking of different structural parts, etc. Performance is the relative performance compared to
that is, bottomshell or topshell) the maximum possible performance. Quality is the relative
•• high wear spots on the spider and head number of accepted parts of production. This is a difficult
FIG 3 – Uneven and non-choke fed conditions with consequent cyclic pressure loading in the crusher.
FIG 4 – Uneven loading of the high pressure grinding rolls screens.

M S Powell et al
parameter to retrieve in a crushing process, so is assumed to

be 100 per cent in this study; meaning that all production is
accepted. However, this can be set to a desired product size
with a sliding scale as this is not achieved. Figure 5 shows the
key OEE parameters for CR003, and demonstrates how they
are calculated.
For CR003, the availability and performance were calculated
to be 87 per cent and 56 per cent respectively. The overall OEE
index is therefore 49 per cent. It can therefore be concluded
that the crusher has a theoretical limit of 50 per cent spare
capacity available for production, and that the bottlenecks
stem from elsewhere in the circuit.
High pressure grinding rolls set-up

The HPGR was operating well below its potential, acting FIG 6 – HPGR product at the standard operating pressure of 110 bar.
more in the form of a tertiary rolls crusher than a compressed
bed machine. This was clear from the considerable fraction competent material enters the rolls, forcing them to open up.
of coarse rocks in the product that look like standard crusher A 10 mm increase in gap will spike the pressure up to about
product, and the complete absence of flake, as can be seen in 200 bar. This leads to stud damage and rolls skewing, as the
Figure 6. rolls are unable to move rapidly enough. Bearing in mind that
The HPGR also showed considerable spare capacity, as per the throughput is directly proportional to the gap width, if
design expectation. However, the limitation of a fixed rolls the nitrogen ‘spring’ is softened (indicated by ‘softer spring’
speed resulted in a frequent stop-start operation, having in Figure 7), the rolls cannot achieve the required operating
a number of unfavourable knock-on wear effects. This pressure at a feasible operating gap. The third option of
motivated the upgrade to variable speed rolls, to leverage the starting at a lower gap and having a softer nitrogen spring,
rapid throughput response potential of the HPGR. allows the rolls to operate at the same gap as is currently used,
It is standard practice in mineral processing operations but without the steep pressure penalty. The pressure builds
to provide a relatively coarse feed to the HPGR. This is up sooner at a lower gap, so the rolls are better able to form
principally to save on crushing capacity. The feed tends to a competent bed. Consequently, this option allows for a safer
drive a large zero gap, to limit stress on the rolls surface at and more controlled rolls operation.
start-up and to protect the studs from point loading as large The physical set-up and the control system were not well
particles pass through. suited to HPGR operation, resulting in a varying feed and
However, the large zero gap has a significant knock-on stop-start operation. The HPGR feed rate was controlled by
effect on the operation of the HPGR. The interaction of the the level in the feed bin which is undesirable since the feed
gap with the nitrogen and oil pressures in the loading system, size distribution changes as the ratio of recycle to new feed
dictate the operating pressure of the HPGR. The outcome is varied. This variation is due to the fluctuating topping up
of modelling this interaction is most informative. Figure 7 of the bin with crushed feed to achieve level control. The
illustrates an example of the interplay of these factors. For recycle around the HPGR should be kept near equilibrium,
a zero gap of 48 mm and a desired operating pressure of by matching its rate in and out of the feed bin. This can be
140 bar, the rolls open up to 68 mm, on a rather steep pressure achieved by adjusting the fresh feed rate to provide the
response curve. The down side of this response is that the steep difference between the recycle rate and the feed rate being
pressure curve results in massive pressure spikes, if large or drawn by the HPGR.
FIG 5 – Overall equipment effectiveness calculations for CR003.

FIG 7 – High pressure grinding rolls pressure gap response.
Feed into bins processing conditions and equipment set-up. The secondary
The screen bin was identified as the primary bottleneck in and tertiary duty crushers were each operated at four closed
the circuit. This was reflected in the narrow band within side settings; the screen feeds were varied by a factor of two;
which the bin filling had to be controlled in order to ensure and the HPGR was run at a wide range of pressures, different
delivery of feed to both screens, dramatically uneven feed zero gap settings, and with different fractions of fines in the
to each screen, and rapid overfilling of the bins necessitating recycle. These measures of performance related to product,
frequent feed stoppages. A vertical wedge of compacted ore provide a range of useful information in themselves, such as
had formed between the two outlets. This extended about half screening efficiencies, reduction ratios, power utilisation etc.
way up the bin and limited the control window to between These measures were conducted over three separate periods
70 per cent per cent and 90 per cent of bin filling, thus reducing of surveying, driven by access, availability and the operating
the effective surge capacity to about one-third of achievable, regime of the plant. This effort allowed a wide range of
leaving about 50 t live surge capacity, or about 10 s. operating data to be collected. Standard product sizing and
The underlying cause is that the main conveyor feeds into ore testing was conducted on the survey samples.
the bin at 90° to the discharge feeding the screens. The natural
segregation of coarse material to the top of the conveyor and Crushers
the overthrow of the discharge stream, resulted in a strong The Sandvik Hydrocone CH880 crushers were controlled to
bias of coarse material and total flow to the further (No 1) a range of closed side settings (CSS) and the feed rate was
screen. The high fines and clay content in the start-up phase adjusted to provide full choke-feed conditions. Once the
of the plant led to the wedge. However, a vertical and central power and pressure were stable, markers were thrown onto
falling feed stream would have kept this mainly clear and the product belt to indicate the start and end of an appropriate
prevented this extreme build-up. section of product. This was repeated for each CSS. The
Based on the individual performance of the No 1 screen, resultant set of response curves are shown in Figure 8 for
it was estimated that the screening capacity would increase CR001 – the crusher acting in secondary duty receiving the
by 30–50 per cent if both screens were evenly fed. Based on grizzly oversize material. In addition to the throughput
this analysis the site shifted the conveyor head-pulley and and power relationships, it can be observed that the specific
installed a deflector rock box to force a more vertical and energy drops off from 1.1 to an exceedingly low 0.5 kWh/t,
central rock drop into the bin thereby helping redistribute the as the CSS is increased from 30 to 60 mm. Thus, although
feed. Circuit capacity was noted to markedly improve. the throughput increases with CSS, the product will become
The issues outlined above are in no way unique to this coarser, so the circuit capacity will be a balance of the degree
operation. They were presented as a reference source in of reduction per pass and the crusher capacity.
the work of Powell et al (2011). As part of the upgrade A similar set of response curves were derived for CR003 – the
and expansion process, the site systematically addressed crusher acting in a tertiary capacity treating the majority of the
the localised issues and bottlenecks, and thereby ensured recycle feed from the crusher screens, Figure 9. As with CR001,
consistent performance at design tonnage. This initial success the specific energy drops off with increasing CSS; down to a low
emphasizes that both simulation and operation knowledge value of only 0.4 kWh/t. With an installed power of 600 kW,
should be used in concert to achieve operational improvement. operating at 250 kW clearly leaves considerable spare power
to increase the degree of reduction around this circuit. This
DETAILED MEASUREMENT provides a clear path toward increasing equipment utilisation
The next stage was a more thorough performance assessment through changing the operating regime of the crusher.
based on detailed surveys of each piece of equipment and
section of the dry processing plant. The survey data, taken Screen response
over a range of operating conditions, supplies the data The screens were sampled over a range of operating
required to model the response of the equipment to changes in conditions as the circuit operation was varied over a few

M S Powell et al
FIG 8 – Power draw, average hydroset pressure and specific power input for CR001.
days. The Crusher screens were found to adequately cope of circuit feed rate (+25 per cent), power draw (+50 per cent)
with the full range of feed rates (The feed system had by then and specific energy (+30 per cent) applied to the feed.
been modified to improve distribution between the screens). The sizing data is presented in Figure 11 or the A1 and C
However, the HPGR screens exhibited severe segregation set of surveys. The C trial feed was purposely made finer,
and overloading. Comparative data from the parallel HPGR through recycle screen overloading and bypass, around
screens are presented in Figure 10, in the form of the efficiency 20 per cent -1 mm. It is clear that the C trials had a markedly
curves of the fraction of feed reporting to oversize as a finer HPGR product of around 35 per cent -1 mm compared to
function of particle size. Screen 2 shows a steep classification 20 per cent for the A trial.
at the 10 mm aperture size. Screen 1 has a substantial bypass
of fines, with 25 per cent to 40 per cent of <1 mm material
reporting to the oversize; a clear symptom of overloading and MODELLING
consequent loss of efficiency. The next stage was to model each unit in the circuit and then
The difference in screen operation is driven by the combine them into a single circuit simulation. This was not
segregation in the feed bin. This arises from the feed belt straightforward, as the project team had a range of models
discharging perpendicular to the screen feed direction. The that were not on the same simulation platform, including
screen response curve presents a clear quantification of the some still under development.
screen inefficiency. Initially four independent simulation platforms were used,
with data copied over to the master simulator – JKSimMet.
High pressure grinding roll Parallel to this, the researcher’s simulator, under development
The HPGR was surveyed on three occasions. The feed and in the AMIRA P9 project, was utilised to add new modelling
product data for the first and last set of tests, each conducted capability. This simulator is known as the model developers
at a range of three pressures, are detailed in Table 1. In the kit, Andrusiewicz et al (2011). It has a flexible interface and
first set, the zero gap was about 48 mm, the pressures were structure, allowing new models to be added as they are
quite low and the operating gap was over 60 mm. For the developed, without too high an overhead. During the project,
second set, conducted six months later, the plant had reset three new models were added to the simulator so that the
the HPGR based on recommendations to reduce the zero gap entire circuit can be simulated on a single platform, an essential
and increase the operating pressure. The zero gap was around element for usable outcomes. Within the AMIRA P9 project,
38 mm, the pressure 130 bar and operating gap around 50 mm. these have since been transferred to the integrated extraction
The C series of surveys show a marked improvement in terms simulator so as to move onto a commercial platform.

FIG 9 – Capacity, power draw, average hydroset pressure and specific power for CR003.
overloads, the fines begin to carry over with the oversize and
the undersize flow rate drops off with increasing feed rate
beyond this peak. The crusher screen capacity exceeds the
required duty range. For the HPGR screens, Figure 13 shows
that the capacity in the as-surveyed condition (given by the
unbalanced screens response) is inadequate for the required
duty. With both screens being fed evenly, the peak capacity
increases by 400 t/h to around 2250 t/h. This is adequate
for current needs, but may be insufficient for higher future
demands.
Crushers
Although Chalmers has a mechanistic crusher model
published by Evertsson (1999, 2000), it was not applied at
this stage, as changes to chamber geometry could be assessed
external to the full simulations. Instead, the standard JKMRC
power-based crusher model (Napier Munn et al, 1996) was
used, as this proved satisfactory for product scaling once
fitted to the range of CSS test data. However, the crusher
capacity limits had to be read off the survey data, as the
JKMRC models have no capacity relationships. Capacity
of the crushers were modelled in the MDK by applying the
measured CSS-throughput relationships to the feeders of each
crusher by utilising the new bin model.
FIG 10 – Efficiency curves for the two high pressure grinding rolls screens. The
data represent three separate tests with associated fitted efficiency curves. Screens
The Karra screen model (Karra, 1979) was coded up and It was critical in this work to have a screen model that
responded to overloading, as screen capacity is a critical
included in the MDK. This model proved particularly suitable
bottleneck in any circuit of this nature.
(given the wide range of survey data including overload
conditions) for the troublesome HPGR screens. High pressure grinding roll
The fitted curve for the crusher screens is presented in The newly implemented Hacettepe HPGR model, originally
Figure 12. When the screen has sufficient capacity, the split to developed at Hacettepe University, Aydogan, Ergun and
undersize is linear with an increasing feed rate. Once the screen Benzer (2006) and Dundar et al (2011a), then further developed

M S Powell et al
Table 1
High pressure grinding rolls third survey operating conditions.
Parameter A1 A2 A3 C1 C2 C3
Drive side pressure (bar) 116 97 134 126 152 164
Total power draw (kW) 2736 2377 3056 3180 3650 4500
Gap (mm) 60 65 63 49 47 53
Circuit product (t/h) 1184 1182 1290 1126 1250 1556
Feedrate to HPGR (t/h) 1953 2187 2198 2357 2307 2550
Screen O/S – recycle (t/h) 995 1175 986 1303 1114 1057
Specific energy, kWh/t HPGR feed 1.4 1.1 1.4 1.3 1.6 1.8
Specific energy, kWh/t circuit product 2.3 2.0 2.4 2.8 2.9 2.9
in the AMIRA P9 project by Dundar et al (2011b), was

utilised. This model was specifically chosen over the existing
JKMRC model, as the old model is weaker at dealing with
high circulating loads and fine feeds and does not provide a
pressure throughput response.
Bins
One issue with utilising steady state simulation is that the
crushing circuit does not operate in steady state. Instead,
there is a permanent ongoing build-up and depletion of
bins around the circuit. It is incorrect to try and simulate
this in an ‘average’ steady state mode, as all-important flow
fluctuations and associated peak loadings are then missed. A
simple bin model has been implemented that can limit feed
rate out of the bin while providing an output of time to fill or
empty. This non-steady state model has proven essential to
adequately simulating crushing circuits.
Dynamic simulations
As the MDK does not in any way deal with dynamics, a
separate dynamic simulation platform was used to test the
instantaneous fluctuations about the circuit and the influence
of control on these. Simplified response models and standard
equipment models were utilised with selected parts of the
control system implemented as well for the evaluation
of the suggested process alteration. An actual control
algorithm was later implemented in the process model for
the current configuration. Chalmers has been developing a
dynamic simulation capability based on Matlab/Simulink,
FIG 11 – High pressure grinding rolls feed and product size distributions. Asbjörnsson and Evertsson (2012), which was used for this
purpose.
MODELLING AND SIMULATION OUTPUTS

Once all the process units were modelled, the entire circuit was
simulated in the MDK. A summary output table is presented
in Table 2. This gives stream flow rates in terms of the solids
capacity, and size distribution as P80 and as per cent passing
1 mm. The belt data shows the percentage of belt capacity
utilised. The bin capacity data is the time, in minutes, the bin
takes to fill up under this operating scenario. This scenario
cannot be maintained as the crusher bins are steadily filling up.
When emptying the crusher bins by switching on crusher 2,
the screen bin fills and the added recycle to crusher 3 fills that
bin in 20 minutes – more rapidly than the 40 minutes it takes
to empty the bins. Thus the circuit shifts into an unsustainable
surging mode and the run-of-mine feed has to be switched off
FIG 12 – Crusher screen performance as function of total feed rate. The for at least 20 minutes for the circuit to recover (not shown in
green band represents the typical operating range of the circuit. the table).

FIG 13 – High pressure grinding rolls screen performance as a function of total feed rate.
Table 2
Simulation of standard operating conditions.
Condition: 1400 t/h fresh feed

Comments: bins filling up, feedrate unsustainable
Equipment t/h P80 mm % -1 mm Belt or bin Surge time, min % belt capacity
ROM Feed 1400 182 7.3
Grizzly O/S 674 218 0.1 405-CV-002 63
Grizzly U/S 726 68 14.0
Crush 1 bin 236 218 0.1 crush 1 bin 37
Crusher 1 prod 438 60 4.1
Crusher 2 prod 0
Crusher 3 prod 471 32 5.4
Main feed belt 1635 55 8.9 405-CV-001 48
Crush screen feed 1635 55 8.9 crusher scrn 55
Crush screen U/S 1081 28 13.4 405-CV-006 58
Crush screen O/S 554 78 0.0 405-CV-003 21
Crush 3 bin 83 78 0.0 crush 3 bin 91
High pressure grinding rolls storage silo 1081 28 13.4 HPGR Silo 4854
HPGR recycle feed 1517 23 8.5 406-CV-004 97
HPGR feed 2469 25 10.4 406-CV-002 88
HPGR prod 2469 19 20.3 406-CV-003 95
HPGR screen feed 815 19 20.3
HPGR SC-001 O/S 1142 22 11.3 821
HPGR SC-002 O/S 375 25 0.0
HPGR SC U/S - mill feed 951 5 39.1 406-CV-007 63
Scenarios to increase throughput were sought that didn’t require the use of bin capacity in the
long-term average.
Based on the limitations noted about the circuit, and the surging
caused by the standard feed rate, a number of alternative The crusher screen aperture was varied to reduce the top-
operating scenarios were trialled with the simulator. These size feed to the HPGR. This allows the zero gap to be reduced,
assumed that critical physical design constraints had been allowing a higher operating pressure and a safer damper
addressed. These are all fairly minor engineering changes, stiffness in the hydraulic system.
most likely totaling less than $2 M. A bypass was trialled to recycle a portion of HPGR product
By closing the CSS of CR001, this would relieve the load on back to the HPGR in order to supply a finer feed. This increases
the tertiary crusher (CR003), plus reduce the circulating load bed density and allows the HPGR to have a higher power draw,
about the screen bins. thus producing a finer product. This is attractive as the HPGR
has been operating at around 50 per cent of installed power.
The simulator was used to explore the maximum feed
rate that could be maintained without needing to utilise bin The HPGR was simulated at higher pressures, increasing from
capacity. If the bins begin to fill in any steady state simulation 130 to 170 bar – about ten bar below the design recommendation.
scenario, then this is likely to be unsustainable in a real The feed to the HPGR screens was evened out in both tonnage
operating regime that suffers from short-term fluctuations in and size distribution. This simulates the feeding arrangement
feed size and flow splits throughout the circuit. Thus scenarios to the HPGR feed bin being redesigned to avoid segregation.

M S Powell et al
Potential capacity with current equipment improving the control algorithms used on the plant. Scenarios
Some of the simulated outputs are summarised in Table 3. that could not be tested on the operating plant, such as varying
The simulations show that even at a feed rate of 1100 t/h, the speed of the crushers and increasing the capacity of the
the circuit would not stabilise under the standard operating final feed belt, were assessed to identify critical bottlenecks
conditions. The stable operating feed rate is nearer 1000 t/h. and prioritise expenditure on releasing full circuit capacity.
The crushers were then closed down to increase their
utilisation. This brings in crusher 2 for 40 per cent of the time NEW DESIGNS
in the secondary crushing mode. It was confirmed that the Having established the optimal operating conditions, Anglo
circuit did not overload when crusher 2 was switched on, as American Platinum chose to assess expansion options, while
the crusher bin emptied out faster than the others filled. Thus actively pursuing implementation of the circuit optimisation
the circuit could handle a steady feed rate of 1100 t/h. over the period of a few years. This design study would
In the third simulation the plant is pushed to its throughput address the engineering constraints, while adding capacity to
limit with the current equipment. It is demonstrated that the maximise the total circuit utilisation, provide robust standby
reduction and screening equipment and the bins could cope, capacity and increase overall circuit capacity.
but that some upgraded conveyors would be required. In
reality this feed rate is unlikely to be sustainable, but 1400 t/h Examples of potential plant layouts were simulated. These
may be feasible. are not necessarily related to a design that may in future be
selected by Anglo American Platinum, but serve to illustrate
Based on sound simulations, utilising the equipment
the capability of simulation in assisting in circuit selection.
within the limits to which it was tested, it can be concluded
that the crushing and HPGR plant can be expanded by up to A possible circuit option consists of a stage-crushing
30 per cent with only minor engineering upgrades, with the configuration, as illustrated in Figure 14:
addition of some upgraded conveyors. •• The grizzly is converted to a scalping screen with
The work of Brown, Steyn and Fouchee (2016) reports on around 80 mm cut size – this provides positive top-size
application of the dynamic control simulator to testing and protection to the screens and tertiary crushers.
Table 3
Simulated circuit capacity with a range of operating and engineering improvements.
Condition 1100 t/h fresh feed 1100 t/h, reduced crushers 1500 t/h, coarse feed, crushers at
CSS: CR001 to 35 mm; CR003 to best settings: 1&2 at 40 mm, 3 at
22 mm; use CR002 40% of time 22 mm; CSS: all fully utilised
Comments Crusher bins slowly filling Circuit stable Belts overloaded, crusher and
in about 1.5 h screens at maximum capacity
Equipment t/h P80 % Surge % belt t/h P80 % Surge % belt t/h P80 % Surge % belt
mm -1 mm time cap mm -1 mm time cap mm -1 mm time cap
(min) (min) (min)
ROM feed 1100 182 7.3 1100 182 7.3 1500 182 7.3
Grizzly O/S 530 218 0.1 50 530 218 0.1 50 723 218 0.1 67
Grizzly U/S 570 68 14.0 570 68 14.0 778 68 14.0
Crush 1 bin 79 218 0.1 109 0 0 100.0 0 0 0 100.0 0
Crusher 1 prod 450 60 4.1 371 34 5.8 361 39 5.5
Crusher 2 prod 0 159 45 4.9 361 51 4.4
Crusher 3 prod 482 34 4.9 282 26 6.4 483 26 6.4
Main feed belt 1502 53 8.1 44 1381 44 9.2 41 1983 48 8.8 58
Crush screen feed 1502 53 8.1 50 1381 44 9.2 46 1983 48 8.8 0 66
Crush screen U/S 936 27 13.1 50 1100 30 11.6 59 1500 29 11.7 80
Crush screen O/S 567 74 0.0 22 282 82 0.0 11 483 79 0.0 19
Crush 3 bin 85 74 0.0 89 0 0 0 0 0 0
HPGR storage silo 936 27 13.1 5610 1100 30 11.6 4773 1500 29 11.7 3500
High pressure grinding roll 1305 22 8.6 84 1676 23 7.6 107 2060 26 0.0 132
recycle feed
HPGR feed 2129 24 10.3 76 2644 26 9.1 94 3380 27 4.6 120
HPGR prod 2129 18 20.4 82 2644 19 18.8 102 3380 22 13.9 130
HPGR screen feed 703 18 20.4 873 19 18.8 1115 22 13.9
HPGR SC-001 O/S 983 21 11.4 1251 23 10.2 1380 26 0.0 900
HPGR SC-002 O/S 322 24 0.0 425 25 0.0 680 26 0.0
HPGR scrn U/S – mill 823 5 39.2 -3563 55 968 5 38.0 -20 000 65 1320 6 35.5 2000 88

FIG 14 – Expansion option flow sheet with new equipment: highlighted dashed arrows indicate optional flows; blue indicates new equipment.
•• The three existing crushers are converted to a common Potential capacity with a new circuit configuration
secondary crushing duty, in open circuit with crushers The circuit was set up in the new MDK software and a
operated with a smaller gap to increase size reduction, wide range of scenarios were simulated. The outcomes are
and an increased throw to increase capacity and allow tabulated in Table 4. This summarises the flow rates and
them to draw more of their installed power. A feed- size of the key streams about the circuit; the time to fill or
tripper is installed to feed all three crushers evenly. empty storage bins; the percentage of conveyor capacity in
•• As the secondary crushers are in open circuit, this use; and the key equipment settings. When bins are shown to
removes the overloading of the main common feed belt be empty for a given scenario, this indicates that the second
and screen bin. crusher would be switched on and off to hold the required
•• The existing recycle belt from the screens is left in place average feed rate. This results in a lower product rate feeding
to allow the crushers to be closed-circuited if required. the downstream units, which then have a chance to catch up
The product belt from the screens is made bi-directional, if they have experienced a gradual filling up. These scenarios
to enable product to be fed to the new tertiary crushing can be studied in detail with the simulator and can be used to
plant or to the existing recycle belt. select the likely bin size requirements.
Three scenarios are presented. These are described below
•• Secondary crusher screen oversize is crushed in a new
and detailed in Table 4.
set of tertiary crushers. For commonality of maintenance
spares, these could be identical to the secondary crushers The new configuration and additional crushing capacity
but with finer chambers. ensures that the capacity constraints are determined by the
HPGR circuit, in particular the HPGR screens. Crushers
•• Tertiary crusher product is screened on a new double-
operating in on-off mode would cause the HPGR silos to fill
deck screen, with oversize returning to the crushers and
and empty. If operated with two duty secondary and two duty
fines bypassed to the milling circuit. The double-deck
tertiary crushers, allowing one of each in a standby mode, the
configuration removes load from the HPGR circuit as
well as several conveyors which would otherwise need ‘Base Case’ circuit capacity will increase to around 1540 t/h.
to be upgraded. An increase in throughput requires a comparable increase
in comminution energy to prevent coarsening of the feed to
•• The key bottleneck in the HPGR circuit is the screens, so the milling circuit. This option requires the specific energy
removing feed to the circuit reduces the feed rate to the in the crushing section to be increased to ensure a finer
HPGR screens; thereby opening up circuit capacity. The feed to the HPGR. To achieve this and to allow the crusher
option of using the secondary screens in full double-deck to run in an improved and safer compressed bed mode, the
mode is allowed for, to further reduce feed to the HPGR. crushing screens are reduced from an aperture of over 45 mm
Currently the top deck is just for protection, but with a to 35 mm. If all of the crushing area screens are converted to
smaller top-size reporting here, the deck can be reduced a double-deck configuration, around 30 per cent of the plant
and the bottom deck converted to the same aperture as feed can be bypassed around the HPGR, thereby avoiding the
the HPGR circuit screens. The screen discharge would need to upgrade the HPGR screen capacity (which would be
need to be modified to allow the three product streams difficult to achieve due to layout constraints). The capacity of
to be removed independently. the plant increases to around 1760 t/h with the addition of the
•• An HPGR screen bypass is added to recirculate the fine screen decks.
HPGR product to the HPGR. This gives the HPGR a Changes in feed size distribution shift the load from one set
finer feed; thereby, improving the size reduction without of crushers to the other; a coarser feed increases the demand
overloading the screens. on the secondary crushers, while a finer feed puts a greater

M S Powell et al
Table 4
Simulation results for the new circuit option.
Base case Double deck screening Base case with finer feed
Stream t/h P80 % -4 mm t/h P80 % -4 mm t/h P80 % -4 mm
Main feed 1540 181 9.7 1760 181 9.7 1570 151 10.4
Scalper O/S 882 209 0.0 1024 209 0.0 840 182 0.0
Scalper U/S 658 50 22.6 736 48 23.1 730 49 22.3
Secondary crusher feed 1237 209 0.0 1283 209 0.0 1283 182 0.0
Secondary crusher prod 1237 46 8.1 1283 51 6.9 1283 51 6.7
Tertiary crusher feed 1311 53 0.0 1311 56 0.0 1311 56 0.0
Tertiary crush prod 1311 33 11.1 1311 33 11.4 1311 33 11.4
High pressure grinding roll feed 2398 21 13.4 2373 22 4.7 2349 22 11.8
HPGR prod 2398 14 45.1 2373 15 40.2 2349 14 45.2
HPGR bypass 250 14 45.1 250 15 40.2 150 14 45.2
HPRG screen feed 2148 14 45.1 2123 15 40.2 2199 14 45.2
HPGR screen O/S 801 21 0.4 890 21 0.3 825 21 0.4
Mill feed 1540 5.2 71.2 1760 5.2 70.3 1570 5.2 71.7
Bin fill times min min min
Secondary crusher -25 -34 -20
Secondary crusher screens CONST CONST CONST
Tertiary crusher 1014 278 285
HPGR silo 333 756 327
HPGR screens CONST CONST CONST
Conveyor capacity % % %
Secondary crush screen feed 56 59 59
Secondary screen O/S 35 39 39
HPGR silo feed 72 74 74
HPGR silo prod 96 87 97
HPGR product 92 91 90
HPGR screen O/S 51 57 53
Mill silo feed 103 117 105
Equipment settings
# CSS # CSS # CSS
Secondary crushers 2 45 2 50 2 50
Tertiary crushers 2 21 2 21 2 21
# top bottom # top bottom # top bottom
Scalper 1 80 1 80 1 80
Secondary crusher screens 2 35 2 35 7 2 35
Tertiary crusher screens 1 35 7 1 35 7 1 35 7
HPGR screens 2 11 2 11 2 11
load on the tertiary crushers. This imbalance can be countered for extended periods of time, either due to maintenance
by incorporating sufficient surge capacity ahead of the new requirements or during building of the expansion. These
tertiary crushers, or by varying the crusher settings. The effect possibilities are incorporated into the simulations by inserting
of a 20 per cent change in F80 to the circuit has only a small splitters to divert streams to a different part of the circuit.
change in overall plant throughput in the HPGR circuit, if the The following scenarios, presented in Table 5, have been
crushing circuit is upgraded to have excess capacity. considered.
Operating flexibility Secondary crusher bypass

The circuit can be reconfigured to allow the plant to be Bypassing of the secondary crusher circuit can be achieved
operated at a reduced rate if parts of the circuit are unavailable by coarsening the scalper screen cut size and removing

Table 5
Simulation results for flexible operating scenarios.
Secondary crusher bypass Tertiary crusher bypass HPGR bypass

Stream t/h P80 % -4 mm t/h P80 % -4 mm t/h P80 % -4 mm
Main feed 1850 181 9.7 1100 181 9.7 550 181 9.7
Scalper O/S 748 225 0.0 663 207 0.0 288 214 0.0
Scalper U/S 1102 83 16.2 437 45 24.3 262 60 20.3
Secondary crusher feed 0 0 100.0 589 183 0.0 553 197 0.0
Secondary crusher prod 0 0 100.0 589 44 6.8 553 38 8.4
Tertiary crusher feed 1311 86 0.0 0 0 100.0 1120 34 1.2
Tertiary crush prod 1311 32 12.3 0 0 100.0 1120 22 9.3
High pressure grinding roll feed 2179 18 33.4 2419 25 28.0 0 0 100.0
HPGR prod 2179 11 58.1 2419 15 52.1 0 0 100.0
HPGR bypass 1000 11 58.1 800 15 52.1 0 0 100.0
HPRG screen feed 1179 11 58.1 1619 15 52.1 0 0 100.0
HPGR screen O/S 283 21 0.5 519 26 0.4 0
Mill feed 1102 4.6 75.6 1100 4.7 75.8 554 12.2 33.8
Bin fill times min min min
Secondary crusher -20 19 -57
Secondary crusher screens CONST CONST CONST
Tertiary crusher -112 -112 215
HPGR silo 374 -230 CONST
HPGR screens CONST CONST CONST
Conveyor capacity % % %
Secondary crush screen feed 32 30 24
Secondary screen O/S 24 15 21
HPGR silo feed 52 24 -
HPGR silo prod 64 78 -
HPGR product 84 93 -
HPGR screen O/S 18 33 -
Mill silo feed 73 73 37
Equipment settings
# CSS # CSS # CSS
Secondary crushers 0 50 1 40 1 35
Tertiary crushers 2 21 0 21 2 15
# top bottom # top bottom # top bottom
Scalper 1 130 1 70 1 80
Secondary crusher screens 2 35 2 45 2 18
Tertiary crusher screens 1 35 7 1 7 1 18
HPGR screens 2 11 2 11 2 15
the oversize to a temporary stockpile (or possibly even a Tertiary crusher bypass
temporary contractor crushing station) and crushing all of the
These are bypassed by reversing the direction of the secondary
undersize in the tertiary crushers.
screen discharge conveyor, set-up to feed the new circuit,
Capacity is constrained by the capacity of the tertiary crushers or by recycling to the secondary crushers via the existing
and screens, as well as by their ability to handle the coarse
conveyor system.
oversize. In theory, a feed rate of 1850 t/h may be possible (of
which 750 t/h is coarse reject material), allowing for a wider A smaller crusher gap and a larger screen aperture is selected
opening on the scalping screen. The capacity is limited by to shift more of the size reduction work to the HPGR. A plant
the ability of the HPGR circuit to produce a sufficiently fine feed rate of around 1100 t/h is achievable in this mode. A
product: a large recirculating load of HPGR product back to the large HPGR bypass recirculation (similar to scenario 1) is
HPGR is required to achieve the required size reduction. required to maintain a fine grind.

M S Powell et al
High pressure grinding roll bypass Due to the extensive internal and external disturbances in
Bypassing the entire HPGR can be achieved by tightening the circuit the improvements are difficult to obtain by control
down the crusher and screen gaps and sending the crushed alterations alone. Changes in operational variables were also
product directly to the mill. investigated in a simulate environment. An increased stability
by altering the conveyor capacity allowed the APC more
All of the size reduction needs to be performed by two
room to stabilise and optimise the circuit closer to its dynamic
stages of crushing, with the crushed product fed directly to
and physical constraints. Indicating that more stability of the
the milling circuit. This circuit requires the lower screen decks
circuit implies increase throughput.
to be changed to around 15 mm and the crushers to be set as
fine as possible. A feed rate of 550 t/h is predicted with a mill
F80 of around 12 mm. CONCLUSIONS
The collaborative GCC research team joined Anglo Platinum
DYNAMIC SIMULATION in debottlenecking the MNC crushing and HPGR circuit, then
The purpose of the dynamic simulations was to find the measuring circuit performance to quantify the influence of
plant performance for a given condition for the suggested proposed changes to upgrade capacity and finally applying
expansion. The circuit represented by the process units and this to potential future expansion designs.
control algorithms is presented in Figure 15. Two scenarios A staged approach was taken to progressively build
are illustrated here – 1800 and 2000 t/h. The plant simulation knowledge and address the increased throughput objectives:
represents 35 hours of operation. •• Scoping review – a two-day site visit with a detailed
In the first scenario (1800 t/h; Figure 16) the feed is operated review of equipment performance, operating data and
with a 100 t/h disturbance fluctuation. The operation is still direct observation. This led to an initial list of equipment,
stable but disturbances can be observed on the conveyors and operational and control recommendations.
levels in bins, specifically related to the on-off operation of •• Detailed measurements – three sets of surveys were
crushers. In the second scenario of 2000 t/h, the process starts conducted over a range of operating conditions for the
to fluctuate because secondary crushers are unable to handle key equipment. Extensive operating data was collected
the circulating load, that is, the oversize from the screens. The and time was spent in the control room. The survey
feed is forced into an on-off switching mode as bins overfill
data was used to develop process models. The control
and a crusher feed bin becomes overloaded, so the circuit will
data provided performance assessment and dynamic
not be able to maintain this feed rate. From this it is evident
response data.
that the plant reaches saturation somewhere between 1800–
2000 t/h, which is above the set target value of 1600 t/h. •• Simulations – the models were used to develop a full
circuit simulation in the MDK simulator, then used to
An implementation of the process model for the current
quantify how different changes would influence circuit
process configuration was performed at the Anglo Platinum’s
performance, and to advise on operational settings.
control and instrumentation department in Johannesburg.
The implementation was a part of performance testing Dynamic simulations were used to test control algorithms.
of the simulation platform to verify system performance •• Expansion design – the simulators were used to assess
characteristics. The objective was to verify the start-up expansion design options, including best use of extra
sequences of the plant and simulate different operating equipment, relieving bottlenecks, storage capacity,
conditions with the same advanced process control (APC) conveying requirements, response to changes in feed
as the actual process to investigate additional development size and the ability of the circuit design to cope with
opportunities for the process units and process controls. equipment being offline.
FIG 15 – The expansion layout in the simulation software.

FIG 16 – The simulated results from 1800 and 2000 t/h scenarios.
The site progressively implemented these recommendations The HPGR-crushing upgrade project demonstrates the
and have addressed circuit design issues over the years to potential benefits of nurturing research groups to build a
improve plant reliability and dramatically increase production strong inter-disciplinary technical team that can support
from 20 per cent below design to well above without any new operational needs while developing future technology.
processing equipment. The control system has been a key part
of this improvement, supported and informed by the dynamic
process simulator that can act as a test bed for control logic
ACKNOWLEDGEMENTS
and algorithms. This project is carried out with substantial on-site support
from the MNC staff and funding from Anglo American
The outcomes of this process have been substantial and
Platinum. We acknowledge their permission in publishing
rewarding to the operation.
this work. A number of staff and students of the University
Based on sound simulation that utilises equipment within research groups have actively contributed to this work. We
the limits to which is was tested, it was concluded that are particularly grateful for their enthusiasm, dedication
the crushing and HPGR plant can be expanded by over and high-quality of work. These include Hakan Dundar,
30 per cent with only minor engineering upgrades, including Johannes Quist, Erik Hulthén, Angus Robertson, André van
the addition of some upgraded conveyors. der Westhuizen and Rodrigo de Carvalho.
A high-quality simulation of all the relevant equipment
provides a route to confidently size the equipment for an
expansion, and to trial a wide range of potential design and
REFERENCES
circuit layouts. Enhanced steady state simulations using the Andrusiewicz, M, Brennan, M, Evans, C, Morrison, R, Perkins, T and
Wightman, E, 2011. Multicomponent modelling and simulation.
MDK allow crusher capacity and some aspects of the circuit
AMIRA Project P9O Final Technical Report, pp 119–203.
dynamics to be modelled, but not as completely as true
dynamic simulations would allow. The simulations show that Asbjörnsson, G and Evertsson, C M, 2012. Modelling and simulation
the addition of two tertiary crushers and a double-deck screen of dynamic crushing plant behaviour with MATLAB/Simulink,
in Proceedings Comminution 12 (Minerals Eng Int: Cape Town).
(along with the required bin and conveyors) would allow the
circuit to expand by 70 per cent from the original throughput. Aydogan, N A, Ergun, L and Benzer, A H, 2006. High pressure
grinding rolls (HPGR) applications in the cement industry, Miner
Dynamic simulation adds the capability to specify storage Eng, 19(2):130–139.
capacity, to trial circuit operation in terms of operating
strategy and fine-tune control algorithms in order to maximise Banini, G, Villanueva, A, Hollow, J and Mosher, J, 2011. Evaluation
of scale up effect on high pressure grinding roll (HPGR)
the circuit performance. This was achieved by overlaying
implementation at PT Freeport Indonesia, in Proceedings
the control system on the dynamic simulator. The dynamic International Autogenous and Semiautogenous Grinding Technology
simulation together with the actual APC indicate that higher 2011 (eds: K Major, B C Flintoff, B Klein and K McLeod), 16 p
stability will increase process throughput, which is possible (Canadian Institute of Mining, Metallurgy and Petroleum:
with smaller changes without redesigning the process. Westmount).

M S Powell et al
Brown, R P, Steyn, C W and Fouchee, R J, 2016. Improving crusher Karra, V, 1979. Development of a model for predicting the screening
performance by comparing various control strategies using performance of a vibrating screen, CIM Bulletin, pp 168–171.
a validated simulation, in Proceedings Comminution ‘16, 9 p,
Koski, S, Vanderbeek, J and Enriquez, J, 2011. Cerro Verde
(Minerals Engineering International).
concentrator – four years operating HPGRs, in Proceedings
Dundar, H, Benzer, A H and Aydogan, N, 2011a. Development of International Autogenous and Semiautogenous Grinding Technology
the HPGR model, AMIRA P9O final report, pp 45–72 (AMIRA 2011 (eds: K Major, B C Flintoff, B Klein and K McLeod) 19 p
International). (Canadian Institute of Mining, Metallurgy and Petroleum:
Westmount).
Dundar, H, Benzer, H, Aydogan, N A, Powell, M S and Mainza, A
N, 2011b. Importance of the feed size distribution and recycle on Napier-Munn, T J, Morrell, S, Morrison, R D and Kojovic, T, 1996.
the HPGR performance, in Proceedings International Autogenous Mineral comminution circuits: their operation and optimisation,
and Semiautogenous Grinding Technology 2011 (eds: K Major, Julius Kruttschnitt Mineral Research Centre, Indooroopilly, Qld.
B C Flintoff, B Klein and K McLeod) (Canadian Institute of
Powell, M S, Benzer, H, Mainza, A N, Evertsson, M, Tavares, L M,
Mining, Metallurgy and Petroleum: Westmount).
Potgieter, M, Davis, B, Plint, N and Rule, C, 2011. Transforming the
Evertsson, C M, 1999. Modelling of flow in cone crushers, Minerals effectiveness of a HPGR circuit at Anglo Platinum Mogalakwena,
Engineering, 12(12):1479–1499. in Proceedings International Autogenous and Semiautogenous
Grinding Technology 2011 (eds: K Major, B C Flintoff, B Klein
Evertsson, C M, 2000. Cone crusher performance, PhD thesis,
and K McLeod) (Canadian Institute of Mining, Metallurgy and
Göteborg: Chalmers University of Technology.
Petroleum: Westmount).
Hart, S, Parker, B, Rees, T, Manesh, A and Mcgaffin, I, 2011a.
Villanueva, A, Banini, G, Hollow, J, Butar-Butar, R and Mosher, J,
Commissioning and ramp up of the HPGR circuit at Newmont
2011. Effects of HPGR introduction on grinding performance at
Boddington Gold, in Proceedings International Autogenous
PT Freeport Indonesia’s concentrator, in Proceedings International
and Semiautogenous Grinding Technology 2011 (eds: K Major,
Autogenous and Semiautogenous Grinding Technology 2011 (eds:
B C Flintoff, B Klein and K McLeod) (Canadian Institute of
K Major, B C Flintoff, B Klein and K McLeod) 16 p (Canadian
Mining, Metallurgy and Petroleum: Westmount).
Institute of Mining, Metallurgy and Petroleum: Westmount).
Hart, S, Rees, T, Tavani, S, Valery, W and Jankovic, A, 2011b. Process
integration and optimisation of the Boddington HPGR circuit,
in Proceedings International Autogenous and Semiautogenous
Grinding Technology 2011, (eds: K Major, B C Flintoff, B Klein and
K McLeod) 13 p (Canadian Institute of Mining, Metallurgy and
Petroleum: Westmount).

Contents
An update on applications of high frequency

screens in closed grinding circuits
S B Valine1, J E Wheeler2, B Packer3 and A N Cavendor4
ABSTRACT
The objective of this paper is to examine the operation of high frequency screens in
closed circuit grinding, where the outcomes achieved indicate favourable economic
gains at the comminution stage and in downstream processes, including flotation,
dewatering and filtration operations. It has long been known that closed circuit grinding
with a classifier improves grinding efficiency, reduces unit specific power consumption
and reduces over-grinding. The classification device removes material from the circuit
that has reached the desired size and returns coarse material back to the mill for further
grinding. The efficiency of the classifier is an important factor in the overall performance
of a grinding circuit. The classifier affects the capacity of the grinding circuit and the
metallurgical performance of downstream operations. A review of historical research
and modern case studies is used to demonstrate the benefits of high frequency screens
for classification in grinding circuits (Valine et al, 2013).
INTRODUCTION
While screens have long been commonplace in coarser size reduction circuits, the
use of screens for finer grinding circuits was often considered impractical. Decades
of research have supported the concept of closing circuit grinding with screens. As
far back as 1925, E W Davis at the University of Minnesota Mines Experiment Station
conducted closed circuit grinding tests with a rake classifier, a screw classifier and a
screen. He concluded that the higher separation efficiencies associated with screen
classification would result in higher mill capacities with less over-grinding (Davis,
1925). However, due to limitations in screening technology at that time, screw and
rake classifiers were more commonly applied in finer grinding circuits.
By the 1950s, with the advent of high capacity slurry pumps, hydrocyclones became
the standard classification device for high throughput fine-grinding circuits. The
transition to hydrocyclones did not offer significant improvements in classification
efficiencies but did lower capital costs and floor space requirements. Hydrocyclones
make separations based on particle size as well as specific gravity and therefore do
not provide a true size separation. This often results in liberated dense minerals at the
target grind size returned to the mill resulting in over-grinding and the generation of
slimes. However, by the 1960s, researchers forecasted a need for improved grinding
efficiencies based on increasing energy costs and lowering ore grades. Two studies
by Hukki and Eland (1965) and Hukki (1967) concluded that ‘the master key for great
improvements in capacity and in energy consumption in closed grinding circuits is
improved sharpness of separation’. These studies also concluded that to reduce over-
grinding and increase separating efficiency an accurate sizing device was required.
While the concept of classification by physical size was clearly understood, at that
time there were no cost-effective technologies available to meet these objectives. As
demonstrated in the following seven case studies, improved high frequency screening
1. President, Separation Technology Group, technologies including high capacity multifeed point machines and durable blinding
Derrick Corporation, Buffalo NY 14225 USA.
resistant fine mesh urethane screen surfaces, closed circuit grinding with fine screens
Email: svaline@derrickcorp.com
is now a practical reality with substantial benefits.
2. Senior Product Application Manager,
Separation Technology Group, Derrick
Corporation, Buffalo NY 14225 USA. DEFINITIONS
Email: jewheeler@derrickcorp.com As provided by (Dündar et al, 2014), a classifier is described by its partition curve
3. Director, WPE Process Equipment, which shows the per cent of each size fraction in the feed which enters the coarse
Welshpool WA 6106. Email: product. Three variables define the curve as shown in Figure 1: the slope or sharpness
brian.p@wpeprocessequipment.com.au
of the linear section of the curve, the d50 value which defines its location on the graph,
4. Product Application Manager, Separation and the bypass which is the fraction of fine misplaced particles in the coarse fraction.
Technology Group, Derrick Corporation,
Buffalo NY 14225 USA. The breakage rate can be used to be able to compare the efficiency of closed grinding
Email: ancavendor@derrickcorp.com circuits using cyclones or high frequency screens. Breakage rate is defined by r/d,
499
S B Valine et al
100 2002). Currently, fine screening (wet or dry) is normally

90
Ideal accomplished with high frequency, low amplitude, vibrating
separation Real screening machines with either elliptical or straight-line
80
separation motion. Various types of wet and dry fine screening machines
70
and the factors affecting their operation have been discussed
partition %
60
previously.
50
In fine particle wet screening, the undersize particles are
40
transported through the screen openings by the fluid and the
30
fraction of fluid in the slurry will therefore affect the efficiency
20
Fish Hook
of the separation. From a practical standpoint, the feed slurry
10 to a fine screen should be around 20 per cent solids by volume
Bypass
0
d50 to achieve reasonable separation efficiency. As most of the
1 10 100 1000 fluid passes through the screen openings rather quickly, the
mean particle size (µm)
fine screening process can be completed in a short screen
FIG 1 – Partition or Tromp curve for a classifier (Dündar et al, 2014). length. Therefore screen width, rather than screen area, is an
important design consideration for fine wet screening.
where r is the breakage rate of particles and d is the discharge Recognition of this concept led to the development of
rate of particles, which relates feed size to product size multiple feed point, fine, wet screening machines. For
(Whiten, 1974). example, the Multifeed screen consists of three screen panels
mounted within a rectangular vibrating frame and is actually
BACKGROUND three short screens operating in parallel. Each screen panel has
Cyclones have been used for decades in both dry and wet its own feed box and the oversize from each panel flows into
grinding circuits to separate fine from coarse particles in a common launder and then to the oversize chute. Similarly,
the grinding mill discharge, with the coarse particles being the undersize from each of the three panels flows into the
recycled to the mill for more breakage. The recirculating undersize hopper. The popular 1.2 m (4 ft) wide by 2.4 m
stream contains the fine bypass as well as the coarse particles (8 ft) long version has a total effective width of 3.0 m (10 ft).
lead to high recirculating stream loads. Dry cyclones In general, multiple feed point machines have been shown to
have been used for more than 100 years and have shown have 1.5 to two times more capacity than a single feed point
steady improvements in classification, reducing bypass by machine of equivalent size and screen area.
minimising the d50 and increasing sharpness of the partition
curve, which improves productivity. ADVANCEMENTS IN SCREENING TECHNOLOGY
In contrast, wet classification using hydrocyclones has
made little progress in improving efficiency and this is Stack Sizer screening machine
mainly due to two reasons. The coarse particle apex stream Expanding further on this concept, the Stack Sizer screening
will also carry fine particles in the water causing bypass, and machine was introduced in 2001. With a capacity considerably
the fine vortex finder stream will carry coarse particles due to greater than any other type of fine wet screening machine
misplacement of low specific gravity particles and turbulence. previously available, the Stack Sizer has up to five vibrating
However, advancements in non-blinding high-throughput screen decks operating in parallel for a total effective width
screening technology have led to drastic improvements in of 5.1 m (17 ft). As shown in Figures 2 and 3, the decks are
classification by screening, improving the efficiency of fine positioned one above the other and each deck has its own
wet grinding circuits. According to Albert (1945) and Hukki feed box. A custom-engineered single or multiple-stage
and Allenius (1968), the most substantial changes that occur flow distribution system is normally included in the scope
when substituting hydrocyclones with screens in closing of supply to representatively split the feed slurry to each
circuit grinding are: Stack Sizer screen and then to the decks on each machine.
•• increased productivity/throughput Ample space is provided between each of the screen decks
•• reduced circulating load for clear observation during operation and easy access for
maintenance and replacement of screen surfaces. Each screen
•• lower power consumption per tonne
•• reduced over-grinding of valuable minerals/generation
of slimes/tailings
•• consistent product size distribution and density
•• increased recovery of valuable minerals
•• reduced reagent consumption in flotation
•• reduction in overall operating costs.
Another major benefit that screen size classification offers is
control over the top size of the particles that leave the grinding
circuit. For example, copper flotation circuits often have poor
recovery of particles coarser than 70 mesh. A screen can be
used to reduce the amount of particles coarser than 70 mesh
that leave the grinding circuit. Plant operators should look at
all unit operations to take full advantage of the improvements
that screens offer in the grinding circuit. Other processes,
such as filtration often benefit from a reduction of fines.
Fine screening typically refers to separations in the range FIG 2 – One of 38 Stack Sizer screening machines
of 10 mm (3/8 in) to 38 µm (400 mesh) (Valine and Wennen, operating at an iron ore mine in Brazil.

An update on applications of high frequency screens in closed grinding circuits
through the use of high frequency screening machines and

fine urethane screen surfaces in place of conventional particle
classification equipment.
CASE STUDIES
As noted in the introduction to these seven case studies that
all describe the use of fine screens in closed circuit grinding,
it is interesting to note that research work over many decades
has supported the concept of closing grinding circuits with
screens. In 1925, E W Davis at the University of Minnesota
Mines Experiment Station conducted closed circuit grinding
tests with a rake classifier, a screw classifier and a screen.
He concluded that the use of screens would result in higher
mill capacities with less over-grinding (Davis, 1925). Other
studies concluded with statements such as ‘lower grinding
costs are possible with screen circuits’ (Albert, 1945) and
‘the master key for great improvements in capacity and in
energy consumption in closed grinding circuits in improved
sharpness of classification’ (Hukke and Eland, 1965). A recent
presentation on flow sheet development for the planned
Essar Steel iron ore concentrator project in Minnesota which
included results from pilot scale tests and simulation work
FIG 3 – Principle of operation. demonstrated a significant coarsening of the grind/grade
relationship through the use of screens in the grinding circuit
deck, consisting of two screen panels in series, is equipped
instead of hydrocyclones (Murr, Wennen and Nordstrom,
with an undersize collection pan which discharges into a
2009). This study predicted that a high-grade concentrate
common launder with a single outlet. Similarly, the oversize could be produced at a reasonable but significantly coarser
from each of the screen decks collects in a single hopper grind than would be required if the mills were closed by
with a common outlet. Two Super G® vibrating motors rated hydrocyclones. In fact, this study concluded that the desired
at 1.9 kW (2.5 HP) each and rotating in opposite directions product grade could not be obtained with hydrocyclones,
produce a uniform high frequency linear motion throughout even with extremely fine grinding. This coarsening of the
the entire length and width of all screen decks for superior grind/grade relationship with improved classification in
oversize conveyance. grinding was also demonstrated on a commercial scale with
As mentioned above, the fluid passing through the openings the addition of screens to all secondary mill lines at the nearby
carries the undersize particles through the screen openings. National Steel Pellet Company, now US Steel Keetac, resulting
The screening process is essentially complete when most of in the production of the same product grade at a significantly
the fluid has passed through the openings. Any remaining coarser grind, specifically, 72 per cent passing 45 µm with
undersize particles adhere to the coarse particles and are screens compared to 79 per cent passing 45 µm in the previous
misdirected to the oversize product. An optional ‘repulping’ circuit (Wennen, Nordstrom and Murr, 1997).
system is available for the Stack Sizer in which spray water is
directed into a rubber-lined trough located between the two PhosAgro – OJSC Apatit
panels on each deck. With this feature, oversize from the first OJSC Apatit (Apatit), a division of PhosAgro, is a phosphate
panel is re-slurried and screened again on the second panel. producer located in the Murmansk Region of the Kola
This repulping action maximises the correct placement of Peninsula in Russia. At Apatit, multiple parallel lines of single
undersize particles and its use will depend upon the particular stage ball mills in closed circuit with either screw classifiers or
objective of the screening machine. single and double-stage hydrocyclones grind crushed apatite-
To date, over 1800 Stack Sizer screening machines are in nepheline ore. The product from multiple grinding lines
operation at mineral processing plants worldwide. Dry mass is combined and pumped to the flotation plant. Over time,
flow capacity typically ranges from 100 to 350 tn/h (US Apatit has experimented with various flow sheet options to
tonne). This is roughly equivalent to three or four of the older improve classification efficiency and increase mill production
style Multifeed screens discussed above. Like all screening rates. In 1996 and 1997, tests were conducted with a full-scale
machines, capacity depends upon many factors such as screen Multifeed screen and various urethane screen panel openings.
panel opening, weight recovery to oversize, the amount of Although the test results were encouraging, screens were
near-size particles, particle shape and slurry viscosity. eliminated as a possible alternative due to the high number
of machines required and the limited space available in the
Polyurethane screen surfaces vicinity of the grinding mills.
The availability of high open area (typically ranging from 30 With the development of the high capacity Stack Sizer,
to 45 per cent) long wearing polyurethane screen surfaces Apatit desired to re-evaluate screens as a possible method to
with openings as fine as 45 µm (320 mesh) has made fine improve grinding efficiency and increase production rates.
wet screening more feasible than previously thought They began in 2005 with continuous pilot scale grinding
possible. In the past, plant designers typically avoided the tests with a ball mill in closed circuit with a hydrocyclone.
use of fine screens due their relatively low capacity, the high This was to establish a correlation between the pilot plant
consumption rate of woven steel wire screen panels and the and commercial plant operation. Next, a single-deck Stack
tendency of these types of panels to blind with near-size Sizer screen was installed in place of the hydrocyclone in the
particles. This view has now changed as demonstrated by pilot plant grinding circuit. The pilot tests demonstrated that
the following case studies describing process improvements the use of screens in place of hydrocyclones would increase

S B Valine et al
production rates and reduce over-grinding and justified the

next phase of evaluation, the modification of an existing
production line to screen classification.
Mill 24, 3.6 × 4.0 m, of the apatite-nepheline ore dressing
factory (ANOF-2) was selected for the modification because
it was a smaller line and this would minimise the cost of
the industrial-scale test. Typical performance of Mill 24
with hydrocyclones is illustrated in Figure 4. The normal
production rate was 95 t/h with a circulating load of
430 per cent. Hydrocyclone feed and product particle size
distributions are provided in Table 1.
In 2006, four 5-deck Stack Size screens fitted with 0.39 mm
urethane panels were installed with Mill 24 in place of the
hydrocyclones. Upon the completion of industrial-scale
testing, Apatit concluded that grinding mills closed with
Stack Sizer screens will see production rate increases of at
FIG 5 – OJSC Apatit Mill 24 (ANOF-2) grinding circuit with screen classification.
least 30 per cent compared to mill lines using conventional
classification equipment. The modified circuit is shown 100
in Figure 5. The production rate of Mill 24 increased from BM-screen / 297tph BM feed
95 t/h when closed with hydrocyclones to 124 t/h when the BM-cyc. / 448tph BM feed
mill circuit was closed with screens. The circulating load 10
decreased from well over 400 per cent to about 135 per cent. R/D
Power consumption per t dropped about 20 to 25 per cent
as total mill power was about the same for both circuits. 1
Likewise, the breakage rate improved while using screens,
as seen in Figure 6. Urethane screen panel life was in the
range of ten to 12 months. Particle size data for the screen 0.1
feed and products is provided in Table 2 and the partition 0.1 1 10 100
curve in Figure 7. p.size (mm)
The coarsening of the grind/grade relationship is illustrated FIG 6 – Breakage rates in the 3.6 m × 4.0 m ball mill in ANOF-2
in Figure 8 with a comparison of the original circuit product when cyclones and screens were used (Lynch, 2015).
(hydrocyclone overflow) and that produced in the screen
TABLE 2
OJSC Apatit Mill 24 (ANOF-2) screen feed and product particle size distributions.
Size (µm) Cumulative % passing

Screen feed Screen oversize Screen undersize
450 61.5 32.8 100.0
320 49.3 18.0 91.4
200 38.2 9.9 76.2
160 31.8 6.9 65.2
100 24.4 4.7 50.9
71 19.3 3.6 40.5
circuit (screen undersize). Final product grade from the

FIG 4 – OJSC Apatit Mill 24 (ANOF-2) grinding flotation circuit was similar even though the flotation feed is
circuit with hydrocyclone classification. measurably coarser.
As provided by (Dündar et al, 2014) the improvement in
TABLE 1 results experienced in Mill 24 indicated that the feed rate
OJSC Apatit Mill 24 (ANOF-2) hydrocyclone feed would be increased in the larger mill circuits when screens
and product particle size distributions. were used, so the cyclones were replaced by screens in one
circuit in which a 4.5 × 5.0 m ball mill was used. The result
Size (µm) Cumulative % passing was an increase of 25 per cent in feed rate. Screens were then
Cyclone feed Cyclone underflow Cyclone overflow used in six circuits at ANOF-3 for mills of size 5.5 × 6.5 m with
450 68.8 61.5 100.0 similar results. To verify the results a circuit was sampled
using cyclones as classifiers and then screens. The partition
320 57.3 49.0 93.0 curves and breakage rates are shown in Figure 9 and 10. The
200 41.1 31.7 81.3 result of the work is that high frequency screens are now
160 32.3 23.0 72.1 being installed in ANOF-3.
100 21.2 13.3 55.3
Cia Minera Condestable SA
71 16.8 9.9 46.4 The following section is a case study by (Dündar et al, 2014).

100 100
screen - 297 tph Feed
BM-screen/400tph BM feed
80 cyclone - 447 tph Feed
BM-cyclone/320tph BM feed
10
partition %
60
R/D
40
1
20
0
0.1
0.01 0.1 1
mean p.size (mm) 0.1 1 10 100
p.size (mm)
FIG 7 – Partition curves for hydrocyclones and screens in the circuit FIG 10 – Breakage rates in the 5.5 m × 6.5 m ball mills in ANOF-3
containing the 3.6 m × 4.0 m mill in ANOF-2 (Lynch, 2015). when cyclones and screens were used (Dündar et al, 2014).
•• with the cyclone installed the feed rate of fresh ore to the
circuit was 59 t/h and the flow rate of ore to the ball mill
was 177 t/h
•• when the high frequency screen was installed, the rates
were 68 and 131 t/h respectively
•• the partition curve for the screen was sharper than for
the cyclone and the bypass was smaller; this caused the
reduction in recycling load
•• the screen product (15.2 per cent +212 µm) was coarser
than the cyclone product (7.4 per cent +212 µm) but this
did not adversely affect the flotation result.
Figures 11 and 12 show the partition curves and breakage
rates when the cyclone and screen were used. The higher
breakage rates with the screen indicated that higher circuit
throughputs were available if required.
FIG 8 – Comparison of circuit products from cyclone Additionally, information about this study comes from
and screen circuits at OJSC Apatit. the case study by Albuquerque et al (2008) on Delgado, Diaz
and Chambi (2007). For this application, the control of the
100 percentage of solids in the screen feed has been determined
screen as the most important parameter to produce the best screen
80 cyclone performance. Table 3 shows the technical comparisons
between the two classification systems, prior to and after the
partition %
60 expansion. Other important results obtained at Condestable

beneficiation plant were a 17 per cent increase in the amount
40
of ore processed, better operation control, more uniform
20
particle size distribution produced by the grinding circuit,
and a reduction in power consumption per tonne.
0
0.01 0.10 1.00 OJSC KMAruda
mean p.size(mm)
KMAruda operates an iron ore mine in Russia and run-of-
FIG 9 – Partition curves for hydrocyclones and screens in the circuit mine ore contains about 34 per cent total iron (Pelevin and
containing 5.5 m × 6.5 m ball mills in ANOF-3 (Dündar et al, 2014). Lazebnaya, 2009). The concentrator flow sheet includes two
stages of grinding and three stages of magnetic separation.
The Condestable and Raúl mines are owned by Condestable The primary grinding circuit on each line consists of two
SA, which is a unit of Southern Peaks Mining SA. They are
located in the Province of Cañete, Department of Lima, Peru. 100
The principal sulfides are chalcopyrite, bornite, pyrite and cyclone
80
pyrrhotite and conventional grinding and flotation circuits screen
are used to produce copper concentrate containing gold and
partition %
60
silver by-products. The Condestable mine was expanded
during 2000 to 2007 to increase plant capacity from 1569 t/d 40
to 5500 t/d and by 2006 the plant was operating at 4100 t/d.
At this time cyclones were replaced by high frequency screens 20
and the capacity of the plant increased to 6000 t/d in 2013. The
reasons were the better partition curves of the screens and the 0
higher breakage rates in the mills and these will be discussed 10 100 1000
mean p.size (micron)
using data from one of four ball mill-classifier grinding
circuits. The ball mill was 3.81 m by 4.72 m, the cyclone was FIG 11 – Cyclone and screen partition curves at
26’’ diameter, and the screen had 300 µm apertures: Condestable (Dündar et al, 2014).

S B Valine et al
100
BM-cyclone
BM-screen
10
R/D
0.1
0.01
10 100 1000 10000
p.size (µm)
FIG 12 – Breakage rates for the 3.81 m × 4.72 m ball mill at Condestable
when the cyclone and screen were used (Dündar et al, 2014).
FIG 14 – KMAruda second stage grinding circuit with screen classification.
TABLE 3
Technical Comparisons Hydrocyclone – Screen (Albuquerque, 2008). TABLE 4
KMAruda concentrator performance with hydrocyclones
Operational parameters Hydrocyclone Stack sizer and screens in the secondary mill circuit.
Circulating load 204% 96%
Parameter With hydrocyclones With screens
Classification efficiency 62–64% 85%
Capacity (t/h) 120 140 to 150
Processes TMS/Hr 59 68
Secondary mill power draw (kW) 1260 630
Size distribution control Variable Uniform
Secondary mill circulating load (%) 210 to 220 130 to 160
ball mills in parallel operating in closed circuit with spiral Secondary mill circulating load size 51.0 22.9
classifiers. Similarly, the secondary circuit has two ball mills (% -71 μm)
in parallel in closed circuit with hydrocyclones. All four Final concentrate size (% -71 μm) 92.5 82.0
mills are identical in size. The secondary circuit is illustrated Final concentrate grade (%Fe) 65.85 66.00
in Figure 13. The first stage magnetic concentrate feeds the
Final concentrate iron recovery (%) 82.2 84.6
cyclones and the cyclone underflow is distributed to the two
secondary mills. The mill discharge is fed to the second stage Final tails grade (%Fe) 10.9 9.09
magnetic separators and the magnetic concentrate circulated
back to the cyclones. The circuit product goes on to the third -71 µm compared to 51 per cent -71 µm in the cyclone circuit.
stage of magnetic separation. While the goal was to maintain the same final product grade,
Two high frequency screens fitted with 100 µm urethane the final concentrate from the screen circuit was also coarser
panels were installed in place of the hydrocyclones on at 82 per cent 71 µm compared to 92.5 per cent -71 µm on a
one concentrator production line. Subsequent testing and typical cyclone line. KMAruda also noted that while cyclone
optimisation of the screen circuit resulted in the shutdown of underflow was normally around 65 per cent iron, in the screen
one of the two second stage ball mills. The screen circuit is circuit the screen oversize ranged from 36 to 45 per cent iron,
illustrated in Figure 14. leading to lower iron losses in the second stage of magnetic
separation.
As summarised in Table 4, improved classification efficiency
in the second stage circuit resulted in greater line capacity and
Mina Colquijirca – Sociedad Minera El Brocal SAA
a significant reduction in power consumption per tonne with
Sociedad Minera El Brocal, located at Tinyahuarco, Pasco
the elimination of one mill. The circulating load in the screen
Province, Peru, produces lead, zinc, copper and gold
circuit was lower and considerably coarser at 22.9 per cent
(Aquino and Vizcarra, 2007). The concentrator flow sheet
begins with three rod mills operating in parallel and in open
circuit followed by three ball mills operating in parallel and
in closed circuit with a bank of ten 10’’ hydrocyclones. As
illustrated in Figure 15, rod mill discharge and ball mill
discharge are combined in the same sump and pumped to
the cyclones. The circulating load is about 350 per cent and
over-grinding of the high specific gravity minerals such as
galena results in significant slime losses before flotation.
Cyclone feed and product particle size distributions are
provided in Table 5.
Brocal desired to increase capacity and was considering
the addition of a fourth ball mill. As an alternative, Brocal
considered ways to improve grinding efficiency and conducted
full-scale high frequency screen tests. Encouraging test results
led to the installation of two high frequency screens fitted with
FIG 13 – KMAruda second stage grinding circuit 0.50 mm urethane panels in place of the ten hydrocyclones.
with hydrocyclone classification. The effect was immediate as production increased 11 per cent

FIG 15 – Brocal grinding circuit with hydrocyclone classification.

FIG 16 – Brocal grinding circuit with screen classification.
TABLE 5
1000
Brocal hydrocyclone feed and product particle size distributions.
BM-cyclone
100
Size (µm) Cumulative % passing BM-screen
Cyclone feed Cyclone underflow Cyclone overflow 10

R/D
500 72.4 71.6 98.6

1
300 62.3 59.9 96.2
212 49.2 45.0 91.7 0.1
150 36.3 28.4 85.4 0.01

75 20.5 11.4 67.7 10 100 1000 10000
p.size (µm)
and lead recovery increased 9 per cent even though Brocal FIG 17 – Breakage rates in the 8 ft × 10 ft ball mills at El Brocal
shutdown two of the three operating ball mills. when cyclones and screens were used (Dündar et al, 2014).
To take advantage of the increased grinding capacity,
Brocal added additional crushing, flotation and filtration TABLE 6
capacity, as well as three additional high frequency screens. Brocal screen feed and product particle size distributions.
One of the two ball mills originally shutdown was put back
into operation and rod mill rotational speed was increased Size (µm) Cumulative % passing
(Aquino and Vizcarra, 2007). As shown on Figure 16, Screen feed Screen oversize Screen undersize
production has increased from 138 t/h with the cyclone
500 69.5 22.2 98.4
circuit to 245 t/h with the screen circuit, an increase of
over 75 per cent. Additionally, the breakage rate increased 300 57.1 11.5 85.2
requiring less total grinding energy as shown in Figure 17. 212 52.8 10.5 78.0
The circulating load is about 60 per cent and over-grinding
150 42.1 8.7 68.9
has been minimised; the slime content (particles finer than
10 µm) in the flotation feed has been reduced from 18 per cent 75 38.1 8.2 56.3
to 10 per cent, resulting in signification increases in metal
recovery. Screen feed and product particle size distributions The Eczacibasi Esan Mine is located in the province of
are provided in Table 6. Balikesir, Turkey, and produces lead and zinc concentrates.
A comparison of the particle size distributions of The grinding circuit at Eczacibasi Esan consists of five parallel
the grinding circuit product (flotation feed) from the ball mill circuits. The ball mills were formerly operated with
hydrocyclone circuit and the screen circuit is shown cyclones, then they were replaced by high frequency screens
in Figure 18. Note the coarsening of the grind/grade with 210 µm apertures. The data were collected from a ball
relationship as the concentrate grade in both cases was mill (3.2 m × 5.0 m) screen circuit and a ball mill (3.2 m ×
similar. In addition, the partition curve for the cyclone 4.2 m) cyclone circuit. Feed rate to the ball mill cyclone circuit
versus the screen is shown in Figure 19. was about 32 t/h and to the ball mill screen circuit was about
52 t/h which was much higher than would be expected from
Eczacibasi Esan Mine the mill lengths. Flotation results were similar. The partition
The following section is a case study by (Dündar et al, 2014). curves and breakage rates are shown in Figure s20 and 21.

S B Valine et al
100
screen
80 cyclone
partition %
60
40
20
0
1 10 100 1000
mean p.size (µm)
FIG 20 – Partition curves for cyclones and screens

in the ball mill-classifier circuit at Esan.
10000
FIG 18 – Comparison of circuit products from BM-screen
1000
hydrocyclone and screen circuits at Brocal. BM-cyclone
100
R/D
100
cyclone -
538 tph 10
80 Feed
screen - 1
385 tph
60 Feed
partition %
0.1
100 1000 10000 100000
40 p.size (µm)
20 FIG 21 – Breakage rates in the ball mills at Esan

when cyclones and screens were used.
0
10 100 1000
mean p.size (µm)
FIG 19 – Partition curves for cyclones and screens

at El Brocal (Dündar et al, 2014).
Minera Cerro Lindo

Minera Cerro Lindo is located south-west of Lima, Peru and
produces copper, lead, and zinc concentrates (Ticse, 2009).
As shown in Figure 22, the grinding circuit at Cerro Lindo
consists of a 14.5 ft × 23.5 ft ball mill operating in closed
circuit with a bank of 26’’ hydrocyclones. The circulating load
is about 260 per cent. Cyclone feed and product particle size
distributions are provided in Table 7.
Following the success of screen classification in grinding
circuits at Brocal and other polymetallic mining operations in FIG 22 – Cerro Lindo grinding circuit with hydrocyclone classification.
Peru, Cerro Lindo initiated a program to improve concentrator
performance. Following full-scale testing, four high Catalina Huanca Mine
frequency screens were installed in place of the 26’’ diameter
The following section is a case study by (Dündar et al, 2014).
hydrocyclones. Three screens are fitted with 0.23 mm urethane
panels and one with 0.18 mm urethane panels. As shown in The Catalina Huanca mine is located in the province of
Figure 23, the circulating load has decreased to 108 per cent Victor Fajardo, Department of Ayacucho, Peru, south-east of
and line capacity has increased 13.6 per cent to 275 t/h. Screen Lima. Mining has been carried out at the site during many
centuries and the present operation started in 1964. The ore is
feed and product particle size distributions are provided in
polymetallic with the current head grade being approximately
Table 8 and the partition curves for the cyclone and screen are
5.5 per cent zinc (Zn), 1.4 per cent lead (Pb), 0.2 per cent copper
shown in Figure 24.
(Cu) and 1.25 oz/t silver (Ag). Processing is by conventional
Table 9 highlights some of the benefits of the change to grinding and flotation. The capacity of the San Jerónimo plant
screen classification. In the screen circuit, while the mill has been increased on three occasions since 1964, each time by
discharge product has more minus 75 µm material, flotation the addition of a ball mill and flotation cells and the current
feed has less -75 µm. The P80 is coarser in the screen circuit throughput is 1800 t/d. For many years primary grinding was
at 160 µm compared to 141 µm with the cyclone circuit. Steel in a 2.90 m × 4.27 m rod mill and secondary grinding was in
consumption per kwh has decreased due to the improved three parallel ball mills with sizes of 1.83 m × 1.83 m, 2.13 m ×
breakage rates seen in Figure 25. Tailings filtration capacity 2.44 m, and 2.44 m × 1.83 m. In 2011 the circuit was rearranged
also increased as the tailings are now coarser and more so that the three smaller mills became open circuit primary
homogenous. mills and discharges went to a high frequency screen. The

TABLE 7 TABLE 9
Cerro Lindo hydrocyclone feed and product particle size distributions. Comparison of Cerro Lindo concentrator performance
with hydrocyclones and screens.
Cyclone feed Cyclone underflow Cyclone overflow Parameter With With screens
hydrocyclones
500 100.0 100.0 100.0
Capacity (t/h) 242 275
300 100.0 100.0 100.0
Circulating load (%) 244 108
212 63.9 52.1 94.4
Steel consumption (g/kwh) 778 696
150 46.1 31.0 85.4
Mill discharge particle size (% -75 μm) 24 30
75 22.5 8.4 58.9
Mill solids (%) 83 79
Flotation feed particle size (% -75 μm) 55 47
Flotation feed P80 (μm) 141 160
100
10
R/D
1
0.1 BM-cyclone
BM-screen
0.01
10 100 1000 10000
FIG 23 – Cerro Lindo grinding circuit with screen classification. p.size (micron)
TABLE 8 FIG 25 – Breakage rates in the 4.42 m × 7.16 m ball mill at Cerro
Cerro Lindo screen feed and product particle size distributions. Lindo when cyclones and screens were used (Dündar et al, 2014).

Screen feed Screen oversize Screen undersize
500 91.0 88.0 100
300 75.3 54.9 99.3
212 59.4 31.1 90.6
150 47.8 20.5 75.0
75 30.4 9.9 47.0
100
cyclone
80 screen
partition %
60
FIG 26 – Flow sheet of the Catalina Huanca grinding circuit.
40
breakage rates when cyclones and screens were used, are
20 given in Figures 27–29.
0
10 100 1000 CONCLUSIONS
mean p.size (micron)
Grinding is one of the most critical and costly operations in
FIG 24 – Partition curves for cyclones and screens in the ore processing and efficient classification is vital to efficient
grinding circuit at Cerro Lindo (Dündar et al, 2014). grinding. Classification improvements in classification are a
necessity with the decline in orebody head grades, increased
oversize was ground in a 2.44 m × 1.83 m ball mill which was orebody complexity and rising costs of energy. The benefits of
operated in closed circuit with the classifier (Figure 26). high frequency screens can meet these demands. However, the
Cyclones and screens were tested in the closed circuit and size and capital cost of the classifier is regularly given priority
it was found that the better partition curve obtained with over classification efficiency. There is often a preconceived idea
the screen reduced the circulating load and increased the that the classifier must fit in a ‘small box’ and at the same time
breakage rates in the mill. Partition curves for the open circuit most concentrators include large building filled with all types
screen and for the screens and cyclones in closed circuit, and of separation equipment such as flotation cells, spirals, jigs

S B Valine et al
100 rates. The benefits seen in the case studies provided include:
300 mic - 44 tph Feed increased productivity throughput, reduction in circulating
300 mic - 45 tph Feed
80 300 mic - 42 tph Feed load, decreased power consumption per tonne, reduced
300 mic - 42 tph Feed over-grinding of valuable minerals/generation of slimes/
partition %
60 230 mic - 43 tph Feed

230 mic - 42 tph Feed
tailings, and increased recovery of valuable minerals. Other
40 230 mic - 44 tph Feed benefits not discussed in these case studies may have been:
230 mic - 43 tph Feed consistent product size distribution and density, reduced
20 reagent consumption in flotation, reduced dewatering costs
and reduction in overall operational costs.
0
10 100 1000
mean p.size (µm)
REFERENCES
FIG 27 – Partition curves for open circuit screens classifying Albert, E, 1945. Characteristics of screen-circuit products, Min Tech
TP, 1820.
primary mill discharge at Catalina Huanca.
Albuquerque, L, Wheeler, J, Valine, S and Ganahl, B, 2008.
Application of high frequency screens in closing grinding circuits,
100 in Proceedings Fifth International Mineral Processing Seminar
(Procemin 2008) Santiago, Chile.
80
Aquino, B and Vizcarra, J, 2007. Re-engineering the metallurgical
partition %
60 processes at Colquirjirca Mine, Socidad Mineral El Brocal SAA

cyc. - 126 tph -
568% C.L.
Mineria, pp 30–38.
40 cyc. - 75 tph -
Davis, W E, 1925. Ball mill crushed in closed circuits with screens,
396% C.L.
sc. 230 mic - 41 University of Minnesota, Mines Experiment Station Bulletin,
20 tph - 129% C.L. 10:2842.
sc. 230 mic - 27
0 tph - 88% C.L. Delgado, M, Diaz, G and Chambi, R, 2007. Production expansion
10 100 1000 in condestable with technologies innovations in grinding
mean p.size (µm) classification, XXVIII Mining Convention Extemin 2007, Peru.
Dündar, H, Kalugin, A, Delgado, M, Palomino, A, Türkistanlı, A,
FIG 28 – Partition curves for cyclones and screens in
Aquino, B and Lynch, A, 2014. Screens and cyclones in closed
the circuit with ball mill at Catalina Huanca. grinding circuits, XXVII International Mineral Processing
Congress.
10 Hukki, R T, 1967. An analysis of mill and classifier performance in a
BM-cyclone / 126 tph BM feed
BM-cyclone / 75 tph BM feed
closed grinding circuit, Society of Mining Engineers, Transactions.
BM-screen / 41 tph BM feed Hukki, R T and Allenius, H, 1968. A quantitative investigation of
BM-screen / 27 tph BM feed
1 closed grinding circuit, Society of Mining Engineers, AIME –
Transactions Volt 241, December.
R/D
Hukke, R T and Eland, H, 1965. The relationship between sharpness

0.1 of classification and circulating load in closed grinding circuits,
Transactions, Society of Mining Engineers.
Lynch, A, 2015. Comminution Handbook, Spectrum 21, pp 157–159 (The
0.01 Australasian Institute of Mining and Metallurgy: Melbourne).
10 100 1000 Murr, D L, Wennen, J E and Nordstrom, W J, 2009. Essar Steel
p.size (µm)
Minnesota, concentrator flowsheet development, 70th Annual
FIG 29 – Breakage rates in the ball mill at Catalina University of Minnesota Mining Symposium, Duluth, Minnesota.
Huanca when cyclones and screens were used. OJSC Apatit, 2009. Written communication.
Pelevin, A E and Lazebnaya, M V, 2009. Application of Derrick
and magnetic separators. Although grinding circuits closed Screens in locked grinding circuit at the KMAruda Mining
with screens often require a ‘large box’, this concept should Complex Concentrating Plant, Mineral Processing Journal (Russia),
be judged on its own merits without prejudice of the volume 2:4–8.
and initial cost of the equipment. High circulating loads are Ticse, F, 2009. Milpo-Cerro Lindo, Information provided during
often employed to reduce over-grinding in grinding circuits plant visit.
with hydrocyclone classification. Even so, over-grinding is a Valine, S B and Wennen, J E, 2002. Fine screening in mineral
common characteristic in cyclone circuits. Circulating loads processing operations, in Mineral Processing Plant Design, Practice,
around 100 per cent are common in screen circuits and over- And Control (eds: A L Mular, D N Halbe and D J Barratt), (Society
grinding is minimised. for Mining, Metallurgy and Exploration).
Replacing hydrocyclones with high frequency screening Wennen, J E, Nordstrom, W J and Murr, D L, 1997. National steel
systems, such as the Stack Sizer, in grinding circuits has pellet company’s secondary grinding circuit modifications,
demonstrated significant metallurgical and economic benefits Comminution Practices, 99 19–25.
due to the improvements in partition curve and breakage Whiten, W J, 1974. A matrix theory of comminution machines, Chem
Eng Sci, 29:589–599.

Contents
The influence of liner wear on milling efficiency

P Toor1, T Perkins2, M S Powell3 and J Franke4
ABSTRACT
Milling is an energy intensive process and any increases in efficiency are highly
desirable. One option to reduce energy and material consumption while optimising
production yield in mill operation is through balanced design and selection of liners.
Overdesign of liners leads to increased life but at a cost to grinding performance,
a relationship that to date has not been studied closely. This paper provides a
description of this process of liner life cycle optimisation and presents the results
from a 32 foot semi-autogenus grinding (SAG) mill. The use of high resolution laser
scanning based liner shape information provided by the MillMapper® software
correlated with well controlled site surveys at various points in the liner life was used
as the basis for comparison. The survey data was modelled in JKSimMet to provide
direct comparisons of mill performance. This paper provides a methodology for the
design of liners that maintain a favourable shape for the majority of liner life, yielding
desirable production parameters such as lower specific energy consumption and
higher throughput, while at the same time maintaining practical liner life cycles.
INTRODUCTION
A key area of interest in mineral processing has been research into the establishment of
grinding processes and more recently, of representative grinding simulations. This is
due to grinding being an important unit operation in a processing plant, both in terms
of energy consumption and overall performance. It is not uncommon for grinding
circuits to constitute up to 40 per cent of the plant power usage and operating costs in
a processing plant (Herbst, Lo and Flintoff, 2003).
The interest in research and simulation of grinding and comminution in general
has been triggered by many current issues such as those listed by Norgate and
Haque (2010):
•• The increasing cost of fuel and materials. The comminution of ore does not only
consume significant amounts of energy but also large amounts of steel due to
wear of grinding media and liners.
•• Climate change is a pressing global environmental issue and has increased the
need to make comminution more energy efficient beyond purely economic
arguments.
•• Milling uses about 90 per cent of the comminution energy, or 40 per cent of
total plant energy. In the case of copper ore, the grinding stages comprise
the largest contribution to the total greenhouse gas emissions for copper
concentrate production.
•• The general trend of falling grade of orebodies which are more complex and
finer grained. This will require finer grinding to liberate the valuable material
to achieve separation and concentration and inherently make comminution
processes more energy intensive.
1. MAusIMM, Development Manager/Grinder •• The ever-increasing global demand for industrial and consumer goods means
Wear Solutions, Metso Australia, Eagle Farm that the production of primary metals can be expected to increase well into the
Qld 4009. Email: paul.toor@metso.com future.
2. MAusIMM(CP), Senior Metallurgist, Many research and industry practitioners are working on addressing the energy
Rio Tinto, Brisbane Qld 4000. issue in comminution with new equipment designs or novel circuit solutions using
Email: thomas.perkins@riotinto.com
existing equipment. These approaches can overcome the inherent limitations in the
3. FAusIMM, Professorial Research Fellow, existing grinding equipment designs and circuits that cause high energy consumption.
Chair in Sustainable Comminution, Julius
While they are expected to provide solutions for new projects, it is important to note
Sustainable Minerals Institute, The University that existing plants will continue to run with traditional grinding circuits because
of Queensland, Indooroopilly Qld 4068. of the capital investment costs already incurred. It will not be commercially viable
Email: malcolm.powell@uq.edu.au to refit more than a small percentage of existing plants with the new equipment
4. R&D Manager – PCMP, Outotec Pty Ltd, West and circuit designs. Therefore, solutions to existing traditional grinding circuits are
Perth WA 6005. needed if progress is to be made in the industry as a whole.
509
P Toor et al
One possible way to reduce energy and material based on manual single point distance measurements that
consumption in milling is through the appropriate design cannot account for 3D shape changes. MillMapper technology
and selection of liners. Liners play an important role in can track changes in mill volume, arising from the wear of
grinding due to their strong influence on load motion and shell lifters, as well as feed and discharge cones at a high
behaviour (Makokha et al, 2006; Powell and Verneulen, 1993). degree of accuracy.
Generally the milling efficiency depends on the behaviour of In the case of the large SAG mill referred to in Figure 1, the
the load inside the mill which governs the nature of the ore mill’s initial volume is 757 m3 when new liners are installed,
presentation to breakage (Makokha et al, 2006), yet liner shape with the volume on average increasing to 792 m3 at the end of
is dynamic as it is continually changing because of wear. shell liner life over three liner lives for the 2010 calendar year,
It is a common practice in the mineral processing industry which equates to an average increase of 4.4 per cent. However,
to attempt to overdesign mill liners with the aim of achieving historical data demonstrates the period from new to old liner
a maximum possible liner life. Overdesigning liners is a trade- stage yields on average an increase in throughput in the order
off between liner life and grinding performance that to date of ten per cent (Table 1). Notably, this increase in throughput
has not been studied closely. A common observation over is achieved for the six observed subsequent shell life cycles
the life of a shell liner life cycle is an unplanned but welcome to varying degrees, and therefore is a systematic trend rather
throughput increase and specific energy (kWh/t) decrease than an outlier. The reline schedule of this mill is for all the
as wear progresses. Without the use of appropriate liner shell and feed end to be relined simultaneously. Only the
measurement tools to track and quantify wear, an intentional discharge liner is relined in an intermediate sequence. This
increase in throughput and decrease in power draw with loss routine imposes less distorting influence on the assessment of
of liner profile as the conclusively identified root cause is throughput over shell liner life cycles. The reline schedule for
difficult to validate. the large SAG mill is summarised in Figure 2.
The observations of the authors of several mills show an Assuming constant transport rates and percentage fill levels
increase as great as 20 per cent in throughput from new to for new and completely worn liners, the increase of only
old liner and commonly in the order of ten per cent, Figure 1 4–5 per cent in mill volume does not account for the observed
shows an example for a large SAG mill. It illustrates that ten per cent increase in throughput. Lacking other conclusive
there is a consistent trend of increased throughput over shell evidence that could explain any potential change in transport
liner life. Since mill speed increases as shown were used to rate or fill level, the remaining 5–6 per cent in observed
compensate for the decrease in lift as the liners wear, it is increased throughput may be attributed to an increase in
likely that the highlighted increase in throughput over shell efficiency of liner shape as it wears as well as grate aperture
liner life is due to both an increase in internal mill volume wear. The data suggests that new liners, as used, are inefficient
as the liner wears, as well as an optimum liner shape being and at the other end of the scale, worn liners are efficient. This
achieved during liner life. paper outlines a robust technique to capture and quantify this
optimum liner shape for improved milling efficiency.
PROBLEM STATEMENT – LOW EARLY LINER LIFE throughput This can aid the design of liners to maintain a favourable
The hypothesis of this work is that the increase in throughput shape for the majority of liner life, yielding desirable
in SAG milling over a shell liner life cycle witnessed in production parameters such as lower energy consumption
industry is due to both an increase in internal mill volume and higher throughput, while at the same time maintaining
due to liner wear, and secondly, to an optimum liner shape practical liner life cycles and not compromising product sizes.
being achieved during liner life.
The increase in internal mill volume over liner life cycles METHODOLOGY
is easily tracked from the MillMapper® software as it is a This section outlines the methodology used for this study –
standard survey output. The net mill volume for each survey to design liners which maintain a favourable shape for the
is calculated by accounting for the exact laser scanned 3D majority of their life, yielding lower energy consumption
liner shape at the wear stage in question for feed, shell and and higher throughput, while at the same time maintaining
SAG Mill Throughput as a Function of Liner Life
discharge liners. Traditional mill volume estimations are practical liner life cycles.
2500 0.75
Specific Energy (kWhrs/t); Mill Speed/10 (RPM)
Average Throughput (tph)
2000 0.6
1500 0.45
1000 0.3
500 0.15
0 0
4/06/2007
8/07/2007
9/11/2007
1/02/2008
2/12/2008
2/04/2009
2/05/2009
9/06/2009
30/04/2007
11/08/2007
29/09/2007
25/12/2007
14/03/2008
30/04/2008
14/06/2008
21/07/2008
25/08/2008
12/10/2008
19/10/2008
11/01/2009
21/02/2009
25/07/2009
tph Specific Energy Shell Reline Mill Speed/20
FIG 1 – Increase in throughput over liner life for a large semi-autogenous grinding (SAG) mill.

Table 1
Per cent increased throughput from old to new shell liner.
New shell liner Very worn shell liner Increased throughput %

(worn–new)
Period Tonnage Period Tonnage
23/04/2007–07/05/2007 1939 11/08/2007–25/08/2007 2154 11.1%
15/09/2007–29/09/2007 1912 18/01/2008–1/02/2008 2183 14.2%
14/02/2008–28/02/2008 1972 14/06/2008–28/06/2008 2126 7.8%
07/07/2008–21/07/2008 1939 30/10/2008–13/11/2008 2267 16.9%
25/11/2008–09/12/2008 2181 19/03/2008–2/04/2009 2242 2.8%
03/04/2009–17/04/2009 2081 12/08/2009–26/08/2009 2268 9.0%
Average Average Average
2004 2207 10.3%
MillMapper liner shapes. With knowledge of optimum liner

shape, a new liner design can be proposed which maintains
a favourable liner shape for most of its life. The functionality
of the new liner design can be further tested in a subsequent
DEM simulation.
To summarise, the liner optimisation methodology is:
•• Conduct controlled grinding surveys in conjunction
with accurate laser scanning during a mill liner life
cycle to make direct comparisons between various liner
shapes and their impact on mill performance.
•• Measure the liner shapes and mill fill volume levels
at four (4) times during the mill liner life cycle using
established laser scanning techniques.
•• Conduct carefully controlled surveys to measure mill
performance directly prior to each liner measurement.
•• Use mill surveys, mill performance and production data
to calculate key mill performance parameters over the
life of the liner. These parameters include:
•• throughput
FIG 2 – Reline schedule for large semi-autogenous grinding (SAG) mill. •• grind size – fineness of product desired to feed
downstream processes
To aid in determining this optimum liner shape the
following tools were used: •• specific energy consumption.
•• Controlled Grinding Surveys •• Use JKSimMet to normalise the data to keep parameters
such as feed hardness and cyclone performance constant
•• MillMapper across all surveys to highlight the effect of change in
•• JKSimMet. liner shape on mill performance.
It is also possible to add Discrete Element Method (DEM)
Models at four different liner life stages based on the survey Liner Shape Tracking
work and MillMapper scans. Furthermore, it is also possible MillMapper provides high resolution three-dimensional mill
to develop DEM for appropriate ‘what if’ scenarios aimed liner thickness information inside a mill by mapping tens of
at liner redesign. Figure 3 is the suggested methodology for millions of individual survey points. Aside from tracking liner
testing the hypothesis. shape and wear, other variables such as liner weight, net mill
Grinding surveys are conducted at various stages in the volume, charge volume, ball size distribution and discharge
liner life (four times) to gain snap shots of milling efficiency grate open area are also quantified. After completion of a
at each stage over two life cycles. In conjunction with the condition monitoring series, high wear zones are automatically
grinding surveys MillMapper provides detailed and accurate identified and used for reline forecasting, as illustrated in the
3D liner shapes at each stage. liner thickness plots of Figure 4. More detailed descriptions
of the MillMapper methodology and industry case studies
Due to the large statistical noise associated with commercial
have been provided in Franke, Lichti and Stewart (2006) and
processing plants, JKSimMet is used to filter noise in
Franke (2008). Variables such as effective lifter height and face
the grinding surveys by normalising feed and operating
angle which significantly affect the trajectory of the charge
conditions and only leaving the breakage rates as the variables are measured and provide the basis for representative DEM
in the steady state models. modelling. Critically this can be done not just for new liner
JKSimMet model results are used to identify the most shapes, but for worn shapes as well, which is the only way
productive operating period of the life cycle for the present to identify actual grinding behaviour at times other than the
liner design before correlating them to the corresponding very initial period of a liner life cycle.

P Toor et al
FIG 3 – Conceptual methodology for determining optimum liner shape (DEM – Discrete Element Method).
FIG 4 – 3D thickness model as produced by MillMapper®.
Figure 5 shows the stages of the liner wear when undertaking obtain liner shape, ball size distribution, liner volume and
the surveys. Each line represents the liner profile measured at liner mass (suited for load cell calibration) and ball/ore charge
a given survey date with the darkest line being the most recent volume information. The surveys coincided with when the
(06/12/2010 reline) and the lightest being the most dated liners were new, half worn, 80 per cent worn and fully worn
(08/07/2010 new liner). Peak performance periods therefore (prior to reline).
can be correlated to a distinct liner shape. While Figure 6 shows The SAG circuit being studied is shown in Figure 11. The
the specific 2D profile of the highest wearing zone on the shell SAG was in closed circuit with cyclones and the feed material
liners, the identification of optimum new liner shapes needs was typically minus 75 mm. This material was a combination of
to be considered in 3D when assessing all liner areas. This is crushed run-of-mine material and a recycle stream of crushed
relevant for all cases where liner wear varies significantly along SAG mill pebbles. The SAG mill discharged onto a trommel,
the mill axis, something which holds true for the vast majority the coarse material from the trommel then passed over a single
of SAG and AG mills in operation (Franke, 2008). deck vibrating screen before being conveyed to the pebble
stockpile. The pebble stockpile material was fed to one or both
PLANT SURVEYS cone crushers that discharged onto the SAG feed belt.
Four grinding surveys were conducted during the liner life The SAG trommel undersize flowed into a common sump
alongside MillMapper laser scanning of the mill liners to shared with the ball mill. The material in this sump was

FIG 5 – Highest wearing profile progression at a central mill section over liner life.
FIG 6 – Size distributions of repeat semi-autogenous grinding (SAG) trommel undersize samples – the original sample point.
pumped to a cluster of cyclones for classification. Portions •• Recycle Crusher Feed

of the cyclone underflow stream were used to feed the •• SAG Pebble Recycle
flash flotation cell and the gravity concentrators, with the
•• SAG Trommel Undersize
remainder of the material being fed into the SAG, ball and
Vertimills. The gravity concentrate reported directly to the •• SAG Screen Undersize
gold room and the cyclone overflow formed the feed to the •• Cyclone Underflow
flotation circuit. •• Cyclone Overflow
In order to properly sample the mill product, samples were •• Cyclone Feed.
taken from the SAG trommel undersize, the SAG screen
Of the above samples, the SAG trommel undersize (US)
undersize and the SAG screen oversize. The full feed to the
sample was inarguably the most crucial sample point.
mill was sampled through a SAG feed belt cut and sampling The sample from this point was initially shown to be non-
of the cyclone underflow streams. The feed and product representative. To test the repeatability of the SAG trommel
samples from the recycle cone crushers were sampled directly sample point, three samples were taken immediately
at the conclusion of the first survey. The cyclone overflow was following each other while the circuit was relatively stable.
sampled to provide circuit product information and allow The samples were sized and produced the results as shown
an estimation of the effect of the reline on the total circuit in Figure 6.
performance to be determined.
It can be seen there is a large amount of variability in the
Samples were taken around all relevant points with the goal sample point. The modifications required to improve the
of fully specifying the inputs and outputs of the SAG mill, sample point were not possible to implement without a major
recycle crushers and cyclones. A summary of the sample shutdown, thus this refit was not completed until the second
points is given below: life cycle.
•• SAG Feed The original sample was taken through the door in the
•• Recycle Crusher Product trommel casing shown in Figure 7 and collected with a very

P Toor et al
The new sample point allowed for a representative sample

to be taken. As can be seen in Figure 9, the improvements
resulted in the distributions in repeat sampling being
virtually identical. Additionally, the distribution is different
to the previous reproducibility test, indicating a likely bias in
the original sample.
PLANT DATA
Table 2 and Figure 10 summarise the mill performance over
the six-month liner life based on weekly averages. There is a
trend of increased throughput and reduced specific energy as
the liner wears. Initially the mill throughput is approximately
650 t/h at the start of liner life while at the end of liner life
it is about 730 t/h. This equates to an increase of 12 per cent
throughput when comparing a fully worn liner with a new
liner, and it also coincides with a decrease in specific energy
of 12 per cent.
FIG 7 – Semi-autogenous grinding (SAG) trommel undersize door sample point. A noticeable reduction in throughput occurs at half-life
when the discharge grates (½ set) and the feed outer liners
long arm sample cutter. The capacity of the sample cutter was are relined. This causes the regressed throughput curve
enough to ensure that overflowing was not likely, but the reach to taper but still increase as the shell liners wear further. It
of the sample cutter was not enough to allow the entire stream should be noted that the throughput post grate reline is still
to be sampled. The highest flow section of the stream was not approximately three per cent higher on average than the
reachable, introducing a strong possibility of bias in the sample. throughput prior to the grate reline, indicating the increase in
During a reline shutdown, the opportunity was taken to throughput is not only due to an increased open area of the
improve the sample points by cutting three hatches on the grates but also due to the wear in the shell liners (see Table 3).
opposite side of the trommel casing closer to the high flow Another observation is that throughput at the end of liner life
stream from the trommel. The hatches were spaced such that is slightly reduced for this mill, which could be explained by
the entire length of the stream could be sampled with a short excessive loss of lift caused by severe liner wear at the end of
arm sample cutter with each hatch allowing for one-third of the life cycle (see most worn profile in Figure 6). However, the
the sample/stream to be cut. The new sample point is shown decrease in throughput is minor and could therefore potentially
in Figure 8. be explained by other operational parameters and plant noise.
There are also some noticeable outliers where there is a
sudden drop in throughput that are caused by lack of ore
availability. These outliers have been highlighted in red in
Table 2.
Finally, grinding surveys conducted during the course of the
project indicated that SAG product becomes finer as the liner
wears. This is highlighted in Table 4, which alongside product
sizes shows the fresh tonnes per hour, recycled tonnes per
hour (pebble crusher), total tonnes per hour and power (kW).
Thus, from all data available it appears that as the liners wear,
the three primary performance indicators of throughput,
power and product size all improved considerably, with
greater throughput, reduced power and finer product.
JKSIMMET MODELLING
JKSimMet is a steady state simulation program for
FIG 8 – New SAG trommel undersize point with multiple hatches. comminution and classification circuits which is widely used
FIG 9 – Size distributions of repeat semi-autogenous grinding (SAG) trommel undersize samples after installation of improved sample point.

Table 2
Mill performance summary over liner life; ore availability constrained throughput periods highlighted in red.
Period Throughput (t/h) Power (kW) Specific energy (kWh/t)

6/12/2010–17/12/2010 652 6744 10.34
17/12/2010–24/12/2010 673 6835 10.16
24/12/2010–31/12/2010 658 6810 10.35
31/12/2010–7/01/2011 682 6669 9.77
7/01/2011–18/01/2011 651 6538 10.04
18/01/2011–26/01/2011 676 6643 9.82
26/01/2011–3/02/2011 711 6629 9.32
3/02/2011–11/02/2011 714 6765 9.48
11/02/2011–18/02/2011 708 6713 9.48
18/02/2011–25/02/2011 746 6607 8.86
25/02/2011–4/03/2011 735 6691 9.10
4/03/2011–9/03/2011 720 6829 9.49
Grates reline
9/03/2011–14/03/2011 672 6760 10.07
14/03/2011–21/03/2011 676 6712 9.93
21/03/2011–28/03/2011 741 6754 9.12
28/03/2011–4/04/2011 716 6776 9.46
4/04/2011–13/04/2011 683 6742 9.88
13/04/2011–20/04/2011 714 6718 9.41
20/04/2011–27/04/2011 743 6632 8.92
27/04/2011–7/05/2011 709 6662 9.40
7/05/2011–14/05/2011 710 6575 9.26
14/05/2011–21/05/2011 740 6526 8.82
21/05/2011–01/06/2011 737 6483 8.80
28/05/2011 732 6379 8.71
in industry. The JKSimMet circuit flow diagram used in this by the pebble ports – a discrepancy being addressed in
project is given in Figure 11. separate work on upgrading the SAG mill model – so is
JKSimMet was used in this project to compare breakage a relative number.
rates over the liner life. This was done by performing well Current values were chosen based on experience and
controlled grinding surveys at four stages of liner life and were allowed to vary slightly around their starting points to
creating subsequent models. To make direct comparisons determine the stability of the solution. All these best fit values
between surveys and gauge the effect of change in liner were then locked into the simulation to calculate the discharge
shape, the data was normalised. This was done by holding functions for each survey.
other variables constant such as the ‘A’ and ‘b’ ore impact
The comparison of breakage rates across the liner life is
breakage parameters from the JK drop weight test. By doing
of particular interest as such a comparison will assist in
this, any change in the breakage rates and or mass transfer can
describing the progression of mill performance over liner life.
be attributed to the change in liner shape rather than changes
in feed or operating conditions. From Figure 12 it can be seen the rates follow similar
Table 5 presents the model parameters for the surveys. breakage regimes, though there is discrepancy for the rates for
Actual differences in the data from those listed in this table size classes greater than 75 mm. As the feed size to the SAG
are as follows. mill was typically minus 75 mm, thus the breakage function
past this size should be expected to be inaccurate and should
•• The diameter varies throughout the liner life and this
not be used to make comparisons or predictions.
value was taken from the MillMapper outputs.
Issues with the half-life survey which may explain the
•• The fine classifying size for the discharge function was
inconsistent results include:
defined to be approximately 1.1 mm, because this was
the best overall fit for all the surveys and there should •• During the belt cuts conducted for the first survey it
be very little variance between surveys. began to rain heavily, washing away a portion of the
•• The pebble port fraction varied over the period analysed, fine fractions.
so must change in the models to allow for this impact on •• The mill was running on a relatively high filling of
mill hold-up. The fraction used in the model does not 20.7 per cent compared to a typical filling between
represent the true fraction of the open area made up 16–19 per cent.

P Toor et al
FIG 10 – Throughput over liner life for 32 foot case study semi-autogenous grinding (SAG) mill.
Table 3
Effect of key relines on mill performance.
Period Throughput (t/h) % Throughput change Specific energy (kWh/t/h) % Specific energy
Shell reline 6/12/2010–17/12/2010 652 0% 9.6 0%
Pre grate reline 4/03/2011–9/03/2011 720 10% 9.1 -5%
Grate reline
Post grate reline 9/03/2011–14/03/2011 672 3% 9.5 -1%
Pre shell reline 21/05/2011–01/06/2011 732 12% 8.4 -12%
Table 4
Summary of production parameters at survey times.
Liner life Fresh feed t/h Recycle crusher t/h Total t/h Power (kW) Product size p80 (microns)
New 635 156 791 6015 1630
Half (post grate reline) 725 117 842 6338 1140
80% 581 136 717 6085 1373
Fully worn 754 151 905 5861 1395
FIG 11 – JKSimMet model flow diagram.

Table 5
Mill data.
Dimension New Half life 80% worn Fully worn

Int mill dia (m) 9.45 9.5 9.55 9.56
Belly mill len (m) 4.5 4.5 4.5 4.5
Feed trunion dia (m) 2 2 2 2
Feed cone angle (dg) 15 15 15 15
Disch cone angle (dg) 15 15 15 15
Grate size xg (mm) 12 12 12 12
Fine size xm (mm) 0.6 0.4 0.4 0.644
Grate OA fraction 0.068 0.068 0.075 0.0828
Pebble port open fr 0.65 0.291 0.314 0.661
Pebble port size (mm) 70 70 70 70
Wt radius 0.79 0.797 0.79 0.79
Ball load (%) 13.9 13.9 13.6 13.1
Ball top size (mm) 125 125 125 125
10
9
8
7
Breakage Rate
6
5
4
3
2
1
0
0.1 1 10 100 1000
Size, mm
New March May Fully Worn
FIG 12 – Semi-autogenous grinding (SAG) breakage rates liner life 2.
•• The secondary crusher was bypassed during the end approximately 80 per cent of the material entering the mill is
of the survey causing large rock (+106 mm) to enter the experiencing increased breakage.
mill. Typically the SAG mill does not see feed larger Though the data is consistent in the sense that it shows a
than +75 mm. Thus, the feed size distribution from the similar breakage regime and that the critical turning point to
May survey was used.
be at approximately 16 mm the data may still be too noisy
Figure 13 shows the breakage rates for the two liner to allow any meaningful analytical approaches. Thus for
conditions of fully worn and new. The plot illustrates that completeness the new and fully worn liners were taken and
there is an increased breakage rate for the size classes below the half-life and 80 per cent worn life were interpolated to
16 mm for the worn liner. Conversely, the new liner exhibits give the following breakage knots shown in Table 6 and the
an increased breakage rates for particles greater than 16 mm. breakage rates shown in Figure 14, with values greater than
This result can be explained by new liners imparting more 75 mm not shown. These rates give a better representation
lift to the charge thereby causing comparatively high impact
of how the breakage rates progress as the liner wears from
breakage of the coarse particles whereas the worn liner
new to fully worn and this approach is recommended for
causes a greater cascading motion of the charge. Thus, a
any future work as opposed to conducting the intermediate
larger amount of breakage would occur through attrition and
surveys (for example, half worn and 80 per cent worn).
abrasion of the finer material for the worn liner case.
This intersection is a likely explanation to why the mill
experiences an increase in throughput as the liner wears. POSSIBLE INDUSTRIAL APPLICATIONS
This may be due to the SAG feed, which is comprised of fresh One possible method of improving SAG performance through
feed, recycle crusher product and cyclone underflow having a better understanding of liner wear is to design liners for
a P80 in the order of 15–20 mm. Thus, as the liners wear, performance and not life (Toor et al, 2013, 2015).

P Toor et al
FIG 13 – Semi-autogenous grinding (SAG) breakage rates liner life 2 new and fully worn.
Table 6
Breakage knots for new and fully worn liners with intermediates interpolated.
Breakage knots
Knot (mm) New, actual Half life, interpolated 80% worn, interpolated Fully worn, actual
0.25 2.184 2.203 2.207 2.222
4 3.638 3.844 3.885 4.050
16 5.033 5.079 5.088 5.125
44.8 4.273 3.935 3.867 3.596
128 7.396 7.290 7.269 7.184
Plant data, survey results and JKSimMet modelling show Table 7

that key mill performance indicators of throughput, specific Summary of improved performance with shorter life liners.
energy/power and product size all improve as liners wear.
This suggests that liners with a reduced thickness and mass No of relines/ Throughput Power (kW) Specific energy
out perform their larger and heavier counterparts. This 6 month cycle (t/h) (kWh/t)
section highlights the potential benefits of having shorter
1 707 6684 9.47
life liners which deliver more tonnes at a lower power and
produce a finer grind. Importantly, it is envisaged the increase 2 719 6638 9.25
in downtime due to the additional relines is overcome and 3 733 6548 8.93
exceeded due to the increase in mill performance.
4 738 6505 8.81
It is possible to predict the throughput of liners with
reduced mass by analysing existing plant data. For example, 6 736 6463 8.78
to estimate the throughput for a liner which requires a 8 736 6463 8.78
reline every three months (two relines/six month cycle) and
hence resembles the half-life shape we could simply average month cycle would deliver the most tonnes. This equates to a
the throughput from the half-life to the end of liner life. shell reline every two months. However, if relines are taking
Using this approach, it is possible to predict the resulting approximately 60 hours, then there is little advantage in relining
increase in mill performance by having thinner liners from thinner liners more frequently, this highlights the importance
existing plant data. Table 7 summarises the increase in mill of relining efficiency in relation to overall mill performance.
performance with shorter liner lives and additional relines.
It should be noted that periods where the feed rate was
reduced due to a lack of ore supply were excluded from DESIGNING LINERS FOR PERFORMANCE CASE STUDY
analysis alongside any other outliers. As stated earlier the above scenarios are simplified to
Using simplified examples of each reline taking an extra 20, highlight the concepts of shorter liner lives with improved
30, 40 and 50 hours Figure 16 was produced. As can be seen performance. Relining every two months or less may be
in all scenarios the current set-up of a reline every six months impractical for many sites and not an option. The above
is not optimal. At present the shell reline of the 32 foot mill examples also do not take into account the relining of non-
takes approximately 30–40 hours. Interestingly the throughput shell components. Figure 17 outlines the relining schedule
curves for the 30 hour and 40 hour relines indicate that thinner for the main liner pieces which are relined either every six
optimised liners which require a reline three times per six- months or less. Currently the Shell liners are relined every six

10
9
8
7
Breakage Rate 6
5
4
3
2
1
0
0.1 1 10 100 1000
Size, mm
New Liner Half 80% Worn
FIG 14 – Semi-autogenous grinding (SAG) breakage rates for new and fully worn liners with intermediates interpolated.
FIG 15 – Worn liner profiles for design basis.
FIG 16 – Effect of reline shutdown hours on throughput.
months as is the Feed Outer which is staggered with the Shell conclusion that the mill does not require any additional shuts;
Liner. That is when the Shell liner is fully worn the Feed Outer rather that the shutdown may only be required to be extended
is half worn and vice versa. Half a set of discharge grates are to allow for the extra liners to be relined. This makes the
relined every three months as well. concept of using an optimised liner shape with reduced mass
The fact that there is a reline every three months provides a even more appealing. To make comparisons between reline
window to redesign the liner shape to match that provided by duration, the estimation of an extra 15 hours to reline the
MillMapper at half-life (as per Figure 15) and reline the shell Shell and Feed Outer simultaneously is made. The proposed
liner at three month intervals. Importantly this leads to the new reline schedule with half-life shaped liners in shown in

P Toor et al
Feed Discharge Feed Discharge

Months Shell Outer Grates Months Shell Outer Grates
1 half worn 1 half worn
2 2
3 30 3 50
4 4
5 5
6 30 6 30
Total Reline Duration 6 Months = 60 Total Reline Duration 6 Months = 75
7 7
8 8
9 30 9 50
10 10
11 11
12 30 12 30
Total Reline Duration 12 Months = 120 Total Reline Duration 12 Months =150
13 13
14 14
15 30 15 50
16 16
17 17
18 30 half worn 18 30 half worn
Total Reline Duration 18 Months = 180 Total Reline Duration 12 Months =225
FIG 17 – Current reline schedule. FIG 18 – Proposed reline schedule.
Figure 18. Thus, in every six-month cycle the reline of the structure required for the very high new lifter bar and is not
current liner design requires 60 hours, 30 hours for the Feed required for a lower lifter design. Improvements along these
Outer and Grates and 30 hours for the shell and grates. While lines would further decrease the liner cost, furthermore there
for the proposed design the reline duration would be in the is added potential of decreasing the number of liner elements
order of 75 hours: 45 hours relining the Shell, Grates and Feed in line with the capacity of the liner handler. A detailed study
Outer and 30 hours for relining the Shell and Grates only. was conducted by Yahyaei et al (2015) which extended on this
As stated earlier, the fact that the extra reline could be work by incorporating the above recommendations to the
incorporated into a non-shell reline shut and minimise the liner design as well as utilising discrete event simulations to
amount of downtime makes a redesign far more appealing accurately determine new reline times. It yielded a potential
as Figure 18 illustrates, reducing the additional downtime increase in throughput of 3.7 per cent, highlighting the
increases the throughput gained exponentially. Assuming potential of utilising a similar approach of designing liners
that using the profile of half worn liners will only require an for performance not life.
extra 20 hours of reline per six-month cycle a comparison can
be made between the two designs on overall tonnes processed CONCLUSIONS
at the power required. The comparison of the two designs is A methodology for quantifying the influence of liner wear
summarised in Table 8. As it can be seen by using the proposed has been provided. The methodology could be utilised for
design an increase of 35 000 t is predicted or 1.25 per cent with optimising liner design and reline scheduling. An analysis of
a power reduction 226 MWh or 0.85 per cent. fully worn and new liners was carried out as this represented
There is clearly further opportunity to optimise the liner the greatest step change in the mill operating conditions.
profile if a shorter liner life is targeted. Optimisation would Initial results for the mill indicate that the worn mill is better
aim at minimising any wasted shell plate thickness evident at at producing fines at a higher rate, resulting in increased
the end of the liner life, plus changing lifter width to reduce throughput. The increased throughput is achieved at a lower
metal wastage there. The excessive width is a function of the power, possibly due to a decrease in high impact energy
Table 8
Comparison of throughput and power for current and proposed liner design.
Current design Proposed design Proposed design Difference % Difference

(3 months) (6 months)
t/h 707 719 719
Liner life days 165 82.9 164.38
Total throughput (kt) 2801 1418 2836 35 1.25%
Total power (MWh) 26 607 13 191 26 348 -226 -0.85%

collisions occurring in a worn mill due to the reduction in Franke, J, Lichti, D and Stewart, M P, 2006. MillMapper: a new tool
lift. This is based on findings from JKSimMet breakage rates, for grinding mill thickness gauging, in Proceedings International
which show that the worn mill has increased breakage rates Autogenous and Semiautogenous Grinding Technology 2006,
Volume III, pp III-75–III-87 (Department of Mining Engineering,
for particles below 30 mm. The result of worn liners showing
University of British Columbia: Vancouver).
a measurable increase in mill performance (t/h, kW, p80)
begs the question of whether a liner with a reduced profile Herbst, J A, Lo, Y A and Flintoff, B, 2003. Size reduction and
(that is replicating a worn liner) and shorter life outperforms liberation, in Principles of Minerals Processing (eds: K N Han and
M C Fuerstenau), pp 61–115 (Society for Mining, Metallurgy, and
the oversized incumbent. This was indeed the result of the
Exploration: Littleton).
analysis for this case study. A preliminary methodology has
been provided to compare the effectiveness of a reduced Makokha, A B, Moys, M H, Bwalya, M M and Kimera, K, 2006. A new
approach to optimising the life and performance of worn liners
profile liner with the incumbent design. This methodology
in ball mills: experimental study and DEM simulation, Minerals
can be easily used as a first pass analysis to validate the Engineering, pp 1439–1445.
approach followed by a more detailed analysis incorporating
discrete event simulations, liner redesign and DEM. Norgate, T and Haque, N, 2010. Energy and greenhouse gas impacts
of mining and mineral processing operations, Journal of Cleaner
Production, pp 266–274.
ACKNOWLEDGEMENTS Powell, M S and Verneulen, L A, 1993. The influence of liner design
The authors would like to acknowledge the support and co- on the rate of production of fines in a rotary mill, Minerals
operation of site personnel. Financial support for this project Engineering, pp 169–183.
was provided by AusIndustry through the Climate Ready Toor, P, Franke, J, Powell, M P, Bird, M and Waters, T, 2013. Designing
grant scheme. liners for performance not life, Minerals Engineering, pp 22–28,
pp 43–44.
REFERENCES Toor, P, Powell, M, Hilden, M and Weerasekera, N, 2015.
Franke, J, 2008. MillMapper experiences – a mill condition Understanding the effects of liner wear on sag mill performance,
monitoring and operational improvement case study, in in Proceedings Metplant 2015, pp 150–161 (The Australasian
Proceedings MetPlant 2008 pp 65–80 (The Australasian Institute Institute of Mining and Metallurgy: Melbourne).
of Mining and Metallurgy: Melbourne). Yahyaei, M, Powell, M P, Toor, P, Tuxford, A and Limpus, A,
2015. Relining efficiency and liner design for improved plant
performance, Minerals Engineering, 83:64–77.

Contents
Development of the Fosterville

gold mine heated leach process
M Binks1 and P Wemyss2
ABSTRACT
The Fosterville gold mine is located approximately 20 km east of Bendigo in central
Victoria. The gold at Fosterville predominantly occurs as solid solution within
disseminated arsenopyrite and pyrite grains. In recent times, more of the gold has
been mined from quartz-carbonate veins. The Fosterville orebodies contain various
amounts of native carbon in the form of bituminous coal. This carbon (non-carbonate
carbon or NCC) has been the predominant mechanism for gold loss from the
processing facility through ‘preg-robbing’.
Processing of the Fosterville ore begins with crushing, grinding and flotation to
extract the sulfides. The sulfide concentrate is oxidised using bacterial oxidation, before
being leached in a conventional carbon-in-leach (CIL) circuit. Due to an increased
proportion of coarse gold in the mill feed, a gravity circuit was added in early 2016
to the flotation concentrate regrind circuit to recover free gold before reporting to the
bacterial oxidation circuit. The gravity recovered gold is direct smelted.
A high portion of the native carbon (NCC) in the mine ore is naturally hydrophobic
in nature, and subsequently reports to the flotation concentrate stream and ultimately
into the CIL circuit. This NCC has a notable preg-robbing ability. Treatment of black
shale ores, which have elevated NCC levels, has historically resulted in CIL recoveries
as low as 35 per cent, with around 60 to 80 per cent of the gold loss from the leach
circuit attributed to preg-robbing.
Standard technologies for dealing with a carbonaceous leach feed was trialled
with limited success. The discovery of the significance of heat on leach recoveries,
triggered extensive test work on a range of leach feed and tails samples. Pilot plant
testing demonstrated that ‘heated leaching’ of the CIL tailings was the best practical
process with an average recovery increase of 7.5 per cent being achieved.
Following a successful pilot study, an engineering feasibility study was completed
by Minerva Engineering, and the project economics determined. With strong
recovery gains evident and a predicted project payback of one to 1.5 years, approval
for the installation of the full-scale plant was granted. Detailed design commenced
in September 2008 with installation commencing in January 2009. The circuit was
successfully commissioned in April 2009 and payback achieved in 1.5 years
Following commissioning, the heated leach circuit has achieved recovery gains of
four to 14 per cent, and has proven itself to be a significant contributor to the overall
plant performance. The heated leach process was also successfully applied to treating
samples of the historic CIL tails. A repulping circuit was installed to allow retreatment
of the CIL tails through the heated leach circuit, capitalising on any spare capacity.
INTRODUCTION
The Fosterville gold mine was developed by Perseverance Pty Ltd. Ownership of the
mine has changed several times over the past 11 years of operation and is currently
wholly owned by Kirkland Lake Gold, a Canadian based mining company. Mining
at Fosterville has taken place intermittently since 1894. Contemporary exploration
and heap leach operations commenced in the 1980s and up to 2001 produced a total
of 240 k troy ounces (oz) of gold from these operations. Following the completion
of a successful deeper drilling program in 2001/02, a detailed feasibility study was
1. MAusIMM, Metallurgy Consultant, formerly undertaken based on mining and processing refractory ore at a nominal rate of 800 k
Processing Manager, Fosterville Gold Mine, metric t/a. Engineering for the ore processing plant commenced in November 2003
Fosterville Vic 3557.
with plant construction commencing in March 2004. Installation was completed and
Email: marcus.binks@bigpond.com
the first sulfide gold bar was poured in May 2005.
2. Processing Manager, Centennial Mining,
formerly Processing Manager at Stawell The Fosterville mineralisation occurs within Lower Ordovician sediments comprising
Gold Mine. of interbedded sandstones, siltstones and shales. The predominant feature in the
Email: peter.wemyss@centennialmining.com area is the Fosterville Fault System, a north–south striking, steep westerly dipping,
523
reverse fault comprising of numerous subparallel faults. Table 1

Gold is typically located in disseminated arsenopyrite and Carbon-in-leach (CIL) tail diagnostic analysis.
pyrite forming a halo to veins in a quartz-carbonate veinlet
stockwork, which is in turn controlled by late brittle faults. Diagnostic Average
The arsenopyrite occurs as fine-grained acicular needles with Carbon-in-leach (CIL) gold feed grade (g/t) 58.5
no preferred orientation. The disseminated pyrite associated
with gold mineralisation occurs as crystalline pyritohedrons Carbon-in-leach (CIL) tail grade gold grade (g/t) 8.9
(Hitchman, Holland and Evans, 2008). Preg-robbed gold (%) 61.7
More recently, the mine has intersected later stage quartz- Native carbon (non-carbonate carbon) (%) 1.2
carbonate veins which carry a notable amount of coarse Gold grade (g/t) 5.5
gold. Massive stibnite veining also periodically occurs
in conjunction with the quartz-carbonate all of which is Sulfide locked gold (%) 35.7
considered to be related to the later-stage geological event. Sulfide sulfur (%) 0.9
The Fosterville orebodies contain various amounts of Silicate locked gold (%) 2.5
native carbon in the form of bituminous coal. This carbon
(referred to as non-carbonate carbon or NCC) occurs through
hydrothermal alteration and as a significant sedimentary Table 2
structure along the Fosterville fault line. The carbonaceous Average elemental and weight distributions within the CIL tails stream.
minerals in the ore are the predominant mechanism for gold
loss from the processing facility through ‘preg-robbing’. Size Carbon-in-leach (CIL) tail distributions
Processing of the Fosterville ore is achieved through a Wt (%) Au (g/t) S2- (%) Non-carbonate carbon (%)
simple single stage jaw crushing circuit followed by a semi-
autogenous grinding (SAG) mill. The sulfides are separated +45 micron 12 14 47 3
using a flash flotation cell and a three-stage secondary flotation +20 micron 8 10 35 2
circuit. Due to a recent mineralogical change in the mill feed, a +9 micron 8 4 1 5
gravity circuit consisting of a Knelson concentrator and Gemini
tables, was added to the circuit in early 2016 to recover coarse +2 micron 30 26 1 46
gold from the flotation concentrate regrind circuit. <2 micron 42 46 16 44
The reground sulfide concentrate is oxidised using bacterial Total (%) 100 100 100 100
oxidation, with the residue washed through a counter current
decant (CCD) circuit to remove solubilised iron and sulfur Analysis of the size by assay showed that on average
species, before being leached in a conventional CIL circuit. 70 per cent of the gold and 90 per cent of the native carbon is
The tails from the CIL circuit are heated (heated leach) to below 9 micron in size. The fine nature of the gold and native
recover a portion of the preg-robbed gold, before being carbon limits the ability to study the surface of the mineral
pumped to a CIL residue storage dam. The flotation residue particles to more firmly establish the gold/NCC association.
and the neutralised liquor from the CCD circuit are combined The gold/NCC association is subsequently indirectly
and pumped to either an in-pit, or surface paddock style determined through diagnostic test work.
residue storage facility.
Figure 1 is a plot of the gold and NCC levels obtained
A high portion of the native carbon in the mine ore is from composite samples taken over a two year period. It can
naturally hydrophobic in nature, and subsequently reports be clearly seen that a reasonably strong correlation exists
to the flotation concentrate stream and ultimately into the between the level of native carbon and gold within the CIL
CIL circuit. This NCC has been shown to have a high a preg- residue stream.
robbing ability, with gold loss of six to seven per cent from
even the cleanest of ores. Treatment of black shale ores, which
have elevated NCC levels, has historically resulted in CIL
COMBATING ‘PREG-ROBBING’
recoveries as low as 35 per cent, with around 70–80 per cent of During the prefeasibility study for the Fosterville project, the
the gold loss from the leach circuit attributed to preg-robbing presence of native carbon and losses through preg-robbing
(Wemyss, 2007). were identified, but the degree to which the orebodies
contained native carbon was underestimated. During the
prefeasibility, initial tests on the carbonaceous samples
DEFINING FOSTERVILLE ‘PREG-ROBBING’ used the standard method of kerosene blanking in a bid to
Both plant and laboratory leaching demonstrate that the leach address the prerobbing loss, however this proved relatively
rate of the gold in oxidised Fosterville ores is extremely fast. unsuccessful.
Kinetic leach testing, however, demonstrates a continued Following commissioning of the Fosterville processing
drop in solid grades over a 48 hour period and indicates a facility, CIL gold recoveries struggled to achieve the levels
steady release (desorption) of weakly preg-robbed gold back predicted in the prefeasibility study. Although leach gold
into solution for adsorption onto activated carbon. The rate of losses were initially exacerbated by low oxidation levels, the
gold desorption from the preg-robbing NCC is, as expected, loss of gold to preg-robbing was identified as the key recovery
solution grade and time dependent. influence.
Diagnostic analysis (Table 1) and size by size assays A myriad of test work was conducted in a bid to establish a
(Table 2) provides an indication as to which size fraction the treatment method, including but not limited to:
gold reports, and the mechanism by which it is reporting. The
•• elevated carbon concentrations
historical average level of preg-robbed gold in CIL tails is
approximately 62 per cent, with 36 per cent locked in sulfides •• low density leaching
and the balance in silicates or precipitates. •• kerosene blanking

Development of the Fosterville gold mine Heated Leach process
FIG 1 – Carbon-in-leach (CIL) tail gold grade versus per cent non-carbonate carbon (NCC).
•• synthetic blanking subsequently indicates that the ‘preg-robbing’ is initiated in

•• pressure oxidation the bacterial oxidation circuit.
•• thiosulfate leaching This understanding moved the focus of test work to
reversing or breaking the gold/NCC association developed
•• chlorination
within the bacterial oxidation circuit. The application of heat
•• thiourea leaching. to the pulp was tested as a method to break the gold/NCC
Outside of the treatment of the leach feed to combat preg- bond. Initial leach temperatures of 40°C were adopted with
robbing, other tests focused on the flotation circuit and the view to using waste heat from the bacterial oxidation
the rejection of native carbon at this stage of the process. circuit which operates at a similar temperature. ‘Heated leach’
Chemical suppression was unsuccessfully trialled several tests conducted on CIL feed at 40°C immediately showed
times in the plant. Through the trials however, the use of benefits with a step increase in CIL recoveries over both plant
Guardisperse (naphthalene sulfonate) was found to assist in and previous test work results and provided the foundation
the management of tenacious carbonaceous froth experienced for further work.
during the blending of black shale ores.
Due to the fine nature of the native carbon and its lower HEATED LEACH PILOT PLANT STUDIES
density relative to the sulfide minerals, physical separation To fully evaluate the heated leach process, establish a potential
from the secondary circuit flotation concentrate was processing circuit flow sheet and to enable identification of
assessed. The use of 1 inch Mozley cyclones was evaluated any idiosyncrasies of operating such a circuit, a pilot plant was
(Binks, 2006) and ultimately applied to remove fine native constructed. Como Engineering of Perth, Western Australia,
carbon from the stream. This project proved successful in its were engaged to construct the plant based on site design.
own right with 50 to 60 per cent carbon rejection achieved The pilot plant (Figure 2) consisted of two trains of six tanks.
with an associated three to five per cent gold loss from the Each train was different in volume to enable a comparison of
flotation circuit. The cleaned cyclone underflow produces residence time and to conduct two tests simultaneously.
vastly improved leach recoveries.
Heating of the pulp was achieved by circulating hot water
Ore containing high levels of NCC are currently stockpiled through jackets around the tanks. Each tank was fitted with a
and batched through the plant. During the batch treatment, basket to retain the carbon and allow for easy carbon transfer.
the Mozley cyclones were operated to reject the native carbon. The circuit was also setup with chemical, oxygen and air
The operation of the cyclones not only reduces the demand on
dosing to the tanks.
the heated leach circuit but also reduces the level of surface
foam created in the bacterial oxidation tanks. Standard carbon management in the pilot plant was designed
around having 20 g/L carbon concentration in each tank, with
Roasting is clearly the best metallurgical method for the
basket movements in each train timed to approximate carbon
treatment of Fosterville flotation concentrates; however,
movement rates and dwell times achieved in the existing
the location of the site in a highly populated area, coupled
CIL plant. Free cyanide levels were kept at historical plant
with the local weather conditions meant that the process was
levels and pH was kept in line with the plant. Any variation
considered environmentally unsound and therefore not a
from the above was performed specifically to test different
viable proposition.
treatment scenarios.
Sampling of streams was performed every six or 12 hours
A LEFT-FIELD APPROACH in order to give good variation around the nine hour and
Standard technologies and known methods for dealing with 18 hour basket movement regime. This way assay results
the Fosterville carbonaceous leach feed have been trialled were available from the start, middle and end of basket dwell
with limited success. The lack of success from laboratory tests duration in order to determine variation effects.
provided evidence that a portion of the gold must already be
The heated leach pilot plant was operated for a period of
associated with the native carbon prior to entering the leach
four months on a continuous roster to evaluate the effects of:
circuit. Mozley cyclone classification test work on flotation
concentrate (Binks, 2006) demonstrated that a gold/carbon •• heated leach process on CIL feed
association is not present within the flotation concentrate and •• heated leach process on CIL tails

The detailed design identified the necessary cooling of

the secondary stages (tanks 4–6) could be achieved through
simple aeration of the tanks. However, this cooling method
proved unsuccessful during the commissioning phase and
was up until recently, replaced by the use of cold recycled
water from the CIL residue dam to achieve the temperature
adjustment, at the sacrifice of density and tank residence time.
The addition of a new slurry / slurry heat exchanger has since
negated the need for the addition of cooling water.
In stages 1 to 3, recessed impeller pumps are used for carbon
transfer to prevent unnecessary cooling of the pulp that would
occur if using airlifts. Airlifts are used for carbon transfer in
stages 4 to 6 where lower operating temperatures are required.
In all cases, the carbon advance slurry is screened to allow
the carbon only to progress forward and the slurry to return
to the tank from which it originated. Screening of the carbon
is also important for preventing the flow of cold pulp being
transferred from stage 4 into stage 3 which is running at 60°C.
Barren carbon is continuously added to both stages 1 and
6 in the circuit. The loaded carbon from the heated leach
circuit is recovered to a carbon column, regenerated and
FIG 2 – Heated leach pilot plant. then returned to the CIL circuit. Figure 3 depicts the carbon
movement at nominal rates and grades.
•• various temperature profiles
Following commissioning, a stable foam was evident on
•• carbon addition to the circuit including location, activity
the first three tanks which appeared to originate from bubble
and rate
creation following the steam injection. Control of the foam was
•• cyanide addition necessary to manage the circuit, subsequently an antifoam
•• caustic addition. was dosed to the circuit feed pump to bring the foam levels
Pilot plant testing clearly demonstrated that heated leaching under control.
of the tailings from the CIL plant was the best method The circuit operates continuously with an availability of
for increasing overall gold recovery from the ore, with an 95 to 98 per cent. In 2010, the first full calendar year post-
average recovery increase of 7.5 per cent being achieved. The commissioning, the heated leach circuit (Figure 4) provided
pilot work also identified that the recovery increase through an average recovery gain of 7.4 per cent. This was deemed
heated leaching is possible without any further additions of an excellent result particularly given the level of circuit
cyanide, and can be performed in the absence of air or oxygen downtime and reduced efficiency associated with poor
injection into the pulp (Wemyss, 2007). availability of the steam boiler at the time. During the year, a
Pilot studies were later extended to evaluate the ability of peak monthly gold recovery gain of 14 per cent was achieved
the process to recover gold from historic leach tails which during the processing of a blend containing high NCC levels.
had an average gold grade of 10 g/t. The test work indicated Through changes in the ore sources and mineralogy, NCC
that 35 to 50 per cent of the gold within the tail residue could levels and in some instances the activity of the NCC has
be recovered, further supporting the implementation of the declined considerably from when the heated leach circuit was
process. Additional capacity was factored into the design to first commissioned. NCC levels in the mill feed are currently
allow for reclaimed tails to be processed concurrently. around 0.2 per cent. When coupled with batching and Mozley
Board approval for the project was received in October 2008. cyclone operation on the high NCC ores, the leach circuit gold
Minerva Engineering of Melbourne were engaged to complete recovery gain through the heated leach circuit is subsequently
the detailed design. Site construction commenced in January at a lower average of around 4 per cent.
2009 with commissioning underway by April of that year.
HEAT RECOVERY
FULL-SCALE OPERATION The greatest unit cost for the heated leach process is the
The six stages used during the pilot plant studies on the CIL slurry heating. Opportunities to save costs through heat
tail were ultimately applied to the full-scale plant. The initial recovery were considered during the initial design however,
three stages of the heated leach circuit are operated up to 70°C they were not implemented due to capital constraints and
with the remaining three stages cooled to aid the adsorption the focus being on the main prize of gold recovery. Several
of gold from solution. slurry/slurry heater exchangers were evaluated from a cost
Tails from the CIL circuit gravitate to a pump hopper. perspective, including spiral, plate and tube units. The risk
The CIL tails slurry is pumped through a slurry/slurry heat of scaling was considered and formed part of the evaluation
exchanger for preheating, taking the temperature from 20°C process. The use of site designed simple straight-through
to approximately 40°C. Final heating of the CIL tails stream to tube heat exchangers were initially adopted based on price,
nominally 65°C is achieved through inline injection of steam availability and accessibility for scale cleaning.
which is provided by a 4 megawatt liquefied petroleum gas The tube heat exchangers were installed to recover the heat
(LPG) fired boiler. Tanks on the first three stages are rubber from stage 3 which needed to be cooled to facilitate gold
lined and insulated to reduce radiant heat loss. Stainless adsorption onto the plant carbon. Slurry, post the carbon safety
steel lids are fitted to four of the six stages to retain heat and screen, was pumped through the tubes where it was contacted
reduce steam entering the operator working zone on the top with the counter current flowing new feed. The tubes were able
of the tanks. to exchange approximately 10°C between the new feed and

Development of the Fosterville gold mine Heated Leach process
FIG 3 – Carbon movement through the heated leach and carbon-in-leach (CIL) circuits.
FIG 4 – Fosterville gold mine heated leach circuit.
the tank 3 discharge slurry. The original tube heat exchanger was successful and subsequently a tails repulping circuit was
was later replaced with spiral tubes in a bid to increase the constructed in 2010. The repulping plant consists of a single
level of heat exchange. Fouling of the tubes and the inability 11 m3 tank fitted with two 30 kW agitators. Reclaimed water
to remove the centre tube following fouling with scale, meant is injected into the base of the tank for dilution while the semi-
that an alternate option was required. Recently, a spiral heat dry tails are fed by front end loader directly into the tank. The
exchanger, manufactured by Tranter (Figure 5), a German overflow of the tank is screened to remove any tails lumps
engineering company, was installed. The heat exchanger and tramp oversize before reporting to a surge tank. The
consists of two counter current spirals which have a surface repulped tails are combined with the plant CIL tails prior to
area of 75 m2. The heat exchanger is providing a nominal 20°C treatment through the heated leach circuit.
temperature exchange which translates into a 40 to 50 per cent
reduction in LPG consumption per tonne of heated leach feed.
ADDED BENEFITS
Not only has the heat exchanger significantly reduced the
heating demand, it has cooled the tank 3 discharge slurry The heated leach circuit was efficient at destroying any free
enough to negate the need for the addition of cooling water in cyanide leaving the CIL circuit and notably reduced the
stages 4–6. The unit is designed to allow access to the spirals weak and dissociable (WAD) cyanide levels. These benefits
for cleaning by removal of a bolt on the cover plate. One spiral were identified during the pilot phase, and confirmed during
is accessed from one side, while the other spiral is accessed operation. The performance was substantial enough to replace
from the opposite side. Cleaning of the spirals is currently the existing cyanide destruction circuit which utilised ferric
monthly to maintain optimum heat exchange efficiency. sulfate for the precipitation of free cyanide and dissociation of
the WAD cyanide species.
TAILS RETREAT
With the accumulation of a considerable amount of high-grade ENHANCING THE TECHNOLOGY
CIL tails, test work was completed to confirm the suitability In 2016, Fosterville Gold Mine sold the technology rights of
for the heated leach circuit to recover gold. The test work the heated leach process to Outotec. The process has now been

CONCLUSION
The heated leach circuit remains the most notable addition to
the Fosterville processing facility that has been made since the
operation of the plant commenced. The heated leach circuit
continues to contribute significantly to process economics
and the long-term future of the site. The installation of a
tails retreatment circuit has enabled the spare capacity of
the heated leach circuit to be utilised as required, increasing
gold production in the process. The recent introduction of a
spiral heat exchanger has significantly reduced the operating
costs and in turn, enhances the circuits overall economic
contribution. The recent acquisition and rebranding of
the process by Outotec will enhance the exposure of the
technology for the benefit of others.
ACKNOWLEDGEMENTS
The authors would like to acknowledge Northgate Minerals
for the support and confidence in firstly piloting the process
and then establishing a full-scale operation. The metallurgy
staff at Kirkland Lake Gold must also be acknowledged for
providing updated operating information.
FIG 5 – Tranter spiral heat exchanger. REFERENCES

Binks, M, 2006. Cyclone classification report February 2006,
renamed as HiTeCC, high temperature caustic conditioning. Fosterville Gold Mine.
The adoption of the HiTeCC process by Outotec will aid in
Hitchman, S, Holland, I and Evans, B, 2008. Technical report on
the optimisation of the process and is complimentary to their
Fosterville gold mine, Victoria Australia.
existing suite of other processing technologies. It will also
significantly boost its exposure and makes it more readily Wemyss, P, 2007. CIL tail heated leaching, Fosterville gold mine.
available to other end users.

Contents
Collector – addition point and consumption

C J Greet1, W J Bruckard2 and D MacKay3
ABSTRACT
One of the mining world’s greatest challenges is to effectively use consumables (that
is, reagents and grinding media) to maximise metallurgical response, while minimising
operating costs. This paper provides an example of a logical approach to achieve this.
The current study at Perilya’s Broken Hill concentrator in the lead primary grinding/
roughing circuits examines the dependency of collector adsorption on Eh. The data
demonstrates that the addition of a xanthate collector during grinding can improve
galena flotation. Further, operating the ball mill under a superior grinding chemistry
regime can lead to significant reductions in collector consumption while maintaining
or improving the metallurgical results.
INTRODUCTION
In a grinding media/sulfide mineral system, due to the differences in rest potentials,
complex galvanic cells are set-up where the sulfide mineral can act as an anode or
cathode, depending on its contact with other sulfide minerals, media and reagents
(Cheng, Smith and Iwasaki, 1993). Generally, the forged steel balls act as anodes
relative to the sulfide mineral cathodes:
Cathode (sulfide mineral): 1 O + H O + 2e " 2OH

- -
(1)
2 2 2
Anode (grinding media): Fe " Fe 2+ + 2e (2)

-
The iron ions dissolved from the grinding media bond with the hydroxyl species to
form iron oxy-hydroxy species which are hydrophilic and deposit indiscriminately on
the sulfide mineral surfaces. This surface contamination can affect the floatability of
the valuable sulfide minerals in two ways. Firstly, by the deposition of a hydrophilic
species on the surface of the particle, thereby decreasing its hydrophobicity and
secondly by hindering effective collector adsorption. Adam, Natarajan and Iwasaki
(1984) noted that in a forged steel media/sulfide mineral system the mixed potential
of the pulp is generally very negative, and under such reducing conditions xanthate
adsorption is inefficient. Guy and Trahar (1984) demonstrated clearly that galena
flotation rate and recovery are dependent on pulp Eh, with a flotation edge at
about 50 mV, and strong flotation up to 350 mV (Figure 1). It is suggested that this
range of pulp potentials corresponds to a region of stability for lead ethyl xanthate,
which provides the mineral with its hydrophobicity. It would appear that xanthate
adsorption is dependent on pulp potential.
Xanthate additions in plants are usually made in a relatively arbitrary fashion, based
largely on experience, and are often made to the grinding as well as to the flotation
circuits. The major reasons cited for splitting the collector dose between the grinding
and flotation circuits are that this facilitates collector adsorption when new surfaces
are being cleaved or fractured, and to promote the dispersion of oily collectors which
may not be completely water soluble.
Unfortunately, the metallurgical benefits of such reagent splitting are rarely
quantified and excess reagent dosing often results. To confirm any benefit requires
not only a clearer understanding of the complex reagent-mineral-media interactions
occurring during milling in terms of the pulp chemistry, but also a means by which
reagent levels and other key chemical parameters can be monitored online. Advances
1. FAusIMM(CP), Manager Metallurgy – in the research into this area are gaining momentum (Greet et al, 2003; Greet, Kinal
Minerals Processing Research, Magotteaux
and Steinier, 2005; Greet and Steinier, 2005) and need to be based on laboratory as
Australia Pty Ltd, West Perth WA 6005.
Email: christopher.greet@magotteaux.com
well as plant studies such as this one.
2. MAusIMM, Project Manager, CSIRO Minerals, An opportunity to test these options at plant scale became available when Perilya’s
Clayton South Vic 3169. Broken Hill concentrator was converting from forged steel to high chrome media. To
Email: warren.bruckard@csiro.com maximise the potential benefits of grinding with an inert media Perilya had decided
3. Formerly Metallurgical Superintendent, to dump their eight ball mills (three primary and three secondary ball mills in the
Perilya Broken Hill Operations, Broken Hill primary grinding circuit and two in zinc regrinding duties), and load them with
NSW 2880. graded charges of high chrome grinding media. This process, which occurred over
529
(Watters, 1993). All three parallel lines are ‘identical’, with

the following configuration. Ore from the fine ore bin feeds
a rod mill. The rod mill discharge flows in to a sump where
it combines with the primary ball mill discharge, and is
pumped to the primary cyclone bank. The primary cyclone
underflow feeds the primary ball mill, and the overflow
gravity feeds to the lead primary rougher flotation cells. The
lead primary rougher tailing discharges into a sump and
combines with the secondary ball mill discharge. This slurry
is pumped to the secondary cyclone pack. The secondary
cyclone underflow reports to the secondary ball mill feed,
while the secondary cyclone overflow (from the three
lines) feeds lead secondary rougher/scavenger flotation. A
simplified flow sheet is given in Figure 2.
The current experimental work was completed in Line 1
and 2 primary ball mill/primary lead rougher flotation banks.
During the test period Line 1 primary ball mill was operated
with forged steel grinding media, while Line 2 was loaded
with high chrome.
FIG 1 – The relationship between galena recovery after one minute and
pulp potential at pH 8 in the presence of xanthate and alcohol frother.
The galena was ground in a steel mill (Guy and Trahar, 1984) with a
EXPERIMENTAL
collector dosage of 20 grams per tonne added to the flotation cell. Plant conditions
several shutdowns, presented a scenario whereby Line 1 was During these tests, the feed rates to the rod mills of both
loaded with forged steel media, and Line 2 operated with Lines 1 and 2 were maintained at 110 t/h. The mills were
high chrome grinding media. This situation would allow the operated steadily during the tests. The primary cyclone
authors the opportunity to collect data examining the impact overflow particle size distribution was nominally the same
of collector addition on galena flotation prepared using for both lines, recording a P80 of 300 microns.
different grinding media types and to investigate the impact Before commencing the test program the froth depth for
of collector addition point location on galena flotation. the lead primary rougher flotation cells was altered such that
The Commonwealth Scientific and Industrial Research it was maintained at 100 mm for all tests for the duration of
Organisation, Australia (CSIRO) have developed an ultra the program.
violet (UV) detection system (Hao et al, 2008), which allows In the first series of experiments the collector addition
the levels of species such as xanthate and cyanide in a mineral (sodium ethyl xanthate) was to the primary rougher feed
pulp to be measured continuously. It was decided to utilise box. The collector addition was physically measured, and
the CSIRO’s UV detection system in the investigation to maintained during each sampling period. At the completion
provide additional information on the pulp chemistry. of sampling the collector dosage rate was changed to a new
level and the process repeated.
PERILYA BROKEN HILL MINE In the second series of tests the collector addition point
was changed to the primary ball mill feed. These data were
Process flow sheet compared to the previous tests completed where the collector
Perilya’s Broken Hill concentrator (previously known as New was added to the lead primary rougher feed box. In this work,
Broken Hill Consolidated, NBHC) is described elsewhere the collector dosage was kept constant.
Grinding Line1
Pri
Sec Pb Rghr1 Sec Pb Rghr2 Pb Scav
Pb Rghr
Pri Sec
Rod Mill
Ball Mill Ball Mill
De-Leaded
Tailing
Grinding Line2
Pri
Pb Rghr Pb Clnr
Pri Sec
Rod Mill
Ball Mill Ball Mill
Grinding Line3
Pri
Pb Rghr
Pri Sec
Rod Mill Pb ReClnr
Ball Mill Ball Mill
Pb Con
FIG 2 – A simplified flow sheet of Perilya’s Broken Hill lead/zinc concentrator.

Pulp chemical surveys

A preliminary survey of Lines 1 and 2 was completed to
determine if there were measurable differences in the pulp
chemistry of the ore ground with forged steel and high
chrome grinding media. The collector addition during these
surveys was 24 grams per tonne of xanthate.
For each further test condition, pulp chemical surveys
measuring the Eh, pH, dissolved oxygen and the
Ethylenediaminetetraacetic acid (EDTA) extractable iron
were completed. These data were collected from the primary
ball mill discharge and the lead primary rougher tailing
process streams only. FIG 3 – Schematic diagram of CSIRO online reagent monitoring system. It consists
of an online filter, a pump and ultra violet spectrometer. Signal (spectra) is
Metallurgical surveys displayed directly on a screen and data recorded/logged for further processing.
For each test condition a metallurgical survey was conducted,
in laboratory flotation cells (Hao et al, 2008). Xanthate shows
where samples of lead primary rougher feed, concentrate
strong absorption at 301 nm, as shown in Figure 4. Generally,
and tailing were collected. For each metallurgical survey, a
the absorption background for most flotation pulps is very
composite sample of each process stream was prepared by
low, so that xanthate in the pulp can be measured directly
‘cutting’ each sampling point three times over a 20 minute
period. As the lead primary rougher only consists of two with a UV-V is spectrometer after filtration (Sullivan and
flotation cells, the nominal residence time in this part of the Woodcock, 1973).
circuit is seven minutes. So, the sampling period corresponds The filter head was placed in the tailing box of the lead
to approximately three times the residence time. Therefore, primary rougher, and solution extracted directly from the
the sampling regime was considered to be adequate. industrial cell. The solution was then pumped through the UV
The survey samples were weighed wet and dry, prepared detector. The resultant UV absorbance data was displayed on
and submitted for assay in Perilya’s onsite chemical laboratory. a computer monitor in the plant.
Each sample was assayed for: silver, copper, lead, zinc and To determine if the differences in collector adsorption can
iron. The head assays for the 12 surveys completed are listed in be detected when grinding with an electrochemically reactive
Table 1. While there is some variation, as would be expected in grinding media (forged steel) and an inert media (high
an operating plant, the feed grade is reasonably consistent for chrome), a series of experiments were completed using the
most surveys. Although surveys 10 and 11 have a noticeably CSIRO’s online UV monitoring system.
higher feed grade compared with the other surveys.
A multiple element mass balance using MATBAL Version 8.0 RESULTS
was complete for each metallurgical survey.
It is recognised that the data presented is not comprehensive, Preliminary pulp chemical surveys
and had time permitted more surveys would have been The Eh, dissolved oxygen and percentage EDTA extractable
completed at each collector addition rate to determine the iron profiles through the two circuits are presented in Figure 5.
statistical significant on the metallurgical performance.
The Eh profiles for Line 1 and 2 (Figure 5a) demonstrate
CSIRO online UV monitoring equipment clearly that the pulp potential throughout the circuit for
the forged steel line was significantly more reducing than
The CSIRO online reagent monitoring system consists of an in
that recorded for Line 2 (high chrome grinding media). In
line filter, a pump, and a UV-V is spectrometer equipped with
particular, the primary ball mill discharge for Line 1 was
a flow cell (Figure 3). This monitoring system was originally
developed for online measurement of xanthate concentrations -270 mV (SHE) compared with +80 mV (SHE) for the mill
charged with high chrome media in Line 2.
Table 1 The dissolved oxygen profiles, pictured in Figure 5b, follow
Lead primary rougher feed grades from each metallurgical survey. similar trends to those noted for the Eh. Broadly speaking, the
dissolved oxygen content of pulps prepared with forged steel
Survey Assay are considerably lower than those reported for the high chrome
Ag, ppm Cu, % Pb, % Zn, % Fe, % case. Again, focusing on the primary ball mill discharge, the
1 37 0.12 3.7 5.8 3.8

2 44 0.14 4.5 6.4 4.1
3 43 0.14 4.6 6.6 4.1
4 37 0.13 4.1 6.1 3.9
5 40 0.15 4.3 7.2 4.3
6 38 0.14 4.4 6.4 4.3
7 82 0.16 4.6 7.4 4.2
8 44 0.14 4.2 6.3 3.9
9 61 0.14 4.0 6.6 3.9
10 107 0.17 5.2 7.0 3.8
11 117 0.15 5.2 6.9 3.6
12 55 0.14 4.2 6.3 3.9
Average 59 0.14 4.4 6.6 4.0 FIG 4 – Ultra violet spectrum of xanthate (0.05 mM EX-).

oxygen content in the pulp, and low Eh values (Figure 5). By

changing from forged steel to high chrome grinding media,
the corrosion mechanisms at play change dramatically, such
that the media’s corrosion resistance increase as the chrome
content increases. Therefore, as the corrosion component
of media wear decreases the levels of EDTA extractable
iron decrease. This in turn causes the dissolved oxygen to
increase and the Eh of the pulp to shift to more oxidising
values. Thus, by changing to a more inert grinding media the
system becomes cleaner (ie a reduction in the levels of iron
oxy-hydroxy species), which should result in more efficient
collector adsorption and an improvement in metallurgical
A response.
Xanthate addition to lead primary rougher feed box

Pulp chemistry
Table 2 contains the pulp chemical data for the primary ball
mill discharge for both Lines 1 and 2, collected over a range
of xanthate addition rates. It is apparent that each parameter
varies in magnitude with time. For example, the Eh of Line 1
primary ball mill discharge shifted from -78 to -222 mV (SHE)
with time. These changes can be linked to variation in the
mineralogy of the feed entering the plant.
B
Not withstanding this observation, the key point to be
emphasized from this data is that Line 1 primary ball
mill discharge had consistently more reducing Eh values,
contained considerably less dissolved oxygen and more
EDTA extractable iron than Line 2. Therefore, one would
expect that the sulfide mineral surfaces prepared in Line 1
would have a higher proportion of grinding media corrosion
C products contaminating their surfaces than Line 2. Further, it
is likely that this surface contamination would interfere with
xanthate adsorption.
UV monitoring
Figures 6 and 7 contain plots of the xanthate addition to the
lead primary rougher and residual xanthate measured using
the online UV monitor in the lead primary rougher tailing,
FIG 5 – Pulp chemical profiles for: (A) Eh; (B) dissolved oxygen; and for Line 1 (forged steel) and 2 (high chrome), respectively.
The first observation to be made from these plots is that using
(C) EDTA extractable iron through the Broken Hill grinding circuit.
this technique it was possible to measure comparatively low
Collector dosage for these surveys was 24 grams per tonne. The levels of ‘free’ xanthate in solution. For example, the UV
collector was added to the lead primary rougher feed box. monitor was able to measure down to 1.5 grams per tonne of
‘free’ xanthate in solution. The second significant observation
pulp leaving the mill in Line 1 recorded a dissolved oxygen to be made is that as the xanthate addition rate changed, the
concentration very close to zero, while Line 2 was 1.4 ppm.
The EDTA extractable iron data (Figure 5c) shows that the Table 2
amount of grinding media corrosion products within the pulp The primary ball mill discharge pulp chemical readings from Line 1 (forged
was higher for the forged steel case than for the high chrome steel) and Line 2 (high chrome) over a range of xanthate addition rates.
alloy. This indicated that there is less iron hydroxide species
derived from the grinding media when the high chrome alloy Xanthate addition, g/t Pulp chemical parameter
is employed. In changing from forged steel to high chrome Eh, mV (SHE) pH DO, ppm EDTA Fe, %
grinding media the following general trends were observed:
Forged steel
•• the Eh shifted to more oxidising pulp potentials
6.5 -222 8.3 0.7 0.60
•• the dissolved oxygen content of the pulp increased 14.7 -200 8.3 0.3 0.54
•• the levels of EDTA extractable iron decreased. 18.5 -159 8.4 0.7 0.58
20.2 -89 8.6 0.5 0.50
It is no coincidence that both the Eh and dissolved oxygen
24.0 -78 8.5 0.0 0.55
are related to the levels of EDTA extractable iron present in
the pulp, and that variations in these parameters have an High chrome
impact on flotation response. The EDTA extractable iron data 4.4 112 8.0 2.9 0.39
provides a reasonable indication of the magnitude of grinding 8.4 92 8.0 1.9 0.32
media corrosion. In the case of forged steel the corrosion rates 12.0 120 8.3 3.6 0.39
are high, which manifest themselves as high EDTA extractable 17.5 123 8.2 4.1 0.31
iron readings. High corrosion results in lower dissolved 20.2 64 9.4 2.6 0.45

FIG 6 – Trace of the xanthate addition to the lead primary rougher feed box and residual xanthate in the lead primary
rougher tailing (as measured using the CSIRO’s ultra violet monitoring system) for Line 1 (forged steel).
FIG 7 – Trace of the xanthate addition to the lead primary rougher feed box and residual xanthate in the lead primary
rougher tailing (as measured using the CSIRO’s ultra violet monitoring system) for Line 2 (high chrome).
residual xanthate measured in the lead primary rougher there is an equilibrium with some ‘free’ xanthate always in
tailing also changed. That is, as the xanthate addition rate solution. The significance of the ‘free’ xanthate value, and
decreased so did the xanthate residual. the parameters which influence it, would merit further
With these encouraging observations, the test program investigation.
began in earnest. A comparison of Figures 6 and 7 reveals
that the value of the residual xanthate in the lead primary Metallurgical response
rougher tailing of Line 1 (forged steel) across all xanthate The metallurgical surveys collected for each xanthate addition
addition rates was significantly higher than the values were mass balanced, and the lead recovery, lead concentrate
recorded for Line 2 (high chrome). For example, for grade and zinc recovery data versus xanthate addition are
approximately the same xanthate addition rate (18 grams per provided in Figures 8 to 10, respectively. While these data
tonne) the residual ‘free’ xanthate for Line 1 was nominally are by no means exhaustive, and further work is required
8.2 grams per tonne compared with 3.2 grams per tonne to validate the observations made, there are some positive
for Line 2. This data would suggest that the pulp chemical trends to be noted.
conditions generated when grinding the ore with high In the first instance, as would be expected, as the xanthate
chrome grinding media were more conducive for improved addition was decreased the lead recovery decreased (Figure 8).
xanthate adsorption and would therefore be expected to For xanthate additions of less than about 15 grams per tonne
lead to improved metallurgical response. the lead recovery to the lead primary rougher concentrate
It was very interesting to note that at no point in these were significantly lower than 50 per cent. At xanthate addition
plant trials did the residual xanthate decrease to zero. This rates of greater than 15 grams per tonne the lead recoveries in
may suggest that the adsorption process operates such that both Line 1 and 2 ranged between 50 and 65 per cent. The data

FIG 8 – Lead recovery versus xanthate addition for lead primary rougher flotation.
FIG 9 – Lead concentrate grade versus xanthate addition for lead primary rougher flotation.
indicates that it may be possible to use less xanthate when In any event, the use of high chrome grinding media appears
grinding with high chrome media to get nominally the same to have resulted in a positive shift in the lead grade/recovery
lead recovery to that achieved using forged steel. However, position.
more work is needed to confirm this finding.
It is also interesting that the lead concentrate grades Comparison of xanthate addition points
(Figure 9) were considerably greater for ore ground with In these tests the xanthate was initially added to the lead
high chrome media when compared with the forged steel primary rougher feed, and then once all the appropriate
case. The improved concentrate grade can be attributed to an measurements had been collected, the addition point was
improvement in the selectivity for galena against sphalerite changed to the primary ball mill feed. The survey procedure
(Figure 10), iron sulfides and non-sulfide gangue. was then repeated. This brief analysis will compare the data
obtained for the two addition points, and the two different
It is speculated that the reasons for the improved selectivity media types. These tests were completed at nominally the
for galena against sphalerite are related to more efficient same collector dosage of 24 grams per tonne.
xanthate adsorption onto the surfaces of galena in the case
where the ore is ground in a more inert grinding environment. Pulp chemistry
This may be related to the more oxidising Eh and higher Similar variations in the pulp chemistry were noted during
dissolved oxygen values recorded for Line 2, when compared these experiments to those shown in Table 2. That is, when
with Line 1. Improvements in selectivity for galena against the xanthate was added to the lead primary rougher feed
iron sulfides when grinding with high chrome grinding media box Line 1 primary ball mill discharge had consistently more
may be related to higher levels of dissolved oxygen present reducing Eh values, contained considerably less dissolved
in the pulp, which are then available to aid flotation and for oxygen and more EDTA extractable iron than Line 2. The
gangue sulfide mineral oxidation rather than corrosion (as is addition of xanthate to the primary ball mill did see a subtle
the case in the forged steel system). That is, the iron sulfides increase in the primary ball mill discharge Eh by some 40 mV.
are oxidised in the high chrome system rendering their That is, the Eh of Line 1 primary ball mill discharge increased
surfaces hydrophilic and depressing their flotation. from nominally – 100 mV (SHE) when the collector was added

FIG 10 – Zinc recovery versus xanthate addition for lead primary rougher flotation.
to the lead primary rougher feed to -60 mV (SHE) when the particularly when you consider that Line 1 was operating
collector was added to the primary ball mill feed. with forged steel grinding media, Line 2 with a high chrome
alloy, and the lead primary rougher concentrates and tailing
UV monitoring process streams combining to feed into other parts of the
Pulp samples from the primary ball mill discharge were circuit. That is, the three lead primary rougher concentrates
collected from Lines 1 and 2, and the solutions passed report to the lead re-cleaner feed, and the three lead primary
through the UV monitor to determine the levels of residual rougher tailing eventually combine to form the lead secondary
xanthate in the pulp for each condition. These data suggest rougher feed. So, under the current circumstance the only
that the level of ‘free’ xanthate in solution for Line 1 primary
comparison available was to complete the flotation response
ball mill discharge was significantly higher than that obtained
of the lead primary rougher circuits of Line 1 and Line 2.
for Line 2. The readings are nominally double those noted
for the high chrome case. The higher residual xanthate in Given this limitation, the mass balanced data comparing the
Line 1 primary ball mill discharge may be related to the more addition of collector to the primary ball mill feed with adding
reducing environment experienced in the mill (nominally -60 xanthate to the primary rougher feed box is provided in
to -100 mV (SHE)), where lead xanthate is not stable. Line 2, Table 3. Again, while these data are by no means exhaustive,
operating with high chrome grinding media does not suffer and further work is required to validate the observations
from this effect, as the Eh of the primary ball mill discharge in made, there are some positive trends to be noted. The addition
this case was around 100 mV (SHE) well into the region where
of 24 grams per tonne of collector to both mills appears to
lead xanthate is stable.
have improved the recovery of the copper, lead, zinc and
iron sulfides, and confirms the visual observation discussed
Metallurgical response
above. The addition of collector to the primary ball mill feed
When the xanthate addition (24 grams per tonne) was
appears to have had a bigger impact online 1 than Line 2,
changed to the primary ball mill feed, the rate at which froth
flowed from the lead primary rougher flotation cells visibly when considering lead recovery. However, it is worth noting
increased in both lines. This was a fair indication that the that prior to moving the xanthate addition point from the
mass pull and probably the lead recovery had increased. This feed box to the ball mill feed the lead recovery for Line 2 was
could be investigated further. higher than Line 1. The lead concentrate grade did change as
Ideally, the best test to determine if the metallurgical the collector addition was shifted from the rougher feed box
performance has improved significantly is to compare the to the primary ball mill feed, and is probably indicative of the
final concentrate grades and recoveries. Unfortunately, the recovery of galena, as well as the diluent minerals (sphalerite,
plant circuit configuration does not allow a comparison, chalcopyrite and iron sulfides).
Table 3
Lead concentrate grade and recovery data for the lead primary rougher circuit online 1 (forged steel) and Line 2 (high chrome),
where the xanthate was added to either the lead primary rougher feed box or to the primary ball mill feed.
Condition Pb grade, % Recovery, %

Ag Pb Zn Cu IS NSG
Line 1 – forged steel
Rougher feed box 40.1 40.1 58.7 5.3 34.1 4.5 2.7
Ball mill feed 48.8 40.1 68.3 6.0 45.9 4.8 2.7
Line 2 – high chrome
Rougher feed box 51.8 39.8 66.9 4.7 35.0 1.7 2.4
Ball mill feed 46.4 59.3 69.5 6.2 52.5 3.4 2.4
IS = iron sulfides, NSG = non-sulfide gangue.

FURTHER WORK REFERENCES

Further work is planned, initially in the laboratory, to Adam, K, Natarajan, K A and Iwasaki, I, 1984. Grinding media wear
determine dependence of collector adsorption on Eh during and its effect on the flotation of sulphide minerals, International
grinding and how this affects the flotation response. Journal of Mineral Processing, 12:39–54.
Cheng, X, Smith, K A and Iwasaki, I, 1993. Electrochemistry of
chalcopyrite-pyrrhotite-mild steel interactions and its relevance
CONCLUSIONS to the flotation of complex sulphide ores, in Proceedings Paul E
Converting from forged steel to high chrome grinding media Queneau International Symposium: Extractive Metallurgy of Copper,
resulted in the pulp chemistry of the system shifting to: less Nickel and Cobalt (eds: R G Reddy and R N Weizenback), Volume I:
reducing Eh values, higher dissolved oxygen content and Fundamental Aspects, pp 971–991 (The Minerals, Metals and
lower levels of EDTA extractable iron. Materials Society: New York).
Using the CSIRO’s online UV monitor it was possible Greet, C J, Kinal, J and Steinier, P, 2005. Grinding media – its effect
on pulp chemistry – fact or fiction?, in Proceedings Centenary of
to measure comparatively low levels of ‘free’ xanthate in
Flotation 2005 Symposium, pp 967–972 (The Australasian Institute
solution. Further, it was possible to monitor changes in the of Mining and Metallurgy: Melbourne).
residual xanthate measured in the lead primary rougher
tailing as the xanthate addition rate was varied. Greet, C J, Small, G L, Steinier, P and Grano, S R, 2003. The Magotteaux
Mill: investigating the effect of grinding media on pulp chemistry
With xanthate added to the lead primary rougher feed box and flotation performance, Minerals Engineering, 17(7/8):891–896.
it was noted that the level of ‘free’ xanthate present in the
Greet, C J and Steinier, P, 2004. Grinding – the primary conditioner,
lead primary rougher tailing process stream was significantly in Proceedings Metallurgical Plant Design and Operating Strategies
lower in Line 2 (high chrome) than Line 1 (forged steel). Conference, pp 319–336 (The Australasian Institute of Mining and
This suggested that xanthate adsorption was more efficient Metallurgy: Melbourne).
when using high chrome grinding media. The metallurgical
Guy, P J and Trahar, W J, 1984. The influence of grinding and flotation
response indicated that the high chrome line produced environments on the laboratory batch flotation of galena,
significantly higher lead concentrate grades through greater International Journal of Mineral Processing, 12:15–38 (Elsevier:
selectivity against sphalerite, iron sulfide and non-sulfide Amsterdam).
gangue. Hao, F, Davey, K J, Bruckard, W J and Woodcock, J T, 2008. Online
Shifting the xanthate addition to the primary ball mill analysis for xanthates in laboratory flotation pulps with a UV
feed appeared to increase the recovery of galena, sphalerite, monitor, International Journal of Mineral Processing, 89(1):71–75.
chalcopyrite and iron sulfides in both lines. Sullivan, J and Woodcock, J T, 1973. A simple on-stream xanthate
Throughout the whole range of tests the high chrome monitor, in Proceedings Australasian Institute of Mining and
line was generally superior to the forged steel case with an Metallurgy, No 248 (The Australasian Institute of Mining and
indication that economies in collector usage were possible.
Watters, T, 1993. Lead-zinc ore concentration practice at the South
concentrator of Pasminco Mining – Broken Hill, Broken Hill
ACKNOWLEDGEMENTS NSW, in Australasian Mining and Metallurgy, The Sir Maurice
The authors gratefully acknowledge the support of Perilya Mawby Memorial Volume, second edition (eds: J T Woodcock and
Broken Hill, CSIRO and Magotteaux in providing them with J K Hamilton), Vol 1, pp 511–515 (The Australasian Institute of
the opportunity to complete this work, and allowing it to be Mining and Metallurgy: Melbourne).
published.

Contents
The Ausenco carbon reactivation kiln initiative

J K Claflin1 and S R La Brooy2
INTRODUCTION
In early 2012, Drs Stephen La Brooy and Jeff Claflin were approached to help improve
the reactivation of carbon used in flotation concentrate and tails leach circuits. The
study identified that conventional carbon kilns, while suitable to regenerate carbon
used in oxide ore carbon in leach (CIL) circuits with no fouling agents, were unsuitable
for reactivating carbon in most (if not all) flotation leach carbon circuits. The kiln was
upgraded using the only suitable kiln available and the plant recovery is now on par
with world’s best practice.
Since the 1980s the drive to minimise capital expenditure has driven the industry
away from fit-for-purpose carbon reactivation kilns. Many metallurgists and engineers
have lost touch with the process chemistry and have specified the lower capital cost,
conventional carbon regeneration kilns. There has been a lack of awareness of the lost
opportunity cost associated with not selecting a fit-for-purpose carbon reactivation kiln.
A fit-for-purpose carbon reactivation kiln is a high temperature gas-solid reactor and
one of the most complex and potentially hazardous pieces of equipment in most gold
plants. It is as different to the conventional carbon regeneration kiln as a modern car is
to Ford’s model T of 1912. Like the model T car, the conventional carbon regeneration
kiln is inexpensive, robust and performed well at the time. The model T was replaced
by ever more sophisticated cars that perform far better and are very safe and easy to
operate. A safe, easy to use, fit-for-purpose carbon reactivation kiln is the goal.
In the gold industry, a very robust gold plant can be engineered and delivered
quickly, but recovery is not optimal. Of all gold plants processing more than 2 Mt/a
of ore 50 per cent can replace their kilns and payback the capital in the first year of
operation. Large gold plants will generally have a payback period of just a couple
months. However, if a carbon reactivation kiln is specified in the first place, then
world’s best practice gold recoveries may be achieved for no additional capital cost.
Unfortunately, conventional carbon regeneration kilns do not have the ability to be
operated as carbon reactivation kilns without significant modifications.
To turn the situation around, the industry needed re-education and industrial scale
demonstrations of the benefits of fit-for-purpose carbon reactivation kilns. A series of
six conference papers were planned:
1. Reviewing the chemistry associated with reactivating activated carbon fouled
with various chemicals (La Brooy, S R and Claflin, J K – Carbon Management in a
High Gold Price Environment, MetPlant 2013).
2. Dissect an inappropriately selected kiln and build-up a design criteria based on
chemistry to address the deficiencies (Claflin, J K and La Brooy, S R – Ausenco
Carbon Reactivation Design Criteria, World Gold 2013).
3. (Claflin, J K, La Brooy, S R and Preedy, D R – The Ausenco Carbon Reactivation
Kiln – Lessons Learnt and Performance, MetPlant 2015).
4. Demonstrate the recovered lost opportunity cost associated with a fit-for-
purpose carbon reactivation kiln and update the design criteria (Claflin, J K, La
Brooy, S R, Preedy, D R, Slater, A and Urrutia, F – Fast-Payback Reactivation of
Carbon from a Flotation Tails CIL Circuit, World Gold 2015).
5. Re-evaluate carbon circuit in light of kiln chemistry and new laboratory findings
to optimise the carbon advance rate and reactivation kiln operation (Claflin, J
K, La Brooy, S R, Wardell-Johnson, G M and Claflin, A M – Optimising Carbon
Reactivation Kiln Operation, MetPlant 2017).
6. Combine previous findings with operating costs and benefits to form a practical
guide to support gold producers, engineers and vendors in their decision-
1. Consulting Chief Process Engineer,
making (working title: Carbon Reactivation Kiln Design, Selection and Operation,
Earendil Pty Ltd, Perth WA 6000.
Email: dr.claflin@iinet.net.au Mill Operators’ Conference, 2018).
2. FAusIMM, Principal Process Consultant, Because a full kiln replacement is costly, clients are enquiring how to modify
Ausenco, Perth WA 6000. existing conventional regeneration kilns to meet the reactivation design criteria. At
Email: stephen.labrooy@ausenco.com the time of writing one project is underway and two others will begin in early 2017.
537
The MetPlant 2013 and 2015 papers have been selected for thereby reducing the installed capital cost significantly. For
this publication. The following discussion presents these two a brownfield installation, a shorter residence time permits
MetPlant papers in context with the other published and much higher throughput rates and shorter operating times,
unpublished papers in the series. Readers are directed to the hence lower operating costs (MetPlant, 2017).
relevant paper of the series for further information. The required steam to carbon ratio has been reported as 0.5
to 1.0 kg of water to 1.0 kg of carbon (MetPlant, 2013). This
CHEMISTRY is not correct. The referenced test work referred to the water
Process chemistry must drive the selection of the carbon content of the activated carbon as it was introduced into the
kiln (MetPlant, 2013). If there are high molecular weight test reactor. On heating the carbon sample to 850°C almost all
contaminants, for example flotation reagents, oil, naturally the water evaporates so that the water vapour to carbon mass
occurring dissolved organics, they will adsorb onto the ratio in the test reactor is only about 0.001 kg of water/kg of
carbon and will not be removed just by heating. Fouling will carbon. If the drum seals are good and the discharge chute is
reduce both the rate of adsorption of precious metals and carefully designed, then the water vapour raised by cooling
their ultimate loading onto the carbon. The fouling must be the carbon is more than sufficient for the reaction. Some
chemically removed. A carbon reactivation kiln pyrolyses venders purposefully generate water vapour for the reaction
these chemicals and then reacts the residual carbon with water by evaporating the water in the feed carbon inside the reactor.
vapour at high temperatures. The reaction is endothermic This increases the reactor’s heat load almost threefold and
and results in primarily carbon monoxide and hydrogen introduces cool 100°C water vapour which quenches the
(‘syngas’). If no oxygen is present, the reaction is controlled water gas reaction (MetPlant, 2017). There is no benefit to
generating steam in this manner and it requires the kiln to be
by the amount of heat added. The conventional rule of thumb
far larger and consume more fuel.
was to hold activated carbon for over 30 minutes at 750°C
or 20 minutes at 850°C in a steam atmosphere to reactivate If the carbon is dry, the heat required to bring 1000 kg/h to
it (World Gold, 2015). This rule of thumb only applies to a reactivation temperature is just 400 kW. If the carbon is wet,
rotary kiln with lifters and wet carbon feed. Water evaporates the process heat load increases by over threefold to 1300 kW
at 100°C and fills the kiln with relatively cool (100°C) water (MetPlant, 2017). If carbon drying is performed within the kiln
vapour. Lifting hot carbon and dropping it through cool water drum, then the heat supplied by the fuel is 2750 kW including
vapour, cools the carbon and quenches the reaction (World equipment and sensible heat losses. If drying is performed
Gold, 2013). The above rule of thumb is based on laboratory externally using the exhaust combustion gases from the kiln,
test work presented in Figure 1 (MetPlant, 2013) and practical then the exhaust temperatures are far lower and the total heat
industrial experience with conventional regeneration kilns. supplied from the fuel drops to 1500 kW – almost a 50 per cent
Figure 1 clearly shows that a temperature over 700°C is reduction in heating requirement (MetPlant, 2017). Of note is
required and that a residence time of some sort is required, that a direct contact dryer is physically quite small relative to
but increasing the residence time at temperature over ten the indirectly heated rotary kiln. The dryer volume required
minutes has little benefit. is only tens of litres if designed properly, hence the capital
cost of the dryer/kiln package overall is significantly reduced.
The reactivation reaction follows the Arrhenius equation
and reaches a practical maximum around 800°C (World Gold, Oxygen ingress into the reactor is a very common problem.
2015). The significance of the Arrhenius reaction is that as the Oxygen burns the carbon and does not reactivate it. The
temperature increases, the required residence time decreases. consequence is increased carbon consumption, smaller
Given that the ten and 20 minute residence timelines in carbon particles in the carbon circuit and carbon losses (with
Figure 1 are nearly identical, it is likely that a residence time of adsorbed gold) to tails. Oxygen is usually present due to
five minutes or less at 850°C in a reactivation kiln will perform deteriorated gas seals allowing air into the kiln and poor
similarly to the 30 minute curves at 750°C in a conventional pressure control (World Gold, 2013). To control the kiln
carbon regeneration kiln. Carbon reactivation test results atmosphere, keep off-gases in, and air out of the kiln, the
show reactivation to 90 per cent for a carbon residence pressure in the kiln must be controlled. Pressure differences
time at temperature of less than one minute (MetPlant, can be 5 to 10 Pa from the bottom to the top of the kiln, hence
2017). Reducing the design residence time from 30 minutes a neutral pressure everywhere is not an option. Typically,
to, say, 5 minutes shortens the kiln by about 40 per cent the decision is to operate at a slight positive pressure (10 to
15 Pa) to keep air out, unless the off-gases are particularly
hazardous, (for example mercury) and require treatment.
Carbon kilns are adaptations of kilns that were in use when
carbon in pulp processing was first developed in the 1950s
(World Gold, 2015). Rotary kilns had been developed initially
for the cement industry and adapted for drying applications by
including lifters to enhance the gas-solids contact. Indirect kilns
permitted excellent control of the kiln atmosphere and lifters
permitted good gas-solid contact for reactions, but they were
thermally inefficient. The direct/indirect kiln was a means
of improving thermal efficiency for drying when the kiln
atmosphere was not critical. Both indirect and direct/indirect
kilns were used for carbon regeneration. Only an indirectly
heated kiln should be used for carbon reactivation because
the kiln atmosphere is critical. If the mode of heat transfer is
by radiation from the drum (or flame) to the surface of the
bed of solids or by convection from the gas to the individual
FIG 1 – Relative activity of carbon with reaction particles or by conduction from the drum to the bottom of the
temperature and residence time. solids bed, then the fill ratio must be kept small to achieve a

The Ausenco carbon reactivation kiln initiative
reasonable rate of heat transfer, typically 8 to 14 per cent of

the drum cross-section. If heat transfer fins are used internally
for an indirectly kiln, then the mode of heat transfer is almost
entirely by conduction to the first layer of particles touching the
drum or heat transfer fins. The result is a much shorter, smaller
diameter drum to perform the same duty.
Drum rotation speed combined with drum inclination is
used to convey the solids through the kiln. Keeping in mind
that the fill fraction is usually very small, the drum rotation
is usually very slow to give the solids time to heat up. When
heat transfer fins are used, the drum speed is increased
significantly to deliver this hot surface into the carbon. The
result is that the carbon heats up far faster than a conventional
kiln and the kiln can be smaller for the same heat transfer
duty. Increased drum speed also reduces the time the heat
transfer fins are exposed to the reducing atmosphere inside
the bed and reduces fin corrosion.
FIG 3 – Over-firing drum distortion.
A high temperature kiln performs an oxidising gas-solids
reaction while producing a reducing atmosphere in a bed
of carbon that often contains chlorides and other inorganic
material must allow for extreme thermal stress points, fatigue
cracking, corrosion and erosion. The materials of construction
are in the end a compromise and ‘you get what you pay
for’. The drum experiences the majority of the thermal and
chemical wear. Sandvik 253 MA is currently the industry
standard material of construction. It was designed for high
sulfur, high temperature environments and if properly
welded and destressed, it does not distort or sag even when
crash stopped (although this is not recommended). Sandvik
253 MA can withstand extreme temperatures to 1100°C, but
if abused, especially by localised overheating, its life will
shorten significantly (Figures 2 and 3). Note that over-firing
with cool water vapour in the drum can result in lifters tearing
holes in drums (Figure 2) whereas over-firing at the wet feed
end distorts the drum (Figure 3).
The combustion chamber is operated at a higher temperature
than the drum. If an insufficient number of burners and
high sulfur diesel is used corrosion can result in extensive
degradation, Figures 4 and 5 (World Gold, 2013). FIG 4 – Over-fired internal baffle.
To ensure process robustness, carbon circuits are usually
designed based on virgin carbon used repeatedly with no
reactivation (MetPlant, 2017). The resulting plant solution
losses are on average 0.02 ppm compared to 0.01 ppm for a
well operated carbon circuit. The cost of gold lost in solution is
significant and is presented in Figure 6. Many carbon circuits
with solution losses over 0.04 ppm can have their regeneration
kilns replaced with a reactivation kiln and achieve payback
FIG 5 – Internal baffle broken off by hand.
on the installed capital within a year (Figure 7). Conventional

carbon regeneration kilns can be modified to achieve
carbon reactivation design criteria and the payback of these
modifications will be discussed in a future paper.
The operating cost benefit of a carbon reactivation kiln is
not just improved gold recovery, but can include reduced fuel
usage, reagent usage, carbon attrition and associated gold
losses on fine carbon reporting to tails. An economic optimum
FIG 2 – Lifter expansion damage. carbon advance rate exists and will be reported in the future.

FIG 6 – Estimated lost opportunity costs.
FIG 7 – Kiln replacement payback.

Contents
Lessons learnt and performance –

installing and commissioning an Ausenco
carbon reactivation kiln in Africa
J K Claflin1, S R La Brooy2 and D Preedy3
ABSTRACT
A gold mine in north-eastern Africa has expanded throughput from 5 to 10 Mt/a in
recent years. The original carbon regeneration kiln was incapable of achieving the
required throughput or meeting the process conditions to reactivate the carbon, due
to heavy fouling caused by high molecular weight frothers and collectors. A new high
temperature kiln was required.
The design criteria for the carbon reactivation kiln and associated equipment
were revised, then implemented in close cooperation with a kiln vender who had
previously supplied kilns operating successfully on highly fouled carbon. The result
is a whole facility, holistically designed to achieve the technical criteria.
The new carbon reactivation facility has been operating since 1 July, 2014. This
paper presents lessons learnt regarding the design, implementation, operation and
maintenance of the Ausenco carbon reactivation kiln. ‘Lessons learnt’ are associated
with threats and risks that have been realised. The importance of identifying and
managing risks for even small straightforward projects is demonstrated.
INTRODUCTION
‘If the use of some new … type of equipment … is absolutely essential to the
economic viability of the plant, go ahead’ (Halbe, 2009). Innovation is not optional,
but essential in creating value from increasingly low-grade ore deposits (Marsden,
2009). Marsden has also suggested that the only way to avoid prior mistakes is to
incorporate and embrace innovation in the design process, and provide a process
to evaluate innovation in terms of risk and reward. An innovation with high risk/
low reward is a ‘no-go’. Innovations with high risk/high reward or a low risk/low
reward need careful evaluation. A low risk/high reward innovation is considered a
‘no brainer – yes’ decision. For a brownfield operation, new equipment affords the
highest opportunity for innovation and value creation, but can also be associated with
high risk (Marsden, 2009).
The replacement of a low temperature carbon regeneration kiln with a high
temperature reactivation kiln to cope with a major mill upgrade and carbon from
both flotation tails and flotation concentrate leach circuits was deemed a ‘no brainer’.
In 1985, the principle author began using the values in Table 1 to guide projects
and ensure elements of a project were not overlooked. The priority of the middle
and bottom values shift somewhat, but generally the priorities are as presented in
the table. Seven of these values are identified by Marsden as ‘drivers for innovation’
(Marsden, 2009). Innovation introduces risk and the economic benefit to the client
must warrant taking that risk.
When the client values/drivers are qualitatively applied to the project in question,
the following drops out:
•• health and safety issues included process off-gases, materials of construction,
1. Consulting Chief Process Engineer, equipment high temperature protection, access for maintenance and the need
Earendil Pty Ltd, Perth WA 6000.
Email: dr.claflin@iinet.net.au
for a special purpose safety inductions
2. FAusIMM, Principal Process Consultant, •• compliance with environmental licence conditions was the only other legal issue
Ausenco, Perth WA 6000. considered relevant to the kiln upgrade
Email: stephen.labrooy@ausenco.com •• no specific corporate initiatives were identified
3. MAusIMM, Consulting Process Manager,
Dave Preedy Metallurgy.
•• quality activated carbon to improve yield was a high priority
Email: drpreedy@yahoo.com.au •• capital expenditure was not, initially, a major driver
541
Table 1
Generic client project values.
Priority Value Clarification

1 Health and safety The health and safety of people is not to be compromised. Safety risks to assets are to be identified and an informed decision made.
2 Legal and regulatory Identify and comply with all laws and regulations.
3 Corporate initiatives The directors are empowered to apply assets and usually have good reasons – support their initiatives.
4 Product quality If intermediate or end products are not to specification, then equipment is not fit-for-purpose (no matter how cheap it was to buy!)
5 Capital expenditure Plant must be fit-for-purpose for minimum capital cost.
6 Throughput Increases saleable product and reduces unit cost.
7 Yield Increases saleable product and reduces unit cost.
8 Operating cost Important as head grades drop and processing becomes more complex – costs add up.
9 Operability Improves almost all of the above – do not ignore!
10 Maintainability Improves almost all of the above – do not ignore!
11 Constructability Design and engineer keeping delivery and commissioning in mind in addition to other drivers.
•• an increase in mill throughput was the major driver as •• constructability focused on the need to ‘shoe-horn’
the original kiln would be overwhelmed the new facility into the very small land area available
•• as plant throughput overwhelmed the original kiln, gold within the existing processing plant.
recovery would be reduced due to the insufficient gold Independently, the engineer identified that cooperation
loading capacity of the carbon between the client, the engineer and the kiln supplier was a
•• operating expenditure of the new kiln would be high key risk because of their inherent conflict of interests (refer
if heat losses were not minimised and no recovery to Figure 1). Significant energy was invested into the client–
achieved, but increased operating cost was deemed to engineer and engineer–kiln supplier relationships.
be less than the cost of replacement carbon as required if The ‘lessons learnt’ presented in this paper are threats or
the old kiln remained in service risks that were realised. Design reviews, hazard identification
•• operability issues focused primarily on the ability to (HAZID) and hazard and operability (HAZOP) studies were
achieve the required process conditions performed, but note that no overarching project risk review
•• maintainability was considered to be relative to the was carried out. The threats that were realised were largely
original kiln and was not a focus risks not identified and therefore not managed.
Main Engineer Drivers

• Deliver to client expectations
• A fit-for-purpose plant
Main Client Drivers
• Build on extensive gold plant
• Quality activated carbon
engineering experience
• Carbon throughput
• Innovate only as required
adequate to maintain or
improve gold recovery • Only use proven technology
(though new to gold processing)
• An operable plant
• Minimise risk of non-
• A maintainable plant
performance
• Commission before mill
• Be acknowledged as industry
upgrade is completed
leader in activated carbon
processing
Main Vendor Drivers

• Design a state-of-the-art kiln
• Supply at standard mark-up
• Be acknowledged as a
quality supplier
• Open up a new market
• Increase future sales
FIG 1 – Stakeholder drivers.

Lessons learnt and performance – installing and commissioning an Ausenco carbon reactivation kiln in Africa
BACKGROUND 110
As a gold mine moves from processing oxide ore to sulfide 100 Plant kiln
ore, flotation is often used to concentrate the gold containing
Relative Activity, %
90 20 min
sulfide material before leaching. Some of the reagents used are 80 10 min
adsorbed onto the activated carbon in the adsorption circuit, 0 min
70
reducing the carbon activity and taking up adsorption-sites.
60
More activated carbon is added to the circuit to offset the
reduced carbon activity and provide more sites for the gold to 50
adsorb onto. If the reagents are not removed, then the carbon 40
activity continues to degrade. Ultimately, fouled carbon is 30
removed from the circuit and fresh virgin activated carbon is 20
added to maintain overall recovery. 300 400 500 600 700 800 900
Carbon Temperature, °C
The engineer was approached by a mine that was processing
low-grade primary sulfide ore through flotation and carbon- FIG 3 – Effect of time at temperature in thermal
in-pulp (CIP), as they were planning an expansion to increase
reactivation in steam atmosphere.
nominal throughput from 5 Mt/a to 10 Mt/a. The original
kiln was only suitable for treating lightly fouled carbon, such For carbon temperatures up to 700°C, regeneration (stripping)
as from oxide ore processing. The kiln operating conditions
occurs while above that temperature carbon reactivation
were taken to their design limit, but the kiln could not provide
occurs (La Brooy and Claflin, 2013).
the residence time at the extreme operating conditions
required for heavily fouled carbon. Loaded carbon activities The reagents used by the client do not volatilise from
dropped to around ten per cent of virgin carbon, improving the carbon, but rather decompose by pyrolysis at kiln
to around 40 per cent after regeneration (Claflin, La Brooy and temperatures used for oxide ore carbon regeneration, resulting
Preedy, 2013). The kiln could not sustain operation at these in a carbonaceous residue on the activated carbon surface
extreme conditions and required a near complete rebuild. (La Brooy and Claflin, 2013). The carbon particles must be
To compensate for the inadequate kiln performance many held for at least ten minutes at over 750°C in a water vapour
hundreds of tonnes of fouled carbon was removed from the atmosphere for selective oxidation of the pyrolysed residue
circuit and accumulated on-site. and therefore higher reactivation to occur (Avraamides and
The plant in question had both flotation concentrate and La Brooy, 1987).
flotation tails leach circuits. Carbon that spends a long time in
The plant in question uses sea water for plant raw water.
the circuit, such as in a flotation tails leach, has the opportunity
Carbon samples from a plant operating with saline water will
to load flotation reagents that will not be displaced by gold.
show a higher relative activity in saline water than if tested
Hence the gold loading of a flotation tails leach will be more
severely affected than in a flotation concentrate leach circuit, in pure water (La Brooy and Muir, 1994). Furthermore the
where the carbon would be exposed to the same reagents for a site uses an in-house carbon activity test. In-house laboratory
much shorter time (La Brooy and Bax, 1985; Mahapatra, 2009). results will therefore not be comparable with external
laboratory results, but may be comparable to other in-house
Potassium amyl xanthate is used at this mine site as a
collector. Figure 2 (after La Brooy et al, 1986) shows the laboratory results (within the limitations of the procedure).
residual carbon activity after 24 h exposure to 20 mg/L of In-house carbon activity results for 2011 are presented in
different collectors. Increasing the collector concentration Figure 4 prior to a major mill upgrade. Towards the end of
further reduces the carbon activity. Frothers can have an even that period, the mill throughput increased from 3 to 5 Mt/a.
more severe effect on carbon activity, with 24 hour exposure The relative carbon activity dropped from 80 to 90 per cent to
to 20 mg/L frother reducing carbon activity to ten per cent of become variable between 50 and 70 per cent. It is worth noting
fresh carbon (La Brooy et al, 1986). Mahapatra (2009) observed that in June of 2011 carbon samples were sent for external
frother induced reductions in carbon kinetic activity to around analysis. External testing reported the regenerated carbon
four per cent of the control!
activity to be 34 to 43 per cent (Wardell-Johnson and Barbetti,
Avraamides and La Brooy (1987) have summarised the 2011a, 2011b) at a time when in-house testing suggested
required kiln temperature and atmospheric requirements for activities of the order of 80 per cent.
effective thermal reactivation. Figure 3 (after Avraamides and
La Brooy, 1987) illustrates the effect of temperature and time.
50%
40%
Carbon Activity, %
30%
20%
10%
0%
SEX SIBX PAX
FIG 2 – Relative carbon activity after 24 hours in

20 mg/L of different xanthate solutions. FIG 4 – Carbon activity prior to major mill upgrade (in-house test procedure).

PROJECT RISK LESSONS LEARNT The automation system, typically the most complex
aspect of any project, experienced numerous and significant
Project risk assessment problems. Re-wiring of drives, repairs to cabling and repairs
The project was small and expected to be of a short duration, to the Profibus communication system all occurred after
hence a broad project risk assessment did not take place. This commissioning commenced. The automation program was
was an oversight. ‘factory acceptance tested’ based on a centralised system, but it
was decided immediately prior to commissioning to control the
A project risk assessment should not be optional. It should facility from a local programmable logic controller the software
be initiated early in the project life cycle and updated as for which had not been written or tested. Operations had a
the project progresses. Its purpose is to raise awareness predilection for manual control and would use only the most
of potential threats and opportunities so that value can be basic of automatic control in any case. Manual control works,
maximised by implementing opportunities and avoiding
but it is much more difficult and is not as efficient or effective.
or mitigating threats. The depth and breadth of each
assessment is commensurate with the number and severity of Had even a high-level risk assessment taken place, the
consequences remaining in the project. design would have been significantly different. Project
complexity and automation would have been minimised to
The kiln upgrade project consisted of an investigative
be in line with the sophistication of the workforce and site
study, a scoping study and detailed engineering followed
management practices.
by ongoing support of the client during procurement,
construction and commissioning phases. The initial project
risk review could have been performed as part of the scoping TECHNICAL LESSONS LEARNT
study and reported in a single day.
A project risk assessment would generally cover the
Sand trap
following issues: An extraordinary amount of sand and grit existed in the carbon
circuit, nominally 20 to 30 per cent by volume or 50 per cent
•• location – country, weather, communications, logistics
by mass (dry basis). The inorganic material displaces carbon
•• sociopolitical – stability, rule of law, changeability, in the acid washing and elution columns, thereby requiring
education, culture, religion more batches to be processed in order to provide sufficient
•• environmental – regulations, pressure groups, carbon for adsorption. Consequently the kiln must treat
enforcement much higher mass throughput rates to produce the amount of
•• client – ownership, management, experience, carbon needed for processing.
involvement, culture A simple gravity separator was designed. The criteria was
•• commercial – funding, conditions, liabilities to remove as much sand as possible and lose essentially no
carbon. Larger grit settled in a 10 m tall teeter column where
•• contractual – strategy, scope, major contractors, liabilities,
a rising water flow elutriated carbon upwards. Carbon and
responsibilities
smaller grit particles reported to the overflow and passed to
•• project program – planning, project management, the dewatering screen. Sand and grit was discharged via a
integration, performance double dump valve immediately under the elutriation leg.
•• resources – management, continuity, industrial relations, The flow rate to the sand trap was greater than design
materials, capability in an attempt to prevent sanding in the supply piping. An
•• definition phases – suitability, novelty, design additional 100 mm of freeboard was required on the sand trap
•• delivery – procurement, construction, schedule, budget, to prevent overflowing.
commissioning During commissioning, the sand trap did separate grit
•• operation – operability, maintainability, operational from the carbon very successfully, but not for very long.
readiness, security. The excessive sand content required the double dump
Certainly the technical novelty was a known and managed valves to work continuously on a very short cycle in order
risk, however, many other project risks were either not to discharge sand. When the valves closed, they closed on a
identified or not adequately mitigated. packed column of grit and ceramic mill media. The ceramic
media eventually got caught in the seat of the valve (between
Of significant impact was a change in the political situation the ball and the valve housing) and jammed the valves open,
and associated change in law, which impacted both fiscal
dumping sand and carbon into the bund. Pinch valves were
and logistic aspects of the project and contributed to project
tried without success.
delays. While this change in the political situation of the
country could not be known in advance, the possible delay to The sand trap was bypassed during commissioning
the project as a consequence of civil unrest could have been and operations have not yet returned to it. An airlift on
surmised. A suspected delay in the project would have raised the discharge may avoid the problems and be capable of
the possibility of losing key personnel and the importance of operating continuously with the consumption of around
having strong corporate and site management support for 5 m3/h of additional water.
the project through implementation. In fact, for the project The downstream equipment was designed to have
in question, the project champion left the site for personal significant sand associated with the carbon, but not to the
reasons before the kiln was constructed. degree experienced. Acquiring representative samples of the
Workforce technical capabilities, or lack thereof, were not feed and characterising them in terms of composition, solids
sufficiently identified for the engineering principals of the handling and pumping was a lesson learnt.
project. Assembly of the kiln was performed in-house but
was not performed well, with most of the unit operations Carbon dewatering screen
requiring some kind of modification or repair during the The overflow from the sand trap to the dewatering screen
commissioning process. did not have sufficient fall and sanded up. Construction

verification should be performed by commissioning to ensure feed was maintained at 2 m3/h. The equivalent dry mass rate
that what is constructed is as per design and fit-for-purpose. is 1500 kg/h (50 per cent grit and 50 per cent carbon by mass).
Aside from construction issues, the dewatering screen Despite doubling the mass rate, the reactivation temperature
received, washed and discharged carbon (with sand) is still being achieved. The key is that conventional lifters were
satisfactorily. The instantaneous rate was 65 m3/h and not used, rather numerous heat transfer fins were installed
contained 35 t/h of carbon and sand. Screen apertures were and the fill ratio increased to 40 per cent. This permits the
0.63 × 18.5 mm. The screen width was 1.2 m and the length heat from the 950 to 1050°C combustion chamber to transfer
1.8 m. Had only carbon been in the feed, the instantaneous by conduction into the carbon bed at an optimal rate.
specific feed rate would have been 10 t/h (dry, 20 t/h wet). Due to damage caused when unpacking the kiln,
The carbon was not fully dewatered as additional dewatering realignment was required. No procedure was supplied and
was allowed for via the hoppers. on inspection it was found that the kiln seals could not be
realigned as supplied and modifications were required.
Kiln feed hoppers The fuel supply train was supplied using poor quality
The design incorporated two kiln feed hoppers. A slide gate components, for example the main fuel filter was supplied
at the bottom of the hoppers was removed as it was not with a gasket that disintegrated in diesel and blocked all five
necessary. Solids simply built up on the small screens under downstream burners. All burners were identical even though
the hoppers and the drain. After the charge had completed they all operated in different duties – this was a good practice
draining, the screen could be turned on/off to discharge as spares and method of operation are shared despite slightly
solids to maintain level in the feeder hopper below. different set-ups.
Ideally the feeder screens would have had their vibrating The fluid bed cooler hung off the kiln discharge hood to
weights adjusted or variable speed drive (VSD) tuned so allow for kiln expansion while maintaining the hot end gas
that solids would discharge almost continuously to allow seal. Two springs were used to pull the hood and cooler up
functioning of downstream equipment. to the drum and they were insufficient for the task. Even so,
each spring was very dangerous to install – a pneumatic ram
Predryer with pressure relief to allow for drum expansion would have
If the feed is steady, a small flash dryer would receive, dry been a better option. Having the cooler suspended with, and
and discharge to the feeder hopper. Tuning of the feed system travelling with, the discharge hood as the kiln expands was a
was not completed during commissioning and operations good lesson, if the drum seal could be kept pulled up. If not,
have not returned to commission this feature. They were, the seal would fall out and delays would be incurred until the
instead, considering a completely separate predryer. kiln was cool enough to re-install the seal rope. If a floating
seal was used, then in order to keep oxygen out, the kiln
Kiln feed system should be under slightly positive pressure, which introduces
The feeder hopper maintained a level above the vibrating health risks.
tube feeder. The level was important to prevent off-gases back The discharge hood design was intended to permit easy
flowing into the feeder hopper. The tube feeder should have access to the hot end of the kiln. The discharge hood was on a
been inclined slightly upwards to also prevent gas flowing trolley that was pulled by springs to effect a gas seal with the
back into the feeder hopper and overheating (melting) the rotating drum. If the springs were relaxed, the hood could be
vibrating gas seals. rolled 500 mm away from the drum. Access was lost when a
Because the predryer was not commissioned, the bypass vapour duct was located in the path of the trolley during design,
was always used. The height at which the bypass entered restricting the trolley movement. In a similar example of poor
the feeder hopper was low and the level sensors were not design, viewing windows were supplied but made unusable
suitably placed to measure the feeder hopper level when by the discharge hood’s subframe (which was added to the
being supplied by the bypass. Furthermore, the bottom most design because of the need to support the mass of fluid bed
level sensor was on the cone and entered the feeder hopper at cooler). There must be access for operations and maintenance.
an upwards incline. Saltwater and carbon accumulated in the Temperature of the carbon in the middle of the reactivation
socket which, on heating, corroded the instrument. The socket zone and at the discharge of the kiln was measured, but
was made horizontal and self-draining to avoid the problem. this did not provide good information on the temperature
If a significant amount of sand were present, the sand built profile. Installing a longitudinal pipe to permit temperature
up in the tube feeder and the weight dampened the vibration, measurement along the whole length of the kiln would permit
which reduced the flow rate and exacerbated sand build-up. more efficient and effective operation.
Furthermore, off-gases would flow outside the tube feeder Steam from an independent boiler was used to cool the
inside the kiln. The off-gas temperature was very high and carbon in the fluid bed cooler. Steam flow was excellent, but
the tube feeder would heat up resulting in evaporation inside during manual operation, too much steam could be added,
the tube feeder. If sand were present, then salt crystals would which could cool the kiln and interfere with reactivation.
develop, cementing the sand and leading to blockages. To aid operation, much of the steam from the cooler was
Tube feeders are used successfully to feed reactivation kilns permitted to bypass the kiln directly to the stack.
elsewhere, but given the quantity of sand present and the
preference for manual operation, a screw feeder should have Fluid bed cooler
been used despite the damage to the carbon. The fluid bed cooler was a vertical chamber where steam was
introduced from the periphery. Cooling steam feed rate was
Reactivation kiln a function of the solids temperature. If operating manually
The reactivation kiln itself has performed well above and the conditions are changing, this was very hard to set
design despite the inadequate operation of the surrounding and stabilise.
equipment. The kiln was designed to reactivate 1.5 m3/h The cooler was also difficult to operate during commissioning
(750 kg/h) of carbon. To keep the feeder from blocking, the due to incorrect construction with level and temperature

instruments being supplied and located incorrectly. Rotating solution was to install a separate exhaust stack and expel the
paddle type level switches suitable for 600°C were specified, off-gases to atmosphere. Incineration of the gases reporting to
but paddle switches suitable for 350°C were supplied by the the separate exhaust stack were still required to prevent any
vendor. One paddle switch was supplied so that it interfered potential health hazards associated with off-gases. Needless
with the internal wall. It could not be accessed for inspection to say, inadequate design of the exhaust system was another
or repair and always showed a level. A temperature probe serious lesson learnt.
intended for measuring steam temperature leaving the
cooler was located in the falling solids stream from the Carbon activity
kiln. Another temperature probe, intended to measure the Figure 5 presents relative carbon activities of the regenerated
solids temperature being discharged to protect downstream carbon, measured using the in-house procedure immediately
equipment, was far too short and permitted overly hot following kiln commissioning while the 10 Mt/a upgrade
material to cause downstream damage. was ramping up. Importantly the site increased the kiln
If the fluid bed cooler stopped discharging, then carbon would throughput to 1500 kg/h of carbon plus sand (design was
overflow directly to the quench tank. This design worked quite 750 kg/h of carbon) in October, hence the drop in activity
well and indeed, could be modified to provide gentle cooling from 100 per cent to 90 per cent. Increasing throughput
without a separate boiler (see Recommendations section). stabilised facility operation and provided more capacity for
gold loading in the leach circuit.
Fines screen The data is not of high quality and at best we could only
The fine carbon screen is to capture fine carbon before it conclude that carbon activity appeared to be similar to virgin
reports to the last tank, adsorbs gold and escapes to tails. carbon and this was despite very high throughputs. The
Despite being intended to operate at 350°C, the vendor measured activity was not comparable to activities measured
supplied plastic and cloth screen skirts and gas seals, which by other methods or from other laboratories, as there were no
did not last long and allowed a localised fire when overly hot particle size or moisture corrections.
material was discharge from the cooler (because the wrongly
sized temperature probe did not sense the temperature). CONCLUSIONS
The fines screen was a 1 mm aperture wedge wire screen, The Ausenco carbon reactivation kiln is performing above
600 mm wide by 900 mm long. The screen was designed to expectations, however:
have a bed depth of 7 mm. The actual depth experienced
was closer to 40 mm as a result of the height of the fluid bed •• for this site the high temperature rotary carbon
downcomer above the screen deck. More attention should be reactivation kiln facility, as a whole, was too complex
paid to the height of the feed above the screen. The screen was •• for this site the level of automation was inappropriate
not appropriate for sand. The sand caused severe pegging. •• a project risk assessment should always be performed
The screen should have been at least 50 per cent wider and and acted upon
100 per cent longer to permit the bed to thin out, fine material •• a project champion (sponsor) and client management
to pass through and to allow for pegging. In the future, the support is critical
screen should be designed to a specific area of 400 kg/h/m2.
•• construction verification should be performed by an
The fines screen oversize discharge reports to a static independent commissioning team to ensure construction
accretion screen located in the quench tank. When modified is to design and design is fit-for-purpose
with the correct screen aperture and area, the accretion screen
worked quite well. The undersize screen was bolted to the •• feed must be characterised; in this case there was as
structure, which dampened vibration and affected control. much sand and mill media as there was carbon
The underflow was modified to reflect the original design. •• elevated temperatures greatly increase hydrogen and
The carbon return pump drew carbon and water from the carbon monoxide off-gases, which need engineering
quench tank. Because of the sand content, the recessed vane solutions to avoid becoming health and safety incidents.
impellor pump had to operate at extreme speeds to produce
the required discharge head and this resulted in severe pump REFERENCES
erosion and carbon breakdown. The recessed vane pump Avraamides, J and La Brooy, S R, 1987. Evaluation of different
was replaced with a larger, slower speed slurry pump which reactivation systems available for carbons used for gold recovery,
overcame the sand problem, but drew slightly more water in Proceedings Equipment in the Minerals Industry: Exploration
than design from the quench tank. Mining and Processing, pp 1–7 (The Australasian Institute of
Off-gas scrubber
In order to achieve stable operation, the kiln temperature was
raised over the initial design temperature, which increases
the water gas reaction rate. The amount of off-gas produced
went up considerably and consisted of carbon monoxide and
hydrogen. The off-gases report to a venturi scrubber and are
drawn by a fan and discharged directly into the same stack as
the products of combustion. At higher operating temperatures,
a flashback occurred, which caused minor equipment damage
and caused a plastic gas seal to catch fire. Though an incinerator
was included in the stack design as a future option to deal with
excessive non-condensable gases, the flashback was not picked
up as a concern during the HAZOP.
The long-term correct solution would have been to modify the
ducting and install the incinerator. However, the expeditious FIG 5 – Carbon activity post commissioning (in-house test procedure).

Claflin, J K, La Brooy, S R and Preedy, D R, 2013. Ausenco carbon La Brooy, S R and Claflin, J, 2013. Carbon management in a high
reactivation design criteria, in Proceedings World Gold 2013 gold price environment, in Proceedings MetPlant 2013 Conference,
Conference, pp 401–410 (The Australasian Institute of Mining and pp 455–467 (The Australasian Institute of Mining and Metallurgy:
Metallurgy: Melbourne). Melbourne).
Halbe, D N, 2009, A few general rules for plant design, in Recent La Brooy, S R and Muir, D M, 1994. Gold Processing with saline
Advances in Mineral Processing Plant Design (eds: D Malhotra, water, The AusIMM Proceedings, 303(2):81–88.
P R Taylor, E Spiller and M LeVier), pp 9–14 (Society for Mining,
Mahapatra, A, 2009. Characterization of activated carbon and the
Metallurgy and Exploration: Littleton).
effect of organic fouling on gold adsorption by activated carbon,
La Brooy, S R and Bax, A R, 1985. The fouling of activated carbon MSc thesis, University of Utah, Salt Lake City.
by organic reagents, in Proceedings Chemeca 85: Innovations in the
Marsden, J O, 2009. The role of innovation in mineral processing
Process and Resource Industry, pp 187–191 (Institution of Chemical
and metallurgical plant design, in Recent Advances in Mineral
Engineers: Barton).
Processing Plant Design (eds: D Malhotra, P R Taylor, E Spiller
La Brooy, S R, Bax, A R, Muir, D M, Hosking, J W, Hughes, H C and M LeVier), pp 25–33 (Society for Mining, Metallurgy and
and Parentich, A, 1986. Fouling of activated carbon by circuit Exploration: Littleton).
organics, in Proceedings Gold 100, Vol. 2 Extractive Metallurgy of
Wardell-Johnson, G and Barbetti, K, 2011a. Assessment of activated
Gold, pp 123–132 (Southern African Institute of Mining and
carbon samples from the Sukari Gold Mine, Gold Technology
Metallurgy: Johannesburg).
Group Report, Curtin University, June 2011.
Wardell-Johnson, G and Barbetti, K, 2011b. Assessment of activated
carbon samples from the Sukari Gold Mine, Gold Technology
Group Report, Curtin University, November 2011.

Contents
Carbon management in a
high gold price environment
S R La Brooy1 and J K Claflin2
ABSTRACT
A case study from an Ausenco project is presented to demonstrate the importance of
maintaining high carbon activity to illustrate the driving force for plant improvements
such as a carbon regeneration kiln upgrade.
This paper revisits the factors that influence carbon activity for gold loading, with
emphasis on organic solution components such as flotation reagents.
The process design criteria of the kiln for effective thermal regeneration are reviewed
including the interaction of parameters such as temperature, kiln atmosphere and
residence time.
The kiln design’s impact on the design criteria of ancillary equipment is discussed
and the resultant flow sheets presented.
To assist with kiln selection, the applicability of available kiln types in relation to the
kiln design criteria for the different applications and plant experience are reviewed,
with particular focus on a demanding duty involving flotation in high salinity water.
INTRODUCTION
Maximising gold recovery by minimising solution losses in the leach tails slurry, is an
example of ‘low hanging fruit’ in the quest to improve overall profitability. However,
typically the factors that govern carbon activity usually get much less attention
than are merited. Reactivation kilns and reagents are usually chosen on a cost basis
without much consideration of the process implications. While this approach is quite
adequate for clean oxide ores treated with reasonable quality water, it can be a recipe
for increased gold losses in circuits with flotation, particularly with a float tails leach
and when processing is carried out with saline water. Natural organics in the feed
water and preg-robbing ores are other factors that also elevate the importance of
maintaining the carbon activity.
Ausenco was approached by an operating mine experiencing higher than desirable
gold solution losses. A closer look at the situation revealed:
•• processing with sea water in an arid environment
•• flotation circuit maximised for performance relying on potassium amyl xanthate
(PAX) as collector and a moderately strong ‘alcohol blend’ frother
•• float tails leach in addition to concentrate leach
•• industry standard capital ‘cost-effective’ work horse regeneration kiln
•• kiln operation and maintenance issues.
Together this resulted in loaded carbon activities of approximately ten per cent of
virgin carbon, increasing to approximately 40 per cent after regeneration. A site visit
revealed that the kiln had been pushed beyond its limits to improve carbon activity.
A complete kiln rebuild was able to address some of the most serious maintenance
issues, however, it could not overcome design limitations or address the severe wear
that had occurred. A current plant throughput upgrade would further compound
the issue. The recommendation was to install a new kiln suited to the new duty, then
install an additional kiln, or replace with an identical kiln.
The operation of the regeneration kiln is the first thing to investigate if the gold
solution tail is high and the carbon activity is low, however, this is only an end of
1. FAusIMM, Principal Process Consultant, pipe solution. It is also worth reviewing the choice of flotation reagents and whether
Ausenco, Perth WA 6000. there is potential to minimise overdosing to reduce the carbon deactivation in the
Email: stephen.labrooy@ausenco.com first place. This paper will review the front-end reagent factors leading to carbon
2. Manager Process Technology, Ausenco, Perth deactivation as well as outline what is required to reactivate the carbon to achieve an
WA 6000. Email: jeffrey.claflin@ausenco.com optimum solution.
549
BACKGROUND pH adjusted to 10 with caustic at the beginning of the ageing.

Distilled water was used in all the screening tests. A control
Factors affecting gold loading onto activated carbon sample aged in just distilled water was used for comparison. The
tests showed that organic species such as viscosity modifiers,
Species in solution grinding aids and flocculants, had little effect on carbon
activity, being too large to enter the micropores. However,
Previous research (La Brooy et al, 1984, 1986; La Brooy and
species such as frothers, flotation collectors, lubricating oils
Bax, 1985; Ross, 1993) established that many inorganic foulants
and natural organics (as humates) had a much more significant
typically have a minimal effect on carbon activity provided effect on carbon activity.
that regular acid washing is conducted to remove build-up of
La Brooy and Bax (1985) also observed that the gold loading
species such as carbonates that could block the carbon pores.
rate was less severely affected by xanthate, if the carbon
La Brooy and Robinson (1989) showed in laboratory tests that
had been aged in 20 mg/L sodium ethyl xanthate in the
calcium loadings up to 0.5 per cent had no effect on the kinetic
presence of 200 mg/L sodium cyanide, compared to 20 mg/L
activity of carbon, but higher levels were more detrimental,
xanthate without cyanide or pH control. On the other hand
with the relative activity falling to around 52 per cent by the the glycol frothers showed more deactivation in the presence
time the calcium loading reached 2.3 per cent. Acid washing of 200 mg/L sodium cyanide than frother alone in distilled
is typically able to maintain plant carbon calcium levels within water. It is possible that the reduction in xanthate fouling
the 0.5–1 per cent range, this will ensure that calcium carbonate in the presence of cyanide could be related to charge or pH
fouling does not have a significant effect on the carbon activity. effects, while the enhancement in loading by the uncharged
La Brooy et al (1986) and La Brooy and Bax (1985) investigated frother is probably due to ionic strength, since tests were
the effect of a range of organic species that plant carbon might carried out in distilled water.
be exposed to as shown in Table 1. In each case, prior to the As a reality check, La Brooy et al (1986) investigated the effect
activity test, the carbon was aged for 24 hours by bottle rolling of ageing the carbon for 24 hours in a sample of clear Kalgoorlie
in the foulant solution (20 mg/L concentration unless otherwise plant process water containing around 2 ppm (as sodium ethyl
indicated in Table 1), with 200 mg/L sodium cyanide and with xanthate or SEX) xanthate and around 15–50 mg/L Teric 402
TABLE 1
Relative activity of carbon after 24 hours ageing in various reagents (after La Brooy et al, 1986; La Brooy, 1991).
Reagent Concentration (mg/L) Carbon activity (k) Relative activity (%)

Grinding aid
Dow XFS 4272.00 10 650 100%
Viscosity modifier
Sodium pyrophosphate 2 × 10-3 M 600 94%
Sodium hexametaphosphate 130 540 84%
Freevis 528 20 750 117%
Flocculants
Magnafloc E24 20 560 88%
Cyanamid A2120 20 560 88%
Collectors
SEX 20 280 44%
SIBX 20 140 22%
PAX 20 40 6%
Aeroflot 208 20 270 42%
Frothers
Teric 401 20 110 17%
Teric 402 20 140 22%
Dowfroth 200 20 130 20%
Methyl isobutyl carbinol (MIBC) 20 270 42%
Oils
Mobil ALMO 527 20 200 31%
Diesel 20 190 30%
Multigrade 20 390 61%
Natural organic from vegetation
Humate 20 120 19%
Plant water 50 8%
Distilled water control 640 100%

Carbon management in a high gold price environment
polyglycol frother. The reduction in the carbon activity to only Frothers

eight per cent of the control sample showed that the observations
Table 1 shows that the glycol ether frothers had similar effects
in Table 1 were similar to what might be occurring in plants.
with much more significant deactivation than methyl isobutyl
When comparing the relative activity of carbon samples, it is
necessary to carry out tests in the same water (salinity, pH etc). carbinol (MIBC). Mahapatra (2009) looked at the effect of
Carbon samples from a plant operating with saline water will Oreprep X-133, a commercial frother blend that reportedly
show a higher relative activity in saline water than if tested in contained a mixture of alcohols, heavy aldehydes, esters
pure water (La Brooy and Muir, 1994). and glycols. Mahapatra aged the carbon for 16 hours at a
La Brooy and Bax (1985) and more recently Mahapatra concentration of 1 g/L in the presence of 100 mg/L X-133,
(2009) observed that the gold loading rate was more severely 250 mg/L NaCN and at a pH of 10.5. This resulted in a
affected if the carbon had been aged in foulant solution before reduction of the kinetic activity, which was reduced to around
the gold loading test compared to simultaneous exposure to one per cent of the control!
gold and foulant during the loading test. Mahapatra (2009) The main ingredient in Oreprep X-133 is 2 ethylhexanol
took the test a stage further and showed that no frother was with a molecular weight of 130 and a boiling point of 185°C.
displaced from the carbon when the gold was loaded in a However, being a blend the reagent will have a boiling point
frother-free solution onto carbon that had been previously range. For instance, Orica report that 2 ethylhexanol is the main
aged in the presence of frother and cyanide. Thus, carbon
ingredient in its product DSF 802, which has a boiling range of
spending a long time in the circuit, such as in a float tails
180–350°C. Compare this to MIBC with a molecular weight of
leach, has the opportunity to load flotation reagents that will
102 and a boiling point of 132°C.
not be displaced by gold. Hence the gold loading will be more
severely affected than in a circuit without a float tails leach,
where the carbon would be exposed to the same reagents for Ionic strength
a much shorter time in the concentrate leach, before elution The screening tests reported in Table 1 from La Brooy et al
and thermal treatment. (1986) were conducted in distilled water. However, as already
noted ionic strength and pH will have an effect as well as the
Flotation collectors reagent strength. Dai et al (2010) quantified the enhancement
Figure 1 summarises the effect of an increase in chain in the equilibrium gold loading onto carbon in the presence
length of different xanthate collectors on relative carbon of various cations. Ion pairing and ionic strength both have
activity, while holding the collector concentration constant an effect, for example calcium concentration of 0.1 M had a
at 20 mg/L. As expected, higher molecular weight organics similar effect to 1 M sodium, with little further enhancement
load more strongly onto the carbon, resulting in the relative at higher concentrations of either. For monovalent ions the
activity of the aged carbon falling from 44 per cent down to effects of sodium, lithium and potassium were similar at the
six per cent as the organic chain length increases from ethyl to same molar concentration, while it was found that for divalent
amyl. Hence it is clear that a stronger collector will also have
ions, loading was more enhanced by barium than calcium at
a greater effect on carbon activity.
the same concentration.
Mahapatra (2009) looked at the effect on carbon activity of
ageing for 16 hours in distilled water containing PAX and Hence loading of both gold and organic foulants will be
250 mg/L sodium cyanide at pH 10.5. This work also confirmed enhanced when the plant operates with saline water, eg sea
that as expected increasing reagent concentration in solution water or hypersaline waters in some areas of the Western
increased the carbon deactivation going from a relative activity Australian Goldfields.
of 63 per cent at 20 mg/L PAX down to 23 per cent relative Dai et al (2010) also looked at the effect on gold loading in
activity at 200 mg/L PAX (note xanthate deactivation would the presence of copper cyanide. Their work found that the
have reduced in Mahapatra’s tests compared to those tests in equilibrium gold loading reduced approximately 50 per cent
Table 1, due to the shorter ageing time, higher pH and higher by the time there was 15 000 g/t on the carbon, with essentially
cyanide concentration). This test work also looked at the effect no gold loading by 90 000 g/t copper.
on the equilibrium gold loading and showed that ageing
in 20 mg/L xanthate reduced the equilibrium gold loading The warning message that comes out of Mahapatra (2009)
capacity to 40 per cent of the control, while 200 mg/L xanthate and Dai et al (2010) is that one should only assume the carbon
reduced it to 25 per cent of the control. Hence these reagents are selectivity series applies when the species are loading at the
not only affecting the rate of adsorption but also the equilibrium same time. Gold can not necessarily displace species like
loading capacity of the carbon. frothers or CuCN that are already on the carbon.
50%
40%
Carbon Activity, %
30%
20%
10%
0%
SEX SIBX PAX
FIG 1 – Relative carbon activity after 24 hours in 20 mg/L of different xanthate solutions.

Restoration of carbon activity Reactivation

In view of the deactivation by organic foulants, thermal Avraamides and La Brooy (1987) have summarised the
treatment is an essential part of most circuits using carbon required kiln temperature and atmosphere requirements for
for gold recovery. Unless the plant is processing clean oxide effective thermal reactivation. Figure 2 (after Avraamides and
ore with no vegetation decomposition products in the ore, La Brooy, 1987) shows the effect of time at temperature in
attempting to run without a thermal treatment step results in a steam atmosphere in the thermal reactivation of a sample
significant gold solution losses. For example, one small North of plant carbon under controlled laboratory conditions. The
American operation leaching a copper concentrator float tail figure shows that there was little benefit in extending the
and operating without a carbon kiln reported gold solution time spent for temperatures up to 700°C. However, at higher
losses of 0.5–1 mg/L (Woloschuk, 2013). At current gold temperatures when the endothermic water gas reaction kicks
prices eliminating such losses can quickly repay the cost of a in, the final carbon activity depended on the time spent at
kiln in normal operations. temperature. For this sample there did not seem to be any
benefit in going much beyond ten minutes at temperatures
above 750°C.
Regeneration
In the gold industry the terms regeneration and reactivation C + H2O " H2 + CO
are often used interchangeably. However, regeneration
Hence for carbon, temperatures up to 700°C regeneration
implies simple displacement of the adsorbed species that
occurs, while above that the carbon is reactivated with the
restores the carbon activity. This would be appropriate if
creation of a fresh carbon surface. This process only needs to
the carbon activity could be restored by chemical means,
remove the pyrolysis products (eg carbon black) as further
such as acid washing, steam treatment or thermal treatment
reactivation will destroy the carbon structure reducing the
that simply required the foulants to boil off. It is usually hardness and reducing the overall surface area by ultimately
found that regeneration conditions are not able to restore growing micropores into mesopores.
the carbon activity, especially when the plant incorporates
Laboratory test work has demonstrated the need for steam
flotation with strong collectors. Restoration of activity
in the kiln atmosphere for the water gas reaction to strip away
usually requires reactivation to remove carbonaceous
carbonaceous material on the carbon surface. Steam injection at
residues deposited on the activated carbon surface after
the discharge end of a kiln with counter-current gas and carbon
pyrolysis of adsorbed organics.
flows will ensure that steam is present in the high temperature
Acid washing certainly removes some adsorbed species reactivation zone. Depending on the kiln design and gas
(La Brooy and Bax, 1985), but it is generally unable to restore tightness of the system, quenching or water injected for cooling
the gold loading activity. Jula et al (1985) reported that at one ahead of quenching may well provide sufficient steam.
plant acid washing of lightly fouled loaded carbon could Plants routinely report regeneration of their carbon at
raise the carbon activity from 40 per cent of virgin carbon to temperatures of 500–750°C, but the actual temperatures
60 per cent. Johnson and Third (1991) reported one of the few reached by the carbon may be more modest. In the case of
known examples of acid regeneration at the Copperhead plant. the central Kalgoorlie plant that provided the samples for
The plant operated with hypersaline 200 000 mg/L process the test work in Figure 2, the relative activity of the loaded
water and the activity of the loaded carbon was reduced to carbon was 12 per cent and the plant kiln only raised it to
17 per cent of virgin carbon. However, hot acid washing at around 69 per cent despite a control setting of 650°C ± 50°C.
90°C raised this to 40 per cent, with further improvement to Compared to what could be achieved under laboratory
73 per cent after pressure Zadra elution with three per cent conditions it looked as though the carbon in the plant kiln was
caustic at 140°C. A second acid wash after elution using being heated to less than 400°C. As noted in the introduction
cold acid restored the carbon activity to 89 per cent of virgin the loaded carbon from the plant that prompted the paper
carbon so that there was no need for thermal reactivation. only had a relative activity of ten per cent and the regenerated
The experience at Copperhead suggests that acid washing carbon an activity of 40 per cent, very similar to the sample
before and after elution and higher elution temperatures examined by Avraamides and La Brooy (1987).
may be worth wider consideration for processing clean oxide While operating a kiln above the design throughput will
ores without the use of flotation or presence of significant obviously limit the temperature achieved by the carbon,
quantities of natural organics in the process water. the usual reason for low temperatures is the feed moisture
110
100 Plant kiln
Relative Activity, %
90 20 min
80 10 min
70 0 min
60
50
40
30
20
300 400 500 600 700 800 900
Carbon Temperature, °C
FIG 2 – Effect of time at temperature in thermal reactivation in steam atmosphere.

content exceeding the kiln manufacturer’s design values. disulfide and carbonyl sulfide, along with various organic
Despite advances in other plant areas over the years, problems species that depended on the original organic side chain.
with kiln performance still exist. To understand the kiln The result was that the main decomposition products were
performance at the mine that prompted this paper, Ausenco created over the same three temperature ranges (138–209°C,
looked at the first derivative of the thermogravimetric 209–263°C and 263–376°C) for sodium ethyl xanthate (SEX),
analysis (TGA) data, as Urbanic, Jula and Semple (1985) had sodium isobutyl xanthate (SIBX) and potassium xanthate
done previously. Comparing the traces for the loaded, barren (PAX). The differential thermal analysis (DTG) traces in
and plant regenerated samples (Figure 3), the results align Figure 3 show a single exothermic peak covering this whole
around 550°C which would suggest that this is the maximum temperature range, possibly due to other species like the
temperature to which the carbon was exposed to in the kiln. frother being removed in the intermediate range.
Some other samples taken at other times suggested that the Unfortunately, Fisher (2000) did not attempt any corrections
carbon might only have reached around 450°C. for the water loss from the carbon samples used and due to the
A site visit to supervise maintenance and optimise water adsorbed when the xanthate was loaded from solution
operation of the existing kiln until it was replaced with a so the work was unable to estimate how much of the xanthate
more effective kiln, provided an opportunity do test the TGA was left on the carbon despite heating to 1000°C. It would have
data. Fortunately, the replacement drum was fitted with a been interesting to have an estimate of the organic pyrolysis
sight glass making it was possible to observe the colour of residue that needed to be removed in a reactivation step. It also
the carbon in the kiln and provide more accurate information needs to be noted that since these tests were done under an
than could be achieved by any fixed thermocouple. Figure 4 argon atmosphere they are only indicative and not necessarily
shows the view down the drum looking from the feed end. representative of what happens in a steam atmosphere.
By counting the flights from the discharge end it was possible Fisher (2000) showed that in the pyrolysis of PAX, 40 per cent
to estimate that around 3 m of the drum was above 600°C, of the original mass was still present at 500°C. If we assume
with peak temperatures of 750–800°C towards the discharge
end. However, the carbon only reached 550–600°C (dark
red) just prior to discharge. Given that this was achieved
after maintenance which had allowed the temperature to be
increased and the feed rate reduced without stalling the unit,
it is likely that the lower temperatures 450–550°C suggested
from TGA observations may well have been correct.
Xanthate flotation collectors

A PhD research project by Fisher (2000) used thermogravimetric
analysis coupled with mass spectrometry (TG-MS) or Fourier
transform infra-red (TG-FTIR) to try and identify the species
that were created when various flotation collectors and
frothers were pyrolysed. The pyrolysis products created
when the same collectors were adsorbed onto the activated
carbon and the temperature was increased under an inert
atmosphere was also examined. Results indicated that the
pyrolysis products were complicated when the species
had been adsorbed onto activated carbon. Adsorption onto
carbon appeared to have affected the decomposition process,
resulting in a third amount of weight loss in addition to the
two amounts related to alkali and alkyl decomposition in the
absence of carbon.
With xanthate loaded carbon Fisher (2000) observed the FIG 4 – Picture down kiln tube with kiln set to maximum temperature
major gas species created during pyrolysis were carbon of 750–775°C and kiln operating at design feed rate.
FIG 3 – First derivative of TGA traces for plant carbon samples.

that all that was left was potassium sulfide (and ignoring any the carbon after heating to 1000°C under inert conditions.
formation of potassium hydroxide) then the residue would be The corresponding mass retained with MIBC was around
around 27 per cent of the original PAX mass. Hence it can be five per cent and essentially nothing when pine oil is present.
concluded that 40 per cent of the mass left is sulfur if present Hence any frother apart from pine oil is likely to result in
as organic sulfur compounds and as sulfide, while around a build-up in degradation products on the carbon surface
half the original sulfur has been converted to carbon disulfide unless the carbon is regularly reactivated by spending some
and carbonyl sulfide. This suggests that around 12 per cent time at temperatures above 700°C in a steam atmosphere.
of the organic amyl side chain was still present at 500°C. As This build-up is going to be much less with MIBC than with
previously pointed out adsorption onto carbon affects the glycol frothers, in addition to the higher deactivation caused
decomposition process and products, so mass retention when by glycol frothers in the first place.
PAX is adsorbed on carbon and heated to 500°C is likely to be When Fisher (2000) compared the information on the
even greater than the 40 per cent observed for PAX itself. If the speciation of the evolved gases using different techniques
carbon is only subjected to regeneration conditions without he concluded that the transfer line to the mass spectrometer
reactivation in a steam atmosphere at higher temperatures, resulted in significant losses due to condensation of species.
when the surface deposits can be stripped off by the water The line was maintained at 200°C, but this was clearly not
gas reaction, these low volatility organics will build-up on the hot enough to keep the higher molecular weight species as
carbon over time. gases. This suggests the potential for re-adsorption of higher
The plant that prompted this paper used PAX as their molecular weight species at the back end of a reactivation kiln
flotation collector. Hence since the carbon was only being with concurrent gas flow.
heated to less than 700°C it was not too surprising that the
plant was experiencing low carbon activity as the xanthate
KILN REQUIREMENTS
pyrolysis products would have built up over time. Going
back to Figure 3, it can be seen that acid washing and elution
Choice of kiln
did remove some of the foulants that came off the loaded
carbon in the 200–300°C temperature range. The TGA traces Over the years several kiln designs have been proposed for
also suggest that there is perhaps some caustic digestion of the treatment of carbon for gold recovery. Generally these
organics during the elution process as there is even more fall into two categories being either horizontal or vertical,
reaction in the 700–800°C temperature range with the barren with indirectly-heated horizontal rotating drum systems
sample than the loaded sample. Since the TGA trace is carried accounting for most installations. Von Beckman and Semple
out under an inert atmosphere then only pyrolysis processes (2002) considered two types of horizontal kiln, with and
are occurring as there is no stream atmosphere to allow without an integral predryer.
reactivation. Jula et al (1985) showed that carbonate destruction
shows up as peaks in the TGA first derivative traces in the Vertical systems
range of 600–750°C, with peaks in the 600–670°C range from Some of the earliest indirectly-heated vertical kilns for the gold
laboratory impregnation of the carbon with calcium carbonate. industry used an annular design with the moving carbon bed
The barren carbon peak in Figure 3 is not due to carbonate in the annulus and combustion gases either side of the annulus.
destruction, as there should be less carbonate on the barren Uptake was limited due to tube deformation over time resulting
carbon than the loaded carbon. Calcium assays showed that in carbon flow problems as well as corrosion issues.
acid washing had reduced the calcium concentration on the In the late 1980s and 1990s multiple tube falling packed
loaded carbon from around 7800 g/t to around 5200 g/t with bed carbon reactivation kilns, such as those developed by
a further reduction to around 5000 g/t after elution. Combustion Air, were installed in several mines, mainly
Even with reactivation under optimum conditions hydroxide in Australasia. These offered a lower capital cost approach
and alkali sulfides will be left on the carbon pore surface. When than the alternative horizontal rotating drum systems, with
the carbon is returned to the circuit these sulfides could react reasonable thermal efficiency using the combustion gases for
with cyanide to form thiosulfate/thiocyanate on the carbon predrying. The 200–250 kg/h units could be made to work
surface which may hinder gold adsorption as suggested by effectively with due care and attention to ensure that the
Davidson, Veronese and Nkosi (1979). More recent research moisture content was low enough in the feed to the kiln tubes.
(Mpinga, 2012) suggests thiocyanate in solution may not The Macraes gold mine in New Zealand used two Combustion
be an issue as suggested by Davidson, Veronese and Nkosi Air WCC 200 units in parallel and was able to successfully
(1979) but the fact that it is present at the carbon surface might maintain carbon activity (Ross, 1993) at high levels to combat
make it more of an issue. the preg-robbing ore, while operating a float concentrate
leach with MIBC and SIBX used in the float circuit. Provided
Frothers the moisture content was not too high WCC 250 units were
Fisher (2000) also looked at the gases created when frothers able to get the carbon up to reactivation temperatures within
were pyrolysed compared to what was created when the the first 1–2 m of the packed tubes. Water injection into the
pyrolysis experiments were repeated with samples of bottom section of the kiln ensured a steam atmosphere with
carbon loaded with the individual frothers. Pyrolysis of the counter-current gas and carbon flows in the tubes. However,
frothers in the absence of carbon showed that MIBC boiled as gold plants grew in size, higher throughput kilns were
off without decomposition, while the poly glycol ethers required. As a result, Combustion Air developed larger
underwent thermal destruction. When adsorbed onto carbon diameter tubes in their 500 series kilns. Unfortunately, this
even MIBC was subject to some thermal destruction into four created a problem due to the limited thermal conductivity of
other species, with around 88 per cent coming off as MIBC, carbon, carbon in the centre of the tube failed to dry out, let
compared to 12–14 breakdown products with the poly glycol alone reach reactivation temperatures. Vertical units fell out
frothers. Fisher loaded the frothers directly on to dry carbon, of favour because of less reliable reactivation performance.
so the mass loss results were not complicated by water on Moving packed beds using direct electric heating by current
the carbon. The mass balance showed that for the poly glycol flowing through the carbon, have been proposed. One or two
ethers 20–25 per cent of the initial frother mass remained on small-scale plants were trialled in South Africa. Scale-up issues

were encountered, as well as electrical resistance of the carbon Carbon dewatering screen
reduces with temperature, amplifying any inhomogeneity in Aperture the same as that required for the loaded carbon
current conduction with hot spots continuing to get hotter. screen.
Horizontal systems Kiln feed hoppers

Horizontal rotating drum systems have the advantage that A second hopper enables continuous kiln operation as one
the basic technology was well developed to much larger can be receiving feed and draining while the other is feeding
scale for other applications. Systems for gold applications the kiln. Continuous operation of the kiln will be more
are currently offered by at least six suppliers namely Ansac, efficient as well as minimising issues with brittleness of drum
Consep, FLSmidth, Kemix, Metso and Nutec Bickley (formerly alloys at temperatures below 600°C after exposure to sulfur
GFC). While use of combustion off-gases to predry the feed is compounds. It is also worth considering whether combustion
a useful feature to ensure the feed is dry enough and reduce gases can be used for predrying (von Beckman and Semple,
the required heat input, it is perhaps even more important to 2002). The feed should be fully dewatered and ideally
consider whether carbon and off-gas flow is co- or counter- predried before entering the drum. If combustion gases are
current within the kiln. not used for predrying ahead of the drum, then extra drum
length is required to allow this predrying to happen in the
Many kilns are designed for co-current flow of carbon and
kiln. The feed hopper discharge system needs to be designed
off-gases on the basis that this simplifies the design as water so that it minimises damage to the carbon.
vapour driven off as the carbon dries at the feed end of the
kiln will provide all the required moisture to ensure a steam Kiln feed system
atmosphere throughout the kiln. At first sight this seems a
Devices like screw feeders that can damage the carbon should
neat solution to providing a steam atmosphere, however, it
be avoided in favour of belt/vibratory systems.
neglects the danger of pyrolysis products being re-adsorbed
on the carbon as it cools before leaving the kiln. It is clear Reactivation kiln
from the work of Fisher (2000) that there are high molecular
All equipment surfaces must be safe to touch or adequately
weight organic pyrolysis products that can condense out at
guarded.
temperatures higher than 200°C. Even if these compounds are
As well as specifying the design throughput, there is a need
still volatile it is possible for them to adsorb to some extent.
to consider things such as:
Hence to ensure maximum reactivation of heavily fouled
carbon a counter-current system is recommended such as in •• need for flights or lifters
the Nutec Bickley design. •• inclination, (eg 3°)
•• fill ratio, (eg 8–12 per cent)
Off-gas treatment •• steam injection to discharge end, (eg 0.5 kg/kg carbon).
In addition to carbon dioxide, carbon monoxide and For a plant that includes flotation, especially if there is a
hydrogen (from the water gas reaction), research (Fisher, float tails leach as well as the concentrate leach, the kiln needs
2000) has shown that xanthates will decompose to yield to be heat the carbon to at least 750–850°C and hold it there for
carbon disulfide, carbonyl sulfide and a host of complex ten to 20 minutes at the design feed rate. Lower temperatures
organic species. It is not exactly an environmentally friendly may be adequate depending on the level of fouling and the
cocktail so that at the very least it needs disposal up a stack, foulants involved, but effective reactivation will require the
with, preferably, a gas scrubber or even an incinerator as a ability to get the carbon to at least 750°C.
safer option. Temperature measurement of the carbon within the drum
In circuits where mercury is an issue the off-gases from the is required at the beginning of the reactivation zone and end
carbon reactivation kiln should go through a mercury capture of the heat transfer zone before it discharges from the drum.
system, as the electro-winning (EW) sludge retort and the Viewing ports are essential to enable visual inspection of the
carbon reactivation kiln would be the main sources of mercury inside of the drum to provide information on actual drum and
emission from the circuit. Mercury abatement system for the carbon temperatures.
kiln predryer off-gases and final discharge could consist of a Effective sealing is required to ensure a steam atmosphere
spray wet scrubber, a venturi scrubber, a wet gas condenser, and minimise air ingress. All seals must be easy to inspect and
a coalescer, with final gas polishing in sulfur impregnated easy to replace. Separately supplied steam (steam injection
activated carbon columns. Mercury will tend to condense out at the discharge end) can be useful in ensuring the steam
in cooler areas of ducting and equipment so mercury traps atmosphere, but care is required to ensure that steam does not
will need to be included within the gas capture system to over-cool the carbon in the kiln. The carbon is largely heated
allow for this. by conduction of heat from the particles in contact with the
drum surface, as individual particles may not meet the drum
What to look for when purchasing an surface during the reactivation process.
In directly-heated driers and similar equipment, the use of
effective reactivation kiln system lifters is important to increase the contact between the particles
and the gas flowing down the drum. For carbon regeneration,
Sand trap this process is flawed. The vapours flowing down the drum
Depending on the efficacy of grit screening ahead of carbon- are primarily water vapour produced by drying the carbon,
in-pulp (CIP)/carbon-in-leach (CIL) a sand trap system may and while the water vapour is important to control oxidation
be required to minimise the amount of mineral matter fed to and produce a water gas reaction to reactive the surface of the
the kiln along with the carbon leading to increased wear and carbon, the water is generated at 100°C and does not readily
build-up of accretions on cooler surfaces. heat up. In fact, excessive contact with the water vapour can

cool the carbon that has been in contact with the kiln wall MIBC was used as frother in the all the initial test work
(Figure 5). for this plant in which Ausenco was involved. However, the
For a kiln with lifters, or some form of notched exit dam company that subsequently built the plant, noted that MIBC
and water quenching of the discharge carbon, there can be had a flash point of 43°C, so explosion proof motors and
intermittent bursts of steam as each batch of carbon exits the switches were required for the reagent circuit. This company
kiln and is quenched. This can lead to periods of inadequate suggested that it would be safer and more cost-effective to go
steam in the atmosphere until the next batch of carbon is lifted with a frother with a higher flash point and recommended
to the discharge. Huntsman Polyfroth W22. It is interesting to speculate that
had MIBC been used on the plant, the carbon would probably
Fluid bed cooler for discharge carbon be less deactivated and the gold solution losses lower and the
Carbon should be gently cooled to prevent fracture by thermal additional cost for explosion proofing of the reagent delivery
shock. A fluid bed cooler using steam will ensure more gentle system repaid many times over.
treatment of the discharge carbon rather than just quenching
in water. This fluid bed can discharge directly onto a screen CONCLUSIONS
to simplify fines removal from the reactivated carbon and Maximising gold recovery requires maximising carbon
at the same time minimise the risk of carbon burning on the activity across the circuit. Minimising fouling caused by
discharge screen. reagents should be a priority, otherwise even with the best
possible reactivation the carbon activity can still be reduced
Fines screen to low levels while it is still in the circuit.
Fines need to be removed before the carbon is returned to Restoration of carbon activity requires reactivation in most
the circuit. Even with gentle treatment that minimises carbon cases, especially if it is subject to organic foulants such as
breakage through the process, carbonaceous fines will be frothers and flotation collectors. Effective reactivation requires
produced by pyrolysis products that condense onto cooler the carbon to be heated to temperatures above 700°C and held
areas of the drum, to later flake off and be carried through the at there for some time, without air ingress. To achieve this, the
kiln with the carbon and get activated in the reactivation zone. kiln needs to have easily replaced seals; an observation port; a
satisfactory carbon discharge mechanism with water vapour
Off-gas scrubber coverage throughout the kiln and an off-gas collection system
Spray or venturi systems with caustic make the best off-gas to required OH&S standards.
scrubbers.
REFERENCES
DISCUSSION Avraamides, J and La Brooy, S R, 1987. Evaluation of different
reactivation systems available for carbons used for gold recovery,
Plant example in Proceedings Equipment in the Minerals Industry: Exploration
The plant that prompted the paper employed PAX as flotation Mining and Processing 1987, pp 1–7 (The Australasian Institute of
collector and Orica DSF 802 as frother. As reported above from Mining and Metallurgy: Melbourne).
the work of Mahapatra (2009) a combination of 2 ethylhexanol Dai, X, Breuer, P L and Jeffrey, M I, 2010. Modeling the equilibrium
and PAX is an extremely effective way to deactivate carbon. loading of gold onto activated carbon from complex cyanide
The only way to minimise gold solution losses is to ensure solutions, Minerals and Metallurgical Processing, 27(4):190–195.
effective thermal reactivation of the carbon. However, from Davidson, R J, Veronese, V and Nkosi, M V, 1979. The use of
the material presented in this paper it can be seen that this end activated carbon for the recovery of gold and silver from gold
of pipe solution may still be limited in its effectiveness when plant solutions, Journal Southern African Institute of Mining and
faced with this particular reagent consumption and longer Metallurgy, 79(10):281–297.
carbon residence time in a float tails leach. Fisher, N G, 2000. Identification of organic fouling agents on activated
carbon by evolved gas analysis, PhD thesis (unpublished), Curtin
University, Perth.
Johnson, G and Third, B, 1991. Comparison of AARL and high
pressure Zadra stripping, why one carbon elution process works
best, in Proceedings Randol Gold Forum Cairns 1991, pp 373–377
(Randol International: Golden).
Jula, R J, Urbanic, J E and Faulkner, W D, 1985. Regeneration of
activated carbon used for recovery of gold, SME Preprint 85–108,
9 p (Society for Mining, Metallurgy and Exploration: Littleton).
La Brooy, S R, 1991. Carbon fouling, in Carbon Fouling in CIP and
Gold Technology Workshop Course Notes (Australian Mineral
Foundation: Adelaide).
La Brooy, S R and Bax, A R, 1985. The fouling of activated carbon
by organic reagents, in Proceedings Chemeca 85: Innovations in
the Process and Resource Industry 1985, pp 187–191 (Institution of
Chemical Engineers: Barton ACT).
La Brooy, S R, Bax, A R, Muir, D M, Hosking, J W, Hughes, H C and
Parentich, A, 1986. Fouling of activated carbon by circuit organics,
in Proceedings Gold 100 Vol. 2 Extractive Metallurgy of Gold 1986,
pp 123–132 (South African Institute Mining and Metallurgy:
FIG 5 – Carbon spilling from flights can be cooled in water vapour. Johannesburg).

La Brooy, S R, Hosking, J W, Muir, D M, Ruane, M, Smith, I and Mpinga, C N, 2012. The extraction of precious metals from an alkaline
Hinchliffe, W D, 1984. Studies on the fouling and regeneration cyanided medium by granular activated carbon, MSc thesis
of CIP carbon, in Proceedings Gold Mining, Metallurgy and Geology (unpublished), Stellenbosch University, South Africa.
1984, pp 257–269 (The Australasian Institute of Mining and
Ross, A, 1993. The fouling of activated carbon – a study based on
Macraes gold mine, in Proceedings 27th Annual Conference, New
La Brooy, S R and Muir, D M, 1994. Gold processing with saline water, Zealand Branch 1993, pp 175–183 (The Australasian Institute of
The Australasian Institute of Mining and Metallurgy Proceedings, Mining and Metallurgy: New Zealand Branch).
303(2):81–88.
Urbanic, J E, Jula, R J and Faulkner, W D, 1985. Regeneration of
La Brooy, S R and Robinson, J J, 1989. Practical aspects affecting gold activated carbon used for gold recovery, Minerals and Metallurgical
adsorption and carbon fouling, in Proceedings Randol Gold Forum Processing, 2(4):193–198.
Sacramento, 89, pp 275–279 (Randol International: Golden).
Von Beckman, J and Semple, P G, 2002. Selection and design of
Mahapatra, A, 2009. Characterization of activated carbon and the carbon reactivation circuits, in Mineral Processing Plant Design,
effect of organic fouling on gold adsorption by activated carbon, Practice and Control (eds: A L Mular, D J Barratt and D N Halbe),
MSc thesis (unpublished), University of Utah, Salt Lake City. pp 1680–1693 (Society for Mining, Metallurgy, and Exploration:
Littleton).

Contents
The impact of gravity gold recovery at

Kalgoorlie Consolidated Gold Mines
A Giblett1, D Hillier2, K Parker3 and V Ramsell4
ABSTRACT
Kalgoorlie Consolidation Gold Mines (KCGM) operates the Kalgoorlie Super Pit
open cut, Mt Charlotte underground mine, Fimiston process plant and Gidji roaster
facility for joint venture partners Newmont Mining Corporation and Barrick Gold.
In 2004, the Fimiston plant was upgraded by the addition of gravity concentration to
the Fimiston and Mt Charlotte grinding circuits, including the installation of a central
intensive cyanidation reactor to process the gravity concentrates. The gravity project
was justified after detailed analysis demonstrating an expected increase in overall
gold recovery of approximately one per cent, in addition to reduced elution costs and
reduced gold inventory in flotation concentrate stockpiles. The gravity circuits were
commissioned in August 2004, and have since consistently recovered 10 to 20 per cent
of the gold present in the KCGM ore, delivering a reduction in coarse gold losses
equivalent to 0.45 per cent additional metal recovery and contributing to improved
metal accountability.
INTRODUCTION
KCGM is a 50/50 joint venture between Barrick Gold and Newmont Mining
Corporation, managing mining and processing activities in and around the city of
Kalgoorlie–Boulder, 600 km east of Perth, Western Australia. Gold ore is mined
from the Super Pit open cut and the Mt Charlotte underground mine, and processed
through the Fimiston plant, with a refractory telluride/pyrite flotation concentrate
processed at the Gidji roaster facility, situated some 20 km north of the city of
Kalgoorlie–Boulder. Additional concentrate treatment capacity is provided by the
ultra-fine grinding (UFG) circuits installed at both the Fimiston and Gidji facilities,
where the concentrate is ground to 80 per cent passing 10–15 µm ahead of cyanidation.
The first discovery of gold in Kalgoorlie in 1893, by Paddy Hannan, marked the
commencement of the great Western Australian gold rush and the mining of the
famous ‘Golden Mile’ that continues to this day. Since 1893, the Kalgoorlie–Boulder
region has produced over 55 Moz of gold making it one of the richest goldfields in
the world, ranking alongside the Carlin trend in Nevada and the Witwatersrand in
South Africa in terms of total gold production. Following the exhaustion of the rich
alluvial gold deposits in the region, in the early years following Hannans initial gold
discovery, mining activities have focused on the refractory pyrite and gold-telluride
mineralisation that remains the dominant ore source to this day.
The KCGM joint venture was originally formed in 1989, as a management company
representing the interests of the joint venture partners Homestake Gold of Australia
and Gold Mines of Kalgoorlie and consolidating the mining leases and processing
activities of CSR Paringa, North Kalgurli Mines Limited and Kalgoorlie Mining
Associates. Prior to the formation of KCGM, the Golden Mile had been worked during
1. FAusIMM, Senior Technical Advisor – Mineral the 1980s by a series of open pits, including the Kemlo, Croesus, Eclipse, Central,
Processing, Newmont Mining Corporation, Paringa, Great Boulder, South, Judd and Brownhill pits (Cutifani and O’Neill, 1999).
Englewood CO 80112, USA. The KCGM concept was to form a massive ‘superpit’ incorporating these mining
Email: aidan.giblett@newmont.com areas. Following the 2001 acquisition of Homestake Mining by Barrick Gold and the
2. FAusIMM(CP), Senior Manager Metallurgy, 2002 acquisition of Normandy by Newmont Mining Corporation, the current day
Barrick Australia/Pacific, Perth WA 6000. joint venture ownership of KCGM was established. The KCGM operations have
Email: djhillier@barrick.com been Australia’s largest gold-producing mining operation year after year from 1990
3. AAusIMM, Senior Process Metallurgist, through to the present day, cumulatively producing almost 7 Moz more than the next
Kalgoorlie Consolidated Gold Mines, most productive gold mining operation during that time, as shown by Figure 1.
Kalgoorlie WA 6430.
Email: kparker@kalgold.com.au This paper reviews the historical application of gravity concentration for gold
4. MAusIMM, Metallurgist – Projects and recovery on the Golden Mile and the current day use of gravity concentration for gold
Planning, Kalgoorlie Consolidated Gold recovery at the KCGM Fimiston plant. The KCGM gravity circuits are an example of
Mines, Kalgoorlie WA 6430. best practice in both circuit design and the execution of a rigorous assessment of the
Email: vramsell@kalgold.com.au impact of gravity gold recovery on overall plant performance.
559
A Giblett et al
FIG 1 – Australian gold production since 1986 (source: Metals Economics Group).
GRAVITY GOLD AND THE GOLDEN MILE the gold to amalgamation of gravity concentrates. Gravity
Since as early as 1897, the refractory sulfide and telluride gold was recovered from the primary rod mill discharge by a
mineralisation of the Golden Mile has been the primary hydraulic trap and narrow plane tables with primary gravity
focus of mining activities in the region, as a function of the concentrate cleaning by strake tables ahead of amalgamation.
association of high gold grades and larger ore reserves with Blaskett (1952b) reports the recovery of gravity gold at
this style of mineralisation. The occurrence of gold enriched the Croesus plant from the mill discharge by strake tables
pyrite and gold-telluride minerals such as Calaverite (AuTe2), and subsequent concentrate amalgamation to recover
Sylvanite ([Au,Ag]2Te4) and Petzite (Ag3AuTe2) in the Golden approximately 14 per cent of the total gold.
Mile ores are well documented as is the refractory response of The CSR Paringa operation, under construction at the time
these minerals to conventional gold recovery by cyanidation. of the 1983 Randol review, was reported to include a gravity
Numerous references are also found to the occurrence circuit comprised of strake tables as primary concentrators
of native gold in the Golden Mile ores although the with the gravity concentrates to be cleaned by shaking
concentration and size of gold grains have been reported to tables ahead of amalgamation. Blaskett (1952b) reported
vary significantly. Gravity concentration was widely applied historical gold recovery by gravity at the Paringa Mining
for the purpose of producing sulfide-telluride concentrates for and Exploration Co Ltd plant by straking the mill discharges
downstream processing prior to the adaptation of the froth and concentrate amalgamation of 11–23 per cent of total gold
flotation process in the 1930s. However, the application of in feed.
gravity concentration for free gold recovery is given relatively Blaskett (1952c) discussed the operation of the Lake View
limited coverage in the literature in comparison to the wealth
and Star Ltd process plant, at the time the largest operating in
of documentation addressing the metallurgical challenges
Kalgoorlie at 60 000 t/m capacity, recovering approximately
and success related to the processing of the refractory Golden
13 per cent of the total gold in feed by strake tables operating
Mile ores.
on mill discharge ahead of concentrate amalgamation. Stevens
The use of gravity concentration at the Gold Mines of (1933) had previously reported typical gold recoveries by
Kalgoorlie (GMK) is discussed by Penrose (1948), describing gravity of 24 per cent, up to as high as 35 per cent for the Lake
the processing of ores from the eastern end of the Golden View and Star process plant by straking mill discharges ahead
Mile, specifically the Australia East open pit, Iron Duke, New of amalgamation.
North Boulder and Oroya shafts. The gravity recovery circuit
is reported to include a hydraulic gold trap on the primary mill While the literature also contains references to ore types
discharge and strake tables on the secondary mill discharges. devoid of free gold and gold recovery flow sheets not
The strake table concentrate was ground in amalgam barrels including gravity gold recovery, it is of interest and relevant
and retorted to yield approximately seven per cent of the gold to this discussion that numerous examples are cited for
in the mill feed. Blaskett (1952a) reports that strakes were gravity gold recoveries in the range of 5–25 per cent of mill
achieving approximately 20 per cent recovery of the feed gold feed gold for gold recovery plants treating various Golden
at GMK to gravity concentrate in the late 1930s, declining to Mile ores. It is further notable that a 1993 review of KCGM’s
four per cent in 1946 and ultimately the gravity gold recovery mineral processing operations at the time by Cutifani (1993),
was so low by 1952 that gravity concentration was no longer encompassing the Fimiston, Oroya, Croesus, Gidji and
operational. On the addition of a refractory ore circuit to Mt Percy treatment operations, is absent any reference to the
process ore from the Fimiston underground mine in the early use of gravity concentration for gold recovery at any of these
1980s a gravity circuit consisting of strake tables processing operations. This would seem to indicate that the importance
cyclone overflow for coarse gold recovery was included of gravity concentration ahead of cyanidation or flotation
(Randol International Ltd, 1983). had declined over the ten years since the Randol publication.
Randol International Limited (1983) also describes the The Fimiston concentrator, predestined to replace the
operations of North Kalgurli Mines, operating the Croesus durable Oroya and Croesus process plants, was certainly
treatment plant and reporting gold recoveries by gravity and commissioned and operated for a substantial number of
amalgamation ranging from 10–25 per cent. Cashman (1980) years without employing gravity concentration for free gold
describes the same plant recovering typically 9.4 per cent of recovery ahead of flotation and cyanidation.

The impact of gravity gold recovery at Kalgoorlie Consolidated Gold Mines
THE FIMISTON PLANT when these facilities were sequentially decommissioned

commensurate with the expansions of the Fimiston plant
The Fimiston plant was originally commissioned in 1989 as
and Gidji roaster, the depletion of the Mt Percy pits and the
a 2 Mt/a sulfide concentrator based on three-stage crushing,
consolidation of the Golden Mile workings into the current
single-stage ball milling flotation and cyanidation and Super Pit.
expanded in 1991 to 5 Mt/a by the addition of a second ball
In 2000, the Fimiston plant was upgraded again by doubling
mill and second carbon-in-leach (CIL) train. In 1995, the plant
the capacity of the Fimiston flotation circuit and installing
was upgraded again by the addition of the 36 ft Fimiston
the Mt Charlotte flotation circuit. The Gidji UFG circuit was
semi-autogenous grinding (SAG) mill, adding a third CIL
installed in 2001, followed by the installation of the Fimiston
train and introducing the Mt Charlotte grinding circuit. UFG circuit in 2002. The overview level KCGM metal recovery
Total plant capacity was now 12 Mt/a with the ability to process flow sheet is shown in Figure 3.
process 2 Mt/a through the Mt Charlotte circuit and 10 Mt/a
The Fimiston plant is comprised of two circuits, the Fimiston
through the Fimiston plant. Figure 2 demonstrates the
circuit and the Mt Charlotte circuit. The Fimiston circuit is the
annual tonnes processed and gold produced by the KCGM
larger of the two and processes approximately 10 Mt/a of
operations since 1990. open pit ore through the SABC circuit consisting of one 36 ft
KCGM had operated the Paringa, Oroya, Croesus and SAG mill and two 18 ft ball mills. A pyrite-telluride flotation
Mt Percy ore processing facilities until the mid to late 1990s concentrate is produced by rougher and cleaner flotation, and
FIG 2 – Kalgoorlie Consolidation Gold Mines annual production statistics.
FIG 3 – Kalgoorlie Consolidation Gold Mines process flow sheet overview.

A Giblett et al
the flotation tailings is thickened and distributed between the bias remained and a commonly postulated rationale was the
two large CIL circuits, train 2 and train 3. The Mt Charlotte occurrence of coarse gold particles not being recovered by
circuit processes 2 Mt/a of ore from the open pits and the flotation and cyanidation additionally bypassing the Isolok
Mt Charlotte underground mine through a grinding circuit tailings pipe sampler. In any event since 1999 a constant
comprising a single 24 ft SAG mill and one 15 ft diameter ball seven per cent discount was applied to the indicated plant
mill. Scats from the Mt Charlotte SAG mill are transferred to gold production to provide a more reliable estimate of final
the Fimiston mill pebble crushing circuit. The Mt Charlotte end of month gold production.
ball mill cyclone overflow is treated by flotation to produce In May 2002, samples of rougher tailings and cleaner
a rougher-scavenger concentrate for cleaning in the Fimiston scavenger tailings were submitted to Amtel for detailed
cleaner flotation circuit. The Mt Charlotte scavenger tailings analysis of the gold mineralogy and distribution. This analysis
are thickened and then combined with the Fimiston scavenger generated a number of significant observations:
tailings for leaching in CIL trains 2 and 3.
•• 58 per cent of the gold in rougher tailings was free gold,
The combined flotation concentrate is deslimed to produce 22 per cent of which was coarser than 100 µm
a sub10 µm slimes fraction for cleaning to reject the majority
•• 44 per cent of the gold in the scavenger tailings was free
of the slimes ahead of filtration. The plus 10 µm fraction is
gold, although almost exclusively minus 40 µm in size.
further classified to produce a minus 60 µm fraction for UFG.
The plus 60 µm and slimes concentrate fractions are vacuum Further analysis by Amtel of a CIL train 3 tailings sample,
filtered and trucked to the Gidji plant for roasting and calcine observed 19 per cent of the gold present as exposed gold grains,
leaching. The intermediate fraction is ground to 80 per cent averaging 13 µm in size. However, the presence of occasional
passing 12 µm in a M3000 IsaMill before intensive cyanidation large gold flakes, greater than 100 µm in size, was noted.
in CIL train 1, with train 1 tailings reporting to the main CIL Additionally, surface coatings (chloride, phosphate and
trains 2 or 3 for additional residence time and gold extraction. carbonate based) were observed in several of the finer gold
particles thought to have contributed to reduced floatability
QUANTIFYING THE POTENTIAL IMPACT OF GRAVITY RECOVERY of gold particles, which may also have contributed to reduced
amenability of the gold particles to cyanidation. The relative
The KCGM site has over time initiated numerous absence of gold-telluride minerals from the CIL tailings sample
investigations into the nature of gold losses to the Fimiston was considered to demonstrate the enhanced floatability of
plant CIL tailings stream, with several initiatives occurring these minerals above that of native gold particles.
between 1996 through to 2002 (Hillier and Tilly, 2003).
Samples of new ore, grinding circuit streams, CIL tailings In order to better quantify the significance of free gold
and TSF solids were tested typically by processing the losses to CIL tailings a project was implemented to operate
sample through a laboratory three-inch Knelson to produce a 12-inch Knelson concentrator on a bleed of the CIL
a gravity concentrate for analysis by assay, cyanidation or train 3 tailings stream for a continuous period from June to
amalgamation – or in some cases by screen fire assay analysis. September 2003. Knelson concentrates were collected over a
Varying grind sizes and sample masses were used for this test 12-hour shift and then subjected to intensive cyanidation to
work resulting in a significant spread of gravity recoverable determine the free gold content, or alternatively fed through a
gold estimates, but also allowing general agreement on the three-inch Knelson concentrator with the subsequent gravity
presence of a gravity recoverable gold component. concentrate fire assayed to extinction. The feed rate to the
12‑inch Knelson during this time averaged approximately
The primary motivations for these investigations were a 2 t/h, the mass pull to concentrate averaged 0.02–0.03 per cent
sustained focus on continuous improvement by the KCGM and intensive cyanidation recoveries typically ranged from
metallurgical staff, and the presence of a long-standing bias in 85–95 per cent. Initial review of the trial results indicated a
the metallurgical accounting process. Over a number of years 0.027 g/t Au gravity recoverable portion of the CIL tailings
a consistent negative bias in the metallurgical accounting stream, representing approximately one per cent of the gold
system had been observed such that from 1999 to 2002 the content of the mill feed.
actual gold production was lower than the indicated gold
production by an average of seven per cent (Figure 4). During the period of this 12-inch Knelson test campaign the
Best practice, multistage cross stream samplers had been Fimiston plant was fed material representing approximately
installed on the Fimiston circuit flotation feed (1998) and 50 per cent of the known geological domains of the Superpit
the Mt Charlotte flotation feed (2000) in order to improve resource, with mine geology reporting that the remaining
the reliability of the indicated mill feed grade. However, the ore domains were expected to have an even higher coarse
gold content. Mineralogical assessment of the gravity
concentrates by MLA confirmed the dominance of gold and
silver alloys (>80 per cent) with low levels of gold tellurides,
calaverite (<five per cent) and sylvanite (<five per cent), again
supporting the observation of improved floatability of the
telluride species over the native gold-silver alloys.
Further confirmation of the presence of coarse gold in the
mill feed was provided by analysing the circulating gold ratio
in the three ball mill circuits over a three week period, with
the gold assay of the cyclone underflow being approximately
five times higher than the cyclone overflow assay for the
Fimiston ball mills, and nine times higher for the Mt Charlotte
ball mill. Recirculating gold assay ratios of this magnitude are
a reliable indicator of the potential for gravity gold recovery.
Additionally historical data on the gold content of the ball
FIG 4 – Kalgoorlie Consolidation Gold Mines metallurgical mill discharge streams was reviewed indicating consistently
accounting variances (1999–2003). higher gold assays in the ball mill discharge.

A composite of the July 2003 Fimiston SAG mill discharge •• a reduction in CIL gold in solution losses
was sent to the A J Parker CRC for gravity recoverable gold •• a reduction in the Gidji calcine leach tailings grade
(GRG) testing using the methodology developed by the late
•• reduced carbon-stripping costs
Professor Andre Laplante. Additionally, composites were
submitted to Knelson concentrators and Falcon Concentrators •• reduced inventory value of stockpiled flotation
for similar testing. Figure 5 shows the cumulative GRG concentrate.
distribution as determined by the Knelson and A J Parker The optimum project economic value was determined for
Centre testing which are almost identical. This testing was a gravity circuit comprised of two centrifugal concentrators
used as the basis for modelling of the plant gravity recovery dedicated to each of the Fimiston circuit ball mills, one
levels by the equipment vendors and the Parker CRC, centrifugal concentrator for the Mt Charlotte ball mill, and
resulting in estimates of plant scale gravity recovery ranging a central intensive cyanidation reactor. Adding additional
from 24 per cent to 38 per cent. gravity concentrators was determined to marginally increase
the total gold recovery through modelling by Professor
Laplante although the increased recovery was insufficient to
THE BUSINESS CASE FOR GRAVITY overcome the additional capital within the constraints of the
Following the completion of the technical review a capital financial assessment.
request was compiled for the implementation of a permanent
gravity recovery circuit including Knelson concentrators in all
three ball milling circuits and a central intensive cyanidation
GRAVITY CIRCUIT FLOW SHEET AND EQUIPMENT SELECTION
reactor. Total project capital cost was estimated at $9.6 M, and It was determined that each of the three ball mills would have
a payback period of 23 months was determined based on the its own dedicated gravity circuit. For each of the Fimiston
following key deliverables: ball mills a dedicated gravity feed pump would be added to
the mill discharge hopper to feed a portion of the ball mill
•• an increase in overall gold recovery of 0.9 per cent
discharge to a dedicated gravity screen. Screen undersize
reports to a splitter box, where feed flow to the parallel
concentrators is controlled by actuated knife-gate valves.
Gravity screen oversize combines with the concentrator
tailings and returns by gravity to the ball mill discharge
hopper. Gravity concentrate discharges to a pump hopper
and is pumped to a settling cone ahead of the intensive
cyanidation reactor. The Mt Charlotte gravity circuit follows
the same process flow with the fundamental difference being
that only one concentrator is installed. The overall KCGM
gravity flow sheet is shown in Figure 6.
An important design consideration for the intensive
cyanidation facility was the likelihood of significant amounts
of gold-telluride minerals being present in the gravity
FIG 5 – Gravity recoverable gold content of samples concentrate. Gravity testing of ball mill discharge samples at
from Kalgoorlie Consolidation Gold Mine. KCGM (2003) and subsequent mineralogical analysis of the
FIG 6 – Kalgoorlie Consolidation Gold Mines gravity flow sheet.

A Giblett et al
gravity concentrates had identified up to 50 per cent of the FLEX*Flo® system was a new product line for Knelson, and
free gold content being present as calaverite. Although it was has subsequently been suspended.
demonstrated during this test work that intensive cyanidation
of the gravity concentrate yielded similar gold extraction to Knelson feed and bypass slurry valves
samples where native gold dominated the gravity concentrate The installed slurry knife-gate slurry valves failed prematurely
(87 per cent versus 88–90 per cent gold extraction), it was and were replaced with DeZURIK knife-gate slurry valves.
considered necessary to ensure that the intensive cyanidation
reactor could operate at elevated temperatures, and allowed Feed screen panel apertures
the addition of leach accelerants.
The initial high wear rates of the Knelson bowls was
The project design criteria inputs were initially developed significantly relieved by the installation of 2.2 mm aperture
by Roche Mining (2003) and included the following key items: panels on the feed screen, to reduce top-size of the feed to the
•• gravity circuit design feed rate approximately Knelsons.
100 per cent of fresh mill feed rate
•• 300 t/h for Mt Charlotte gravity circuit Knelson rotary unions
•• 600 t/h each for the Fimiston ball mill 1 and ball mill 2 The initially installed standard low-cost mild steel rotary
gravity circuits unions on the Knelson concentrators were replaced with SAF
2205 rotary unions within two to three months of operation.
•• gravity screen aperture size 3.2 mm, design capacity of
The short service-life of the mild steel components was due
approximately 45 m3/h/m2
to the nature of the hypersaline water used at the Fimiston
•• concentrator cycle time of one hour, concentrate batch processing plant.
mass 50 kg per cycle
Since these changes and improvements were made,
•• intensive cyanidation leach reactor capacity of 6 t/d, the Knelson KC XD-48 concentrators have performed
leach extraction of 90 per cent. reasonably well.
Technology considered for the gravity concentrators were
the Knelson and Falcon centrifugal concentrators. At the time Consep acacia reactor
of the analysis the Falcon SB5200 proposed for the project
had only one installation compared to multiple installations Acacia reactor level sensor
for the Knelson XD48. Subsequently largely on the basis of Some initial problems were experienced with the leaching
reduced risk the project opted to move forward with the vessel level sensor, believed to be have been caused by
more established Knelson XD48s as the preferred gravity interference from the vessel lid, to the ultrasonic level sensor.
concentrator. This was evidenced by successful operation of the sensor with
The technologies considered for intensive cyanidation were the lid removed. These issues were resolved by replacement
the ConSep Acacia CS6000 and the Gekko Inline Reactor of the lid, incorporating an elevated sensor position with large
model ILR 10 000, with concentrate leach capacities of 6 t and diameter access for the sensor head.
10 t per batch respectively. At a high level both intensive leach
technologies had demonstrated their ability to leach gravity Diffusion screen
concentrates at the production scale and were considered The diffusion screen proved to have a service life of
appropriate technologies for the application. Tilly and Hall approximately three months, at which point it required
(2003) performed a thorough assessment against the site replacement and refurbishing. A pressure sensor was
specific requirements of the installation, including the slow installed on the Consep acacia reactor unit at KCGM, as a
leach kinetics of the telluride minerals and the salinity of trial, to measure the back-pressure of the feed screen and
the site process water. Based on the demonstrated history of thereby to provide some indication of the current condition of
operating the nominated size of leach reactor, operating at the screen. This has since worked reasonably well.
elevated temperatures in the presence of leach accelerant and
the rubber lined leach chamber as security against corrosion
it was recommended that the Acacia CS6000 be selected for
Leach vessel froth
the project. Intermittently a froth developed on the surface of the leaching
vessel solution during the deslime step. Further to this,
occasionally entrained air within the unit was identified, that
COMMISSIONING AND STABILISATION created flotation surfaces (bubbles) for sulfides to attach to,
The KCGM gravity circuit was commissioned during July to during the deslime step. A survey was conducted of a deslime
August, 2004. A number of improvements were required to stage to measure gold loss and negligible losses were detected.
be made to the circuit, and these were carried out across the A future potential improvement could be the installation of a
subsequent six-month operating period after commissioning. froth ring to permit capture of any floatable species that rise
The following sections describe the various improvements to the surface during the deslime step.
and changes made to the three gravity circuits and intensive
cyanide leach reactor. Double leach test work program
A double leach test work program was embarked upon,
Gravity concentrators which quantifies gold extraction rates per batch, which then
The following changes were required, concerning the KC enables rational optimisation of the Consep acacia reactor
XD‑48 gravity concentrator circuits installed. leaching regime. Gold extraction is quantified by pregnant
solution profiling and re-leaching each batch (that is a
FLEX*Flo® Knelson concentrator bowls double leach). An example of an unoptimised double leach
Due to premature failure of the white FLEX*Flo® strips, these profile is shown in Figure 7. Due to the large capacity of the
bowls were replaced with the standard G4.6 Bowls in all five Consep acacia reactor installed and the available intermediate
KC XD Knelson concentrators. At the time of installation, the gravity concentrate storage capacity, it is possible to carry

out the double 24-hour leach cycle, without compromising rationalised as a desensitising of plant recovery to grind size
the required centrifugal concentrator mass recovery. During as a function of the reduced accumulation of coarser gold
this work, it became obvious that a full 48-hour leach was particles in the recirculating load and the ability to operate at
required to consistently achieve the required levels of gold a coarser classification cut size without coarse gold particles
dissolution, along with an intermediate reagent top-up step. reporting to the flotation feed.
The latest leach recipe and regime currently in use at KCGM,
after a number of years of optimisation consists of: The impact of gravity recovery on metallurgical accounting
•• first leaching stage (16 hours): As stated previously the KCGM metallurgical balance was
•• leach solution mix contains freshwater (12 300 L), subject to an accounting bias for a number of years prior to
caustic (68 L of 50 per cent w/w solution), Cyanide the installation of the gravity circuit. The persistence of the
(1000 L of 30 per cent w/w solution), LeachAid UL bias was such that a seven per cent adjustment was routinely
(15 kg solid) applied to the indicated gold production in order to produce
a more reliable estimate of month end gold production and
•• after six hours leach time, 400 L of additional cyanide
mill feed grades. It was believed in some circles that the cause
solution is added, and 3 kg of LeachAid UL.
of this bias was the presence of coarse gold in the tailings
•• second leaching stage (16 hours) – no rinsing stage: representing an unaccounted loss of gold due to ineffective
•• leach solution mix contains freshwater (12 300 L), sampling. However, the installation of cross stream samplers
caustic (68 L), cyanide (750 L), LeachAid UL (15 kg). for the Fimiston and Mt Charlotte flotation feed streams in
•• leaching temperature: 33°C 1998 and 2000 respectively did not remove the bias. Figure 8
•• gravity concentrate mass: 6–7 t presents a cumulative sum (CuSum) chart of the metal
balance discrepancy illustrating the consistent negative bias
•• leach solution pH 13.0. from 1999 through to early 2004, where it appears that the bias
is effectively eliminated and the CuSum flattens out through
POST-IMPLEMENTATION REVIEWS to mid-2007.
The red line through the chart marks the start-up of the
The impact of gravity recovery on plant performance gravity circuit in August 2004. While it appears that the bias
Following the start-up and stabilisation of the gravity circuit appears to have been rectified immediately prior to the start-
a series of tests to review the quantity of GRG reporting to up of the gravity circuit it is reasonable to attribute some of
the CIL tailings were conducted through July and August the improvement in metal accounting to the gravity circuit.
2006 and the results reviewed by Dombrose (2006a). These Specifically by removing the coarse gold content from the
tests consisted of operating a 12-inch Knelson concentrator plant tailings and also the flotation concentrate to some
on a 3–4 dt/h bleed of the CIL tailings slurry on two-hour extent, these streams can be analysed with greater precision
cycles. The 12-inch Knelson concentrate was either directly due to the absence of spotty gold values.
leached or passed through a three inch Knelson concentrator
to upgrade the concentrate ahead of intense cyanidation.
The 12-inch Knelson concentrate grade was observed to be
PRESENT-DAY OPERATIONS
considerably lower than that achieved prior to the installation The gravity circuit at KCGM has now operated for eight years
of the gravity circuit (2 g/t Au down from 40 g/t Au), and the and consistently recovers 10–20 per cent of the gold present
gold recovered as a percentage of total gold in feed decreased in the mill feed. Gravity gold recovery varies with the coarse
from 0.45 per cent before the trial to 0.01 per cent after the gold content of the feed, consistent with the variable free
trial. On this basis the gravity project had demonstrated a gold contents of the Golden Mile ores that has been observed
reduction in gold losses by approximately 0.44 per cent due since the turn of the last century. A recent audit of the KCGM
to improved coarse gold recovery, and could thereby be gravity circuits by the AMIRA P420 project team confirms
considered a success. that the efficiency of the circuit has not deteriorated over
time, testament to the effective and operable design of the
It was further observed by Dombrose (2006b) that
KCGM gravity circuits. Figure 9 shows the monthly gravity
the grind size of the flotation feed had coarsened in the
months following the commissioning of the gravity circuit, gold recovery achieved at KCGM since start-up, framed by
increasing from a P80 of 173 µm to a P80 of 185 µm. This was the high and low stage 1 GRG test values as indicators of
commensurate with a two per cent increase in milled tonnes minimum and maximum performance indicators.
and no discernible reduction in plant recovery. This may be Gravity recovery for the KCGM circuit is strongly
influenced by the gravity circuit utilisation as can be observed
when comparing Figures 9 and 10, the monthly gravity circuit
utilisation trend. The configuration of the Fimiston gravity
circuits is such that maintenance to the duty cyclone feed
pump requires a gravity plant shutdown as the circuit cannot
be fed by the standby cyclone feed pump. Additionally the
mechanical availability of the gravity circuit feed screens is
a significant influence on the utilisation of the gravity circuit
and thereby the monthly gravity gold recovery.
The gravity circuit utilisation calculation at KCGM includes
concentrator dump cycles as downtime; hence the theoretical
maximum availability of the gravity circuit operating on
45-minute cycles is 90 per cent. It is clear from Figure 10 that
some time was taken to achieve higher levels of gravity plant
availability following start-up. This is largely associated with
FIG 7 – Two-stage acacia leach profile. high wear rates associated with high flows and a moderately

A Giblett et al
FIG 8 – Metal balance discrepancy CuSum (1999–2011).
and gravity concentration circuits, including stream GRG

determination, mass balancing and simulation using the P420
gravity recovery models. This audit was conducted in April
2011, generating a detailed mass and GRG balance around the
gravity circuit. The mass balance indicated that the Fimiston
gravity circuit was treating approximately 50 per cent of the
recirculating load while the Mt Charlotte circuit was treating
40 per cent of the Mt Charlotte ball mill recirculating load.
Given the available screen capacity the gravity circuit was
well utilised and gold recovery was consistent if not better
than the model predictions based on GRG analysis of the
mill feed samples. The cut size of GRG at KCGM is coarser
than observed at other operations, where the recirculating
load is much higher and the classification efficiency (defined
FIG 9 – Monthly gravity recovery performance. by the extent of fines bypass to cyclone underflow) is lower.
However, this is a result of a well operated circuit, where grind
size and throughput is optimised and gold accumulation in
the recirculating load is low, with underflow to overflow gold
grade ratios around 2:1.
CONCLUSIONS
The decision to retro-fit gravity concentration to the Fimiston
and Mt Charlotte grinding circuits was made after a series
of investigations over several years, motivated by a lasting
commitment to continuous improvement by the KCGM
metallurgical group. The economic benefit of gravity gold
recovery was ultimately demonstrated by continuous
sampling and gravity concentration of the plant tailings to
allow quantification of gold losses due to GRG. Following
the commissioning of the gravity circuits at KCGM the same
FIG 10 – Monthly gravity circuit utilisation. methodology was used to demonstrate an improvement in
total gold recovery of the order of 0.45 per cent, validating the
abrasive ore, although in recent years the gravity circuit decision to invest approximately $10 M to retro-fit the three
availability has been stabilised at more acceptable levels. gravity circuits and central intensive cyanidation facility.
The design of the KCGM gravity circuit includes several
AMIRA P420 gravity circuit audit examples of best practice, including dedicated gravity circuit
Both Newmont and Barrick sponsor the AMIRA P420 research feed pumps, a dedicated gravity circuit feed hopper in the
project managed and executed by the Gold Technology case of the Mt Charlotte circuit, large capacity screen decks
Group of Curtin University. Under the project optimisation allowing a high gravity circuit feed rate, parallel concentrators
theme of this project the research team were invited to the in the Fimiston circuit giving an element of redundancy
KCGM operation to perform a full survey of the classification and a single large capacity intensive cyanidation reactor of

sufficient capacity to allow short concentrator cycle times and Cutifani, M and O’Neill, A, 1999. Fimiston open pit, in Proceedings
maximum concentrate mass pull. World Gold ’91 (The Australasian Institute of Mining and
The KCGM gravity circuit project is a solid example of sound
business case development, effective project engineering and Dombrose, E, 2006a. Post GRG 12-inch Knelson Tests at KCGM –
Assessment of results, internal document, Metallurgical Support
project execution, allowing the economic benefit of gravity
Pty Ltd.
concentration to this site to be quantified and realised.
Dombrose, E, 2006b. Review of GRG performance at KCGM, internal
document, Metallurgical Support Pty Ltd.
REFERENCES
Hillier, D and Tilly, S, 2003. Gravity recovery position paper, KCGM
Blaskett, K S, 1952a. Ore treatment at Western Australian Gold
internal document.
Mines – 7, Pre-cyanidation at GMK and Great Boulder, Chemical
Engineering and Mining Review, pp 295–304, May. Kalgoorlie Consolidation Gold Mines (KCGM), 2003. Ammtec
report A8954, KCGM gravity evaluation.
Blaskett, K S, 1952b. Ore treatment at Western Australian Gold Mines
– 4, Straight flotation at Paringa and Croesus, Chemical Engineering Penrose, E K, 1948. Treatment Methods and Plant of Gold Mines
and Mining Review, pp 172–182, February. of Kalgoorlie Ltd (The Australasian Institute of Mining and
Blaskett, K S, 1952c. Ore treatment at Western Australian Gold Mines
– 5, Straight flotation with post cyanidation at Lake View and Randol International Limited, 1983. Randol Gold and Silver
Star, Chemical Engineering and Mining Review, pp 213–221, March. Innovations, Phase II.
Cashman, J O, 1980. Gold ore treatment at North Kalgurli Mines Ltd, Roche Mining (JR) Pty Ltd, 2003. Kalgoorlie Consolidated Gold
Kalgoorlie WA, Mining and Metallurgical Practices in Australasia: Mines Pty Ltd gravity project test work review.
The Sir Maurice Mawby Memorial Volume (ed: J T Woodcock), Stevens, T B, 1933. Ore treatment at the Lake View and Star, The
pp 501–503 (The Australasian Institute of Mining and Mining Magazine, pp 201–212.
Tilly, S and Hall, A, 2003. Intensive cyanidation reactor technical
Cutifani, M, 1993. Gold ore treatment at Kalgoorlie Consolidated evaluation and selection, Kalgoorlie Consolidated Gold Mines
Gold Mines Pty Ltd, Kalgoorlie WA, in Australasian Mining and internal document.
Metallurgy: The Sir Maurice Mawby Memorial Volume, second
edition, pp 966–976.

Contents
Agglomeration – the key to success

for the Murrin Murrin heap leach
D Readett1 and J Fox2
ABSTRACT
The most critical variable in the successful establishment of the Murrin Murrin heap
leach plant was the effective agglomeration of the nickel (Ni) laterite ore.
This paper describes the test work undertaken to establish the key agglomeration
parameters including state-of-the-art testing and analysis procedures. These parameters
were then utilised to establish the key design and operational concepts for the world’s
first commercial scale Ni laterite heap leach.
Following the technical and economic success of the operation, the actual operational
data is compared back to the original test work data.
INTRODUCTION
Agglomeration has been a key operating practice for both the gold and copper heap
leach. McClelland, Pool and Eisele (1983) provided an historical perspective on
precious metals heap leaching. The use of copper heap leaching has now become
an industry standard. In 2005, Bouffard (2005) provided a review of agglomeration
practice both for gold and copper heap leach operations.
Bouffard (2005) concluded that ‘agglomeration was a breakthrough technology...
with ore of high fines or clay content. Achieving up to 80–90 per cent metal recovery
from ores at first thought to be heap-unleachable’. All of the operations that were
reviewed used some form of chemical binder that was dependent on the leach
environment that was subsequently going to be utilised. For copper heap leaching
typical agglomeration was conducted in a drum agglomerator with the addition of
15–25 kg/sulfuric acid per tonne of ore and 60–100 kg of water.
Readett and Miller (1996, 1997) and Readett (1997) highlighted that for copper bio-
heap leach applications, agglomeration using acidified leach liquor is required to
ensure adequate inoculation of the ore with bacterial culture.
As the leaching of Ni laterites is an acid based process (Agatzini-Leonardou and
Dimak, 1994), the data from gold leaching is of limited use. However, the data
generated from acid based copper heap leaching is of relevance.
Ni laterite ores are also known to have unique characteristics and materials handling
of these ores can be problematic. The agglomerated laterite ore can exacerbate these
materials handling characteristics therefore these characteristics need to be considered
in the detailed design of the proposed materials handling system.
LABORATORY TESTING
Testing of Murrin Murrin ore types commenced in 2004. Initially ‘bottle roll’ tests
were performed on samples to determine their amenability to potential heap leach.
Several ore types achieved in excess of +50 per cent Ni recoveries with the addition
of ~300 kg/t of acid.
Preliminary agglomeration testing indicated highly variable results. It was possible
to add up to 200 kg/t of acid solution in agglomeration, due to the very high inherent
moisture content of the laterite ores, which ranged from 15 to 30 per cent. However,
results from 1 m column tests showed permeability issues at the very high acid
solution addition rate to agglomeration (200 kg/t). Acceptable column leaching was
achieved with agglomerates from 50–150 kg/t of acid solution. Unagglomerated ore
was impermeable in column testing.
1. FAusIMM(CP), Principal Consultant,
Mworx Pty Ltd, Coolum Beach Qld 4573. Agglomeration and column testing with water and acidified leach liquor indicated
Email: dreadett@mworx.net.au that it was necessary to ensure optimal overall moisture addition for the best
2. MD and CEO, PNX Metals Pty Ltd, Rose Park SA agglomeration result and that acidified leach liquor was superior to water as the
5067. Email: james.fox@pnxmetals.com.au moisture agent.
569
D Readett and J Fox
In laboratory 1 m columns the Ni rate and extent of extraction

was found to be solely a function of acid solution addition,
with more acid solution added in agglomeration resulting in
earlier release of Ni.
Each ore type, however, appeared to exhibit its own unique
optimal agglomeration condition leading to optimal column
performance.
These preliminary results highlighted that it was necessary
to try and establish laboratory testing techniques that allowed
for the measurement of the effectiveness of agglomeration
prior to conducting column leach testing. Additionally, these
tests were also utilised to try and establish if the effectiveness
of agglomeration (including maximum potential heap height
and maximum potential leach solution application rate) could
be sustained over the leach cycle after the consumption of
300–500 kg/t of acid.
A summary of the testing techniques undertaken is
provided below.
Kappes percolation test

This is an empirical test attributed to Kappes, Cassidy and
Associates and is a laboratory based single pass/fail test
to establish if agglomerated gold ores would perform in
an industrial heap leach application. The test was adopted
for the Ni ores (Table 1). The test measures the drain down
permeability of a saturate column of ore whereby empirical
data indicates if the sample achieves a percolation rate greater
than 10 000 L/h/m2 and a slump of less than 15 per cent then it FIG 1 – Load permeability results.
will perform acceptably in an industrial heap leach application.
Soil water characteristic curve (SWCC)
Load permeability test The SWCC is used to describe the unsaturated hydraulic
The load permeability test was used to establish the likely properties of soil and can be used to estimate the unsaturated
saturated permeability of a sample and can measure how permeability of a material with respect to the degree of
the permeability may change with heap height (Figure 1). saturation (Figure 2).
It provides an indication of the relationship between heap
height and the leach liquor application rate that would result Load percolation test
in saturation of the ore. This test is aimed at determining the maximum loading
The results from this test are considered limited as they that can be applied to a sample, at a constant leach solution
are influenced by the loading and saturation history of the application rate, before flooding (saturation) occurs (Figure 3).
sample tested (Williams, 2005). Results are likely therefore The maximum load can then be converted to an equivalent
to be conservative. heap height for design purposes.
Table 1
Kappes percolation results.
Ore type Agglomeration addition kg/t Bulk density Drain rate Slump
Water Acid solution t/m3 L/m2/h %
Scats 34 0 1.14 4278 1.1
0 25 1.00 41 559 1.1
13 50 1.11 43 545 1.0
0 100 1.17 23 529 1.5
0 150 1.28 1757 1.1
BB 0 246 1.00 6723 3.3
277 50 1.07 9397 3.6
BD 0 421 1.11 7181 0.0
388 50 1.15 16 119 0.5
B Blend 0 242 1.10 1451 3.7
269 50 1.18 7105 1.1

Agglomeration – the key to success for the Murrin Murrin heap leach
In general, the test work indicated that the use of acid

solution, such as counter current decantation leach solution,
as the moisture agent in agglomeration was more effective
that water. A total acid solution addition of 50–100 kg/t was
found to be optimal. The optimal condition for each ore type
was unique. However, even when combined with results from
an extensive column leach program, a definitive answer on
the optimal heap height and solution application rate required
for the successful heap leaching of Ni laterite ores through to
completion could not be established (Scheffel, 2005).
As an example, for the acid solution agglomerated scats:
•• Kappes tests indicated the material was heap leachable
•• load permeability showed ranges of maximum heap
height of:
•• from 4 m at 5 to 45 L/h/m2
•• to 8 m at 3 to 25 L/h/m2.
•• load percolation tests showed ranges of maximum heap
height of:
•• from 19 m at 30 L/h/m2
•• to 7 m at 60 L/h/m2.
•• saturated hydraulic conductivity was 0.14 to 0.24 cm/s
•• stacked density was:
•• 0.9 g/cm3 at 4 m
•• 0.94 g/cm3 at 8 m.
•• column tests indicated ranges of maximum heap
height of:
•• from 4 m at 60 L/h/m2,
•• to 8 m at 0 to 30 L/h/m2.
On the other hand, results for leached scats were:
•• Kappes tests indicated the material was heap leachable.
FIG 2 – Soil water characteristic curve results for scats and scats residue.
•• load permeability showed ranges of maximum heap
Saturated hydraulic conductivity test height of:
The saturated hydraulic conductivity of a porous medium •• from 4 m at 2 to 14 L/h/m2
defines the maximum hydraulic capacity to transmit •• to 8 m at 1 to 7 L/h/m2.
solution once all the void space within a sample is occupied
•• load percolation tests showed ranges of maximum heap
with water (that is, saturated). To simulate a stacked heap
height of:
the saturated hydraulic conductivity tests were conducted
under various loading conditions (Figure 4). These tests •• from 20 m at 30 L/h/m2
were designed to establish the short-term upper percolation •• to 8–20 m at 60 L/h/m2.
capacity (Guzman, 2005). The test however, does not •• saturated hydraulic conductivity was 0.04 to 0.005 cm/s.
quantify the transient behaviour as a result of the interaction
•• stacked density was:
of acid solution with the ore.
•• from 0.6 g/cm3 at 4 m
Stacking (or stacked density) test •• to 0.68 g/cm3 at 8 m.
This test protocol was recommended by Guzman (2005) based •• column tests indicated ranges of maximum heap
on experience that had shown that the hydraulic performance height of:
of a heap is strongly controlled by the density of the ore. •• from 4 m at 60 L/h/m2
Samples are subjected to various static loads to define the
relationship between heap height and dry bulk density, and •• to 8 m at 0 to 15 L/h/m2.
hence provide the density profile throughout the proposed An added complication with the results was that in general,
stacking depth. These tests are conducted under partially if the agglomerated material was exposed to excessive
saturated conditions on the basis that this better represents the mechanical impact or saturated leach conditions, its resultant
actual density profile developed during ore stacking and also permeability approached zero (that is, unheap leachable).
allows determination of the hydraulic conductivity without As expected the testing program highlighted the unique
the destruction of the agglomerated product (Figure 5). nature of the laterite ores and also the unique nature of the
acid heap leach process. Laterite ores are naturally high in
Summary of tests clay content, high in moisture content and have very low
No single test was found to provide a definitive answer density. The stacked ore can have a dry bulk density of less
to the ultimate question regarding true performance of than 1.0 g/cm3 so the impact of solution application (with a
an agglomerated material in a commercial operation. leach solution with a specific gravity of 1.1 to 1.3 g/cm3) and
Each was found to provide indicative data on the optimal solution saturation can be extreme. The ore exhibits extremely
agglomeration conditions. high acid demand (350–500 kg/t) to maximise Ni recovery.

D Readett and J Fox
FIG 3 – Typical load percolation results.
1.0E+03
USAA (0.95 g/cm3) LS (0.65 g/cm3)

Application Rate (L/hr/m2)
1.0E+02
1.0E+01
1.0E+00
1.0E-01
40 50 60 70 80 90 100
Liquid Saturation (%)
FIG 4 – Scats saturated hydraulic conductivity.
As a consequence, heap leaching demands high solution The key design areas were bins, agglomeration drum and
application rates over extended leaching cycle times of one chutes/transfer points.
to two years. The stacked ore exhibits a high degree of initial The run-of-mine feed bin and subsequent surge bin were
slump upon commencement of leaching and then during the designed with very steep angles and were to be lined with
leach cycle the acid can consume 15–30 per cent of the mass
a Teflon-based wear system. This would ensure minimal
of the ore.
likelihood of material compaction and hang up within the bins.
Ultimately the lack of a definitive answer led the project team
Some of the key design features of the agglomeration
to recommend the construction of demonstration industrial/
commercial scale heaps. Then upon successful operation of drum included:
the demonstration heaps, the demonstration plant could be •• a steeply angled and Teflon-lined feed chute
expanded to a full commercial plant. •• an effective control system for moisture and acid
addition
AGGLOMERATION AND MATERIALS HANDLING DESIGN •• the design and placement of moisture and acid
The test work indicated that well controlled effective addition points within the drum to ensure maximum
agglomeration was the key to success. Additionally, it indicated mixing and minimal potential impact with the drum
that the ore and subsequent agglomerates exhibited very poor lining system
materials handling characteristics and rapid deterioration
•• the addition of a scraper bar to allow for removal of
of permeability under saturated conditions, both of which
compacted ore from within the drum.
could adversely affect the heap leach performance of the
agglomerated ores. The agglomerates were relatively low in Post agglomeration, the design incorporated the minimum
strength and agglomeration effectiveness could be destroyed number of transfer points to minimise the potential for
with excessive handling. This combination of characteristics mechanical degradation of the agglomerates. Each transfer
required a very high degree of attention to detail in the design point was also designed to allow the agglomerated ore to free
of the agglomeration and materials handling system. fall onto the proceeding conveyor.

Agglomeration – the key to success for the Murrin Murrin heap leach
1.12 0.64
1.10
Dry Bulk Density Porosity
1.08 0.62
Dry Bulk Density (g/cm3)

1.06
Porosity
1.04 0.6
1.02
1.00 0.58
0.98
0.96 0.56
0 2 4 6 8 10
Estimated Heap Height (m)
FIG 5 – Stacked density test results for ore.
The permeability and subsequent column testing indicated the test work program. Again, the actual operational data has
that, depending on ore types and agglomeration conditions, tended to the conservative conditions determined in testing.
the optimum heap height should be in the range of 4–8 m. As From a metallurgical perspective, the heaps have performed
a consequence, the stacker design allowed for a variable stack satisfactorily and results of the performance has been
height of ore from about 3–9 m. provided in a number of papers (Readett and Fox, 2009, 2010).
Heap drainage was also a key design parameter. The testing
had highlighted that saturation of the heap could result in CONCLUSIONS
the destruction of the agglomerates and effectively reduce
The most critical variable in the successful establishment
permeability to a minimum. As such it was necessary to
of the Murrin Murrin heap leach plant was the effective
ensure heaps were adequately drained. The need for drainage
agglomeration of the Ni laterite ore.
was exacerbated by the high application rates required. To
ensure effective drainage was achieved a 500 mm layer of The test work conducted allowed the optimal agglomeration
free draining acid resistant coarse rock was placed at the base conditions to be established and in general provided an
of each heap. Within the drainage material a ‘half herring optimistic indication of the likely performance in the full-scale
bone’ arrangement of drainage piping was also installed. The operation. Results also provided a focus for key design criteria
drainage piping connected to header drainage pipes that ran and necessary materials handling characteristics that required
detailed attention. In general, the test results overstated the
the full length of the pad and exited into the collection sump
ultimate optimal heap height and application rate that could
at the front of each pad.
be sustained in a full-scale operation over the entire leach cycle.
Use of all the available data ultimately helped to ensure a
OPERATING EXPERIENCE successful project. In hindsight, a conservative approach was
The attention to detail in the design phase ensured minimal warranted considering the unique nature of this project.
commissioning issues and minimal longer term operational
problems.
ACKNOWLEDGEMENTS
To date there have been no significant changes to the overall
The authors would like to thank Minara Resources Limited for
design of the agglomeration and materials handling system
permission to publish and also acknowledge the contribution
(Readett, Meadows and Rodriguez, 2006). The agglomerator
of Michael Rodriguez, David Williams, Randy Scheffel and
feed chute did cause some preliminary compaction and
Amado Guzman.
blockage issues however the addition of a small water spray
directly onto the chute impact surface overcame this.
The ore type that has been predominantly stacked is ‘scats’.
REFERENCES
Agatzini-Leonardou, S and Dimak, D, 1994. Heap leaching of
Testing of this material indicated stack heights of 2–10 m
poor nickel laterites by sulphuric acid at ambient temperature,
could support application rates of 15–30 L/h/m2. Column International Symposium Hydrometallurgy ’94, pp 193–208.
leach testing indicated potential for heap saturation at 8 m
Bouffard, S C, 2005. Review of agglomeration practice fundamentals
and 15–30 L/h/m2. The actual operational data has tended
in heap leaching, Mineral Processing and Extractive Metallurgy,
to the conservative end of the test results. Typically, optimal 26:233–294.
heap heights are 3–5 m with sustainable initial application
McClelland, G E, Pool, D L and Eisele, J A, 1983. Agglomeration-heap
rates of 15–20 L/h/m2 and longer term rates degrading down
leaching operations in the precious metals industry, USBM IC1945.
to 5–10 L/h/m2.
Readett, D J, 1997. Engineering and operational aspects of copper
Solution drainage at the base of the pads has worked heap bioleaching, AMF Application of Biotechnology to Economic
effectively and no drainage related side slope failures have Recovery of Metals from Ores and Concentrates, Sydney.
occurred.
Readett, D J and Fox, J, 2009. Development of Heap Leaching at its
Leaching of fresh laterite ore has also been implemented integration into the Murrin Murrin Operations, in Proceedings
using the optimal agglomeration conditions established in ALTA Nickel Cobalt Conference, Perth, May.

D Readett and J Fox
Readett, D J and Fox, J, 2010. Commercialisation of Ni heap leaching at Readett, D J and Miller, G M, 1996. Practical aspects of heap leaching,
Murrin Murrin operations, in Proceedings IMPC 2010, pp 3611–3616 Copper Hydrometallurgy Forum, ALTA, Brisbane.
Readett, D J and Miller, G M, 1997. Engineering and process
Readett, D J, Meadows, N E and Rodriguez, M, 2006. Murrin Murrin developments associated with industrial scale copper bioleaching,
heap leach project, in Proceedings MetPlant Conference, pp 367–372 IBS Biomine 97 Conference, Sydney.

Contents
Author index
A Davis, B 483
Afewu, K 243 Davis, M 193
Appelhans, D 297 Drinkwater, D 85
Asbjörnsson, G 483 Duffy, K 201
Atasoy, Y 411 Dunlop, J S 173
Dunne, R 297, 327, 355, 365
B
E
Baas, D 411
Bailey, C 235 Ehrig, K 33
Bartsch, P 249 Elwin, D 137
Bax, A 263 Evertsson, C M 483
Becker, G 427
Bennett, D 201 F
Benzer, A H 483 Fleay, J 193
Bessen, B 315 Fountain, C 75
Bickert, G 285 Fox, J 569
Binks, M 523 Franke, J 509
Bourgeois, F S 349
Brett, D 419 G
Bruckard, W J 529 Giblett, A 243, 263, 297, 361, 365, 559
Gibson, G 315
C Greet, C J 529
Cameron, P 85
Canterford, J 183 H
Card, P 65 Hall, S 249
Cavendor, A N 499 Harbort, G 37
Claflin, J K 537, 541, 549 Hilden, M M 483
Clements, B 187 Hill, D J 461
Coleman, R 107 Hillier, D 559
Connelly, D 69, 101 Holmes, R J 333
Corder, G 427 Hoyle, A 201
Crnkovic, I 201 Hunter, T 107
Currey, N 427
J
D
Jankovic, A 307
Dakin, P 137 Johnson, N W 445
Daniel, M J 475 Jones, K 37
David, D 115, 121

K Pease, J D 27, 85, 445
Karageorgos, J 411 Perkins, T 509
King, J 107 Pitard, F F 343
Kirsch, S W 475 Plint, N 483
Klepper, R 273 Powell, M S 401, 483, 509
Preedy, D 541
L
R
La Brooy, S R 75, 537, 541, 549
Lane, G 75, 137, 163, 187, 193, 235 Ramsell, V 559
Länger, B 285 Readett, D 569
La Rosa, D 201 Rule, C 483
Li, X 401
Lyman, G J 349 S
Skinner, E 163
M Sola, C 37
MacKay, D 529 Staples, P 235
Mainza, A N 483 Staunton, W 263
McCaffery, K 365
McCallum, W 355 T
McCarthy, P L 55 Tavares, L M 483
McCurdie, P 273 Thwaites, P 373
McKeague, W 401 Tilyard, P A 11
McLean, E 3, 193 Toor, P 509
McMillan, G 437 Tordoir, A 201
Miller, G 257
Morgan, D 37 V
Morrell, S 131, 235 Valery, W 201, 307
Mulvey, P 437 Valine, S B 499
Munro, P D 11, 19, 91
Munro, S E 91 W
Walker, P 201
N
Wardell-Johnson, G 263
Napier-Munn, T J 315, 361, 455, 469 Wemyss, P 523
Newell, A J H 145 Wheeler, J E 499
Nilsson, K 101 Whincup, P R 155
P Y
Packer, B 499 Young, M F 445
Parker, K 559

We Are Metallurgists

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

We Are Metallurgists

Uploaded by

Copyright:

Available Formats

We are Metallurgists, Not Magicians

Landmark Papers by Practising Metallurgists

Edited by D Pollard, G Dunlop and J Herzig

No part of this publication may be reproduced, stored in a retrieval system or transmitted

ISBN 978 1 925100 62 4

Front cover image (top):

Back cover image:

“A splendid outcome and a worthwhile contribution to the industry.”

“I would be honoured … I see it as a great idea.”

“It is a great honour to be included in a such august company

“Honoured to be included and love the title.”

Follow us on: www.newmont.com

PanAust’s producing assets are the Phu Kham

For more information, visit www.panaust.com.au.

Back to the future – why change doesn’t necessarily mean progress 11

Back to the future – still on the dark side 19

Undermining productivity – when good key performance indicators go bad 27

Integrating geometallurgy with copper concentrator design and operation 37

Integrated mining and metallurgical planning and operation 55

Sensible cost cutting for resource projects 69

When does further processing at the mine site make sense? 75

The ABC of Mine-to-Mill and metal price cycles 85

Cost-effective concentrator design 137

Project management and delivery

Guidelines for mineral process plant development studies 155

Project delivery 163

Keeping projects on the rails 183

Performance testing – when, what and how? 193

Unit design and development

What can go wrong in comminution circuit design? 235

Design of copper-cobalt hydrometallurgical circuits 257

A review of best practice in gravity circuit design and operation 263

Considerations for effective gold process development 297

Energy efficient ball mill circuit – equipment sizing considerations 307

Advances in dense medium cyclone plant design 315

Sampling, metallurgical accounting and control

The importance of sampling in the mineral industry 333

Sampling, corporate governance and risk analysis 349

Metallurgical management tools for continuous improvement 355

Unlocking processing potential by empowering our operators 401

Environmental management and sustainability

Sustainability, made easy – a business improvement case study 427

Process improvements and case studies

Cunning solutions to process improvement 455

How to prioritise process improvements 461

Optimisation opportunities for high pressure grinding rolls circuits 483

An update on applications of high frequency screens in closed grinding circuits 499

The influence of liner wear on milling efficiency 509

Development of the Fosterville gold mine heated leach process 523

Collector – addition point and consumption 529

The Ausenco carbon reactivation kiln initiative 537

Carbon management in a high gold price environment 549

Author index 575

‘We’re metallurgists, not magicians!’

THE DESIGN SEQUENCE

Stage Development activities Description of key activities in development stage

4 we are metallurgists, not magicians

we are metallurgists, not magicians 5

6 we are metallurgists, not magicians

we are metallurgists, not magicians 7

were carried out in ideal, batch conditions, were discounted

Year Pit 1 Pit 2 Pit 3 Total Grade Grade

8 we are metallurgists, not magicians

we are metallurgists, not magicians 9