Professional Documents
Culture Documents
February 2007
© AMIRA International
Following the first release of these documents in October 2005, extensive feedback
and comments were received from Dr GJ Lyman, Sampling Consultant to Anglo
Platinum, which mainly affected the sections of the documents related to sampling,
and which were incorporated into Release 2 of the Code and Guidelines.
Further contributions have also been received from the P754 Research Teams at the
Universities of Cape Town, Queensland, and Stellenbosch and these have assisted
the team preparing the Code and the Guidelines.
During the drafting of these documents, regular discussions were held with
representatives of the South African Accounting Profession and their suggestions
and comments have been incorporated.
On behalf of the members of the P754 Code Development Team, I would like to
express our sincere thanks to all those listed above for their help in compiling the
Code of Practice and the Guidelines.
This release is an update of Release 2 which was seen as a trial or ‘beta’ version of
what we hope will become a set of documents which can be applied throughout the
minerals industry to assist operating management in the design, installation and
operation of sound metal accounting systems, which will meet the requirements of
corporate governance. To achieve this, Release 3 of this Code of Practice and
Guidelines must be introduced and applied in the operating environment of
companies, so that any flaws in the documents can be identified and eliminated as
rapidly as possible.
In addition to the sponsors of P754, Release 3 will also be distributed to people who
have expressed interest in participating in the evaluation of the documents, and to a
wider group of potentially interested parties at workshops provisionally scheduled to
be held in Australia and South Africa during 2007. A further review and revision of
the documents is planned for the end of 2007, after which Release 4 is expected to
be distributed to a wider group of interested parties, including mining companies in
North America, during the first quarter of 2008.
Peter Gaylard
2007-02-19
These drawbacks, together with the need for metal accounting to be conducted to the
level of “financial accounting” standards with respect to auditability and transparency,
for example, have been recognised for some time. Following the symposium entitled
“Challenges in Metallurgical Accounting & Information Management” held in Cape
Town in August 2001 under the auspices of the SAIMM, several interested parties
held discussions on how the situation could be improved. The outcome of these
discussions was a proposal for a project to be coordinated by AMIRA International
entitled “P754 – Metal Accounting and Reconciliation” which was subsequently
accepted and sponsored by Anglo American, Anglo Platinum, BHP Billiton, Kumba
Resources and Rio Tinto. The objective of the project was to develop a Code of
Practice for Metal Accounting, which would be similar in concept to the JORC and
SAMREC Codes produced for the reporting of mineral resources and ore reserves.
At the same time, research projects were established at several universities in South
Africa and Australia, which, in collaboration with industry, will aim to provide solutions
to some of the more intractable aspects of metal accounting.
A team under the leadership of Professor Peter Gaylard, UCT, was commissioned in
August 2004 to develop the Code, which focuses on the need for all activities in the
areas of mass measurement, sampling, sample preparation and analysis to be
carried out accurately and with an acceptable level of precision. In addition,
wherever possible the Code recommends the use of International Standards as the
basis for all evaluation activities. Calculations and reports must be on a full “Check
in-Check out” basis and must be prepared in a completely transparent manner with
clearly defined audit trails. During its development, the Code has been discussed on
many occasions with representatives of the financial auditing profession, who have
offered many helpful suggestions on its content and who are in general agreement
with its aims.
2 The system must be consistent and transparent and the source of all input data
to the system must be clear and understood by all users of the system. The
design and specification of the system must incorporate the outcomes of a risk
assessment of all aspects of the metal accounting process.
3 The accounting procedures must be well documented and user friendly for easy
application by plant personnel, to avoid the system becoming dependent on
one person, and must incorporate clear controls and audit trails. Calculation
procedures must be in line with the requirements set out in this Code and
consistent at all times with clear rules for handling the data.
4 The system must be subject to regular internal and external audits and reviews
as specified in the relevant sections of the Code to ensure compliance with all
aspects of the defined procedures. These reviews must include assessments
of the associated risks and recommendations for their mitigation, when the
agreed risk is exceeded.
7 The system must generate sufficient data to allow for data verification, the
handling of metal/commodity transfers, the reconciliation of metal/commodity
balances, and the measurement of accuracies and error detection, which
should not show any consistent bias. Measurement and computational
procedures must be free of a defined critical level of bias.
8 Target accuracies for the mass measurements and the sampling and analyses
must be identified for each input and output stream used for accounting
10 The metal accounting system must ensure that every effort is made to identify
any bias that may occur, as rapidly as possible and eliminate or reduce to an
acceptable level the source of bias from all measurement, sampling and
analytical procedures, when the source is identified.
In addition to those already mentioned, the Code covers areas such as Safety,
Training, Data Management and Design of Metal Accounting Systems. Two further
very important sections cover the appointment of a Competent Person to coordinate
implementation of the Code and the requirements for internal and external audits of
the Metal Accounting system.
The recipients of the Code and its attachments should include independent Directors,
members of the Risk/Audit Committee, senior corporate and operational
management, project design staff (including outside contractors) and plant operation
and laboratory staff.
• The Code has not at this stage been formally adopted, for internal purposes,
by any of the Sponsors or by any industry associations and hence it is not
binding on any such companies, organisations, associations or on any
directors, officers, agents, employers, advisors or members thereof
• The Code has not been adopted by any legislative or regulatory bodies
(including any stock exchanges), nor has it been incorporated into the listing
or other rules of such exchanges
• The Code does not purport to be, nor should it be construed as constituting,
any form of legal or other standard in relation to metal accounting
• Neither AMIRA, nor the Sponsors nor any of their directors, officers, agents,
employees and advisors make any representation or warranty regarding, nor
shall they have any liability whatsoever in respect of any statement,
recommendation, advice, opinion or information contained in the Code
• To the extent that the Code is made available to third parties by or on behalf
of the Sponsors, such action (or any subsequent interactions) shall not be
construed as giving rise to any contractual relationship or any duty of care, of
whatsoever nature, between AMIRA, the Sponsors (and/or their directors,
officers, agents, employees or advisors) and such third parties.
Disclaimer vi
1 INTRODUCTION .................................................................................1
1.1 SAFETY AND HEALTH 2
1.2 PRINCIPLES OF METAL ACCOUNTING 2
1.3 STANDARDS 3
1.4 TRACEABILITY 3
1.5 ACCURACY AND PRECISION IN METAL ACCOUNTING 4
1.6 ERROR TYPES 4
1.7 METAL ACCOUNTING RISK 4
1.8 TRAINING 5
2 DATA MANAGEMENT........................................................................6
2.1 INTRODUCTION 6
2.2 DATA ACQUISITION / ENTRY 6
2.3 DATA DOCUMENTATION 6
3 MASS MEASUREMENT .....................................................................7
3.1 INTRODUCTION 7
3.2 ACCURACY AND PRECISION OF MASS MEASUREMENT 8
3.2.1 Scales / Weighing Equipment 8
3.2.2 Flow Meters 9
3.3 METHODS OF MEASURING MASS 9
4 SAMPLING........................................................................................ 11
4.1 INTRODUCTION 11
4.2 ACCURACY AND PRECISION OF THE SAMPLING SYSTEM 11
4.3 MULTI-STAGE SAMPLING SYSTEMS 12
4.4 SAMPLE STORAGE 12
4.5 SAMPLE LABELLING 12
5 SAMPLE PREPARATION AND ANALYSIS ..................................... 13
5.1 INTRODUCTION 13
5.2 SAMPLE PREPARATION 13
5.2.1 Moisture determination 13
5.2.2 Sample Reduction 13
Where words are highlighted in bold lettering in the Code, their definitions in terms of
Metal Accounting are included in the Glossary to this Code Document.
Metal Accounting is defined as the system whereby selected process data (pertaining
to metals of economic interest), including mass measurement and analysis are
collected from various sources and transformed into a coherent report format that is
delivered in a timely fashion in order to meet specified reporting requirements.
The ultimate purpose of the Code is to define the best practice for creating and
reconciling process metal balances, although it does not prescribe specific valuation
methods, as these will be dependent on the practices and policies of the company or
operation concerned.
The Code is based on a set of metal accounting principles, which are set out in detail
in Section 1.2. These incorporate the use of the Check in-Check out procedure for
metal balancing, as well as the use of Exception Reports for reporting all deviations
from the Code. These reports, in turn, have to be signed off by the Competent
Person before being submitted to the Audit Committee of the company for ratification.
The Competent Person is also required to approve the final design and installation of
the metal accounting system and to conduct routine audits of its operation.
At present, the Code does not address metal accounting in mines and open cast
operations, as it takes the feed to metallurgical treatment plants as its starting point.
It is envisaged that this aspect will be addressed as the Code is developed. However,
mass measurement in stopes and ore blocks is discussed briefly.
2 The accounting system must be consistent and transparent and the source of
all input data to the system must be clear and understood by all users. The
design and specification of the system must incorporate the outcomes of a risk
assessment of all aspects of the metal accounting process.
3 The accounting procedures must be well documented and user friendly for easy
application by plant personnel, to avoid the system becoming dependent on
one person, and must incorporate clear controls and audit trails. Calculation
procedures must be in line with the requirements set out in this Code and must
be consistent at all times with clear rules for handling the data.
4 The system must be subject to regular internal and external audits and reviews
as specified in the relevant sections of the Code, so as to ensure compliance
with all aspects of the defined procedures. These reviews must include
assessments of the associated risks and recommendations for their mitigation,
where the agreed risk is exceeded.
7 The system must generate sufficient data to allow for data verification, handling
of metal/commodity transfers, reconciliation of metal/commodity balances, and
measurement of accuracies and error detection, which should not show any
consistent bias. Measurement and computational procedures must be free of a
defined critical level of bias.
8 Target accuracies for the mass measurements and the sampling and analyses
must be identified for each input and output stream used for accounting
purposes. The actual accuracies for metal recoveries, based on the actual
accuracies, as determined by statistical analysis of the raw data, achieved over
a company’s reporting period must be stated in the report to the Company’s
Audit Committee. Should these show a bias which the Company considers
material to its results, the fact must be reported to shareholders.
10 The metal accounting system must ensure that every effort is made to identify
any bias which may occur as rapidly as possible and to eliminate or reduce to
an acceptable level the source of the bias from all measurement, sampling and
analytical procedures, once the source has been identified.
1.3 STANDARDS
Wherever possible, procedures used in Metal Accounting must be based on the
appropriate International or National Standards.
1.4 TRACEABILITY
In the areas of mass measurement and analysis, the aim must be to establish
traceability back to the relevant international standard units of mass, distance and
time.
The actual accuracy of the independent operations involved in obtaining data for the
Metallurgical Accounts must be measured on at least an annual basis or, if the nature
of one of the measurement parameters changes, e.g. particle size of the process
stream, sampler design, analytical method etc. The precision of mass measurement,
sampling and analyses must be monitored on an on-going basis by a robust
statistical method and an investigation must be commenced when the monitoring
system indicates that the precision is outside the acceptable range. The overall
expected precision of the Metallurgical Accounts must be calculated from these
values and compared to the actual metal imbalance.
• Random Errors
• Systematic Errors
• Illegitimate (Spurious) Errors
The value of the random errors must be minimised by careful operation and the
causes of systematic and illegitimate errors identified and eliminated.
The use of statistical analysis, and check measurements and analyses provides a
database, which becomes an essential component of the risk management process.
The other fundamental component of the risk management process is regular and
frequent management reviews, incorporating Risk Assessments, of the entire Metal
Accounting system, supported by both internal audits and external audits carried out
by the Competent Person.
1.8 TRAINING
There must be full training procedures available to staff in line with International and
National Standards together with appropriate sections of the Metal Accounting
Guidelines. The level of training and qualification requirements of staff will obviously
vary depending on their role in the accounting system.
2.1 INTRODUCTION
The operation of a Metal Accounting System represents the accumulation and
evaluation of process data (e.g. mass measurements and analyses) from various
sources. The acquisition, processing and management of data require a substantial
portion of the effort associated with Metal Accounting. Efficient, secure and reliable
data management adds considerable value to the Metal Accounting System.
The goal of data management is to create data files from data obtained from various
plant sources in such a way as to prevent systematic and spurious error. Production
(e.g. mass measurement) and laboratory data (e.g. analyses) need to be integrated
to a single user interface to facilitate Metal Accounting. At the same time, however, it
is essential that those responsible for entering plant operating data, and those
responsible for entering analytical results into the system, are able to operate
independently of each other.
3.1 INTRODUCTION
The overriding objective of mass measurement for Metal Accounting purposes is to
establish the mass of the particular material or component present at a specific time,
or the mass flow of that component over a defined time period, to a defined accuracy
suitable for mass balancing, using procedures that take account of Best Practices.
At the same time, the cost/benefit relationship, the difficulty of measuring and
sampling the process stream and the importance of the measurement to the Metal
Balance must also be taken into consideration.
In all weigher and flow meter installations, it is essential that the selection, system
design, installation, operation, maintenance and calibration follow the applicable
standards and manufacturers’ recommendations, in order to achieve the accuracy
and precision levels quoted in Section 3.2, as well as those required by International
and National Standards and National Metrology Acts. All the data from the checks
and calibrations must be collected and stored to enable statistical analysis and
assessment, to establish if bias exists and to calculate the variance and likely error in
the value recorded.
Static scales provide the most accurate and precise method of measuring mass and
must be the choice, where circumstances permit, for Custody Transfer and Primary
and Secondary Metal Accounting purposes.
Certified in-motion weighing is also a suitable choice for Primary Metal Accounting
but for rail transport only.
Certified electromechanical belt weighers must be the second choice for metal
accounting purposes after static scales, if static weighing is not feasible.
Providing proper installation, operation and calibration are adhered to, precisions
obtainable generally range between ± 0.05% to ± 0.2% for platform scales and
between ± 0.1% to ± 0.5% for weighbridges on gross load measurements. For
hopper/bin scales and gantry scales the equivalent figures are ± 0.1% to ± 0.25%
and ± 0.15% to ± 0.4% respectively.
For in-motion weighing, provided that the key principles of taring each truck or
hopper on each occasion and controlling the truck speed are followed, a precision of
± 0.5 % is feasible.
Nuclear belt weighers must not be used for Primary Metal Accounting but they can
be used for Secondary Metal Accounting if the material is homogeneous in size and
composition and the belt loading is uniform.
Mass measurement equipment will measure the wet weight, therefore, accurate
determination of the moisture content is essential.
• Percentage of span
• Percentage of a rate or reading
• Percentage of the upper range limit
The accuracies quoted below are those at 95% confidence levels.
Coriolis mass flow meters are capable of high accuracy, ± 0.2% of reading in certain
cases, but they only have limited application in the extractive industry, as they require
low fluid velocity and they are sensitive to the amount of entrained air.
Ultrasonic transit time meters have limited applicability for specific applications and
they can give an accuracy of 0.5 to 2% of the reading.
Momentum (impact) mass flow meters can be used for free flowing solids and they
are capable of achieving an accuracy of ± 1.0% of the reading.
All methods utilizing a volume or volume flow determination require the measurement
of the bulk density of the material in the stockpile/bin or of the specific gravity/relative
density of the slurry or liquid. This can be performed by nuclear (gamma) gauge, or
preferably by bulk or laboratory tests which are the best methods. In a few cases,
such as for fine coal stockpiles, a penetrometer may be used. The density
determination is often a cause of error in the mass measurement calculation.
The mass may be static in a tank, bin or stockpile or flowing on a conveyor belt, in a
conduit or in an open channel, or it may be moving in railway or road trucks.
Methods for measuring volume and volume flow, density and moisture are also
covered in the Guidelines.
4.1 INTRODUCTION
Correct sampling of a particulate material from a moving stream is achieved when
the expected number fraction of the particles in the sample is equal to the average
number fraction of the particles in the process stream during the time that it takes to
extract the sample increment. To achieve this objective, the sampler must be
correctly designed, installed and operated in accordance with sampling theory. This
issue is dealt with in the Guidelines. A mechanically correct sampler is unbiased.
Where possible, the correctness of the sampler and its freedom from a specified level
of bias should be confirmed by carrying out bias tests by the stopped belt method or
another suitable reference method of sampling.
The sampling from stockpiles and bins for accounting purposes must be avoided
because of the great difficulty in obtaining an accurate sample. The sampling of
other process streams/products must be carried out according to the available
International Standards.
A correct sampling system can show a bias over a short period of time if the sampler
takes increments at fixed time intervals, and this interval is a multiple of the period of
a periodic variation in the mass flow or quality of the process stream. To eliminate
this effect, the sampler must be operated in a random stratified sampling mode.
The only unequivocal method of establishing the precision of a sampling and sample
preparation system is to ensure that the sampling system can carry out inter-leaved
sampling.
The design criteria for multi-stage sampling systems must include the following:
5.1 INTRODUCTION
All operations in sample preparation and analysis must be carried out to provide
unbiased data, with a known and acceptable precision and within a time frame that is
suitable for Metal Accounting purposes.
⎡ 1 1 ⎤
∆σ 2 = a2 K s ⎢ - ⎥
⎣ Mss Ms ⎦
Sample division should be carried out wherever possible using a rotary sample
divider (spinning riffler). This recommendation extends to the extraction of the
analytical aliquot from the mass of sub-sample pulverised in the previous preparation
step. The Incremental Division procedure is acceptable. Stationary rifflers must be
avoided.
5.3 ANALYSIS
Analysis of samples for Metallurgical Accounting must be carried out using methods
that have been accredited to ISO 17025. It is likely that such methods will involve a
twin-stream (parallel analysis) procedure and they are performed using extensive QA
and QC systems and checks. The calibration of instruments should conform to best
practice and, where appropriate, it should comply with the requirements of VAM
document LGC/VAM/2003/032. It is expected that there will be frequent use of
certified reference materials and that an Umpire Laboratory will be used on a regular
basis.
The frequency with which stock-takes or audits of the plant operation are carried out
is a management decision. It will relate to the value of the plant throughput, the in-
process inventories relative to the expected profit at year-end, and the cost of
execution; however, stock-takes must be conducted on at least an annual basis.
Stock-takes may be performed more frequently for plant control and these require a
lesser degree of accuracy than those used in financial reporting. The latter stock-
takes must be performed in conjunction with the company’s auditors.
The cost of executing the stock-takes increases with the demand for accuracy of the
result. The level of financial risk associated with a particular strategy will determine
the final decision regarding stock-take frequency and accuracy. The costs incurred
should be offset against the benefits of faster access to more accurate information
relating to production performance and stock shortages or excesses. The overall
confidence level in the result of the stock-take must be established from the
uncertainties in the component measurements and it must be reported with the
results.
In cases in which the in-process inventory has a value that is large compared to the
value of the net input for the accounting period, the accounting strategy must involve
7.1 INTRODUCTION
Metal Accounting is critical for corporate governance in monitoring, checking and
improving performance and in identifying and quantifying losses and emissions. A
satisfactory metal balance provides a convincing demonstration that:
"What comes into the operation, less any internal accumulation, plus
any draw down from internal stock, is sufficiently similar, i.e. within
acceptable limits in mass and analysed components, to what goes
out as product(s), losses and rejects". (Check in-Check out)
Using the Check in-Check out system, all streams into and out of the Process or
Plant for which the balance is being performed, are measured, sampled and
analysed. In terms of this Code, this is the standard procedure to be adopted for
Metal Accounting in all operations, and all deviations from this approach must be
justified in an Exception Report approved by company management.
The accuracy for each component measured is quantified and the resulting accuracy
of the metal balance calculated and reported in order to determine whether it is within
the acceptable limits, as defined for the operation by the company’s management.
Applying the Check in-Check out system will normally show a difference between net
input and output, due to measurement, sampling and analytical uncertainties, and
this imbalance is treated as an unaccounted gain or loss. The handling of this
unaccounted quantity and techniques for the reconciliation of the data and the
detection of errors and bias are covered in the Code Guidelines (Section 7) with the
essential overriding principles that any procedure or adjustment must be transparent,
fully documented and approved. The primary purpose of the reconciliation
techniques is to assist in pinpointing problem areas, thereby leading to improvements
in applying the Check in-Check out system.
Recovery and Accountability are separate but related parameters, the determination
of both being necessary for the efficient monitoring, over defined boundaries, of the
performance of metallurgical plants or processes, as well as for reliable forecasting
and planning.
Although other definitions for Recovery and Accountability exist, which are currently
used in the industry, as shown in the Guidelines for this Code (Section 7), the
definition in this Code is Actual Recovery, which is calculated on the basis of all
process inputs according to the formula:
Accountability (%) = (sum of all outputs + stock and inventory change) x 100
This provides a check whether over time, inputs, outputs and stock changes
reconcile.
The day-to-day data collected from a metallurgical plant is rarely consistent and it will
usually contain redundant information. In general, any two methods of calculation
will yield different results and, therefore, many factors influence the selection of a
suitable metal accounting strategy. These factors are discussed in the Guidelines
and some practical strategies for different types of metallurgical operations are
indicated.
Care must be taken to ensure that the balance is not compromised by recycle
streams and the return of plant spillage to the circuit, especially if these are of higher
or lower grade than the main stream.
The potential impact of the measurement of plant stocks and in-process inventory on
the overall metal balance has been discussed in Section 6.
As process streams can be made up of either, or both, solids and liquids (organic
and/or aqueous), with both containing significant quantities of metals, the Metal
Accounting system for such plants must make adequate provision for the sampling
and mass measurement of all the components of each feed and output stream and it
must allow for the measurement and assessment of continuous recycle streams and
of process spillage, which may occur on a periodic basis.
Check in-Check out is essential for the feed and product streams including liquid and
gaseous effluents. The precise measurement for effluents is difficult but, usually, the
mass content of these streams is relatively small. Mass balancing of the individual
process steps and circuits (Secondary Accounting) is important for monitoring
process performance and it can provide checks on the Primary Accounting.
Further to this, the fact that the process inventory includes unleached, partially
leached, and fully leached ore, as well as solution with varying metal contents, and
that a significant proportion of the metal is not recoverable, raises issues that are not
encountered in other processes. The main concerns are the inventory of metal in the
leach solution and the determination and prediction of the recoverable metal in the
feed material. Often the Check in-Check out process cannot be used, especially for
Dump Leaching, as the mass and analysis of the dump is not known to a sufficient
degree of accuracy and it is impractical or not cost effective to measure the metal
remaining in the heap independently at the end of the accounting period. The metal
remaining in the heap is then determined by subtracting the quantity that has leached
into solution from the original estimate of what was deposited on the pad. If this
practice is utilised it must be approved by means of an Exception Report.
The principles in this Code thus apply to Heap Leaching only, as they are not
possible to achieve for most Dump Leaching operations.
Feed material must be weighed, sampled and analysed according to the Code and
the material must be tracked to the specific pile or pad. Drained pregnant solution
from each heap and the barren solution returned as irrigation to the heaps must be
accurately measured by suitable flow meters and sampled for analysis.
Reconciliation of these flows is important in order to provide a calculation of the
solution inventory in the leach piles, the metal extraction from all the heaps/modules
under irrigation and the total plant production. To perform this reconciliation requires
estimates to be made of the amount of rainfall, evaporation and the solution content
of each abandoned heap. Rainfall and evaporation must be based on actual site
measurements over the accounting period. The solution content of operating,
abandoned and resting heaps must be periodically established by test work, although
standard figures based on experience are often used. The metal content in the
solution in the heaps is normally the largest proportion of the total solution inventory
and it is usually equivalent to several months’ production.
However, the metal accounting is complicated further by the fact that a significant
proportion of the metal is not recoverable. Thus, it is common practice for process
control, to utilise the recoverable metal placed on the heap, rather than total metal,
for mass balance and performance calculations. If recoverable metal, as opposed to
To achieve a Check in-Check out balance, when a particular pile/heap has reached
the end of its life after leaching and elution has been completed, a closing balance
must be performed and the heap sampled to check the amount of total and
recoverable metal remaining. If the heap is to be left in place, this sampling should
preferably be performed by mechanical drill/augur or sonic drilling. If it is to be
moved, it must be sampled conventionally as it is moved. The metal that was not
recovered must be removed from the plant inventory when a heap is abandoned.
Sampling of heaps is subject to a high level of inaccuracy and due cognisance of this
must be taken in the use of the information generated for stock evaluation and for
use in predicting future performance.
The metal balancing from the output leach solution through the purifying and refining
steps to metal product is similar to that for Hydrometallurgical processes. General
practice is to calculate the recovery of the metal finally produced, after the
purifying/refining process, as a percentage of the metal placed on the heaps.
• All input and output streams must be accounted for, i.e. Check in-Check out
• Accountability must be better than 100% (± 5% at 95% confidence levels) for
the product metals and 100% (±10%) for impurities, for the results to be
utilised for economic and process decisions
• As quantities are much smaller than a full-scale plant, static weighing on
certified weighers only must be used to establish feed and product weights
• Samples are a larger proportion of the process flows than in a full scale plant
and the mass and metal content of the samples must be included in the metal
balances as part of the Check in-Check out
• Similarly, spillage must be collected, weighed, analysed, and accounted for in
the balance at the appropriate stage. It should not be returned to the circuit
as it causes process disturbances. Spillage should be kept to a minimum
In the same way, the measured mass of metal in the ore delivered to the shaft, or the
ore dumping point, from the mining operations is compared with the estimated metal
content in the ore mined, as determined by sampling and surveying, to give the shaft
(or mine section) call factor, also expressed as a percentage. The combined plant
and shaft call factors give the mine call factor for the operation. Typically, these call
factors are calculated monthly and they form a useful means of monitoring
discrepancies between the expected and actual mine output.
The report must include tonnage and analysis figures for all metal accounting
streams, and often a comparison of performance figures against key performance
indicators, and variance from the operating budget or business plan.
The report must contain sufficient information to assist in diagnosing plant operating
problems and for future production planning and project evaluation.
To meet the above objectives, the report must be compiled by adequately trained
personnel and it must use data which has been subjected to the minimum number of
manual entries, with clear traceability to the sources of all inputs. Compilation of the
report must be via the use of approved metal balancing procedures and no data may
be adjusted without written authorisation from the responsible person. The report
should also comply with International Auditing Standards as set out in IAS 2 and it
must be signed off by a Metal Accounting Committee, or another suitable authority,
for subsequent approval by the Audit Committee of the Company’s Board of
Directors.
The design and specification of the system must include the definition of all risks or
consequences of a degradation of the accuracy of the metal/commodity balance
process and it must specify the means by which these risks can be quantified.
The metal accounting system must comply with this Code and it must include the
following:
Where existing plants do not meet the requirements of this Code, either steps must
be taken to comply with these requirements, or non-compliance must be motivated
and approved via an Exception Report.
11.1 INTRODUCTION
A company must have metal accounting rules and a procedure for auditing
compliance with these. These rules are primarily technical to provide management
information and to promote good Corporate Governance but they must also provide
the quantities and the reconciliation necessary for the financial reports.
Accounting Period – The period in which a company would normally report its
financial results to the market. For example in the USA this is quarterly, while in
many other financial markets the requirement is bi-annually.
A measurement that is accurate is one that is free of bias and it has a dispersion
(standard deviation) that is lower than a defined dispersion or indeed a probability
density of a particular nature. The level of dispersion or the nature of the probability
density is defined with the purpose of separating measurements that are entirely fit
for a particular purpose or use from those that are not.
Acid Mine (Rock) Drainage – The low pH effluent which seeps from waste rock, ore
dumps or stockpiles of materials which have an insufficient capacity to neutralise the
acidic products, containing the dissolution products of minerals.
Actual Recovery – The sum of all useful or desired outputs from a process, divided
by the sum of all inputs to that process, expressed as a percentage.
Barren Solution (Tails Solution) – The discard solution from a metal recovery
process.
Battery Limits – The boundaries of a metal accounting system, across which all
inputs to and outputs from that system are measured and balanced.
Best Practices – The most accurate method of performing a measurement with the
information and equipment available at that time.
Bias – Bias is the difference between the mean result of one or more measurements
and the true value of the quantity being measured. It also can be seen as a
consistent difference between two, or more, measuring systems. (e.g. sender’s and
receiver’s weights, the analysis of the same sample by two laboratories etc.)
Bubble Method – A method whereby the flow of material through a process plant is
interrupted to allow for equipment to be emptied in series to facilitate stock-takes
(normally applied to high value materials in the latter stages of processing).
Bulk Density – The density of a granular material normally stored in a bin, other
container or stockpile, defined as the mass of the material divided by the volume
occupied by the material (including the space between the particles).
Calibration Test (weighers) – A test, using known weights and forces to load a
scale, in order to determine the performance and operating tolerance compared to
standard values.
Certified Weight – A weight, which has been certified by the relevant national body,
or an agent authorised by that body, to have a mass within a defined tolerance, to be
used for certifying weighing instruments.
Check in-Check out System – The system whereby all streams into and out of the
Process or Plant, for which the balance is being performed, are measured, sampled
and analysed.
Closure Errors – The difference between the inputs and outputs for a complete
metal balance (that is Check In-Check Out) over a plant or process, including
inventory and stock changes, over a prescribed period, and zero.
Coefficient of Variation – The quotient of the standard deviation and the mean
value of a measurement, expressed as a percentage.
Coriolis mass flow meter – A device consisting of a flow sensor and a transmitter
which primarily measures the mass flow by means of the interaction between a
flowing fluid and the oscillation of a tube or tubes; it may also provide measurements
of the density and the process temperature of the fluid.
Custody Transfer – The transfer of risk and/or ownership of a material from one
party to a second party. This includes ore from a mine to a processing plant,
intermediate products such as concentrate and matte from one plant to another, both
in-house and inter-company and waste products for sale or safe disposal, as well as
final products.
Evaluation – The process by which the data required for Metal Accounting is
acquired.
Gantry scale – A scale where the load measuring elements are attached to a gantry
crane normally used for loading or offloading materials in bulk.
Heap Leaching – Normally refers to the leaching of ores that have been crushed (to
an economic size) prior to placement on an impermeable pad on which suitable
drainage lines have been placed for the collection of lixiviant that has been
percolated through the crushed and prepared ore. The crush size selected may vary
considerably but typically it ranges between 12 and 50 mm.
Home and Away System (Double Sampling and Analysis System) – A system
where batches/lots are weighed and sampled by both the sender and receiver. Data
from the analysis of the samples can be statistically analysed to establish the
presence of biases in the sampling and/or analytical processes. In the “Half Home
and Away System” samples from even numbered batches of material are analysed
by the sampling sites laboratory (i.e. the sender’s sample at the sender’s laboratory
and receiver’s sample at the receiver’s laboratory) and odd numbered batches are
analysed at the other laboratory (i.e. the sender’s sample at the receiver’s laboratory
and receiver’s sample at the sender’s laboratory). Statistical analysis of the data can
establish the level of bias in either the sampling or the analytical procedures.
Hopper Scale – A scale designed for the weighing of granular materials in bulk,
whose load-receiving element is a self-cleaning hopper with an outlet gate.
In-motion weighing – The weighing of objects that are in-motion. Usually applied to
rail or road trucks.
Mine Call Factor – The ratio, expressed as a percentage, which the specific product
accounted for in the product recovered from a plant, plus that in tailings/residues,
bears to the corresponding product called for by the mine’s measuring and evaluation
methods. It is also the product of the Shaft and Plant Call Factors.
Momentum Mass Flow Meter (Impact Mass Flow Meter) – A device that detects
mass flow rate by measuring the momentum of a stream of material falling a definite
distance under gravity.
Plant Historian – A computer system for accurate operational data archiving and its
rapid recovery. The data historian accepts data input only, it does not recognise
structures and procedures, and it is strictly concerned with accurate, efficient data
storage and recovery if required. Data historians perform three key functions:
• Data compression
• Rapid data recovery
• Data save-set facility
Plant Inventory – The inventory can be stockpiles of feed, intermediates or products
(plant stocks) as well as transient and long-term in-process inventory.
Precision – Precise measurements have a dispersion about their mean value which
is lower than a defined dispersion or indeed a probability density of a particular
nature. The level of dispersion or the nature of the probability density is defined with
the purpose of separating measurements that are entirely fit for a particular purpose
or use and those that are not. A precise measurement may not be accurate; its
mean may differ from the true value of the measured quantity by an arbitrary amount.
Standards often quantify the term precision as a value corresponding to the
magnitude of a 95% confidence interval around a result. The interval is ± ts, where t
is the two-sided Student-t value at 95% confidence and s is the estimate of the
standard deviation.
Random Errors – The component of the error of measurement, which in the course
of a number of measurements of the same measurand, varies in an unpredictable
way.
Random Stratified Sampling – The taking of increments whereby the material being
sampled is divided into strata, one increment being taken at random within each
stratum. (AS 4433.1 – 1997). For example, if the sampling interval of a cross stream
cutter is based on time, say every 5 minutes, the cutter is not triggered at 5, 10, 15
etc minutes but rather at a random time within each 5 minute interval.
Recoverable Metal – The proportion of a metal in a feed material that has been
determined to be recoverable, by laboratory and/or pilot plant tests, by the chosen
process, with the remaining metal being defined as Non Recoverable.
Risk – The chance of an event occurring that will have an impact on objectives. A
risk may be quantifiable in terms of the likelihood of loss, less than expected returns
or an undesirable outcome. (The VALMIN Code, 2005 Edition)
Risk Assessment – A formal process for the identification and quantification of the
risks associated with all the processes involved in metal accounting. For example, a
decrease in accuracy of measurement may lead to the true loss of metal exceeding a
threshold value; a risk assessment will quantify the probability that the threshold is
exceeded.
Run of Mine (ROM) Ore – Ore produced from a mining operation, which has not
been subject to any beneficiation.
If the sampling constant is taken to be dependent only on particle top size and the
sampling constant is known at one particular state of comminution of the material, a
'sampling safety line' can be drawn for the particular analyte to which the sampling
constant applies.
Shaft (or Mine Section) Call Factor – The mass of a species delivered to a shaft
compared to the mass of that species calculated from sampling and surveying
procedures, expressed as a percentage.
Sonic Drilling – Sonic drilling involves drilling with a drill stem that is excited at a
relatively high frequency, resulting in a standing wave being transmitted down the
drill stem and back upwards. This vibration substantially reduces the friction between
the stem and the material being drilled. It is claimed that core recovery is better for
sonic drilling than conventional coring methods.
Specific Gravity – The specific gravity of a substance is the ratio of the density of
the substance at a specified temperature to the density of a reference substance at
the same or a specified different temperature (usually the reference substance is
water at a temperature of 4 ºC). Specific gravity is a dimensionless quantity and is
also referred to as 'relative density'.
Triple bottom line reporting – The triple bottom line (TBL) focuses corporations not
just on the economic value they add, but also on the environmental and social value
they add - and destroy.
At its narrowest, the term 'triple bottom line' is used as a framework for measuring
and reporting corporate performance against economic, social and environmental
parameters.
At its broadest, the term is used to capture the whole set of values, issues and
processes that companies must address in order to minimize any harm resulting from
their activities and to create economic, social and environmental value. This involves
being clear about the company's purpose and taking into consideration the needs of
all the company's stakeholders and shareholders, customers, employees, business
partners, governments, local communities and the public.
Ultrasonic Transit Time Flow Meters – A flow meter utilising the principle that the
propagation velocity of sound waves within a fluid is directly influenced by the
velocity of that fluid.
Weighbridges – Weigh scales designed to weigh road trucks or rail wagons where
the lever system is usually below ground in a pit so that the vehicle or rail wagon can
pass over and stop on the scale. In a large capacity scale it also refers to the
Yield – The quantity of coal (of a specified quality) obtained from Run of Mine
Material expressed as a percentage. There are two subsets:
Page
1 INTRODUCTION .................................................................................1
1.1 SAFETY AND HEALTH 2
1.2 PRINCIPLES OF METAL ACCOUNTING 2
1.3 STANDARDS 4
1.4 TRACEABILITY 4
1.5 ACCURACY IN METAL ACCOUNTING 4
1.6 PRECISION IN METAL ACCOUNTING 4
1.7 ERROR TYPES 6
1.7.1 Random Errors 6
1.7.2 Systematic Errors 6
1.7.3 Illegitimate (Spurious) Errors 8
1.8 METAL ACCOUNTING RISK 8
1.8.1 Introduction 8
1.8.2 Misrepresenting the Performance of the Operation 9
1.8.3 Real Losses of an Unknown Nature 10
1.9 TRAINING 10
1.9.1 Plant and Laboratory Operators 10
1.9.2 Plant and Laboratory Management 11
1.9.3 Operations Management 11
2 DATA MANAGEMENT...................................................................... 12
2.1 INTRODUCTION 12
2.2 DATA ACQUISITION AND ENTRY 12
2.3 DATA DOCUMENTATION 13
2.4 DATA PROCESSING 13
2.5 QUALITY CONTROL (QC) AND ASSURANCE (QA) 14
2.6 DATA FORMATS 14
2.7 DATA STORAGE AND BACK-UP 14
3 MASS MEASUREMENT ................................................................... 15
3.1 INTRODUCTION 15
3.2 ACCURACY AND PRECISION OF MASS MEASUREMENT 17
3.2.1 Scales/Weighing Equipment 18
3.2.2 Flow Meters 19
GLOSSARY.................................................................................................... 109
LIST OF ACRONYMS AND ABBREVIATIONS ............................................. 128
APPENDICES................................................................................................. 129
APPENDIX A – ACKNOWLEDGEMENTS..................................................... 129
APPENDIX B – REFERENCES...................................................................... 131
APPENDIX C – STANDARDS........................................................................ 139
APPENDIX D – CHECK-LISTS ...................................................................... 150
APPENDIX E – MASS BALANCING TECHNIQUES ..................................... 159
APPENDIX F – DETERMINATION OF THE SAMPLING CONSTANT.......... 171
APPENDIX G – ERROR AND VARIANCE PROPAGATION IN TOLL
SMELTING OPERATIONS TREATING MULTIPLE
CONCENTRATE STOCKPILES.......................................... 174
APPENDIX H – USEFUL WEB SITES ........................................................... 192
This document is intended to serve as a set of Guidelines for the Code on Metal
Accounting, which will in the future be controlled by suitable professional
organisations closely associated with the industry, and gain industry acceptance
worldwide. It should also be regarded as a living document which will be revised
through contributions from sponsor companies, the related research program and
other initiatives. In this Guidelines document, unless otherwise specifically stated,
references to the “Code” include both the Code and the Guidelines.
A number of sponsor companies and others involved in, or connected to, the metals
production industry have contributed to the contents of this document through the
provision of internal standard procedures and/or specifications, in assisting with
literature searches and by providing advice and information. A list of these
companies and individuals is included in the Appendix A, and their contributions are
gratefully acknowledged.
The aim of this document is to provide standard generic procedures and guidelines
for obtaining credible figures of quantities of materials processed and produced and
further, to be recognised and accepted as the Industry guidelines for best practice.
In cases where companies or operations are unable to comply with the best
practices set out in the Code and these Guidelines, the reasons for this non-
compliance must be documented in the form of an Exception Report. This must
clearly set out the reasons for the non-compliance, the steps taken to mitigate the
effects of the non-compliance, the resultant impact on the overall integrity of the
metal accounting system and plans to ensure compliance in the future. This
Exception Report must form part of the documentation of the metal accounting
system and, as such, must be approved by the Competent Person responsible for
the design of the metal accounting system, and by the Audit Committee of the
company concerned.
There are no differences in principle between base and precious metals and
commodities, but there are differences in the way the Code/Guidelines would be
applied. These Guidelines set out principles for obtaining credible figures, such as
the methods to be used and their accuracy, recommended methods for mass
measurement, sampling and analysis and the applicable International Standards to
achieve this, as well as the basis for the design and management of a Metal
Accounting system. More details are included in the Appendices.
Due to the random variations that are inherent in all measurement processes, the
values derived in metal accounting will not be exact but will have a degree of
uncertainty. However, in correctly designed and operated systems the level of
uncertainty is measurable and should be quoted in the accounting reports. If the
measured precision value is outside of the expected range, and the range will vary
depending on the materiality of the measurement concerned, its cause should be
identified and rectified.
Staff must be trained in the safety and health aspects of their activities and supplied
with the necessary Personal Protective Equipment (PPE). The possibility of potential
exposure to hazardous dusts, gases and chemicals, as well as other hazards, must
be fully understood and adequate precautions implemented.
1.4 TRACEABILITY
In the areas of mass measurement and analysis, the aim should be to establish
traceability to international standard (SI) units.
As defined in this Code, accuracy is related both to random errors (Section 1.7.1),
which are an integral part of the measurement process, and to systematic errors
(Section 1.7.2).
Bias occurs as a result of systematic errors which occur when the mean value of the
random errors is not zero over an extended period which contains many
measurements. Bias is difficult to detect but it can and should be eliminated (or at
least minimised) because its effect accumulates in any cumulative measures of
production or performance.
A metal accounting system, which is performing well, will provide good agreement
between the metal balance, calculated using many data inputs, and the commercial
values used as the basis of commercial transactions. However, the values only need
to be similar to within the precision of the Metal Accounts.
VM can be further divided to show the variance due to the determination of the wet
mass, including the components of the mass measurement process, and that due to
the determination of the moisture of the material. In a multi-stage sampling process,
VS should be expanded to show the variance of each stage.
Where the processes are not independent, (i.e. they are related by a set of
procedures or combined calculations), it is appropriate to use a more general
relationship called the Propagation of Variance which takes covariance of quantities
into account. This relationship allows the variance of calculated values (such as
Recovery) to be estimated in a statistically correct manner.
Because variance is a squared term, (it is equal to the square of the standard
deviation) the operation with the largest variance will have the major impact on the
Total Metal Accounting variance. For this reason, it is important that the precision of
each operation is measured and that it has an acceptable value. An operation in this
context can mean any mass measurement or sampling and analytical procedure.
The precision for mass measurement, sampling and assaying can be established by
carrying out suitable experiments that are described in a number of International
Standards. The operation with the highest relative variance, i.e. the poorest
precision, can be identified and steps taken, if necessary, to reduce the value of the
variance. For example, the number of increments at a sampling stage could be
increased to reduce VS, or the number of replicates taken for the analysis could be
increased to reduce VA. If it was decided that the operation with the highest variance
could not be improved, for example, due to the cost involved, there would be no
purpose in reducing the variance of any other operation to be significantly below this
highest value.
In typical metallurgical plants, where process streams are sampled and analysed on
an 8-hour shift basis and about 90 data sets are produced per month, relative
precision levels (one standard deviation) of 3% for each measurement are
acceptable, because this equates to a monthly precision of 0.32%. However, where
individual shipments/lots are being evaluated, such as in toll treatment, individual
precision levels of < 0.5% will probably be required.
For operations generating a final metal product, commercial transfer precision will
usually be a small fraction of the above targets, but the required precision for coarse
mill feed (f80 of 100 mm or larger) may not be achievable without a substantial
investment in sampling and sample reduction.
and, by convention, it is usual to quote the uncertainty at the 95% confidence level,
which is equal to ± 2 × σ TMA.
Most random errors are assumed to be drawn from a normal distribution. If the
measurement in question is the mean of repeated measurements, then this
assumption will always be approximately correct, the approximation improving as the
number of repeated measurements increases.
Measurement error distributions can be truly random but they have an expected
value or mean that is non-zero and still cause bias as noted in the previous
paragraph. For example, in mass measurement, the most common form of bias is a
small instrument offset. If this offset is small compared with the random error of the
measurement, it is essentially impossible to detect from a single measurement.
Given a situation in which it is known that the expected value of a quantity should be
zero (e.g. the difference in assays of the same sample between two laboratories), it
is possible to detect a small bias by simply summing the differences over time. A
bias component in the difference will accumulate, while the random errors of zero
mean will tend to cancel. A plot of this cumulative sum of the differences is known as
a CUSUM graph.
CUSUM graphs are useful tools for detecting and measuring the level of the
systematic errors, especially if these are relatively small. CUSUMS, or ‘cumulative
sum’ methods, are very valuable for tracking the statistical parameters of a
measurand. They can be used to track the bias between two measured quantities
or they can be used to track the variance of a measurement. In the past, the
application of ‘cumulative sum’ methods implied that the differences between
measurements were simply summed over time. If the cumulative sum value
increased with time, it was interpreted as meaning that there was a real positive
difference between the values.
Such a simple basis for CUSUM methods does not, however, take advantage of
recent advances in statistical development. There are far better methods for
formulating CUSUM procedures. These are described in Hawkins and Olwell. They
describe methods that can be ‘tuned’ to detect changes in the statistical parameters
of measurements and they can also be tuned to provide the desired speed of
response to changes with a given level of risk of ‘false alarms’.
Even without bias, these differences are themselves subject to random error and
they can be expected to be negative or positive for a few times in succession. This
makes the recognition of a bias quite difficult.
In some measurements there are known errors within a system that cannot be
eliminated but they must be quantified and allowed for. These errors, which are
inherent in the method being used, produce a bias from the true value. For practical
reasons, these methods are utilised rather than an alternate method, which would
give a result closer to the true value. Examples in metallurgical accounting systems
are the:
1.8.1 Introduction
Any production operation is exposed to financial risk associated with its metal
accounting process, to the extent that there is uncertainty in the evaluation of the
quantities of metal delivered to that operation and produced by it. In addition, there is
the risk to the reputation of a company and of negative market perceptions resulting
from major metal accounting discrepancies. One of the prime objectives in the
design of a metal accounting system is to enable an operation to quantify, manage,
and minimise the level of risk to which it could become exposed through failures or
shortcomings in its metal accounting system. Each operation must conduct an initial
risk assessment of its entire Metal Accounting system to establish and quantify the
risks. The design of a Metal Accounting system must take these risks into
consideration and set down procedures to be followed when accounting
discrepancies are identified.
The use of statistical analysis and check measurements and analyses provides a
database, which becomes an essential component of this risk management process.
The other fundamental component of the risk management process is regular and
frequent management reviews, incorporating risk assessment, of the entire metal
accounting system, supported by both internal audits and external audits by the
Competent Person.
The main areas of risk to which an operation could be exposed in its metal
accounting system are listed below.
1.9 TRAINING
As stated in Section 1.1 on Safety and Health, it is expected that companies will have
full training procedures, in line with National and International Standards, such as
ISO 9000, 14001 and 17025. The training requirements of staff involved in Metal
Accounting will obviously vary depending on the role they are playing. Appropriate
sections of the Metal Accounting Guidelines must be available to staff so that they
can understand their part in the total system and the requirements necessary to
produce unbiased data to a known and acceptable precision.
Laboratory Management must ensure that their analytical procedures are accurate
and that they are adequately monitored by quality assurance and quality control
procedures. There is a requirement for an in-depth understanding of statistical
analysis techniques.
It is likely that a 4-year tertiary qualification will be required for both Plant and
Laboratory management.
Liaison must be established between the various internal disciplines involved in metal
accounting and the appointed Competent Person. There must be an understanding
of the relevance of Metal Accounting, its relationship to good corporate governance
and the triple bottom line.
2.1 INTRODUCTION
The operation of a Metal Accounting System requires the accumulation and
evaluation of process data (e.g. mass measurements and analyses) from various
sources (e.g. SCADA, Plant Historians, LIMS systems, etc). The acquisition,
processing and management of data require a substantial portion of the effort
associated with Metal Accounting. Efficient and reliable data management adds
considerable value to the Metal Accounting System.
The goal of data management is to create files from data obtained from various plant
sources in such a way as to prevent error during transcription and to enable analysis
of the errors inherent in the raw data. Production (e.g. mass measurement) and
laboratory data (e.g. analyses) must be integrated to a single user interface to
facilitate Metal Accounting. At the same time, it is essential that those responsible for
entering plant operating data, and those entering analytical results into the system,
operate independently of each other.
Many errors can occur when entering or acquiring data and these need to be
guarded against. Common examples of errors are:
• Transpositions
• Copying errors
• Routing errors (e.g. information incorrectly routed from a SCADA system)
• Consistency errors
• Range errors (e.g. a value outside the plausible range for a variable)
Methods to avoid such acquisition or input errors include:
3.1 INTRODUCTION
The overriding objective of mass measurement for Metal Accounting purposes is to
establish the mass of the particular material or component present at a specific time,
or the mass flow of that component over a defined time period, to a defined
accuracy suitable for mass balancing.
For the purposes of these Guidelines, “mass” should be read as “mass or volume,”
as often for practical purposes, volume is measured (e.g. stockpiles or fluid flow). In
these cases the specific gravity/relative density of the slurry, and of the solids
and/or liquids, and/or bulk density, must also be determined by appropriate methods
in order to calculate the mass. Materials often contain moisture, the proportion of
which must also be established in order to obtain the dry mass.
This Code is concerned with the requirements for the first measurement area only.
Secondary Accounting measurements are used for checking and investigation
purposes, in conjunction with other process data, where confirmation of Primary
Accounting information is required.
Feed material to metallurgical plants is often ore, supplied from mining operations in
large quantities, which has a wide and variable size range and moisture content.
Mass measurement of this stream is required for Primary Accounting but, in practice,
it is often weighed in trucks or hoppers, or on conveyor belts, to a lower level of
accuracy than required for the final products. Achieving a high degree of accuracy
for this type of material is more difficult and costly and it requires appropriate design,
selection and installation as emphasised above. If it is felt that the cost and effort to
achieve certification is not justifiable, then this must be motivated by way of an
Exception Report.
In all weigher and flow meter installations, it is essential that the selection, system
design, installation, operation, maintenance and calibration follow the applicable
standards and manufacturers’ recommendations, in order to achieve the accuracy
and precision levels quoted in Section 3.2, and those which are required by
International and National Standards and National Metrology Acts.
In general, the larger the quantity being measured and the more variable in size and
flow rate, the lower the accuracy. In addition, in the case of precious metals and
base metals, the larger the stream or quantity of material, generally the lower the
content of the constituent for which the metal balance is required, which affects the
analytical precision.
Thus, in the mining and metallurgical industry, generally the accuracy of mass
measurement will increase through the processing chain from raw feed or ore to the
final metal, as the materials become more homogeneous in size and composition,
with a higher grade, and they are present in smaller quantities.
The cost of the mass measurement device is often only a small portion of the cost of
the total system required to obtain a mass flow measurement suitable for accounting.
A total system includes equipment for calibration, such as weigh tanks or bins, or
other facilities for check weighing. However, the higher capital cost incurred due to
correct design and installation of the system will often result in a lower total cost of
Static scales provide the most accurate and precise method of measuring mass and
must be the choice, where circumstances permit, for custody transfer and Primary
and Secondary Metal Accounting purposes. They can be easily certified by
standardised procedures. Static scales include platform scales, road and rail
weighbridges, gantry, hopper and bin scales.
Precisions generally range between ± 0.05% to ± 0.2% for platform scales and
± 0.1% to ± 0.5% for weighbridges, on gross load measurements. For hopper/bin
scales and gantry scales the equivalent figures are ± 0.1% to ± 0.25% and ± 0.15%
to ± 0.4% respectively.
The accuracy of in-motion weighing is affected significantly by the truck speed and
acceleration. This effect is more apparent in the weighing of road trucks where the
control of speed and acceleration is much more difficult than for rail. Thus, in-motion
weighers are certifiable for rail use but not for road use. Certified in-motion weighing
is a suitable choice for Primary Metal Accounting for rail transport, provided that the
key principles of taring each truck or hopper on each occasion and controlling the
speed are adhered to. In-motion weighing should give a precision of ± 0.5%.
Certified electromechanical belt weighers must be the second choice for metal
accounting purposes after static scales or in-motion weighing, provided that the
design, installation, calibration, operation and maintenance requirements are
adhered to.
The accuracy of conveyor belt weighing is installation specific and it is affected by:
• belt effects such as tension, stiffness and misalignment of the belt or the
idlers
• flexibility of the weigh carriage
• spillage onto the weigh section
Nuclear belt weighers, although not subject to belt errors, are generally less
accurate because of effects related to the material conveyed. Precisions of ± 1–2%
can be achieved but, in practice, are often much worse. They should not be used for
Primary Metal Accounting but they can be used for Secondary Metal Accounting if
the material is homogeneous in size and composition and the belt loading uniform.
• Percentage of span
• Percentage of a rate or reading
• Percentage of the upper range limit
The accuracies quoted below are those at 95% confidence levels.
Coriolis mass flow meters are capable of high accuracy, ± 0.2% of reading in
certain cases, but they only have limited application in the extractive industry, as they
require low fluid velocity and they are sensitive to the quantity of entrained air.
Ultrasonic transit time meters have limited applicability for specific applications and
in those cases, they can give an accuracy of 0.5 to 2% of the reading.
Similar information for the various types of mass and volume flow meters is given in
various textbooks and publications.
The various methods for measuring mass are listed below. Their applicability and
the recommended guidelines or Standards are covered in the sections which follow.
All methods using volume or volume flow determination require the measurement of
the bulk density of the material in the stockpile or bin, or the specific gravity
(relative density) of the slurry or liquid. This can be performed by using a nuclear
(gamma) gauge (Section 3.7.2), in a few cases by a penetrometer (Section 3.7.4),
or preferably by bulk or laboratory tests (Section 3.7.1). The dry relative density of
the solids and the density of the carrier, in the case of slurry flow which is often water
Material in, or being loaded into or unloaded from, rail or road trucks is usually
weighed on static weighbridges or platform scales, as this is the most
accurate method. The equipment is described by Colijn3.
Hopper/Bin Scales, either singly or in pairs, are also capable of good levels of
accuracy although they require a more elaborate installation for loading or
offloading of the materials. The equipment, design and installation guidelines
and calibration techniques are described by Colijn4 and in literature available
from the equipment suppliers. Gantry scales are usually used for bulk
offloading of ship cargoes. These scales can be calibrated with certified test
weights as for the platform scales and weighbridges and they can be certified
by the relevant authorities.
This is usually used for large quantities such as bulk commodities transported
either by road or by rail. The various types are described in Colijn5, who also
lists installation guidelines and factors affecting the uncertainty of the
measurement.
In theory, the individual weighing of complete, uncoupled trucks should give the most
accurate result. However, modern in-motion weighers normally weigh individual
axles on a coupled basis, as this is simple, more rapid and practical to perform and
the software can compensate for coupling and speed effects. The multiple weighing
of individual axles provides a larger number of readings, which can assist in
evaluating the results and improves the average deviation over a measuring period.
For more accurate installations it is common to have four or eight strain gauges.
Previously train speeds were limited to 8km per hour for certification but with the
latest equipment it is possible to operate up to 20km per hour or higher.
Rail in-motion weighers can be calibrated and certified according to various
standards and national Trade and Metrology Regulations (in South Africa this is
covered in Regulation 61) and used for custody transfer purposes. However,
certification is not possible for road trucks at present because of the variable
accuracy due to the effect of truck speed and the acceleration or deceleration, which
is difficult to control.
The mass flow of feed materials, or inter/intra plant intermediates (e.g. matte),
required for Primary Metal Accounting, must be weighed in the same manner or by
in-motion weighing, even if not required commercially, or by belt conveyor weighing,
as specified below, with the weighing equipment certified, since these streams are
critical for Metal Accounting. However, it may be decided, after due assessment of
the risk to the company and an evaluation of the costs/benefits, company needs and
practical limitations, that certification is unwarranted. In these cases, suitable in-
Where a product is transferred from one plant to another, it is good practice for the
mass to be measured at both loading and unloading points for on-going comparison
and error determination.
Static scales are in general more accurate than belt scales, and they are also
capable of greater sensitivity, because they weigh larger quantities of material at
one time. In addition, the accuracy of belt scales in operation is often considerably
poorer than that expected, based on the specification of the scale, because of poor
choice, system design, or bad operation or calibration.
Certification
Certain makes and types of belt conveyor weighers and installations can be certified
(assized) for trade (custody transfer) purposes, as they are, for example, in the
supply of coal to power stations. Certification is installation specific and procedural
requirements may vary from one country to another. For commercial and Primary
Metal Accounting purposes, belt weighers must be certified in terms of this Code.
Where, after an assessment of the specific application, material and importance of
the process stream, the practicality and cost of achieving an installation of the
required standard and the risk to the company, it is decided that certification is not
warranted, an Exception Report must be motivated and agreed, by the Company’s
approval procedure, to permit uncertified operation. However, even in these cases,
the design, installation, operation and calibration should be in accordance with
suitable and clearly defined in-house procedures, so that a maximum error of ± 0,5 %
can be achieved and maintained.
Most modern weighing installations are fitted with electronics for the
integrator, that facilitate the calibration and resetting of the speed zero and
span and the mass zero and span automatically, and accepting the
instrument readings as correct. However, there is no substitute for
fundamental checks of belt speed by timing a measured distance of travel
and of belt loadings by calibration with check weights. All these results, both
the as found and corrected, must be recorded and analysed to establish the
operating accuracy of the weigher.
They are simple to install and they are unaffected by belt effects, so the
conveyor design is not as important as it is for electro-mechanical weighers.
Nuclear belt weighers are also not susceptible to external forces such as
wind, they are less maintenance intensive and easier to calibrate. However,
they have the same operational limitations with respect to the measurement
of belt speed, plant housekeeping, moisture determination and regular
calibration. Most importantly, they are affected by variation in the size,
composition and shape of the material being measured and they are therefore
normally less accurate than electromechanical weighers, with maximum
errors of ± 1 to ± 2% being achievable, but higher error levels being common.
They can also be used, in certain applications, with nuclear neutron sources
to determine moisture content. Because of the fact that their measurements
are subject to errors from variations in the material being weighed, they
should not be used for metal accounting measurements.
3.4.4.1 Introduction
For convenience and practical reasons, mass flow in closed conduits is often
determined by measuring volume flow rate plus density or specific gravity in
order to calculate the dry mass. This type of measurement is very common in
the extractive industries for process flows and liquid and/or gaseous effluents.
The determination of the relative density or specific gravity is discussed in
3.6. The degree of uncertainty in the mass flow is made up of errors both in
the volume measurement, and in the density determination.
There are over one hundred types of flow meters commercially available,
operating on ten different principles. BS EN 7405 classifies flow meters into
12 different Groups (Strauss18). There is extensive literature available from
suppliers and many technical papers (Appendix B.1.) and standards
(Strauss19) on the selection of the correct flow meter for specific applications.
Strauss20 gives a comparison of the performance of different types of flow
AMIRA P754: Code Guidelines Release 3 20070219 28
Commercial in Confidence
meters and Strauss21 shows the possible applicable areas for the different
types of flow meters for the twelve Groups. However, in the metallurgical
industry, many of the key process flows, in conduits, required for metal
accounting, are abrasive slurries containing a significant amount of solids.
This effectively limits the choice of a suitable meter for these applications to
an electromagnetic flow meter. In a few cases, with low flows and fine
particles, it may be possible to utilise a Coriolis mass flow meter, or for low
consistent solids contents, transit time ultrasonic flow meters may be suitable.
For free flowing solids, momentum type mass flow meters may be used.
Thus, this guide will concentrate on these four types. For relatively clean
solutions, many types of flow meter will operate satisfactorily and give good
accuracy. The measurement of gas flow, with or without particulates, for
metal accounting is normally required for off-gases and stacks, and it needs
specialist equipment but is not covered in detail in this Code.
For critical metal accounting flows, such as thickener or flotation feed, where
the measurement is used for Primary Accounting, the instrument must be
calibrated regularly on the actual material in addition to the initial calibrations
conducted in the test facility.
• Volumetric method
3.4.4.4 Solutions
The measurement of the mass flow of solutions for accounting purposes is
mainly applicable in the industry to some process streams in the refining
sections of the process, and then usually only for intra-company accounting
or for the measurement of effluent streams, to determine losses or
conformance with permitted emissions. Differential pressure type meters will
operate satisfactorily (ISO 5167-1 gives the specifications for geometry,
operation and use), but at a lower level of accuracy than other types such as
magnetic flow meters. Thus they should not be used for Primary Metal
Accounting. The relevant standards, selection and calibration aspects are
covered in the section below on slurries.
3.4.4.5 Slurries
Electromagnetic flow meters
Slurries in the metallurgical industry are abrasive and normally have a range
of particle sizes, operating in the mixed-flow regime. Electromagnetic flow
meters are the most suitable for this application, providing the fluid is
conductive. They are available in a wide range of sizes and materials of
construction to suit the application. The measurement is unaffected by
pressure, temperature or viscosity. The measuring tube has no moving parts
and there are no restrictions to the pipe cross section.
However, they can only be used on conductive liquids and they are affected
by changes in the mineralogy, particle size, velocity and magnetic
constituents in the slurry. Meters of special design are available to handle
these magnetic materials and they give satisfactory results, providing the
magnetic content of the slurry is reasonably constant. Electromagnetic flow
meters are capable of giving an accuracy of ± 0.5% of reading or better on
water or clean solutions, providing they are installed correctly and calibrated
according to standard procedures. For slurries the accuracy is generally
lower.
Calibration
If used for critical metal accounting flows, such as concentrate, thickener or
flotation feed, because static or conveyor belt weighing is not possible, the
flow meter must be calibrated regularly on the actual material as outlined in
3.4.4.3.
3.4.4.6 Solids
Impact mass flow meters can be utilised for free-flowing solids and they
operate by measuring the momentum of the solid stream impacting on a
measuring plate. They are described in Colijn9, who also discusses their
applicability and limitations. They are unaffected by the density of the
material and are pre-calibrated by the manufacturer for the user. Calibration
checks are with test weights and do not require material tests. They measure
wet mass flow and can give accuracies of ± 1%.
The valuable component is normally very small (less than 1% of input) and
thus errors resulting from the measurement are seldom significant in the
overall metal balance calculation. However, because of the risk to a
company’s reputation from unauthorised emissions or from exceeding the
legal regulations and the emphasis on the triple bottom line, these
measurements are important.
For the same reasons as given in 3.4.4.7, the measurement and accounting for
certain trace elements and potential pollutants has great significance for a company’s
reporting. This is discussed further in 7.7.2.
ISO/TR 8363 gives the guidelines for the selection of appropriate methods and lists
28 different methods for measuring flow in open channels that include velocity-area
methods, weirs and flumes. This standard also gives the expected degree of
uncertainty for the various methods varying from ± 10% for some to ± 1% for a thin-
plate, V-notch, sharp crest weir (ISO 1438-1) and it lists the various ISO standards
applying to all the different methods. Thin plate weirs must be used for accounting,
for process streams such as effluents, as they are the most accurate form of
measurement for this purpose.
Draught Surveys
In these cases, a method commonly specified and accepted for material being
transported by bulk carrier is that of the draught survey.
There is not, at this stage, any ISO standard covering draught surveys but, in some
cases, the UN ECE Code for Coal Cargoes is used. There is also no degree of
traceability for the survey technique. This should be contrasted with a certified belt or
other type of weigher, the calibration of which can be traced back through a hierarchy
of testing to internationally accepted weight standards.
Marine surveyors quote an accuracy of ± 0.5% for full cargoes. Merks (1992) has
suggested, in an analysis of the precision of some draught survey data that the
coefficient of variation varies from 1% for full cargoes to 3% for partial cargoes. This
should be compared with that achievable from static scales or more advanced belt
scales as discussed elsewhere in this Code. On critical measurements, it is also
possible to improve the quality of the mass measurement by installing two belt
weighers in series and comparing the results continuously. However, the method of
mass determination for these custody transfers is usually specified in the sales
contract and in many cases, this is still the draught survey. Even where static or belt
conveyor weights are accepted, the draught survey is performed as a check and for
shipping purposes.
If the material is not homogeneous in size and content, sampling of a stockpile to the
accuracy desirable for metal accounting, is almost impossible. The uncertainty of the
metal content could be as poor as ± 15%; the only time when a stockpile value is
correct is when it is empty. In the case of bulk commodities, such as coal or iron ore,
where the material is reasonably consistent, and where the contents are large and
material to the operation, a survey may be useful as an independent check on other
tonnage measurements.
International Standards do not exist and the accuracy of the survey is dependent on
the skill of the surveyor.
For the purposes of metal accounting, often it is the change in stock or inventory that
is significant. If the metal content is critical, such as in the PGM or gold industry,
3.5.1 Transient
This is defined as process inventory (or lock-up) that is of short-term duration
because of some process disturbance and/or where the mass flow through is
continuous and live. It is normally contained in process vessels and equipment such
as mills, furnaces and reaction vessels. This can be estimated and calculated with
the same caveats as for tanks and bins above. It is possible that the accounting
metal can be present in more than one phase (e.g. solid and liquid, concentrate,
matte and slag) in which case estimates must be made for all phases. Normally,
estimates are made monthly, with periodic stock-takes to a more accurate level to
suit company reporting.
• That which is absorbed or trapped in the process, such as gold into concrete
foundations, which can only be recovered when the plant is dismantled and
demolished (permanent inventory)
• That which is recovered at regular intervals such as when mills are relined or
furnaces re-bricked (semi-permanent inventory).
Measuring and assigning quantities to the former is very difficult and must be avoided.
(However, a “rule of thumb” figure used for early generation South African gold
plants of 0.3% of total gold input over the life of Plant has been confirmed by clean
up on closure). Quantities for the latter must be quantified and updated as the
inventory is periodically removed and this information carefully documented for future
use.
In the Coal Industry, for example, the following are examples of the various types of
moisture that may be present in a sample:
• Free moisture
• Surface moisture
• Residual moisture
• Inherent moisture
• Equilibrium moisture
• Moisture holding capacity
• Total moisture
• Moisture in the analysis sample
• Hygroscopic moisture (lignite)
• Moisture of constitution
It is important to ensure that samples for moisture determination are taken in such a
manner, as specified by the Code, to ensure that the moisture in the sample is
representative of the material being weighed and that no evaporation can take place,
or water ingress occur, before the determination takes place. It is also very important
to ensure that any analysis is carried out on the sample that has been used for
moisture determination, or an appropriate correction made, in order that the
analytical result can be related directly to the dry tonnes value.
In the case of stockpiles, the water test can be used whereby pits are dug in the
stockpile, the material removed and accurately weighed. The pit is lined with an
impervious membrane and filled with a measured volume of water. The number and
location of the pits is decided by sampling theory.
In some cases historical information based on previous measurements and tests can
be applied, including analyses and mineralogy of the material, but it should be
regularly reviewed and re-measured.
Nuclear gamma gauges are also used for measurement of the bulk density in
stockpiles or heaps but can only penetrate to about 30 cm. The most common use is
for density measurement of slurry flow in conduits, used in conjunction with an
electromagnetic flow meter. The correct choice and accuracy is installation specific
and guidelines supplied by the vendors must be adhered to.
The initial calibration must be performed in the field at two points on the scale by
measuring the weights of known volumes of water and slurry passed through the
meter. Accuracy is affected by many of the similar aspects that affect flow meters. It
is essential that the relative densities of the solids and the carrier solution are
checked periodically and that the calibration be checked regularly, by correctly hand-
taken samples for SG. The meters themselves are capable of ± 1% accuracy, but
this is often not obtained in plant conditions, due to poor installation, inhomogeneity
and particle size changes.
3.7.4 Penetrometers
These instruments have limited use to determine the bulk density of stockpiles of
material such as coal of a relatively fine uniform size, by measuring the force
required to penetrate a cone to a known depth. They require a skilled operator and
cannot be used on stockpiles with varying size material or containing rocks, such as
ROM ore.
4.1 INTRODUCTION
Within the mineral and metallurgical industries, process streams are very often in the
form of a flow of coarse broken ore or slag, or in the form of a flow of slurried finer
material. Metal accounting balances for such processes demand that these process
flows are sampled accurately with precisions that will cause the accuracy of the final
balance to fall within set target range(s). The ranges, into which the balance
uncertainties for the various target elements or chemical species of interest fall
(referred to in what follows as a measurand), are prescribed by management
considerations of the financial risk associated with the uncertainties. These
uncertainties must be acceptable to any stakeholders in the operations as well as to
outside bodies that are prescriptive of corporate governance standards.
The accuracy of the final result of a sampling of a process flow for a given
measurand depends on four factors:
Intrinsic heterogeneity (IH) which is determined entirely by the size distribution of the
solids and the variation of the concentration of the measurand from one particle to
the next, within the various size fractions of the solids
Distributional heterogeneity (DH) which depends on the variation of the total
measurand concentration from one moment to the next, as the process flow passes
the sampling point.
In most cases, with appropriate investment in equipment and the correct
arrangement of the process layout, it is possible to present the process flow to the
sampler and to take ‘mechanically correct’ sample increments from the process flow,
so that the sample formed by the collection of the increments is unbiased. This will
invariably not apply to sampling of run of mine ore because of the difficulties
presented by the range of, and variations in, the size and grade of individual particles,
moisture, and the problems associated with the design of a multi-stage sampling
system to generate an unbiased laboratory sample from the primary sample. The
accuracy of the sample (the variance between the true average analysis of the
process stream over the sampling period and the final analysis of the sample)
depends on the heterogeneity of the process flow, the accuracy of the analytical
method and the number and mass of the increments taken from the flow.
TEi = FEi + GEi + QE2i + QE3i + WEi + DEi + EEi + PEi [Eqn 4.1]
In terms of the total error in metallurgical accounting, the error due to the mass
measurement including moisture determination, and the error occurring during the
analysis of the measurands, must be added to this equation.
Delimitation errors are caused by the incorrect geometrical definition of the sample
increment, (e.g. a parallel cutter opening in a rotary sampler) and they can be
eliminated if all parts of the stream’s cross-section are diverted by the sample cutter
for the same length of time. Increment extraction errors occur when a sample
increment is not fully extracted, (e.g. part of a process stream is missed by the
sample cutter), and these errors can be eliminated by ensuring that the increment is
completely extracted without any loss of sample.
Weighting errors are experienced when sampling a moving stream that has a high
variation in flow rate. It is not usually a major problem (assuming that both the
minimum and maximum flow rates are correctly sampled) and they can be overcome
by regulating the flow rate prior to sampling. As a general rule, the weighting error is
negligible for variations in flow rate up to 10%, and acceptable for variations in flow
rate up to 20%.
Careful sample preparation practice will ensure that the preparation error PE is
minimised.
The value of the variogram VE(t) corresponding to a lag of t increments is given by:
The resultant variogram is called the ‘experimental’ variogram and it includes the
variance of sample preparation and analysis as well as the sampling variance. To
give the ‘corrected’ variogram, which provides information on the sampling error only,
the sample preparation and analysis errors must be subtracted.
The sampling variance for systematic and random stratified sampling for mass basis
sampling is then given by the following equations:
Systematic Sampling
σS2 = AC + BmL [Eqn 4.6]
n 6n2
where: n = the number of increments
The first and second terms of these last two equations, correspond to the short-range
and long-range quality fluctuations respectively. For systematic sampling, therefore:
σQE12 = AC
n
As described by Gy, the fundamental error variance is proportional to the cube of the
nominal top size of particles in the sample and inversely proportional to the increment
mass.
It must be imagined that the material is separated into various size fractions by
sieving or some other suitable method, and, that for each size fraction, the particles
are sorted into composition classes, as illustrated in the diagram below, and the
particles within each size and composition class are assumed to be identical. The
amount of detail concerning the particles can always be improved by increasing the
number of size and composition classes.
Composition > 1 2 3 4 . . . Nc
Size v 1
2
3
4
.
.
1 2 3 4 5 6 7 8 9 10
.
Class
Ns
⎛ d i3 + d i3+1 ⎞
Vi = f ⎜ ⎟ [Eqn 4.10]
⎝ 2 ⎠
d d
where i and i +1 are the sieve apertures defining the i th size fraction and f is a
shape factor than can be taken to be 0.5 for blocky particles. The particle density is
ρ ij mij = ρijVi
denoted by which gives them a typical mass of . The concentration
a ij
of the measurand in the particles is denoted as .
With the definition of these values for the various size and composition fractions, it is
nij
then possible to describe a sample of material if the numbers of particles, , of
each type are known or specified. The measurand concentration in the sample as a
whole is:
∑∑a
i =1 j =1
ij m ij n ij
as = Ns Nc
∑∑m
i =1 j =1
ij n ij
[Eqn 4.11]
Where Ns is the total number of size classes and Nc is the total number of
composition classes.
and the sample mass is:
Ns Nc
M s = ∑ ∑ mij nij
i =1 j =1
[Eqn 4.12]
In order to understand the statistics of sampling of a lot of material, it may be
imagined that a substantial mass of the material is collected and each particle in the
lot is given a number starting at 1 and going up to some very large figure N. To
‘sample’ the material, a set of S numbers is drawn entirely at random, from the
numbers 1 to N, and the particles whose numbers are drawn are the ones taken as
the ‘sample’.
By selecting the particles in this way, there is no bias in favour of particles of any one
type. In fact, this ‘ideal’ method of sampling can be considered as a standard against
which any other method of particle selection for sampling can be compared. This
method is the most equitable possible method of selecting particles for a sample,
since the numbers of any one type of particle taken into the sample do not depend on
the numbers of particles of any other type taken into the sample, and it is often
referred to as the ‘equiprobable’ sampling method in discussions about sampling.
It follows from this description of the sampling process that the number of particles of
any one type that arrive in a sample follow a Poisson probability density. A number
following a Poisson distribution is characterised simply by its expected or mean value.
The expected value need not be a whole number, but the number of particles must of
course be integers.
The important aspect of the Poisson density is that the variance of the distribution is
equal to the expected value.
1 Ns Nc
Var {a S } = ∑ ∑ ( aij − a L ) m ij2 nij
2
M s2 i =1 j =1
[Eqn 4.13]
where the nominal measurand concentration in the sampled material (the ‘lot’) is:
Ns Nc
∑∑a
i =1 j =1
ij mij nij
aL = Ns Nc
∑∑m
i =1 j =1
ij nij
[Eqn 4.14]
2
⎧a ⎫ 1 Ns Nc
⎛ aij − aL ⎞ 2
Var ⎨ S ⎬ = 2 ∑∑ ⎜ ⎟ mij nij [Eqn 4.15]
⎩ aL ⎭ M s i =1 j =1 ⎝ a L ⎠
This formula in terms of the particle numbers is not very convenient to use. It is
preferable to restate the formula in terms of variables that are more familiar. For the
nominal analysis of the material, the mass fraction in the I th size fraction is denoted
as Xi , and the mass fraction of material in the j th
composition class within the I th
size fraction as Yij . It is then a simple matter to convert Equation 4.15 to the form:
2
⎧aS ⎫ 1 Ns
⎛ aij − aL ⎞ Nc
Var ⎨ ⎬ = ∑XV i i ∑Yij ρij ⎜ ⎟
⎩ L ⎭IH
a MS i=1 j=1 ⎝ L ⎠a [Eqn 4.16]
This formula relates the relative variance (the square root of the left hand side is the
relative standard deviation, expressed as a fraction rather than a percentage) of the
measurand concentration between nominally identical samples of mass M s to the
size distribution and the composition distributions within each size fraction. At a
given state of comminution, the size and composition distributions completely
characterise the material particulate heterogeneity with respect to a particular
measurand.
⎧a ⎫ K
Var ⎨ S ⎬ = s
⎩ aL ⎭IH M S
2
Ns Nc
⎛ aij − aL ⎞
Ks = ∑ X iVi ∑Yij ρij ⎜
where [Eqn 4.17]
⎟
i =1 j =1 ⎝ aL ⎠
The relative variance due to intrinsic heterogeneity is also called the ‘fundamental
sampling error’ (or variance) by Gy (1979).
The sampling constant captures all the necessary information about the intrinsic
heterogeneity of the material to be sampled. Since the variance is a relative variance,
it is dimensionless and the sampling constant has units of mass. If the particle
volume is calculated in cm3, and the particle density has units of g cm-3, the sampling
constant will have units of g. To determine the sample mass that must be retained to
ensure that the relative standard deviation (RSD) between nominally identical
samples does not exceed a fraction r, then:
Ks
Ms = [Eqn 4.18]
r2
If r = 0.02 (2% relative), then:
Ks
Ms = = 2500 K s
0.0004
The sampling constant ‘sets the scale’ of sampling for a material. It is a quantitative
measure of the material intrinsic heterogeneity with respect to a particular measurand.
It is for that reason that the relative variance is given the subscript IH.
⎧a ⎫ ⎧a ⎫ ⎧a ⎫
var ⎨ s ⎬ = var ⎨ s ⎬ + var ⎨ s ⎬
⎩ a L ⎭ total ⎩ a L ⎭ DH ⎩ a L ⎭ IH [Eqn 4.19]
⎧a ⎫ ⎡ 1 1 ⎤
∆ var ⎨ s ⎬ = K s ⎢ − ⎥
⎩ a L ⎭ IH ⎣ M s1 M s ⎦ [Eqn 4.20]
This means that the new variance is:
If the sub-sample is now crushed or pulverised to a new top size D 2 from its
original top size D1 , the sampling constant for the material will be reduced. In
order to understand how the value will change, the expression for the
sampling constant must be examined in more detail. The intrinsic
heterogeneity term is separated out thus:
2
Nc
⎛ aij − aL ⎞
IH i = ∑ Yij ρ ij ⎜ ⎟
j =1 ⎝ aL ⎠ [Eqn 4.22]
This term is equivalent to a variance of the composition distribution over the
composition classes. In fact, if the density of the particles does not change
significantly from one composition class to the next (the phase carrying the
measurand does not differ in density from the remainder of the particle or is
present in low concentration) and the mass of material in each composition
class is constant, then:
Yρ Nc
∑ ( aij − aL )
2
IH i = i 2 i [Eqn 4.23]
aL j =1
NS
Ks = ∑ X iVi IHi
i =1 [Eqn 4.24]
It should be noted that the intrinsic heterogeneity (IH) of each size fraction is
multiplied by the volume of the particles in the size fraction. For many
materials, this means that the value of the sampling constant is controlled by
the IH of the coarsest particles in the material; while the fines may show a
substantially wider composition distribution, their influence on the value of the
sampling constant is suppressed by the particle volume factor.
If the shape of the size distribution remains more or less similar as the
material is comminuted (the cumulative size distribution is simply shifted
sideways on a log size plot), the sampling constant will in most cases (there
are important exceptions) be proportional to the cube of the top size of the
material. Reducing the top size by a factor of 2 reduces the value of the
sampling constant by a factor of 8, if the phases carrying the measurand are
not significantly liberated by the comminution. If the liberation does increase
with reduction in top size, then the sampling constant is reduced by a factor
less than 8.
When dealing with complex materials, where the measurand(s) are carried in
a number of mineralogical phases, it is very difficult to predict the real extent
of reduction in the sampling constant with reduction in top size. This is
especially true when the starting material is coarse and there is little liberation
of the phases carrying the measurand. If the material is crushed past some
critical size, which will depend on the nature of the material, the intrinsic
heterogeneity of the coarser size fractions may show a substantial increase,
possibly by as much as an order of magnitude. In such a case, the sampling
constant reduction with size would be seriously reduced.
2
Ns
⎛ aij − aL ⎞
Nc
Ks = ∑ X iVi ∑Yij ρij ⎜ ⎟ [Eqn 4.17]
i =1 j =1 ⎝ a L ⎠
for the calculation of the sampling constant presupposes that the composition
distribution with respect to the measurand for all the size fractions that have
an impact on the sampling constant value can be determined. In many cases
carrying out such an analysis is simply not a realistic goal; some other means
must be found for estimation of the value of Ks .
• Individual particle analysis, also known as the ‘50 (or more) piece
analysis’ method
• Use of scanning electron microscope data from particle sections
• Multiple sample analysis
These methods all have their limitations, which are discussed below.
These sieve fractions should be chosen so that they each carry about 10% by
mass of the size distribution and the actual mass fractions that they represent
should be calculated and recorded. These fractions do not have to coincide
with the fractions defined in the first sample, but it is convenient if they do.
Then, each fragment is numbered and 50 fragments from each size fraction
are randomly selected (by choosing the 100 numbers at random).
To apply the 50-piece method to the estimation of the sampling constant for a
chosen measurand, the histogram should show a spread (standard deviation)
that is significantly larger than the standard deviation associated with the
analysis. Furthermore, the spread of the analyses of the full set of size
fractions should not be substantially larger than the spread for the particles in
the top size fraction. If there is evidence in the finer fractions of unusual
liberation of phases carrying the measurand(s), one must proceed with more
caution (this issue is illustrated below).
2
Ns Nc
⎛ a ij − a L ⎞
K s = ∑ X i ∑ Y ij V i ρ ij ⎜⎜ ⎟⎟
i =1 j =1 ⎝ aL ⎠
2
Ns Nc
⎛ a ij − a L ⎞
ie , K s = ∑ X i ∑ Y ij m ij ⎜⎜ ⎟⎟
i =1 j =1 ⎝ aL ⎠
2
2 50
⎛ a ij − a L ⎞
∴ K s = ∑ X i ∑ Y ij m ij ⎜⎜ ⎟⎟
i =1 j =1 ⎝ aL ⎠
[Eqn 4.25]
Assuming that the 50 particles in each of the two top size fractions are a
sufficiently good representation of the material, the mass fraction of the jth
particle in the size fraction is:
mij
Yij = 50
∑m
j =1
ij
[Eqn 4.26]
so
AMIRA P754: Code Guidelines Release 3 20070219 51
Commercial in Confidence
2
⎛ aij − aL ⎞
50
2 ∑ m ⎜⎜ 2
ij
⎟⎟
Ks = ∑ X i
j =1 ⎝ L ⎠a
50 [Eqn 4.27]
i =1
∑mik k =1
Note that the analysis aL in the above formula is the average measurand
concentration determined from the size-by-size analysis of all the size
fractions and the a ij are the analyses of the individual particles.
The larger material has a grade that is essentially equal to the head grade of
the ore. Chromite grains dominate the size fractions near 350 microns and
these fractions are relatively low in PGMs (the PGM grade goes down to 0.5
g/t). The heterogeneity with respect to PGMs of the fractions above 350
microns is low, but for sizes below 100 microns, the heterogeneity increases
radically. Given the particle volume weighting, the overall heterogeneity of
the ore with respect to PGMs will have contributions from all but the finest
fractions. The heterogeneity with respect to chromite will be dominated by
the larger particles because their grade differs significantly from the head
grade. This ore provides an example of a material that must be carefully
analysed to discover its full heterogeneity with respect to all measurands of
interest.
25 50
20 40
Cr 2O 3 [%]
Pt [g/t]
15 30
Cr 2 O 3 head
grade
10 20
5 10
Pt head
grade
0 0
1 10 100 1000 10000 100000
Figure 4.2: The grade of particles in a run of mine sample of UG2 ore crushed to a nominal
top size of 3 mm, on a size-by-size basis
When the particles in the material are too small for individual analysis, it is
possible to take advantage of the automated electron microscope
instruments, QEM*SEM and the Mineral Liberation Analyser (MLA) that have
been purpose built to analyse sectioned particle mounts rapidly and
effectively. These instruments are specialised to scan sectioned and polished
mounts of particles of an ore; they have the facility to identify mineral phases
and to analyse by XRF the phases with respect to a given element. They
also have the facility to search the mounts for mineral grains of rare phases
such as gold, electrum and PGMs.
The concept of the method is directly related to the form of Equation 4.17,
which indicates that the variance of measurand concentration between
nominally identical sub-samples of mass M s is inversely proportional to the
sample mass.
This method must, however, be used with common sense. When the sample
mass is large, the measurand variance over the nominally identical samples
will indicate nothing but the analytical variance. The variance over the
samples will not increase until the mass is low enough to expose the material
heterogeneity. The heterogeneity with respect to different measurands may
well appear on different scales.
A correct sampling system can show a bias over a short period of time if the sampler
takes increments at fixed time intervals, and this interval is a multiple of the period of
a periodic variation in the mass flow or quality of the process stream. To eliminate
this effect, the sampler must be operated in a random stratified sampling mode,
preferably based on time intervals.
Stopped belt sampling using a correctly designed former, which is the recognised
reference sampling method for conveyer systems, can obviously be used. But there
is usually a significant cost associated with a stopped belt exercise and this
procedure is usually reserved for establishing the accuracy of alternative sampling
methods.
The sampling from stockpiles and bins for purposes of Metal Accounting must be
avoided because it is almost impossible to obtain a representative sample. The
recommended method is to evaluate the material as it is added to or removed from
the stockpile or bin.
The sampling of trucks, wagons, skips and drums must only be undertaken using
correctly designed auger samplers and in accordance with a Standard such as JIS M
8100.
Solutions, but not slurries, in pipes must be sampled after passing through an
efficient in-line mixer, and in drums they must be thoroughly stirred before sampling.
If there are suspended solids in the drums, the contents must be filtered and the
solution and solids sampled and analysed separately by accepted procedures.
Effluents and discharges through stacks normally only have a minor effect on Metal
Accounting, although this is not the case if accountability of sulphur in a smelter, for
example, is important. Accurate stack sampling requires specialised equipment and
expertise and for this reason, it is often outsourced. Effluents can be accurately
sampled using cross-stream or flexible hose samplers.
It is recommended that the contents of stockpiles and bins only be evaluated by the
accurate sampling of material as it is added to or removed from the stockpile or bin.
The precision of an auger sampler can also be established by carrying out a suitable
number (>30) of inter-leaved sampling experiments on the material being sampled.
Because solutions pumped through pipes can be thoroughly mixed using in-line
mixers, the normal requirement, of repeated sampling of all of the stream part of the
time, can be relaxed and a portion of the stream can be removed under controlled
conditions. There are a number of commercially available sampling systems that can
do this. However, there are examples, for instance in a precious metals refinery,
where the density of a process liquor can be so high that segregation of the liquor
occurs. In these cases it is impossible to take a dip sample and the rule of sampling
a process stream while it is in motion must be applied.
There are areas in Hydrometallurgical Plants where solutions must be sampled when
they are hot and/or under high pressure. In these circumstances, specialised
equipment is usually necessary and care should be taken to ensure the safety of the
plant operators. In addition, care must be taken to avoid the evaporation of water.
The effect of crystallisation when the solution cools to ambient temperature must also
be taken into account.
The two sample containers are collected at the end of the sampling period and they
are processed separately to arrive at a set of analytical results. By collecting a
substantial number of samples in this way, the average absolute difference between
the sample pairs, or the mean squared difference between the sample pairs, can be
used to estimate the total variance of sampling, plus sample preparation, plus
analysis. If it is an objective to break this variance down into its components,
additional work must be undertaken to establish the analytical variance (which should
be known from records of duplicate determinations in the laboratories) and the
sample preparation variance. The Standards are very clear as to how to undertake
inter-leaved sampling work.
If the mass retained at the end of a sampling period (e.g. a shift) is Ms , the sampling
variance due to the fundamental error is:
Ks
σ 2 = a2 [Eqn 4.28]
Ms
where a is the assay of the sample. Note that K s is usually calculated in units of
grams.
The second component of the sampling variance is due to the time variation of the
grade of the process stream during the sampling period. This component of variance
is proportional to the variance of the measurand concentration about the mean for
the sampling period and it is inversely proportional to an increasing function of the
number of increments taken from the process stream during the sampling period.
⎡ 1 1 ⎤
∆σ 2 = a2 K s ⎢ - ⎥ [Eqn 4.29]
⎣ Mss Ms ⎦
where Ks is the sampling constant for the material at its particular state of
comminution and a is the sample assay.
The sampling safety line plot, originally proposed by Gy, is based on the assumption
that the sampling constant is proportional to particle top size cubed. If this is not true,
the sampling safety line is, therefore, correspondingly misleading.
Every stage of sample preparation in which the ore is reduced in size, and
subsequently reduced in mass, causes a change in the sampling constant, and it
From the point of view of the variance of sampling, preparation and analysis, the
preparation stages must be treated in exactly the same way as the sampling stages.
When the material being sampled changes in its mineralogical make-up, or the
intensity of the process grade variations with respect to any measurand of interest
changes, the precision of sampling may need to be re-estimated. New tests of the
sampling precision, using inter-leaved sampling, may be the most effective way of
determining the change in precision. Given that the sampling systems have been
designed to permit inter-leaved sampling, which is highly recommended for metal
accounting samples, this is a simple procedure which can be carried out periodically
and documented on an on-going basis. Such a procedure is a sound component of a
quality assurance program.
Three methods can be used to establish the Sampling Constant and hence the
sampling protocol to achieve a desired level of sampling variance and they are
discussed in Section 4.1.
The bias of a sampling system can be determined only by comparison of the samples
from the installed sampling system with some reference method of sampling
(stopped belt for material flows on conveyors and some other method for other
situations).
Bias is not a statistical quantity and is, therefore, much more difficult to quantify than
sampling precision. A test to estimate bias must be designed to detect a certain level
of bias. As the magnitude of the bias to be detected becomes smaller, in relation to
the sampling precision (or standard deviation), the number of pairs of 'test' and
The best defence against bias in sampling systems is the mechanically correct
sampling of the process stream and the extraction of adequate numbers of
increments at each stage of sampling.
The extraction of the analytical aliquot from the mass of prepared sample must not
be ignored. This is almost universally done by use of a spatula, which is not a
mechanically correct method of sampling and it is prone to bias. Where the material
for analysis has refractory and easily ground phases and density differences between
mineralogical phases, there is a perfect opportunity for phase segregation at the very
point where the phases are best liberated. To reduce bias and variance to their
minima, miniature rotary sample dividers should be used to extract the analytical
aliquot from a mass of sample that has been prepared to a target mass without any
manual 'dipping' of sample increments. To accomplish this task, a 'variable split'
rotary sample divider is recommended.
In the Coal Industry, for example, the particle size of a plant product can be an
important measurement parameter. Particle size reduction, therefore, cannot be
carried out during a multi-stage sampling system on the portion of the sample to be
used for particle size measurement. Systems have been designed, however, where
a separate portion is taken specifically for particle size analysis.
5.1 INTRODUCTION
All operations in sample preparation and analysis must be carried out in order to
provide accurate data with a known and acceptable precision within a time frame
that is suitable for Metal Accounting purposes.
In some operations where a large sample is delivered to the laboratory (>10 kg) this
must be sub-sampled through a correctly designed sample division system, in an
enclosed space (particularly in conditions of low ambient humidity), to provide
material for further treatment within the laboratory.
The drying temperature, which must not exceed 105°C, must be such that the
physical and chemical properties of the sample are unaffected. In the Coal Industry,
for example, it is usual practice to air-dry samples at ambient temperatures.
Dust losses must be avoided because there have been numerous occasions where it
has been demonstrated that the fine material in a sample has a very different
Suitably sized jaw and roll crushers, pulverisers and “puck and ring” mills are
acceptable for sample particle size reduction.
Sample reduction ahead of sample division must always be done to a particle size
that is known to maintain the sample preparation variance at the desired level for the
sample concerned, as determined by an appropriate test and analysis of sample
variance.
Paper or sheet rolling, which is sometimes used in sample preparation facilities, must
be avoided, because of the potential for dust loss, increased segregation and
inadvertent sample loss if the mixing is carried out too vigorously.
The mixing tray procedure specified in JIS M 8100 and by Merks (1985) is
acceptable, where sealed mechanical systems are not available. Because this is a
manual system, there is an upper limit to the sample mass that can be handled in
practice.
Properly designed mechanical sample division systems, such as rotary rifflers, are
preferred provided that the requirements of good sampling practice such as speed of
rotation < 0,6ms-1, radial cutter gaps and a minimum of 30 increments are observed.
In addition these systems should provide completely proportional samples and QC
procedures must be in place to confirm this.
The Incremental Division Procedure specified in JIS M 8100 and Merks (1985) is
acceptable.
Stationary rifflers must be avoided, because, even when the riffler is equipped with a
feed bin to ensure the even distribution of material to both sides of the divider, the
probability of unsatisfactory operation is too high.
After prolonged storage, solid samples must be mixed in a sealed system prior to
extracting an analytical portion.
5.3 ANALYSIS
It is likely that most of the procedures in the following sections will be implemented as
part of an accreditation exercise.
Because of the known losses during the lead collection fire assay of the PGMs,
which are reasonably reproducible if the operating conditions are kept constant, a fire
assay correction factor is often used to correct for these losses. These corrected
values must not be used for Primary Metal Accounting, which must be based on a
technique such as nickel sulphide collection, but they may be used for Secondary
5.3.10 QC Procedures
QC procedures must be established, and strictly adhered to, in relation to the
allowable differences between the accepted and actual values of the Reference
Materials and the differences between replicates in the twin stream (independent
parallel) sets. The use of Shewhart and CUSUM control graphs are recommended
for the monitoring and rapid identification of possible analytical problems. Action to
be taken in these situations must be formalised in the QC procedures.
5.3.11 QA Procedures
Regular multi-laboratory interchanges of samples (Round Robin exercises or
Proficiency Testing) must be conducted with laboratories whose ability to conduct
such analysis has been proven by means such as laboratory accreditation to ISO
17025. The exercise should preferably include laboratories that are independent of
the organisation concerned.
The statistical analysis of accumulated analytical results for reference materials and
for duplicate determinations is a very valuable tool for the on-going improvement of
analytical procedures and it is vital to the maintenance of targets for accuracy and
precision within the analytical systems. These analyses also provide an index of the
efficiency of the laboratories; an increase in the fraction of the samples that require
repetition because they have fallen outside the tolerance for agreement between
duplicates not only indicates that the precision of the analysis is decreasing, but it
also increases the load on the laboratory, limiting its throughput of fresh samples.
It is extremely important that all assay results are collected into a database. The
assay duplicates that failed to meet the tolerance must be included in the database.
If they are not included, the database becomes 'censored' and is therefore very
difficult to use in statistical analysis, since distributions of assay results are then
truncated.
The assay censoring issue also arises in the use of umpire laboratories. External
laboratories may be asked to provide a database that indicates their ability to provide
precise results. The results in this database must not be censored, since this would
make it impossible to understand the risks associated with use of that laboratory.
The frequency with which stock-takes are carried out is a management decision and
it should relate to the value of the plant throughput and the in-process inventories
relative to the expected profit at year-end and the cost of executing the stock-takes,
but it must be carried out on at least an annual basis. Stock-takes can be performed
more frequently, for plant control, to a lesser degree of accuracy than that required
for financial reporting. The latter stock-takes must be performed in conjunction with
the company’s auditors.
The cost of executing the stock-takes increases with the demand for the accuracy of
the result. The final decision regarding stock-take frequency and accuracy will be
determined by the level of financial risk associated with a particular strategy. The
costs incurred should be offset against the benefit achieved by the more timely
availability of accurate information of production performance and of stock shortages
or excesses. The overall confidence level in the result of the stock-take must be
established from the uncertainties in the component measurements and reported
with the results.
Therefore, the policy of eliminating stockpiles on a regular basis must be built into the
general processing strategy.
In cases in which the in-process inventory has a value that is large compared to the
value of the net input for the accounting period, the accounting strategy must involve
7.1 INTRODUCTION
The metals/commodities, alloys or concentrates produced by a metallurgical
processing operation are the source of income which both sustains the operation
financially and justifies the initial capital investment.
Those operations which mine their own deposits as feed stock commonly have
royalty obligations relating to the depletion of the resource and, in some cases, may
attract environmental penalties if levels of, for example, acid generating material or
heavy metals in water, exceed agreed limits.
The metal balance is critical for corporate governance, for monitoring, checking and
improving performance and for identifying and quantifying losses and emissions. A
satisfactory metal balance provides a convincing demonstration that:
"What comes into the operation, less any internal accumulation, plus any draw down
from internal stock, is sufficiently similar, i.e. within acceptable limits in mass and
analysed components, to what goes out as product(s) and reject." (Check in-Check
out)
This is the fundamental purpose of the metal balancing and reconciliation process.
The calculation process will include calculations which verify that the data is as
correct as expected and it will help to identify data which results from intermittent
problems (systematic errors) as well as to identify biases which accumulate within
the process flowsheet. The calculation process should provide the best possible
estimates of product metal recovery, minor element and/or elemental distributions,
ore source grades and tonnages and the accumulation of mass and metal values
within the circuit. It should also identify significant biases in the data over time. In
principle, smoothed data can be used, provided it is used on a consistent and
transparent basis.
Using the Check in-Check out System, all streams into and out of the process or
plant, for which the balance is being performed, are measured, sampled and
analysed. In terms of this Code, this is the standard procedure to be adopted for
Metal Accounting in all operations and all deviations from this approach must be
justified in an Exception Report and approved by company Management.
Applying the Check in-Check out system will normally generate an imbalance, due to
measurement, sampling and analytical errors, or unknown losses. This imbalance is
treated as an unaccounted gain or loss.
Unaccounted = Unknown Losses ± Bias, Random and other Errors [Eqn 7.2]
For each of the accounting streams the precision of the dry mass, sampling and
analytical measurements is calculated and the precision of the overall balance is then
derived. If the imbalance or unaccounted quantity is greater than the limits defined
by this precision, or if the limits of the precision so calculated are excessive, then this
should be regarded as material and the result of an unsatisfactory measurement or
unknown loss.
The manner in which this unaccounted gain or loss is managed and investigated
must be based on procedures laid down in terms of the Code.
Custody transfers of material from one operation to another, (e.g. from a mine to a
concentrator, from a concentrator to a smelter, etc.) follow the generally accepted
practice of utilising the receiver’s mass, sample and analysis. Where these transfers
occur within the same company, the Home and Away System (double sampling and
analysis), whereby both the receiver and the dispatching operation independently
weigh, sample, and analyse the material being transferred, ensures that gross
accounting errors do not occur. The dispatching operation will record both the
dispatched mass and the metal contents and those determined by the receiver. The
receiver’s quantities will apply and any discrepancy is recorded as an unaccounted
loss or gain. Where serious discrepancies between the quantities reported by the
two operations occur, they must be investigated immediately and corrective action
taken.
The last few decades have seen substantial improvement in all areas of
measurement and computational power has ceased to be a constraint. However, the
traditional methods described below are still in common use.
1 Accountability (%) = (sum of all outputs + stock and inventory change) x 100
(sum of all inputs)
[Eqn 7.6]
This, too, must be defined in terms of boundaries and it provides a check as to
whether over time, inputs, outputs and stock changes reconcile. It must be
accompanied by a clear indication of its confidence limits, as calculated from the
constituent measurements.
An alternative form of the above expression for accountability which is also used is:
Any material received by a plant from a previous processing operation, but which has
not yet entered a process stream, or any material produced by that plant to dispatch
standard quality, but which has not yet been dispatched, is generally defined as
plant stock. Any material which has entered the process stream, but which has not
yet been produced as a product for treatment in the next process, is generally
defined as process inventory.
The day-to-day data collected from a metallurgical plant is rarely consistent and it will
usually contain redundant information. In general, any two methods of calculation
will yield different results and, therefore, many factors influence the selection of a
suitable metal accounting strategy. These include the degree to which it is practical
(and economically justified) to weigh and sample accurately the feed to, and the
products from, a processing operation to achieve Check in-Check out, and the
availability of redundant data. An important additional factor is the degree to which
the metallurgical accounting system supports process performance evaluation and
optimisation.
There are some general guidelines. Mining operations and deposit models are
usually relatively sparse in terms of available redundant data. By way of contrast,
and recognising the difficulty in accurate mass measurement and sampling of ROM
ore, a modern mineral processing operation usually enjoys a wealth of redundant
When the material reaches the pyrometallurgical plant, the situation may well reverse
with data becoming scarce and being limited by risk of access to hot metal and by
toxic off-gases.
Table 7.2 below indicates some strategies commonly used for various types of
operations in the metallurgical industry. However, each situation is unique and
should be carefully assessed. While the dominant purpose of metal accounting is to
manage Custody Transfer, it is recommended that the same process be used to
monitor process efficiency and intermediate process inventory.
It is essential to ensure that the Custody Transfer measures are within the expected
limits. For concentrators, if limited data is available, the Two-Product Formula
(Section 7.5.1.1) or the Three-Product Formula (Section 7.5.1.2), as appropriate,
can be used to provide some independent indication of the validity of the selected
measurements. If the measurement accuracies are known and there is sufficient
redundant data, the mass balancing reconciliation measures outlined in 7.4 can
provide a robust means of checking on the measurements used for the balance. If
two measurement methods agree to within the desired statistical criteria, the Custody
Transfer measurements can be accepted as being "correct", within the limitations of
the measuring system. If the measurement methods do not agree to within the
specified criteria, the mass balance can often provide detailed guidance to identify
the source of the problem.
All plant measurements contain some usable information and effective corporate
governance is greatly enhanced by better knowledge of which measurements are the
most valid.
To make statistical statements about the balance and the results that are found, it is
necessary to adopt a specific approach to the solution of the mass balance problem.
A simple approach to the problem in which both mass flow and assays are involved
in a balance around a plant or around units within the plant is given in the paper by
Lyman (Appendix E3). In this paper, the data adjustment problem is formulated as a
mass flow adjustment problem, where the mass flow is the product of a measured
flow rate and a measured assay. The assay adjustments can be back-calculated
from the adjusted flows, but this detail is omitted for brevity.
The changes which are applied to each measured value to achieve this balance are
usually called Adjustments. When adjustments to an analysis or a mass flow are
divided by the estimated standard deviation of the corresponding measurement, they
can be described as 'weighted' adjustments. If all the measurements which are
employed in the metallurgical balance are relatively accurate, the expected value of
any one adjustment is very close to zero. The distribution of the adjustments will also
tend to be normal, but all adjustments calculated from the balance will be statistically
dependent or correlated. One might also think that the weighted adjustments should
have unit variance because they have been divided by a standard deviation, but this
AMIRA P754: Code Guidelines Release 3 20070219 79
Commercial in Confidence
is not correct because the balance equations introduce constraints between the
values of the adjustments.
Any balance problem will have a certain number of degrees of freedom that depend
on the interconnection of the process streams and on what has been measured
(mass flows and number of species involved in the balance and whether or not total
mass flows are balanced as well as the species flows). The total sum of squares for
the adjustment problem is approximately distributed as a chi-squared random
variable with the number of degrees of freedom that the particular adjustment
problem possesses. As such, the expected value of the total sum of squares is
approximately equal to the number of degrees of freedom.
The true expected value of the sum of the weighted adjustments may vary from zero
because, for example, the simultaneous adjustment of mass flows and assays in flow
sheet is a non-linear problem. However, in many cases this expectation can be
taken to be zero.
To test whether or not the weighted adjustments being applied to the data are too big,
advantage may be taken of the fact that the weighted sum of squares of adjustments
is approximately chi-squared. Then, a goodness of fit test can be applied to the
weighted sum of squares value. If it exceeds the chi-squared value at, say, 95%
confidence for the appropriate number of degrees of freedom, there is an unresolved
problem. Either the estimates of the data precision are incorrect or some
combination of measurements has a precision lower than stated. One or more
values used in the balance may be biased. The cause for the excessive weighted
sum of squares should be investigated. Under such circumstances, an Exception
Report must be generated.
It can be difficult to pin down which of the measurements causes the increase in the
chi-squared value. The best approach is to conduct the balance successively with
more and more analyses included in the balance, or with different combinations of
analyses. This strategy may show up the problem quite quickly. The important
statistical results with a case study are provided in Appendix E.3. An alternative
approach to balancing data is provided in Appendix E.1
The examples provided do not consider internal stockpiles (or bin inventories for
example). However, they can be included in the balance by creating an input stream
for draw-down from a stockpile and an output for build-up of a stockpile. It is critical
that the accuracy with which these inputs and outputs can be estimated is
determined.
In many cases, the accuracy on the draw-down or build-up will be low and
adjustments to these mass flows will represent the majority of the overall adjustments
applied to the data. The overall confidence limits attached to the period balance will
be correspondingly wide with a concomitant increase in the risk associated with the
balance. A high risk level may necessitate an improvement in the way these stocks
are handled.
Each change to the raw data must be recorded and justified, and then authorized by
the responsible senior management appointee. An audit trail must be available to
trace the path from the initial balance with its associated recovery and accountability,
7.4.2 Bias
The problem of bias is important and may not become apparent until many balances
have been carried out. The objective is to identify and to eliminate bias, not to adjust
data, and thus conceal the bias. For a given balance structure, it is possible to
calculate the variance of the arithmetic sum of the weighted adjustments for each
analysis type (this calculation is not detailed in E.3, but it can be derived from the
results there; it will be applicable to that type of balance only). With this estimate of
the variance for the analysis type, it is possible to test the significance of the sum of
the adjustments.
Appendix E.3 also provides the covariance matrix of the adjustments and the
diagonal elements of that matrix provide the variance of the adjustments. These two
can be used to test the magnitude of individual adjustments. Note that the
adjustments should be tested against their variance, not the variance of the original
measurement.
The comments about detection of a bias (1.7.2) which is small relative to the
accuracy of measurement for a single accounting period are still valid. The relevant
Cusums can be of adjustments (or of weighted adjustments) for each measurement.
A graph, which shows measured and adjusted data plotted together against
successive accounting periods, will often show a bias very clearly. That is, instead of
crossing on a reasonably frequent basis, one line will lie above or below the other for
many accounting periods.
There are many methods used to account for a plant’s production and most
concentrators produce a metallurgical balance showing the performance of each
shift. These balances are accumulated over a longer period (eg: daily, weekly,
monthly, annually) to show the performance over the period in question. The actual
recovery over the accounting period is declared and any discrepancy between that
and the accountability is regarded as an unaccounted loss or gain. Physical
losses do occur on any plant and every effort should be made to keep these as low
as possible at all times. Care must be taken to ensure that the balance is not
compromised by recycle streams and the return of plant spillage to the circuit
especially if these are of higher or lower grade than the main stream.
Therefore: F=C+T
And: Ff = Cc + Tt
Or: Ff = Cc + (F – C)t
F=P+Z+T
If one of these weights is known, the other three can be calculated, provided
that two component balances are available, i.e.
From these calculations, the concentration ratio and metal recoveries can be
calculated in a similar manner to the two-product formula above.
The three-product formula has its limitations. The equations assume steady-
state conditions and the fundamental assumption that input is equal to output.
This is true over long periods but it may not be true over shift periods. The
three-product formula is subject to appreciable errors when the components
are poorly separated and it is preferable to obtain accurate mass flow
measurements of products rather than to infer them in this way. The formulas
of three or more products are more sensitive to small errors both in data and
calculation, especially when one of the products is of relatively small weight
and low concentration, and this can lead to serious balancing errors (Taggart
1945).
Although sampling and analysing the stockpiles and other areas of accumulation can
assess the metal content of the inventory, it may be more accurate to rely on the
weighted concentrate analysis produced from daily or shift plant balances.
Materials balances are generally completed on an air-dried basis, which tracks ROM
coal as it leaves the pithead, to the point of delivery to the final customer as a final
product. Balances to account for the reject materials can be completed but this is
done to varying levels of detail, depending on the value it is considered to add to the
accounting process.
Feed
The feed material depends on the purpose of the pyrometallurgical plant. This may
be a reducing or an oxidising roast, smelting of sulphidic metal containing materials
to remove gangue materials and iron, melting to reduce oxides (ilmenite smelting),
the oxidation of unwanted species in the metal phase (steel making), or melting to
separate the metal phase from the oxide phase. The gas phases may be air/oxygen
or reducing such as carbon monoxide depending on the purpose of the smelter.
Feed is often mineral concentrate or ore and reducing agent (anthracite/coal) and/or
fluxes from an adjacent plant or from another supplier and it may be from the same
company or it may be for toll treatment. However, many other materials can also be
treated in a furnace owing to the nature of the process. These include recycle
streams from elsewhere in the process and scrap. All these materials may contain
moisture and it is important that appropriate samples are taken to establish the dry
weight.
Pneumatic conveying is sometimes used for transport of the dry material to the
furnaces. In this case the material is usually weighed in batches in weigh bins at the
feed end or in bins on top of the furnace. The weight and analysis of the feed
material should be compared with that from the supplying concentrator or mine. That
weight may have been calculated from the two-product formula and/or measured on
static scales such as weigh bins or weigh tanks. This usually means measuring or
estimating the stock held in thickeners or stockpiles, which, as highlighted previously,
is difficult and inaccurate. However, the comparison can highlight problem areas and
arrangements can be made to reduce or eliminate the stock periodically to provide an
absolute stock value.
Toll treatment materials are processed under contracts, which specify the mass
measurement and sampling methodology for custody transfer and which often
include weighing and sampling in dedicated facilities , after crushing or slurrying, in
order to obtain representative samples. Calculated contents are compared with those
of the supplier and they are subject to agreed maximum tolerances (splitting limits
and umpire analyses). Payment terms, toll refining charges and recoveries are
based on the agreed calculated weight and content. Normally, the toll treatment
company will guarantee a specific recovery within a certain time frame and it will
deliver to the supplier the agreed quantity of product or market it as part of its own
production. Thus any gain or loss resulting from the actual recovery achieved can
reflect on the recovery of in-house materials.
Products
Products are usually furnace or converter matte, metals, metal alloys, and metallic
anodes and they constitute one or more metal phases), an oxide phase and, in some
cases, a gas phase. These are normally sold or sent to a refinery for further
processing. This further treatment may be in-house or on a toll refining basis. In all
cases the product, which may be crushed or granulated, or in the form of ingots or
lumps, must be weighed by the supplier and the consumer on static scales, weigh
bins or on conveyor weigh scales, and sampled/analysed by both parties. The
calculated contents must then be compared and those for metal accounting use must
be agreed according to contractual terms or in-house procedures. This weight is
usually the most accurate mass measurement in a smelter. However, the material
often contains some moisture and this can introduce a significant source of error
depending on the mode of storage, loading/unloading, transport and sampling and
can change during these operations.
Slag
Furnace Slag is usually a discard from smelting operations, although often it is further
processed to recover contained or entrained metals before being discarded.
Normally, losses of valuable constituents are a relatively small percentage of the feed.
Sampling and weighing of furnace slag depends on the manner in which the slag is
handled after it is tapped from the furnace. Samples can be taken during the tapping
operation from the molten material for plant control, but these are not recommended
for metal accounting use. If the slag is granulated or crushed before further
treatment and disposal, sampling and weighing must take place during the conveying
or trucking to treatment/disposal. In some cases, such as in the smelting of ilmenite,
slag can be a product for sale and it should be weighed and sampled as for other
smelter products.
Dusts / Gases
Dusts are recovered from the dust removal equipment treating the off-gases from the
furnaces, converters or other process steps and the dust extraction equipment on the
transfer points. This dust is normally returned to the process but it is not usually
weighed and it is treated as an internal recycle. Some dust is lost to the atmosphere.
This normally contains only a small proportion of the valuable constituents in the
smelter feed and measurement of the mass and sampling on a periodic basis may be
satisfactory, in these cases, to provide an estimate of the losses.
Off-gases may be also recovered to utilise their waste heat or to recover pollutants or
economic constituents.
Sulphur Fixation
In the case of sulphide smelting, a proportion of the sulphur is recovered as sulphuric
acid, or as sulphur fixed in some suitable compound or, in some cases, it passes
directly to the atmosphere as sulphur oxides without any treatment. Sulphur is
usually analysed in the off-gases and the final sulphur product is weighed and sold if
it is sulphuric acid or some other tradeable product.
Smelters operate under permits that limit the mass emission of sulphur to the
atmosphere and they are subject to environmental regulations that specify
permissible ground level concentrations of various pollutants such as sulphur dioxide.
Thus a balance must be performed to calculate recovery and losses of sulphur.
Metal Accounting
Recovery and accountability for smelters and pyrometallurgical plants are
calculated utilising the same formulae as shown in Section 7.3. For Check in-Check
out all input and output streams around the plant must be measured and sampled
leading to the calculation of material balance closure errors for various species.
These balance closure errors reflect the material that is unaccounted. However,
dust and gas losses are normally subject to a high level of uncertainty and slag
quantities are often not accurately measured, thus Check in-Check out cannot be
carried out for species affected by such losses and lack of measurement. Slags are
not homogeneous, they may have metal or alloy inclusions, and may contain varying
amounts of moisture, especially in the case of water-granulated slags. However, the
metal content in gases, dusts and slags is usually only a small proportion of the feed
and, as such, the effect of omissions on the balance should not be significant. The
amount of effort put into improving the accuracy of measurement of dust and slag
losses is dependent on their magnitude and value, risk potential and the
requirements for environmental reporting. However, it is recommended that furnace
slag be weighed before disposal or recycling rather than using truck factors, which
are notoriously inaccurate, as this will provide more accurate information for closure
of the balance.
Suitable tie components such as iron or silica, which report solely to one stream, or
whose distribution is known, which are abundant and which can be readily analysed
to a high precision, can also be used to calculate multiple balances, to provide
redundant data and to facilitate the solving of anomalies, especially where Check in-
Check out is not performed. In some cases indirect measurements such as
temperature can assist in providing additional data.
Stocks are particularly difficult to measure accurately and they should therefore be
kept to a minimum and, ideally, they should periodically be run down to zero since
month-to-month recoveries can give a distorted picture of the metal balance.
One practical way of calculating recovery over a period is to obtain a balance closure
for a minimum of three major elements; this will provide redundant data and thus
enable problem areas to be identified.
Care must be taken to ensure that the system design for the mass measurement and
sampling, allows for recycle streams and for the intermittent return of spillage.
Feed
Hydrometallurgical plants typically treat a solid feed material, such as ore, flotation
concentrate, matte or other intermediate materials, including metal alloys, scrap
metal, metallic concentrates and a variety of recycled materials in solid or liquid form,
all of which may contain some moisture. The materials may be products from an
earlier stage of processing from the same company or they may be purchased for toll
treatment from outside companies. The comments in Section 7.5.4 on Smelters are
equally applicable for measuring, handling and treating these materials.
Products
The feed materials are processed to give product metals, salts, oxides or other metal
containing materials, as well as process effluent streams, and, often, intermediate
products for treatment at some other facility.
The solid products can normally be weighed and sampled to a high degree of
precision by using static weighing methods during the final packaging into drums,
bags or bales and again during dispatch, giving check weights. Terms of sale will
usually specify the acceptable precision and that mass measurement instruments
must be certified for custody transfer. Solutions must likewise be weighed in the
drums or tankers on certified equipment. In some cases, products may be solid
residues sold for further processing. These may also contain moisture or solution
and they may be in the form of slurry or filter cake.
Metal Accounting
As process streams can be made up of either, or both, solids and liquids with both
containing significant quantities of metals, the metal accounting system for such
plants must make adequate provision for the sampling and mass measurement of all
components of each feed and output stream and it must allow for the measurement
and assessment of continuous recycle streams and of process spillage which may
occur on a periodic basis.
Depending on the specific operation, the feed material is usually stockpiled ahead of
the process to ensure continuous operation. Estimation of the metal content of the
stockpile, however, can be a significant factor in overall process metal accounting.
Similarly, hydrometallurgical plants usually have a significant process inventory, in
the form of metals in solutions and in solids, including the inventory in solvent
extraction and electrowinning circuits. Estimation of this inventory is complicated by
the changing metal contents of the solid and liquid phases as they progress through
the plant and it has to incorporate tank level measurements, slurry pulp densities and
solution and solids analyses at each stage of the process. The presence of a large
stockpile of feed (or any other semi-processed intermediate) material complicates
this estimation.
In plants treating high value metals, the process inventory estimate should be verified
at prescribed intervals by physical stock-takes, using the procedures such as the
‘bubble’ method, to reduce the probability of major stock write-offs caused by
undetected errors in the routine accounting system.
Although the principles which follow can be applied in theory to both dump and heap
leaching, in practice they would be impossible or not cost effective to achieve for
many dump leaches where the metal content of the dump is not known with any
degree of accuracy, and which cannot be adequately sampled, where the material is
of low grade or value and high mass and thus Check in-Check out cannot be
achieved or is not economically justified. Therefore, the recommendations in the
following sections will refer to heap leaching only.
In large heap leaching operations, there are often many piles being operated in
parallel but at different stages of the leaching cycle. The large inventory also means
that the effect of input or other changes is subject to long delays and it is not readily
apparent.
Process inventory in heap leach piles or pads has the following components:
The main concerns are the inventory build-up of metal in the leach solution and the
determination and prediction of the recoverable metal in the feed material. The type
of heap construction, as discussed further below, affects the determination and
significance of these inventories.
Samples of feed, crushed to the size deposited on the heaps, should also be
taken for leach tests on weekly or monthly composites, to predict the
recoverable metal and reaction kinetics for the material being deposited and
for comparison with the original feasibility test work.
The solution metal balance for each heap or module is the first indication of
the heap performance. Irrigation solution and drained pregnant solution from
each heap must be accurately measured by suitable flow meters, which can
be calibrated according to recognised techniques and sampled for analysis,
according to this Code. Reconciliation of these flows is important in order to
provide a calculation of the solution inventory in the leach piles, the metal
extraction from all the heaps or modules under irrigation, and total plant
production. Performing this reconciliation requires estimates to be made of
the amount of rainfall, evaporation and the solution content of each
abandoned heap. Rainfall and evaporation should be based on actual site
measurements over the accounting period. The solution content of operating,
abandoned and resting heaps should be periodically established by test work,
although standard figures based on experience are often used. It should be
noted that the metal content in the solution in the heaps is normally the
largest component of the total solution inventory and it can be equivalent to
several months production, thereby emphasising the importance of this figure
to the metal balance.
Metal Accounting
An accurate solution metal balance must be obtained, so that the total metal
recovered from the material abandoned can be calculated. The importance of
an accurate measurement of the metal placed on the heap was highlighted
above. However, the accounting is complicated further by the fact that not all
the metal is recoverable, and this non-recoverable metal can form a
significant percentage of the total, leading to a very large process inventory.
It is common practice, therefore, to use the recoverable metal placed on the
heap, rather than total metal, for process control and for mass balance and
performance calculations. If recoverable metal is also being used for
accounting, then this must be approved by means of an Exception Report.
This recoverable metal should be determined, for the specific material being
placed, by suitable laboratory tests such as rolling bottle or shake tests, or
pilot scale column tests, on representative samples. These values are then
adjusted by factors, based on experience in older heaps, to reflect the
predicted ideal performance of an operating heap. The procedures and
methodology for performing these tests and calculations should be rigorously
designed, documented and monitored. Thus, the balance will reconcile the
output metal in solution with the inputted recoverable metal to give the
recoverable metal still in the heap in both solid and solution forms.
The metal that was not recovered must be removed from the plant inventory
when a heap is abandoned. If variations in feed grade are small, then, as
many heaps are operated at one time, the process inventory will tend to
stabilise as heaps are established and abandoned. Sampling of heaps is
subject to a high level of inaccuracy and due cognisance of this must be
taken in the use of the information generated for stock evaluation and in
predicting future performance.
The metal balancing from the output leach solution through the purifying and
refining steps to metal product is similar to that for hydrometallurgical
processes shown in Section 7.5.5. General practice is to calculate the
recovery as the metal finally produced, after the purifying/ refining process, as
a percentage of the metal placed on the heaps.
7.5.7.1 Introduction
The throughput scale of pilot plant testing covers a large range from Bench
Scale Plants through Pilot Plants to Demonstration Plants.
Pilot plant work is undertaken for a number of different objectives such as:
• the pilot plant feed material is representative of the ore body or the
material to be tested
• the sampling protocol must ensure that there is continuous traceability
of the bulk sample from source to pilot plant
• the mass is measured and monitored according to the recognised
procedures as required by this Code.
7.5.7.3 Design
Metal Accounting for pilot plants must embody all the principles covered in
other sections of this Code. However, in addition, the following factors must
also be considered when designing the plant and the system for deriving the
metal balance:
• The design of the pilot plant must take into account all safety, health and
environmental factors as well as company and regulatory requirements.
Hazops analysis should be carried out
7.5.7.4 Operation
The design and operation of the plant must allow for suitable equipment and
procedures, in line with the Code, for mass measurement, sampling and
analysis that will ensure that meaningful results to a known accuracy can be
obtained for the metal balances and to facilitate the decision process.
Operating and sampling manuals are required.
In the same way, the measured mass of metal in the ore delivered to the shaft from
the mining operations is compared with the estimated metal content in the ore mined,
as determined by sampling and surveying, to give the shaft (or mine section) call
factor, also expressed as a percentage. The combined plant and shaft call factors
give the mine call factor for the operation. Typically, these call factors are
calculated monthly and they form a useful means of monitoring discrepancies
between the expected and actual mine output.
For metal balance purposes measurement of streams from operating plants only is
relevant. However, measurement and quantification of any dissolution of heavy
metals and other contaminants from surface dump material, such as, acid mine
(rock) drainage and percolation of critical compounds into aquifers or watercourses,
or gases emitted and dust blown from slag and tailings dumps, is also necessary for
corporate governance. An overall understanding of the balance of these components
is desirable to provide information in order to take corrective action and make
improvements as required in ISO 14001.
Measurements tend to be more difficult than main process streams and the level of
pollutants and methods of measurement are defined by international agencies as
mentioned above and in relevant national Water Quality Acts and Regulations.
Companies should adhere to these guidelines and/or approved worldwide standards.
Further discussion of the methodology is outside the present scope of the Guidelines.
• provide a metal balance of the desired constituents over the reporting period
(this would normally be monthly, with quarterly or six monthly reports for the
audited company financial reports)
• provide this information promptly to Plant Management and to the Financial
Accountants
• highlight inconsistencies and problem areas in the reported data
• report the level of error in all the reported outputs and their comparison with
target levels
As a secondary objective, the reports should be capable of being used to:
The integrity of home-grown spreadsheets used for metal accounting, which can be
subject to modification by the metal accounting personnel to suit specific reporting
needs, can be a serious concern. The use of such spreadsheets for metal
accounting purposes must be avoided. Various commercial metal accounting
software packages are available and these should be used in preference. Where
home-grown spreadsheets are used, clear rules for setting up such spreadsheets
must be set down, stating that the spreadsheet must be contained in a single file,
with no multiple linked files allowed, and with no incorporation of macros or any other
sub-form of programming. This is to ensure that the system used is uncomplicated
and that it can be serviced and maintained in the absence of the original designer.
The rules must also state that the system must receive original input, which is to be
independently verified, and that it must not receive input via links to any other file or
system, other than data logging systems; this is to ensure that the accounting system
remains uncorrupted by spurious formulae, which may be introduced from other
sources. The rules must further stipulate that no accounting data may be altered
unless the alteration is fully explained and motivated in a written report, which must
also be approved at an appropriate level of authority, defined by the operation
concerned, and retained for record purposes.
• Responsibility for final approval of the design of the Metal Accounting system
• Responsibility for auditing the system once it has been installed, to assess
compliance with the design specifications
• Responsibility for periodic reviews and audits of the system, at pre-
determined intervals, to ensure continued compliance and, as required, to
investigate problems which may occur
• Submission of reports to management in regard to the compliance of the
system and monitoring of the action plans produced periodically to remedy
faults
• Independence from the operation concerned.
The Competent Person should not be required for routine metal accounting, or for the
transfer of materials between organisations, but the Competent Person’s services
should be used to ensure that a complete metal accounting system, or a particular
aspect of it, is working as well as can be expected. Often, in metal accounting, one
is in the hands of a few select people and problems occur, which could be material.
It is recommended that an appropriately experienced person is involved to assist in
solving these problems when they do occur. If the solution results in a change in the
metal accounting system, it is essential the change be reviewed and approved by the
Competent Person.
A Competent Person, for the purposes of this Code, must be a member of the
relevant local, professional registration authority or any other statutory local or
international body that is recognised by the relevant Code administrators. If a
complaint is made in respect of the professional work of a Competent Person, the
complaint will be referred to the recognised professional body with which the
Competent Person is registered.
A Competent Person should have a minimum of five years experience relevant to the
type of metal or mineral under consideration and to the type of operation involved.
The key word ‘relevant’ means that it is not always necessary for a person to have
five years experience in each and every type of metal, mineral or metallurgical
operation, in order to act as a Competent Person. For example, a person with twenty
years experience in metal accounting in a variety of base metals extraction and
recovery operations may not require five years specific experience in a precious
metals operation in order to act as a Competent Person. Experience in the other
In addition to experience in the type of metal or mineral and the operations involved,
a Competent Person assessing metal accounting must have sufficient knowledge of
sampling and analysis techniques, relevant to the metal or mineral under
consideration, in order to be aware of problems that could affect the reliability of the
data. Some appreciation of the mining operation and mineralization applicable to that
deposit type would also be important.
As a general guide, people being called upon to sign as Competent Persons must be
clearly satisfied in their own minds that they can face their peers and demonstrate
competence in the commodity, type of operation and situation under consideration.
The specification of the metal accounting system should comply with this Code and it
should include a clear set of objectives, including:
• agree the performance indicators required for the Metal Accounting System
• decide what data needs to be collected to measure performance against each
indicator
• look at the possible sources of data related to each indicator
• decide who has responsibility for collecting what data
• establish a clear timetable for data collection, so that everyone knows what
they should do and when they should do it
• standardise data collection so that everyone gathers and reports data in the
same way
• agree the appropriate metal balancing procedures, protocols and algorithms
• train and support the data collectors so that they are competent and confident
• check or verify the data that is collected
• set up appropriate and timely communication or reporting systems
• determine the accuracy limits of the data
• analyse the data to support self-assessment and continuous improvement
• evaluate performance using the data
Several commercial metal accounting software packages for the compilation of metal
accounting data, using a central database, are available. These packages are
usually designed to be user-friendly, requiring minimal training of metal accounting
personnel, and they are considered preferable to the use of spreadsheets developed
in-house, as these can easily be corrupted and they are seldom adequately
documented. However, manual entry of data to operator log sheets and
transposition to spreadsheets is common practice on some plants. Spreadsheets do
not allow for controlled data entry; it is easy to inadvertently type over data, lose track
of records, duplicate and improperly enter data in a spreadsheet. Further to this, the
justification of formulae in spreadsheets often lacks clarity, and the justification is not
always recorded with the spreadsheet and it can get lost, especially if the developer
of the spreadsheet moves on, thereby making spreadsheets vulnerable to error and
unjustified computations. For these reasons, spreadsheets must be avoided as a
method of data entry.
11.1 INTRODUCTION
A company must have metal accounting rules and a procedure for auditing these.
These rules are primarily technical to provide management information and to
promote good corporate governance but they must also provide the quantities and
the reconciliation necessary for the financial reports. A set of “Check Lists” is given in
Appendix D, which can form the basis of inspections and audits.
The team’s objective is to obtain a fair and soundly based analysis of an existing
metal balancing and reconciliation operation and to report its findings as appropriate.
A minimum of an annual stock-take is preferred, both to confirm the base line figures
and to lend credibility to the metal accounts.
• witness stock-takes
• verify the mass flow measurements and analytical results
• assess compliance of mass flow, sampling, sample preparation and analytical
procedures.
In addition, an assessment of the accuracy, precision and reconciliation of metal
accounting figures is required. To this end, an audit trail, with clear supporting
documentation, is essential. Reporting should be done to suit the company’s
reporting period, in compliance with international practice and stock exchange
requirements, as applicable. The inclusion of an annual stock-take as part of the
report is recommended.
The use of a Competent Person to sign off the metal accounts is seen as an issue
relating to sound corporate governance, while other critical issues are transparency,
credibility and full disclosure.
The audit must identify and advise on areas that are inadequate and the
recommended actions required to achieve satisfactory performance.
It is essential that the Competent Person conducting the audit should have an
adequate support structure. It should not be necessary for the Competent Person to
sign off the metal accounts every year, but there must be an independent audit of the
metal accounting procedures by a Competent Person at least every three years.
Accounting Period – The period in which a company would normally report its
financial results to the market. For example in the USA this is quarterly, while in
many other financial markets the requirement is bi-annually. In metal accounting,
however, the accounting period is usually one month, but metal accounting reports
also have to be prepared to cover the company’s financial accounting period.
A measurement that is accurate is one that is free of bias and has a dispersion
(standard deviation) that is lower than a defined dispersion or indeed a probability
density of a particular nature. The level of dispersion or the nature of the probability
density is defined with the purpose of separating measurements that are entirely fit
for a particular purpose or use and those that are not.
Acid Mine (Rock) Drainage – The low pH effluent which seeps from waste rock, ore
dumps or stockpiles of materials which have an insufficient capacity to neutralise the
acidic products, containing the dissolution products of minerals.
Actual Recovery – The sum of all useful or desired outputs from a process, divided
by the sum of all inputs to that process, expressed as a percentage.
Analysis Depth – Is the maximum depth in a sample, such as a pressed pellet, from
which the fluorescent X-rays of an element can leave the sample.
Auger Sampling – An increment is collected from, usually, a relatively fine and dry
particulate material using a correctly designed auger sampler.
Barren Solution (Tails Solution) – The discard solution from a metal recovery
process.
Battery Limits – The boundaries of a metal accounting system, across which all
inputs to and outputs from that system are measured and balanced.
Belt travel speed – The average speed of a conveyor belt, which is used in
conjunction with the weigher, to calculate the tonnage of material conveyed over a
period of time.
Belt-speed sensor – A device that generates a signal as a result of the belt speed.
Best Practices – The most precise method of performing a measurement with the
information and equipment available at that time.
Bias – Bias is the difference between the mean result of one or more measurements
and the true value of the quantity being measured. It also can be seen as an
ongoing difference between two, or more, measuring systems. (e.g. sender’s and
receiver’s weights, the analysis of the same sample by two laboratories etc.)
Bubble Method – A method whereby the flow of material through a process plant is
interrupted to allow for equipment to be emptied in series to facilitate stock-takes
(normally applied to high value materials in the latter stages of processing).
Built-up Recovery – The sum of all useful or desired outputs from a process,
divided by the sum of all outputs from that process, expressed as a percentage.
Bulk Density – The density of a granular material normally stored in a bin, other
container or stockpile, defined as the mass of the material divided by the volume
occupied by the material (including the space between the particles).
Calibration Test (weighers) – A test, using known weights and forces to load a
scale, in order to determine the performance and operating tolerance compared to
standard values.
Certified Weight – A weight, which has been certified by the relevant national body,
or an agent authorised by that body, to have a mass within a defined tolerance, to be
used for certifying weighing instruments.
Check in-Check out System – The system whereby all streams into and out of the
Process or Plant, for which the balance is being performed, are measured, sampled
and analysed.
Coefficient of Variation – The quotient of the standard deviation and the mean
value of a measurement, expressed as a percentage.
Concentrate – Term used for the product from a mineral processing operation such
as gravity separation or flotation in which the valuable constituents have been
upgraded and a large proportion of the gangue material discarded in the tailings.
Concentration Ratio – The ratio of the tonnes of ore in the feed to the tonnes
produced as concentrate.
Coriolis Mass Flow Meter – A device, consisting of a flow sensor and a transmitter,
which primarily measures the mass flow by means of the interaction between a
flowing fluid and the oscillation of a tube or tubes; it may also provide measurements
of the density and the process temperature of the fluid.
Custody Transfer – The transfer of risk and/or ownership of a material from one
party to another. This includes ore from a mine to a processing plant, intermediate
products such as concentrate and matte from one plant to another, both in-house
and inter-company and waste products for sale or safe disposal, as well as final
products.
CUSUM Graphs – CUSUM (Cumulative Sum) graphs are constructed by plotting the
cumulative difference between the measurand and its accepted value.
Measurements are “In control” if the plotted line meanders around the zero difference
line. Plotted lines that deviate from the zero difference line are an indication of bias.
Enrichment Ratio – The ratio of the metal grade in the concentrate to the metal
grade in the feed.
Estimated Recovery – Where redundant data are available, the estimated recovery
will depend on which measurement points are chosen as the basis of the calculation.
For example, the mass and grade of the saleable product is usually accepted as the
value agreed as part of a commercial transaction. If the mass and grade of the feed
to the process are known from measurement and sampling, an “input recovery” may
be calculated.
Evaluation – The process by which the data required for Metal Accounting is
acquired.
Exception Report – A report generated, in terms of this Code, whereby each non-
compliance with the requirements of this Code is motivated and approved.
Gantry scale – A scale where the load measuring elements are attached to a gantry
crane normally used for loading or offloading materials in bulk.
Heap Leaching – Normally refers to the leaching of ores that have been crushed (to
an economic size) prior to placement on an impermeable pad on which suitable
Home and Away System (Double Sampling and Analysis) – A system where
batches/lots are weighed and sampled by both the sender and receiver. Data from
the analysis of the samples can then be statistically analysed to establish the
presence of biases in the sampling and/or analytical processes. In the “Half Home
and Away System” samples from even numbered batches of material are analysed
by the sampling sites laboratory (i.e. sender at sender’s laboratory and receiver at
receiver’s laboratory) and odd numbered batches are analysed at the other
laboratory (i.e. sender at receiver’s laboratory and receiver at sender’s laboratory).
Statistical analysis of the data can then establish the level of bias in either the
sampling or analytical procedures.
Hopper Scale – A scale designed for the weighing of granular materials in bulk,
whose load-receiving element is a self-cleaning hopper with an outlet gate.
In-motion weighing – The weighing of objects that are in-motion. Usually applied to
rail or road trucks.
In-situ Leaching – A process whereby the metal contained in an ore body is leached
in-situ by pumping a lixiviant into the ore, which is normally naturally fractured or has
been broken by explosives and the pregnant solution withdrawn
K Factor (Belt Weighers) – The ratio of the result of the mass determined by a bulk
test, arrived at by passing a known weight of material over the weigher, as compared
to that obtained by the use of test weights.
Kyoto Protocol – A Protocol arising from the meeting of 160 nations, in terms of the
International Convention for Climate Change, which took place in Kyoto, Japan in
1997, whereby signatory countries agreed to reduce the emission of greenhouse
gases by certain percentages, relative to levels pertaining in 1990 during the period
2008 - 2012. It came into force on the 16th February 2005.
Level of Confidence – The probability that the value of the measurand lies within
the quoted range of uncertainty.
Load Cell (Sensor) – A device which produces an output signal proportional to the
applied weight or load. The load cell may utilise any physical principle included in the
field of, but not limited to, electricity, electronics, hydraulics, magnetics, mechanics
and pneumatics or combinations thereof.
Lot – A lot is a defined quantity of the material being measured to which the mass
and analysis is applicable. It may be a shipment, a production batch or production
over a defined time period, such as a shift or a day, or a fixed tonnage of material.
Mass Balancing – The process whereby input and process streams, or the
individual components of the streams, stocks and in-process inventory are
measured, sampled and analysed in order to calculate the mass of each and
reconcile the output to the input, plus or minus the stock change.
Matte – The upgraded product from a smelting furnace or converting operation for
sulphidic ores or concentrates which contains the majority of the metal values and a
lower proportion of gangue materials (a large proportion of which have been removed
in the slag). Furnace (green) matte still contains a high proportion of iron whilst in
converter (white) matte the iron content is much lower and the value metals higher.
Mean (of data) – The mean of a set of n items of data is the arithmetic average of
the series of measurements xi. The mean is usually designated by x with a bar
above it.
Metadata – The term used to describe data about data. It describes who collects the
data, what the data contains, where (and how) the data is stored, when (and how
often), and why the data is collected,
Mine Call Factor – The ratio, expressed as a percentage, which the specific product
accounted for in the product recovered, plus that in tailings/residues, bears to the
corresponding product called for by the mine’s measuring and evaluation methods. It
is also the product of the Shaft and Plant Call Factors.
Momentum Mass Flow Meter (Impact Mass Flow Meter) – A device that detects
mass flow rate by measuring the momentum of a stream of material falling a definite
distance under gravity.
Neutron Gauge – A nuclear gauge that utilises an emission of neutrons to effect the
measurement.
Nominal Top Size – The mesh size that will allow 95% of the sample to pass
through.
1 ⎛ ( x - µ )2 ⎞
p(x) = exp ⎜ - ⎟
2πσ ⎜ 2σ 2 ⎟
⎝ ⎠
Paired t-test – A single sample t-test appropriate for comparing two samples in
which individual are paired, and observed values within pairs are positively correlated.
Permanent Inventory – In-process Inventory that has been absorbed into the
fabric of the processing plant, such as gold into concrete foundations, which is only
recovered when the plant is demolished at the end of it’s life.
Plant Call Factor – The mass of a species recovered to product and to tailings
divided by the mass of species fed to the plant, expressed as a percentage.
Plant Historian – A computer system for accurate operational data archiving and its
rapid recovery. The data historian accepts data input only, does not recognise
structures and procedures, and is strictly concerned with the accurate, efficient data
storage and recovery if required. Data historians perform three key functions:
Plant Recovery – The quantity of metal in the product stream from a plant, divided
by the quantity of metal in the feed to that plant, expressed as a percentage.
Precise measurements have a dispersion about their mean value, which is lower
than a defined dispersion or indeed a probability density of a particular nature. The
level of dispersion or the nature of the probability density is defined with the purpose
of separating measurements that are entirely fit for a particular purpose or use and
those that are not. A precise measurement may not be accurate; its mean may differ
from the true value of the measured quantity by an arbitrary amount. Standards
often quantify the term precision as a value corresponding to the magnitude of a 95%
confidence interval around a result. If the n measurements of the quantity to be
estimated are normally distributed, the interval is ± ts, where t is the two-sided
Student-t value at 95% confidence and n-1 degrees of freedom and s is the estimate
of the standard deviation of the estimated quantity.
Quality Control (QC) – The operational techniques and activities that are used to
fulfil the requirements for quality.
Random Stratified Sampling – The taking of increments whereby the material being
sampled is divided into strata, one increment being taken at random within each
stratum. (AS 4433.1 – 1997)
Reconstitution – If the mass and grade of the process rejects are known from
measurement and sampling, then an estimate of the feed to the process can be
made by adding together product and waste. A recovery based on this calculation is
often referred to as an output recovery. Given that all measurements (mass,
sampling and analysis) are subject to random errors (and possibly biases) the
numerical values of input and output recovery are not expected to be equal although
they should agree to within the errors of measurement imposed on each calculation
process. That is, they may be statistically identical.
Recoverable Metal – The proportion of a metal in a feed material that has been
determined to be recoverable, by laboratory and/or pilot plant tests, by the chosen
process, with the remaining metal being defined as Non Recoverable.
Relative Density – The relative density of a substance is the ratio of the density of
the substance at a specified temperature to the density of a reference substance at
the same or a specified different temperature (usually the reference substance is
water at a temperature of 4 ºC). Relative density is a dimensionless quantity and is
also referred to as specific gravity.
Reproducibility (in Mass Measurement) – The measure of the extent that weights
determined by two different instruments or methods will be similar.
Risk – The chance of an event occurring that will have an impact on objectives. A
risk may be quantifiable in terms of the likelihood of loss, less than expected returns
or an undesirable outcome. (The VALMIN Code, 2005 Edition).
Risk Assessment – A formal process for the identification and quantification of the
risks associated with the all the processes involved in metal accounting.
Run of Mine (ROM) Ore – Ore produced from a mining operation, which has not
been subject to any size reduction or other processing.
If the sampling constant is taken to be dependent only on particle top size and the
sampling constant is known at one particular state of comminution of the material, a
'sampling safety line' can be drawn for the particular analyte to which the sampling
constant applies.
SCADA – Acronym for Supervisory Control And Data Acquisition, a computer system
for gathering and analysing real time plant data. SCADA systems are used in
industry to monitor and control plant status and provide logging facilities. As the
name indicates, it is not a full control system, but rather focuses on the supervisory
level. As such, it is a purely software package that is positioned on top of hardware
to which it is interfaced, in general via Programmable Logic Controllers (PLCs), or
other commercial hardware modules.
Shaft (or Mine Section) Call Factor – The mass of a species delivered to a shaft
compared to the mass of that species calculated from sampling and surveying
procedures, expressed as a percentage.
Shewhart (Control) Graphs – These graphs are constructed by plotting the value of
a measurand about a central line of its accepted value. Control and Action lines are
usually drawn at ±2 and ±3 standard deviation levels respectively. Measurements
are said to be “In control” if they are within the 2 standard deviation limits, but
“Action” is required if they are outside 3 standard deviations.
Slag – The product from a furnace, smelting or converting operation which normally
contains the gangue materials such as silica, alumina and magnesia in oxide form
and small quantities of the value metals, which is discarded or recycled. An
exception is the titania slag from the smelting of ilmenite which is a raw material for
titanium oxide production and is sold.
Sonic Drilling – A technique for drilling that utilises high frequency vibrations
generated from sound waves to penetrate the material being drilled and generate
core samples.
Sonic Drilling – A technique for drilling that utilises high frequency vibrations
generated from sound waves to penetrate the material being drilled and generate
core samples.
Splitting Limits – The limits within which values from procedures such as twin-
stream analysis or inter-laboratory analysis should agree.
Spurious Error – The error which invalidates a measurement. They generally have
a single cause such as the incorrect reading of one or more significant digits or the
malfunction of instruments.
Strata – Approximately equal parts of a lot or sub lot based on intervals of time,
mass or space. (AS 4433.1 – 1997)
Tailings (Rejects, Residues) – The material after the extraction of the maximum
economic amount of the value metals from the feed material such as ore,
concentrate or matte, by treatment processes such as concentration or leaching.
Tare – The mass of an empty container such as a drum, bin, or road or rail truck.
Theoretical Recovery – The ratio of the metal input to a plant, minus the measured
losses from that plant, divided by the metal input to the plant, expressed as a
percentage.
Toll Treatment – The treatment, through a metallurgical plant, of material that has
been procured from third parties for processing under agreed contractual conditions.
The terms may include the charges for the treatment, recovery percentages for the
products, the time frame for production and penalties for deleterious impurities.
Totaliser – A device for counting and indicating or printing the algebraic sum of
successive inputs made on a serial entry basis. Inputs may represent counts, shaft
rotations etc.
Transducer – A device for converting energy from one form to another for
information or control purposes.
Triple bottom line reporting – The triple bottom line (TBL) focuses corporations not
just on the economic value that they add, but also on the environmental and social
value that they add and destroy.
At its narrowest, the term 'triple bottom line' is used as a framework for measuring
and reporting corporate performance against economic, social and environmental
parameters.
At its broadest, the term is used to capture the whole set of values, issues and
processes that companies must address in order to minimize any harm resulting from
their activities and to create economic, social and environmental value. This involves
being clear about the company's purpose and taking into consideration the needs of
all the company's stakeholders - shareholders, customers, employees, business
partners, governments, local communities and the public.
t-test (Student’s t test) – A test used on the means of small random samples of one
of the following null hypotheses:
a That the sample has been drawn from a population with a given mean
b That two samples have both been drawn from populations with equal
means
The test compares two sets of results and establishes whether or not there are
statistically significant differences between them.
Variance – The sum of the squares of the differences between each measured value
and the mean of the measured values divided by one less than the number of
measurements in the data set. This formula applies only when the individual
measurements are uncorrelated. Successive measurements during a day or several
days in a plant are often correlated, in which case a different formula must be
employed.
Weigh Carriage – A structure supporting the weigh idlers on a conveyor belt weigher,
which in turn transmits the weight through the carriage to the load sensor.
Weigh Idlers – Idlers positioned in the weigh carriage assembly so that they sense
the weight of material on the conveyor and transmit the weight through the carriage
to the load sensor.
Weighbridges – Weigh scales designed to weigh road trucks or rail wagons where
the lever system is usually below ground in a pit so that the vehicle or rail wagon can
pass over and stop on the scale. In a large capacity scale also refers to the
structural frame carried by the main bearing which supports the load-receiving
element.
Y-blenders – Mechanical mixing devices that are used to mix samples within
enclosed containers.
Yield – The quantity of coal (of the quality that one requires) obtained from Run Of
Mine Material expressed as a percentage. There are two subsets:
AMIRA International
Anglo Research
Anglogold - Ashanti
BHP Billiton
Namakwa Sands
Northparkes Mines
Rio Tinto
University of Cape Town: Drs Dave Deglon and Aninda Chakraborty, Omari
Lukwichi, Paul Bepswa and Erica van Tonder
Development of the original framework for the Code and contributions to the
section on Metal Balancing:
Permission to re-publish:
B.6 ANALYSIS
C.2.1 GENERAL
C.2.2 COAL
1 ASTM E877 – 03: Standard Practice for the Sampling and Sample
Preparation of Iron Ores and Related Materials.
2 ISO 3082:2000 Iron Ores – Sampling and Sample Preparation
Procedures
3 ISO 3083:1973 Iron Ores – Preparation
4 ISO 3085:2002: Iron Ores – Experimental Methods for Checking the
Precision of Sampling, Sample Preparation and Measurement
5 ISO 3086:1998: Iron Ores – Experimental Methods for Checking the Bias
of Sampling
6 ISO 3087:1998 Iron Ores – Determination of the Moisture Content of a Lot
7 ISO 3713:1987 Ferroalloys – Sampling and preparation of samples –
General Rules
8 ISO 3852:1988 Iron Ores – Determination of Bulk Density
9 ISO 4551:1987 Ferroalloys—Sampling and Sieve Analysis
10 ISO 4552-1 Ferroalloys: Sampling and Sample Preparation
11 ISO 4552-2 Ferroalloys: Sampling and Sample Preparation for Chemical
analysis
12 ISO 7087:1984 Ferroalloys – Experimental methods for the evaluation of
the quality variation and methods for checking the precision of sampling
13 ISO 7347:1987 Ferroalloys – Experimental methods for checking the bias
of sampling and sample preparation
14 ISO 7373:1987 Ferroalloys Experimental methods for checking the
precision of sample division
15 ISO 7764:1985 Iron Ores – Preparation of Pre-dried Test Samples for
Chemical Analysis
16 ISO 8049:1988 Ferronickel shot – Sampling for analysis
17 ISO 8050:1988 Ferronickel ingots or pieces – Sampling for analysis
18 ISO 10836:1994: Iron Ores – Method of Sampling and Sample
Preparation for Physical Testing
19 ISO 14284:1996 Steel and Iron – Sampling and preparation of samples
for the determination of chemical composition
C.3 ANALYSIS
D.1.1 GENERAL
CRITERIA EXPLANATION
Certification The procedure, which is a legal requirement, whereby a mass
(Assizing) or volume measuring instrument is approved as usable for
Custody Transfer and trade purposes and, in terms of this
Code, for Primary Metal Accounting, which must be performed
at laid down intervals by the relevant government authority or
their authorised agents. Procedures are laid down in National
Metrology Acts.
Calibration The procedure (as per National Metrology Acts, OIML
Recommendations, the NIST Handbook, ISO or other
applicable standards) for checking a mass or flow measuring
device, which must be performed at periodic intervals and at a
sufficient frequency, to measure the error in the result reported
by the device against a standard known mass or volume. This
is in order to ensure that the reported mass or flow
measurements remain within the required precision stipulated
for Primary Metal Accounting or Custody Transfer.
Exception Report Required for all mass/flow measurements in cases where the
Code requirements cannot be adhered to. Must be fully
motivated with reasons, a plan for improvement and approved
by Senior Management.
Non-Conformance Required for all mass/flow measurements in cases where an
Report internal or independent Metal Accounting Review has found
that the Code requirements have not been met and/or where
there is no Exception Report approved. It must contain the
reasons why the Code requirements have not been met and
recommendations for improvements and it must be signed off
by the Internal audit team leader or the Competent Person (if
as a result of an independent Metal Accounting Review).
Records Records of all certification and calibration test results must be
kept for five years. These records must show all the results and
measured errors and document any alterations and
adjustments made to the instrument and installation.
CRITERIA EXPLANATION
Test Weights Static weights, and in the case of In-Motion weighing, weighed
railway wagons, used for calibration must have been weighed on
a test facility and certified to be accurate to the required
tolerance. In the case of Weigh Bins/Hoppers and Gantry
Scales, facilities for loading and unloading the test weights on to
the framework are required.
Support Structure The structure supporting Weigh bins/Hoppers must be robust
for Bins enough to minimise deflection and to ensure that the load is
transferred evenly onto the load sensors.
In-Motion Truck speed must be limited and controlled to ≤8 km/hr, rail
Weighing trucks to be weighed empty (tared) on the weigher on every
occasion before/after filling or emptying. The rails must be
straight and firmly held in concrete in the vicinity of the weigher.
Track must be well maintained to limit impact, shocks and
vibration
Preferred Method Static weighing is the preferred method for all primary
accounting if feasible, as it is the most precise and it has greater
sensitivity
CRITERIA EXPLANATION
Type of Weigher Whether the weigher is nuclear or electromechanical (approach-
retreat, pivoting or fully suspended), fitted with counterweight or
not, type and number of load sensors, the number of weigh idlers,
the method of belt speed pickup and sensing and the type of
micro processor. Nuclear weighers are not suitable for Primary
Metal Accounting or Custody Transfer.
Calibration Procedures such as bulk (material) tests, test chains, belt cuts,
certified static weights or electronic procedures used to check the
precision of conveyor belt weighers. For Primary Metal
Accounting, bulk tests as per approved standards are mandatory
and thus facilities for accurate check pre- or post-weighing of the
bulk material passing over the weigher are required. Calibration to
include zero and span checks over 20-100% of the operating
range, for ten minutes or three belt revolutions whichever is the
greater, repeated three times to check repeatability as specified in
OIML Recommendations, the NIST Handbook or other applicable
standards. No ISO standard exists for conveyor belt weighers.
Conveyor Belt Endless, with smooth spliced joints and as flexible as possible,
sized to give a minimum loading of 20% and an operating range
CRITERIA EXPLANATION
Type of Flow Careful selection and specification from many types to suit the
meter specific application. For Metal Accounting the choice for slurries is
limited to Electromagnetic flow meters, or in a few cases, Coriolis
or Ultrasonic types. All except the Coriolis, measure volume flow
and thus require relative density/specific gravity measurement
(See E.1.6), which can introduce significant error. Manufacture,
selection and installation must adhere to the applicable standards
(ISO 6817, ISO 9104, ISO10790, ISO 13359, ASME MFC-11M-
2003, ASME MFC-16M-1995)
Specifications The uncertainty of the instrument must be specified in terms of
and Performance ISO 11631 with traceability back to the approved test
meter/facility preferably according to Group A but at least Group
B.
Certification The initial calibration and certification to be in terms of ISO 4185
or ISO 8316, in a facility accredited in accordance with ISO
17025, and the uncertainty assessed in terms of ISO 7066 and
5168.
Calibration For Primary Accounting the meters must be calibrated, on the
material being measured, on a regular basis utilising the
principles in ISO 4185 or 8316. This requires accurate
volumetric/weighing tanks and an efficient diversion system to
obtain flow rate comparisons. Relative density/specific gravity
measurement is as important as the flow (See E.1.6.).
CRITERIA EXPLANATION
Stockpiles Reduce to zero to establish absolute value and/or manage to
measure mass of receipts and withdrawals accurately to calculate
stock by difference. Use of parallel stockpiles to enable regular
emptying. Volumetric surveys only to be used where the above is
impossible on the understanding that the results will only be as
good as sampling and bulk density determination and accuracy
allow. No international standards exist.
Process Inventory “Bubble method” for high value materials and use of parallel
streams or vessels, to facilitate emptying
Audits Internal and Independent audits to check and recommend
corrections, if any, for approval by Competent Person.
CRITERIA EXPLANATION
Moisture Required in all cases as mass measurement instruments measure
wet mass. Sampling for moisture must conform with this Code,
ensuring that the assigned moisture is applicable to the material
being weighed.
Density of Density gauges must be calibrated on the actual material being
Slurries measured and accurate relative densities of both solid and liquid
phases are required.
Bulk Density Variable even for the same material as it is affected by many
factors such as particle size and compaction. Most reliable
method is measurement of the mass of a known volume.
D.2 SAMPLING
CRITERIA EXPLANATION
Type of sampler Whether the primary sampler is a cross-stream sampler from a
conveyor belt or slurry stream, a hammer sampler or an auger
sampler. If it is a cross-stream sampler, what are the design
criteria of the system e.g. gap size, cutter speed, radial or
parallel etc.?
Multi-stage If it is a multi-stage sampling system: The number of stages,
Sampling system whether there is sample reduction between stages and the
percentage cut at each stage, together with the total percentage
cut. Experimentation undertaken to ensure sample division does
not introduce excessive variance.
Sampling Interval The basis of the time interval between increments.
Batch Size The basis of the batch size i.e. is it on a time (shift) or tonnage
(lot) basis.
Sub-sampling Whether any further sample division is done on the plant and, if
so, details of how this is done.
Drying conditions If sample drying is carried out on the plant, the drying
temperature and duration.
Sample labelling Label type and sample identification details.
Quality checks Whether the sample weight at each stage is related to the
expected weight based on process flows, cutter gap, cutter
speed and sampling interval.
Inspections The frequency of inspections and level of employee.
Fault reporting The procedure for reporting faults, their rectification and the
and rectification signing-off by management.
CRITERIA EXPLANATION
Sub-sampling The mass and top particle size of the sample received by the
laboratory. Whether further sample division is necessary before
drying and, if so, details of how this is done.
Drying conditions Drying temperature and duration
Sample reduction Type of equipment used e.g. jaw crushers, pulverisers, ring and
puck mills etc. Action taken to guard against dust loss,
contamination from the equipment e.g. iron, chromium and
cobalt and cross-contamination.
Sample division Type of equipment used e.g. spinning rifflers, turning tube
dividers etc. Experimentation undertaken to ensure that
excessive variance is not introduced by sampling division.
Action taken to guard against dust loss and cross-
contamination.
Sample mixing Type of equipment used e.g. Y-blenders, cube blenders etc.
and duration.
Sample storage Details of container and storage conditions. Duration of storage
and authorisation level for disposal.
Quality checks Contamination from sample reduction equipment. Particle size
analysis after each reduction stage. Splitting ratios from the
sample division equipment.
Inspections The frequency of inspections and the level of employee.
Fault reporting The procedure for reporting faults, their rectification and the
and rectification signing-off by management.
CRITERIA EXPLANATION
Accreditation The nature, quality and appropriateness of the analytical
procedure must be confirmed by accreditation to ISO 17025.
Twin-stream Details of the twin-stream system, preferably different analysts
(Independent on different days, and the measured reproducibility.
Parallel) analysis
Reference Whether the RMs are in-house or CRMs. In particular how the
Materials accepted values were obtained and verified.
Instrument The number of standards used, linearity of the calibration
calibrations range, reproducibility of the calibration, frequency of re-
calibration. Use of internal standards. Frequency of preparation
of the standards and their storage conditions. Source of the
primary standards.
QC procedures Use of QC samples, splitting limits between duplicates.
QA procedures Round-robin exercises.
Umpire laboratory Selection of the umpire laboratory, steps taken to maintain
quality of analysis.
Maintenance Maintenance contracts, maintenance records.
Fault reporting The procedure for reporting faults, their rectification and the
and rectification signing-off by management.
Exception Reports Reports on non-compliance with the Code must contain the
reasons for non-compliance, action plans to reach compliance
and be signed-off by the Competent Person.
CRITERIA EXPLANATION
Metal Balance Desired constituents over the reporting period, timeously to
Plant Management and Financial Accountants
Key Performance Feed and product tonnages and grades; recoveries,
Indicators (KPI’s) accountabilities; reported against target figures; error levels of
all reported outputs; historical comparisons to assist monitoring
of long-term trends
Traceability To all sources of inputs
Data sources Manual entries to be minimised and automatic data collection
from SCADA or LIMS systems maximised
Preparation Adequately trained personnel, who understand source and
meaning of the inputs
Adjustments or Only with written authorisation of responsible person and
alterations recorded for review purposes
Mass balance Only via an approved procedure to identify inputs subject to
reconciliation illegitimate (spurious) errors, to determine bias and to facilitate
remedial action
Approval Reviewed and signed off by Metal Accounting Committee, or
equivalent, for subsequent approval by Audit Committee
CRITERIA EXPLANATION
Compliance with Suitably trained outside of the area they are examining must
Code carry out audits. They must ensure that the requirements of the
Code as listed in Check Lists E.1. to E.6. have been adhered to.
Audit Types There are four types of audit:
Inspections by operating staff – These must be carried out
daily and should cover areas such as cleanliness, operational
problems and calibration.
Internal Audits – These must be carried out at least monthly by
supervisory staff or suitably trained staff from other sections.
External Audits – These must be carried out on a “surprise”
basis at least twice a year by suitably trained and qualified staff,
preferably including a financial auditor, who are independent of
the facility being audited, but more frequently if conditions of
bias or malpractice have been found. These audits must contain
both horizontal audits, where the auditor will examine in detail
single aspects of the quality system e.g. calibration, and vertical
audits, where the auditor will select a sample and follow its
progress from mass measurement to storage and/or disposal.
The management of stockpiles, especially any reduction to zero
or sudden change in process inventory, must be commented
on.
Stock-takes – These must be witnessed by an external audit
team and preferably be based on the bubble method.
Audit Reports Audit Reports should be based on pre-designed hard-copy
checklists, which must be retained as a record for at least 5
years. Recommended examples may be found in the Australian
Standard AS 4433.6 – 2000, for sampling equipment, and
CITAC/Eurachem Guide CG1 – 2002, for laboratories.
Non-Conformance Non-Conformance Reports must be issued, with appropriate
Reports comments, for each instance of non-conformance. The
appropriate operational management must comment on these
reports before they are submitted to higher management.
Exception Reports Exception Reports should be examined and any progress, or
lack of, on the elimination of problems commented on.
Metal Accounting The whole metal accounting system must be reviewed on at
Reviews least an annual basis by senior management to establish that it
is still fulfilling its function. The review findings must be signed-
off by the Audit/Risk Committee and endorsed by the
Competent Person.
One objective of metal balancing calculations is to carry out calculations which can
utilise all of the available data. The calculation process will both expose
inconsistencies in the data and identify the likely source of the inconsistency. It will
also provide a way to link data from many sources, to integrate the data in a
structured manner and to provide a key source of tested, verified or certified
information which can be used as a basis for sound decision-making. The use of
redundant information will necessarily produce differences, raising the question as to
how these differences should be treated.
The following simple example illustrates the principles outlined in Section 7 using
unsophisticated mathematical techniques. The example is drawn from Morrison, Gu
and McCallum (2002).
This example demonstrates how a mass balancing problem can be solved in two
stages. Although it is not necessary to use a two-stage solution, it is often easier to
understand the process in two stages. The first stage estimates the best-fit values of
the flow rates. The second stage estimates the required component assay
adjustments which are compatible with the best fit flow rates. The second stage of
the process is also useful for Reconciliation to the mine or ore body where flow rates
may be well defined and assays only defined by an ore body model. In this case, the
adjustments are essentially an allocation of differences and they are not really a best-
fit solution – simply a check on plausibility.
The general problem looks quite daunting. It will first be defined and then a simple
example will be worked through for illustrative purposes.
The main aim of the balancing process is to achieve a set of self consistent stream
flows and assays, which are in some way a best estimate of the true situation. Self
consistent means that total ore and metal flows in and out of each node must be
equal to within specified numerical criteria.
Consider a flowsheet which has a number of streams indicated by subscript “i”. Some
stream masses “M” are measured over a specified time interval and some streams
are sampled and assayed for “j” assays “a”
Actual assay and mass flow values which are perfectly self consistent but essentially
unknowable M i * and a ij * .
Measured values Mi for streams and aij for assays. These values are always
inconsistent.
The adjusted data M i and aij may estimate the “true” value to within some defined
level. Adjusted data are the outcome of the balancing process.
It is also important to eliminate (or at least measure) bias within accounting balances.
This topic will be considered later.
The usual criteria for this method of analysis is to minimise the weighted sum of the
squares of the difference between each measured and adjusted value divided by our
estimate (standard deviation SD or σ ) of its accuracy. This value is usually called
the weighted sum of squares WSSQ.
There are two common ways to consider this problem. The simplest problem to solve
is to consider the differences between flows of ore and metals in and out of each
node.
2
⎛ M - Mi ⎞
WSSQ = ∑ ⎜⎜ i ⎟⎟ for mass flows
i ⎝ σ Mi ⎠
⎛ (Μ i aij − Μ i aij )2 ⎞
+ ∑⎜ ⎟ for assay flows
⎜ a σ 2
+ Μ σ 2 ⎟
ij
⎝ ij Μ i i a ij ⎠
subject to the constraint that the flowsheet must add up for total stream flows and
component stream flows.
After finding the “best” flows, the adjusted assays can be recalculated.
2
⎛ aij − aij ⎞
WSSQ = ∑ ⎜ ⎟
⎜
ij ⎝ σ aij
⎟
⎠
where aij now refers to all measured mass flows and assays.
As before, adjusted data must be self consistent. That is, it must satisfy the
constraints imposed by the flowsheet network.
For our simple case, let us consider a separator which has a feed stream containing
mass A over a specified time period and which produces a concentrate B and a
tailing stream C.
As noted earlier, three kinds of numbers can be defined for this problem:
• Actual assay and mass flow values which are perfectly self consistent but
essentially unknowable A*,B* and C* and ai* , bi* and ci*
A B
The example process is a flotation bank. Each stream is sampled and assayed for
several metals. These assays are designated ai, bi and ci, where i represents one of
n components.
For any given feed-rate, the product flows can be ratioed to the feed rate to simplify
the calculation.
1 β
1-β
For perfect data:
For measured data, adding a component flow error ∆ balances the equation for a
given value of β
∆i = ai − β bi − (1 − β ) ci
To simplify the calculation, assume the standard deviation estimate (SD) for each ∆
is one. Then the weighted sum of squares is:
SSQ = Σ∆i2 = Σ ( ai − β bi − (1 − β ) ci )
2
If an estimate of SD is available for each component, it can be used to scale the error
for each component (i.e. divide each error by its standard deviation) and it makes the
answer a little more complicated, as is addressed later.
To find the minimum sum of squares, take the derivative of the SSQ with respect to
beta and set the answer to zero.
∂SSQ / ∂β = 2Σ ( ai − β bi − (1 − β ) ci ) ( ai − ci )
which simplifies to
β = Σ ( ai − ci ) Σ ( bi − ci )( ai − ci )
2
(Note that a splitter which works as a splitter still produces Σ0/Σ0 which can take any
value.)
Considering a simple numerical example may clarify some aspects. As β is the only
variable, a direct search provides a way to consider dependencies in the solution.
Figures 1 and 2 show variation in the sum of squares for β values from 0 to 1 in
steps of 0.05 for the data shown in Table 1 below:
1000
100
10 %Cu
1 %Pb
SSQ
%Zn
0.1 %Fe
SSQ
0.01
0.001
0.0001
Beta 0-1
500
450
400
350
%Cu
300
%Pb
SSQ
250 %Zn
%Fe
200 SSQ
150
100
50
Beta 0-1
In mathematical terms, how well β is defined depends on how sharply the SSQ
curves at the minima. Hence, the assay (Zn) with the largest difference between feed
and products and the most accurate data (smallest SDs) will dominate the combined
results and produce the best estimate.
If the separation process encompasses the complete plant, the assay differences will
be larger than for any single separation and they will be better defined. In short, even
if the overall circuit balance is well defined, the internal balances may be poorly, or
not at all, defined.
It is clearly shown in Figure 1 that the smaller the SSQ at the minima, the sharper the
curvature and the better the flows are defined.
Data Adjustment
Given a “best” set of flow rates, the next step is to allocate the flow component errors.
As in the first stage, minimising the SSQ of the adjustments is a useful strategy. For
each assay at the best value of β , there is a mass flow error:
∆ i = ai − β bi − (1 − β )ci
If the small change required in each assay is defined as ∆ai , ∆bi and so on, then
and by difference
subject to the constraint that the adjustments add up to the required component error.
A standard technique to add a constraint to a minimisation is to use a Lagrange
multiplier λ for each constraint. The constraint equation must be arranged to equal
zero. Hence the modified sum of squares becomes:
To find the minimum values, take the derivatives with respect to each adjustment and
the Lagrange multiplier and equate to zero.
∆ai = −λ
∆bi = λβ
(
λ = −∆i 1 + β 2 + (1 − β )
2
)
This implies that apportioning the error according to the flow rate is the least squares
solution when the SDs for each measurement are equal. This is intuitively reasonable.
If each assay has its own SD estimate, the algebra is a bit more cluttered but not
very different and the assay with the smallest SD is adjusted least.
Table 2 shows a set of adjustments at the best-fit mass split of 16.8% to concentrate.
This method gives a set of adjustments for any value of mass split. Therefore, the
numerical derivative of β with respect to SSQ can be taken and iterated to the best
solution. This solution will usually be close to the simply weighted analytical solution
detailed above. If the data is accurate, the minima will closely coincide and the two
methods will provide essentially identical solutions. In this case the solution will be
insensitive to different SD estimates.
There are twelve measurements and four constraints in this case. Following the
comments in Morrison, Gu and McCallum (2002), an expected value of the SSQ with
good SD estimates would be 8 (i.e. 12–4).
Therefore, the measured data in this case probably has much smaller SDs than the
assumed value of unity for each.
A standard way to present this type of data is measured data (flows, assays or
component flows) against adjusted data. If most of the points lie along the diagonal
from zero, then the data and the adjustment procedure and probably working well.
All in all, the method of weighted least squares applied to the mass balancing
problem provides a great deal of useful information.
The weighted least squares case is a little more complex but follows the same path
as the data adjustment method.
If it is assumed that any measurement biases are small, then the required data
adjustments should be drawn from the probability distribution of variation associated
with each measurement as are the experimental or measured values – or at least a
very similar distribution. Therefore, on average, it can be expected that the weighted
adjustments ∆ ai σ ai , ∆ bi σ bi etc, have an average value of about one and the
weighted sum of squares (WSSQ) of adjustments have a value approximating the
number of measurements.
WSSQ = (δ ai σ ai ) + (δ bi σ bi ) + (δ ci σ ci )
2 2 2
The derivatives are also quite similar. The adjustment is scaled by dividing it by the
SD.
∆ai = λσ ai2
∆ci = λ (1 − β )σ ci2
The derivative with respect to the La Grange multiplier stays the same but the scaled
values are substituted back in to calculate it.
(
λ = − ∆i σ ai2 + βσ bi2 + (1 − β ) σ ci2
2
)
AMIRA P754: Code Guidelines Release 3 20070219 166
Commercial in Confidence
NB the denominator for the LaGrange multiplier turns out to be an estimate of the
standard deviation (SD) for ∆ i Therefore, the mass flow error for each measured
component can be weighted and iterated to a weighted best fit value of beta. If the
data is even slightly reasonable, the unweighted estimate of beta cn be used as a
starting point and iterated to a stable solution. Obviously the component with the
largest change in value (between con and tail) and the lowest relative SD for ∆ i will
tend have a flow split very close to the best fit weighted beta.
If the mass has been measured for more than one stream, a weighted stream
adjustment can also be added to the weighted sum of squares. If only measured one
stream has been measured, the flow ratio can be used.
Even for this simple case, the analytical solution is quite complex and it is much
easier to use MS Excel and Solver to find the best values. In this case the weighted
SSQ for all components plus flows can be added together, if appropriate.
There is only one flow split β but there is a λ for each constraint to identify as
variables for Solver to adjust to achieve the minimisation.
However, if the adjustments all lie within plus or minus two SDs, the adjusted balance
is probably as good as can be expected – subject to the checks for bias outlined next.
If the measurement and balancing process were to be repeated many times, the
adjusted data can be expected to exhibit a lower variance than that of the measured
data. For some examples see Deglon and Chakraborty, (2005).
Hence, the adjusted data can be compared with those to be used any Check in-
Check out strategy. If the data adjustments for the streams and masses which are
part commercial transfers, the Check in-Check out values can be assigned a very low
variance and the weighted adjustment procedure run again to derive a set of data
which are both technically self consistent and consistent with commercial value
transfers.
This approach can also integrate the primary and secondary data strategy used by
some precious metal operations.
For example, if the multi-assay based mass split estimate is highly self consistent
and does not match the measured mass split, then each mass measurement system
needs to be checked. Mass measurement systems can very easily acquire bias due
to a build up of material on a weigh frame or a simple fault (such as idler wheel slip)
in belt speed measurement.
For multiple mass or multiple stream cases, in any single data set the data
adjustment procedure will distribute both random variation and bias in one
measurement across others of the same type which share the same constraint.
However, over several data sets the biased measurement can be expected to require
more adjustment in a particular direction than measurements which only exhibit
random variation. In an unbiased system (where the measurement distribution is also
symmetrical), the arithmetic sum of the weighted adjustments should be close to zero
for each set and approximate zero when any selected data point is considered
across multiple data periods. This is most easily checked graphically as many
adjustment can be examined using a single trend graph.
To be detectable at all, a bias must persist long enough for its accumulated effect to
be larger than measurement or balanced data variance. A so called “variable bias” is
a system measurement problem and is not correctable by any mathematical
technique – the underlying measurement problems simply must be rectified.
The best way to guard against “variable bias” is for the person responsible for metal
to visually inspect each mass measurement and sample point at least several times
per week - daily if possible. Mathematical massaging can only conceal these
problems when many of them are very easy for a trained eye to “see”.
The adjusted data provide a self consistent set which can be used to calculate key
performance indicators such as metal recovery in a consistent manner – no matter
which adjusted data is selected as the basis of the calculation.
A second stage of adjustment can be used to align commercial and technical data –
when and only when the data have been fully analysed and the hypothesis that the
only variation is statistical is well supported.
This technique also allows data from many sources within a processing system to be
integrated for performance analysis and reporting.
These methods are powerful but should be supported by cusums of total mass and
key components to detect bias.
Last but not least, regular physical examination of each measurement and sample
point will detect many problems that mathematical techniques will not.
GJ Lyman, AusIMM, Sampling and Blending Conference, Sunshine Coast, QLD 9-12
May 2005
ABSTRACT INTRODUCTION
A metallurgical balance drawn around a mineral processing plant or Large metallurgical processing and refining operations are
smelter must involve all process streams crossing the plant boundary. increasingly concerned with the measurement of their
Modern sampling theory, as formulated by Gy, and sound statistical performance. The availability of powerful on-line measurement
method can be used to determine the uncertainties in the observed assays systems, cheap networked computing power and the ability to
for the set of balance analytes and in the mass flow rates of the process
manage large databases of information collected across the
streams for the balance period. These uncertainties can then be used
together with the material balance equations to arrive at adjusted operation make it possible to construct systems that a manager
component mass flows that satisfy the material balances exactly. To can use to view and analyse any part of the operations at a range
determine whether or not the data is consistent with the level of of scales or levels.
uncertainty in the data and with the material balance equations, a The determination of a material balance around an operation is
goodness of fit test is applied. This goodness of fit criterion is the the most fundamental calculation that can be carried out, but
principal tool to quantify the quality of the material balances. balances and recovery estimates are subject to uncertainties that
The mathematical structure of the statistical material balance problem derive from uncertainties in the measured assays and mass flows.
poses a number of challenges. For example, while the variances of the For high value materials, balances with the corresponding
adjusted component mass flows do not vanish, the variances of the sums statistical analysis may also be used to evaluate the possibility
of the adjusted component mass flows about each balance node do
vanish, simply because the balancing computation demands that the
that theft of material from the process has taken place, if all the
balances close exactly. Even though it can be shown that the variances of measurement processes are well-understood and controlled
the adjusted component mass flows are always lower than the estimates throughout the balance period.
for the corresponding unadjusted component flows, it is still not possible The method of calculating a balance around a particular
to construct a confidence interval on the component mass flow into (and metallurgical operation is usually one that has evolved over time.
out of) a balance node using the adjusted component flows. On the face It is common to find that the balance calculations rely on
of it, such a confidence interval is exactly what is most desired from the
‘assumptions’ about the characteristics of some of the process
balance calculations. The problem arises because the balance problem is
a data adjustment rather than a parameter estimation problem. streams or stock levels within the process. The common origin of
It is possible, however, to calculate the recovery of a component to a
the assumptions is a missing or unreliable sample for the
product stream and determine a confidence interval for the recovery particular process streams or a mass flow measurement that is
value. It is also possible to calculate an uncertainty in the mass flow of considered unreliable, that is, of high variance. For example, the
any balance species in any process stream, before and after balancing. moisture content determination of the material may be deemed
Balance results have little meaning without the confidence interval unreliable and therefore an ‘historically determined’ or ‘agreed’
estimates. Having invested in the manpower and procedures that make a figure for the moisture is used in the calculations. The impact of
statistically valid balance on an operation possible, the management may the assumption is usually unknown.
find that the levels of uncertainty associated with the balances are rather
The traditional methods of balancing also tend to rely upon or
higher than they would like. The obvious solution to the wide confidence
intervals is to reduce the uncertainty in the assays and mass flow rates on focus upon a single chemical or physical species in the
the principal process streams. The cost of this uncertainty reduction may calculations. There are generally two reasons for this. First, the
be quite significant, as it will involve reduction in analytical variances or target species is the one of primary value and second, the use of a
reduction in sampling and preparation variances which may require single species avoids the disconcerting fact that balances based
investment in and development of new sampling hardware or routine on additional species may provide different results and the
duplication of assays to reduce the assaying variance. technically appropriate method of reconciling the results of the
A second strategy for uncertainty reduction can be proposed which disparate balances is not known to the staff of the organisation.
involves the use of additional sampling points to provide material balance
nodes internal to the processing operation. The information required to If an operation has developed a corporate and economic
make these internal balances may well exist within the operation, but it justification for carrying out technically correct material balances
may not be recognised as being of value or the standard of sampling and on an operation, it is necessary to review all sources of data that
assaying may not be deemed to be sufficient to use the data for are used in the balance calculations and then eliminate the
metallurgical accounting purposes. The marginal cost of elevating the ‘assumptions’ involved and replace them with statistically sound
existing sampling systems to metallurgical accounting standards may statements of the uncertainties involved in the measured
well be smaller than that to improve the sampling and assaying on the quantities. The review of the data must scrutinise all sampling
boundary streams.
and analytical procedures that are used to acquire information
This paper presents a statistical analysis of an uncomplicated form of and develop realistic descriptions of the uncertainties involved in
the balance calculations that permits exploration of accuracy issues and
the advantages that come from additional internal balancing nodes for the
all measurements. Ideally, the total variance will be divided into
process flow sheet. The calculations are based on a relatively simple flow two components, the first due to due sampling and preparation
sheet of common structure, but are of general applicability. The objective and the second due to analysis. It is usually necessary to review
is to reveal broad issues. While each process flow sheet and sampling and the operation of the monitoring systems that are in place to
analysis system requires individual analysis to reveal precisely where the ensure that sampling systems functioned as required and that
gains can be made most economically, a general result can be laboratory quality control systems remained in place during the
demonstrated which provides an estimate of the limit to the variance balance period. Ensuring that a large analytical facility carrying
reduction which can be achieved by adding internal balance nodes.
out diverse assays operates within specification accuracy on a
routine basis is a great challenge that demands the goodwill,
1. Sampling Consultant, 17 Diosma Street, Bellbowrie Qld 4070. cooperation and diligence of a large team of skilled persons.
Sampling and Blending Conference Sunshine Coast, QLD, 9 - 12 May 2005 127
G J LYMAN
Competently carried out, the review will usually identify While the theory behind SMB that permits the optimal
effective means of actually increasing the amount of information estimation of balances can become very complex in a general
that can be extracted by the balances. The traditional view of a case and particularly when some uncertainties in assays or mass
mechanical sampler defines it as a device for acquiring a flows do not follow Gaussian (normal) distributions, it can be
representative sample of the material flowing past it; the sample quite simple in some cases and can be used to very great
is sent for analysis to determine the content of various analytes advantage to estimate how accurate balances will be, given the
of interest and the analyte contents are used in the balance level of uncertainty in the data going into the balance and the
calculations. However, as Gy has pointed out many times, a structure of the balance. It can therefore be used to investigate
correct mechanical sampler will also operate in such a way as to alternate balance structures and data quality scenarios and the
provide a sample whose mass is proportional to (that is, quality of the information that will arise from the balance. The
representative of) the total mass of material that passed the theory behind SMB in fact provides a design tool for sampling
sampling point during the period of sampling. The use of and material balance networks. A metallurgical operation would
mechanically correct sampling equipment to determine mass certainly not invest in a new process without making substantial
flow estimates is a means of acquiring information that continues efforts to determine the quality of product it will ultimately
to be grossly overlooked. provide. However, operations often construct sampling and
material balance systems without an appropriate understanding
The review of the sampling systems in place and the data of the quality of the data that will be produced.
available for balance calculation has to be linked to a
This paper has the objective of showing how the simplest SMB
consideration of how the balances should be calculated. In the
theory can be used to estimate the quality of the balance results
simplest case, where all inputs, outputs and changes in stock
that will be obtained, prior to any final decision on a sampling
levels are to be reconciled over an accounting period (month or system and material balancing system design. A simple case is
quarter, say) the structure of the balance is fixed and it remains to studied in which all uncertainties are assumed to follow Gaussian
decide which species will be used in the balance and whether or distributions. The scenario is one in which an overall balance on
not certain mass flows will be computationally estimated at the an operation is needed, but an option exists to include internal
same time as the mass flow rates and/or assays are adjusted to process balances in the SMB calculations. The statistical
provide an exact closure of the balance. advantage that can be obtained by including the internal balances
It would be unusual to find that all the sampling systems and in the calculations is quantified. This scenario is one that is
data monitoring procedures needed to provide the data for a accessible to many operations and the marginal cost of reducing
complete and statistically sound balance were already in place. It sampling variance for the perimeter streams may be far higher
would also be surprising if all systems provided data with than making relatively simple changes in procedures and taking
uncertainty levels that were deemed acceptable. Most operations advantage of samplers that already exist for the internal process
have no quantified knowledge of the accuracy of their sampling streams.
and assaying; the determination of the accuracy of sampling The flow sheet chosen for consideration is also a simple one
is a prerequisite to deciding how to best use the data that but one that is common to flotation as well as smelting
arises from the sampling. An understanding of the constitutional operations. Structurally, it corresponds to the flow sheet that was
and distributional heterogeneity of all process streams at the employed at Bougainville Copper in their flotation circuit
range of levels of comminution involved in the sampling and (Hinckfuss, 1976) and is also a common smelting flow sheet in
analysis for all analytes of interest is not common. which final slags are retreated to recover entrained values. The
Faced with the problem of enlarging the sampling system and flow sheet is shown in Figure 1 together with the stream
doing something to decrease the levels of uncertainty in the data, numbers, process deportments and nominal balance data. For a
the best way of investing in equipment or additional analyses smelting flow sheet, stage 1 corresponds to a furnace producing a
may not be immediately obvious. For some process streams, matte and slag, stage 2 corresponds to a converting process and
there will be little choice; a new sampler will have to be stage 3 to a converter slag scavenging process, recycling a matte
installed. However, the sampling frequency to be employed for material to the converters. The mass flow basis is 100 units of
mass input. In order to keep the problem simple, internal
that stream will dictate the accumulated primary increment
accumulation in the process has been ignored. Stock build-up or
masses and the need for more or less expensive sample handling
drawdown can have a dominant impact on balance variances
equipment and secondary or tertiary sample division at the
when not determined with adequate accuracy.
sample point. The sampling frequency used will also control the
variance component due to distributional heterogeneity (DH),
and, if the material sampled has a high intrinsic (constitutional) STATISTICAL DEVELOPMENT
heterogeneity with respect to analytes of interest, the variance
due to intrinsic heterogeneity (IH) may be important as well. Data model
Statistical material balance (SMB) methods, in which all
The development of a statistical model of the data is the first step
available data are used and are weighted in the calculations
in any statistical analysis problem. The problem considered here
according to their accuracy, have two advantages over traditional
is taken as one in which the processing operation receives
ad hoc balance calculations. First, the method provides a balance
consignments of feed material on a daily basis and a daily
that closes exactly for all species and mass flows involved and
composite sample is prepared for assay. Daily dry solids inputs
second, SMB permits the statistical interpretation of the data and
are determined from a wet tonnage delivered and a moisture
the results of the calculations. In particular, it is possible to
determination on the as-received solids. Outputs are similarly
determine confidence intervals or even full statistical
determined, using daily or weekly composites or assays and
distributions for computed results, where the variables of interest
weights of final product (stream 4) consignments. To keep this
have not had their variance suppressed by the balance
case simple, build-up or draw-down of internal stock of
calculations. It is possible to quantify, for example, the
in-process material will be ignored. Accounting for these stocks
probability that the output of valuable exceeded the balance
would modify external flows and add some internal flows.
value by a given number of grams, ounces or tonnes. It is also
possible to quantify the probability that the recovery of SO2 to At the end of the month, the net input of a species is a sum
the scrubbers fell below a target figure. over the periodic inputs and the net output is a similar sum over
outputs.
128 Sunshine Coast, QLD, 9 - 12 May 2005 Sampling and Blending Conference
SAMPLING, METALLURGICAL ACCOUNTING AND REDUCTION OF BALANCE ESTIMATION VARIANCE
Stage 1 (Rougher)
1 Rec. to 98.00% 2
Conc. 80.00%
Species mass conc 40.00% mass conc
1 1.000 1.00% 10.00% 0.020 0.02%
2 11.000 11.00% 3.00% 2.200 2.64%
3 12.000 12.00% 1.00% 7.200 8.63%
4 13.000 13.00% 11.700 14.03%
5 5.000 5.00% 4.850 5.82%
6 58.000 58.00% mass conc 57.420 68.86%
100.000 100.00% 0.980 5.90% 83.390 100.00%
8.800 52.98%
4.800 28.90%
3 1.300 7.83%
0.150 0.90%
0.580 3.49%
16.610 100.00%
FIG 1 - Process flow sheet for balances, with process unit species deportments and balanced mass flows and concentrations.
When internal flows are also to be taken into account, a similar The measured assay for species k on day i and in stream j is
accounting procedure is common. For example, in a smelting denoted as aijk . The uncertainties in the mass flows are µ ij and
circumstance, furnace matte would be sampled when tapped and in the assays, α kij .
then weighed as charged to the converters; if granulated, it can be Sampling theory combined with knowledge of material IH and
readily sampled and weighed as it is conveyed to the holding DH and testwork can be used to quantify the uncertainties. In the
bins ahead of the converters. In a flotation operation, the ideal present circumstances, these uncertainties will be taken to have
means of determining the mass flows is proportional sampling of Gaussian distributions with zero mean (unbiased sampling and
the slurry streams, with daily or shift samples for assay. For the assaying). The statistical properties of the uncertainties are then
sake of discussion, daily composites and monthly balances completely described by the variance and covariance of the
(30 days) will be assumed. random quantities. Because a species mass flow for the month
This mode of material movement and sampling is modelled by depends on both assays and daily lot masses, all monthly species
measured masses of daily flows denoted by mij for stream j and mass flows within a given stream will be correlated and all
day i. The total measured mass flow of the process stream j over species flows in that stream will correlate with the total mass
the month is then Mj: flow. However, daily mass flow measurement uncertainties will
not tend to correlate with assay uncertainties unless the process
Nj stream assay correlates strongly with the mass flow and the
M j = ∑ mij (1) primary sampling frequency is lower than dictated by sound
i=1 sampling system design.
On the above basis, the monthly total and species mass flows
where: can be written as:
Nj is the number of daily lots involved for the stream Nj
The total mass flow of species k over the month for stream j is M j = ∑(m
~ +µ )
ij ij
given by: i=1 (3)
j N
~
Nj = M j + ∑ µ ij
A kj = ∑ mij aijk (2) i=1
i=1
Sampling and Blending Conference Sunshine Coast, QLD, 9 - 12 May 2005 129
G J LYMAN
Nj
Aijk = ∑ ( m
~ + µ )( a~ k + α k ) MkT is the transpose of a column vector of the flow of species k
ij ij ij ij
i=1
in process streams 1 to Ns
(4)
~
N
j A tilde will again be used to denote the true values, no mark
= Aijk + ∑ ( m
~ α k + a~ kµ + µ α k )
ij ij ij ij ij ij will be used to denote the measured values and the statistically
i=1 adjusted values will be denoted by a caret (^). To write the
material balances concisely, an expanded version of the C matrix
where mathematical symbols with a tilde (~) denote the true, and is needed. Construct an ( N c + 1)( N n ) × ( N c + 1)( N s ) matrix,
thus unknown, value of a flow or assay and Greek symbols where Nn is the number of balance nodes or rows of C, by
denote discrepancies of observed values from true values. The locating copies of the C matrix down the diagonal and call this
variances of the monthly total and species mass flows are then: matrix G:
{ }= N σ
Var M j j
2
µj (5) C
0
0 0 0
C 0 0
G= (11)
and 0 0 O 0
0 0 0 C
{ }= N m 2 σ 2 + a k
( )σ + σ α2 k σ µ2 j
2
Var A kj 2
(6)
j
j α j j µj
k
j
1 −1 −1 0 0 0 0
…
C = 0 0 1 −1 −1 0 1
total flows
(9)
…
0 0 0 0 1 −1 −1
To deal with all the material flows at once (total flow and
species flows), it is convenient to adopt a more compact notation. … species 1
…
z T = [ M T M 1T M 2T K M N c T ]
…
(10)
…
where:
species Nc
MT is the transpose of a column vector of total process flows in
…
streams 1 to Ns
130 Sunshine Coast, QLD, 9 - 12 May 2005 Sampling and Blending Conference
SAMPLING, METALLURGICAL ACCOUNTING AND REDUCTION OF BALANCE ESTIMATION VARIANCE
The weighted least squares formulation requires minimisation therefore subject to less uncertainty than the measured values.
of the quadratic form ξ T Σ −1 ξ, subject to the requirement that For a scalar function of the adjusted flow rates, f ( z$ ), such as a
Equation 12 be satisfied. It is also useful to note that: recovery value, the variance of the function is, approximately:
is a vector of the extent to which the measured values fail to This expression is exact only if the function f is linear in the
satisfy the material balance and is called the vector of node assay adjusted mass flows; ∇f is the gradient vector of the function f.
residuals. Equation 12 is then: Equation 20 shows that the covariance matrix depends only on
the structure of the material balance as captured by the C matrix
Gξ = −δ (15) and the uncertainties in the measurements. The covariance of the
results does not depend on the particular values of the measured
The minimisation of the quadratic form subject to the above mass flows as long as the covariance matrix of the measurements
linear constraints is solved by the method of Lagrange multipliers does not change greatly as the mass flows change. For a
and has a linear solution. The solution is: processing operation that does not have large variations in
relative process species mass flows from one month to the next,
ξ = − ΣG T (GΣG T )−1 δ (16) the covariance matrix of the adjusted values is, to a first
approximation, constant.
If the measurements are unbiased, this result is unbiased. This last result is very significant as it means that once
realistic estimates of the uncertainties in the mass flow and assay
Statistical results measurements have been established, it is possible to study
various balance configurations to see which configuration
It is not difficult to demonstrate that:
provides the greatest reduction in variance between the measured
and adjusted mass flows. While the calculations above may seem
Cov{δ} = E{δδ T } = (GΣG T ) (17) to be daunting, they are actually trivially simple in comparison to
the task of establishing the real measurement accuracies for the
The operator E{} denotes the statistical expectation of the process streams.
quantity.
In the context of this paper, the above result will be used to
Note that this covariance is that of the node assay residuals study the impact of balance configuration and the influence of
before data adjustment. The covariance of the node assay measurement accuracies in key process streams on the accuracy
residuals after adjustment is identically equal to zero. The of adjusted species flow estimates.
method of solution forces the adjusted node assay residuals to be
zero regardless of the data and the outcome of every calculation
is zero node assay residuals. CASE STUDY
The covariance matrix of the adjustments to the mass flows is:
Data adjustment
Cov{ξ} = E{ξξ T } = ΣG T (GΣG T )−1 GΣ (18)
The figures used for the deportments of the species are entirely
fictional but have been developed to include both ‘gangue’ or
and the minimal value of the weighted sum of squares is: ‘slag components’ and valuables in the balances. Table 1 proves
the deportments and comment on the nature of the species it is
WSSQ = ξ T Σ −1 ξ = δ T (GΣG T )−1 δ (19) intended to represent. The species represent a range of
deportments including one that is not well separated in stage 1
This quantity depends on the data and is therefore a random and 2. Species 6 is a mixture of elements for which no assay is
variable. It can be shown to have a chi-squared distribution with determined; it can be regarded as the ‘balance of the sample’.
N n ( N c + 1) degrees of freedom. This fact can be used to carry out
The nominal measurement relative standard deviations for
a test of goodness of fit of the data to the mass balance equations.
If this test fails, it means that either the data do not obey the either total mass flow or analytes one to five within a daily
balance equations or the estimates of the measurement variances sample are provided in Table 2. The figures given are
are incorrect. This test is the principal means by which to percentages and are realistic in the author’s experience.
evaluate the quality of the data and the validity of the balance. To ensure that the estimate of the covariance matrix originates
The weighted sum of squares can be partitioned into components from a realistic data set, that is, one on which the balances do not
due to each species to determine the contribution of each analyte close exactly the base data from Figure 1 were perturbed
to the total. according to the monthly variances and covariances.
Finally, the covariance matrix of the adjusted mass flows is The balance based on all seven streams and three balance
given by: nodes and the first five species leads to the variance estimates of
Table 3. The inclusion of the all six species in the process stream
Cov{z$} = Σ a = Σ − ΣG T (GΣG T )−1 GΣ (20) would lead to a singular solution to the balance, as the sum of all
species must always equal the total mass flow. To determine the
This equation shows that the variance of the adjusted mass mass flow of the last species, the five balance species are
flows must always be less than that of the measured flows as the subtracted from the balanced total mass flow.
matrix being subtracted is a positive definite covariance matrix. To ensure that the estimate of the covariance matrix originates
The fact that the covariance of the adjusted values is a simple from a realistic data set, that is, one on which the balances do not
difference between the covariance matrix of the observed values close exactly the base data from Figure 1 were perturbed
and the covariance of the adjustments is a consequence of their according to the monthly variances and covariances and when
statistical independence which derives from the general structure this perturbed data set was re-balanced, it was taken to be the
of linear least-squares problems. The adjusted mass flows are true base case.
Sampling and Blending Conference Sunshine Coast, QLD, 9 - 12 May 2005 131
G J LYMAN
TABLE 1
Species, deportments and comment.
132 Sunshine Coast, QLD, 9 - 12 May 2005 Sampling and Blending Conference
SAMPLING, METALLURGICAL ACCOUNTING AND REDUCTION OF BALANCE ESTIMATION VARIANCE
TABLE 3
Results for least squares adjustment of component flows with variances using three balance nodes. The WSSQ value is 14.2 and the
average variance ratio is 0.49.
Stream Flows Measured Adjustment Adjusted SDM RSD (%) SDA RSD (%) Var ratio
1 mass 99.5172900 0.3584135 99.8757040 0.4542322 0.45 0.3624343 0.36 0.64
1 0.9975235 0.0070499 1.0045734 0.0120095 1.20 0.0049690 0.49 0.17
2 10.8411280 0.1105074 10.9516350 0.1142302 1.04 0.0533312 0.49 0.22
3 11.9086810 0.1365811 12.0452620 0.1393566 1.16 0.0794091 0.66 0.32
4 12.9694230 0.0134608 12.9828840 0.0943178 0.73 0.0694370 0.53 0.54
5 4.9862954 0.0143167 5.0006122 0.0517201 1.03 0.0373300 0.75 0.52
2 mass 83.6325490 -0.3799723 83.2525770 0.7634572 0.92 0.3581939 0.43 0.22
1 0.0208311 -0.0000987 0.0207324 0.0004256 2.05 0.0003917 1.89 0.85
2 2.2212457 -0.0142280 2.2070177 0.0320104 1.45 0.0265272 1.20 0.69
3 7.3236451 -0.0687750 7.2548701 0.1034831 1.43 0.0759617 1.05 0.54
4 11.7113400 -0.0260468 11.6852930 0.1258338 1.08 0.0692866 0.59 0.30
5 4.8689083 -0.0189101 4.8499982 0.0622700 1.28 0.0373225 0.77 0.36
3 mass 16.6528370 -0.0297098 16.6231270 0.0912113 0.55 0.0465734 0.28 0.26
1 0.9848403 -0.0009993 0.9838410 0.0106426 1.08 0.0049480 0.50 0.22
2 8.8081249 -0.0635076 8.7446173 0.0817230 0.93 0.0475216 0.54 0.34
3 4.8061904 -0.0157982 4.7903922 0.0504319 1.05 0.0313239 0.65 0.39
4 1.3066553 -0.0090642 1.2975910 0.0101236 0.78 0.0063122 0.49 0.39
5 0.1480575 0.0025565 0.1506140 0.0018382 1.22 0.0009678 0.64 0.28
4 mass 13.9745220 -0.0129900 13.9615320 0.0561305 0.40 0.0425430 0.30 0.57
1 0.9835847 -0.0019182 0.9816665 0.0065377 0.67 0.0049468 0.50 0.57
2 8.5839360 0.0087002 8.5926362 0.0715575 0.83 0.0474234 0.55 0.44
3 4.1633568 -0.0065558 4.1568010 0.0422279 1.02 0.0309195 0.74 0.54
4 0.1399171 0.0001824 0.1400995 0.0016089 1.15 0.0015704 1.12 0.95
5 0.0015772 -0.0000019 0.0015754 0.0000211 1.34 0.0000207 1.32 0.96
5 mass 9.9287668 0.0139440 9.9427109 0.0380675 0.38 0.0287047 0.29 0.57
1 0.1082153 -0.0007784 0.1074368 0.0012562 1.17 0.0008919 0.83 0.50
2 1.5189549 0.0074900 1.5264450 0.0152984 1.00 0.0105447 0.69 0.48
3 6.3229750 -0.0363283 6.2866467 0.0723102 1.15 0.0455585 0.72 0.40
4 1.2664532 0.0207463 1.2871995 0.0147014 1.14 0.0062253 0.48 0.18
5 0.1537728 -0.0001663 0.1536065 0.0015228 0.99 0.0009687 0.63 0.40
6 mass 2.6890580 -0.0274629 2.6615950 0.0250386 0.94 0.0151470 0.57 0.37
1 0.0021955 -0.0000210 0.0021745 0.0000523 2.41 0.0000497 2.29 0.90
2 0.1535941 -0.0016130 0.1519811 0.0020041 1.32 0.0016511 1.09 0.68
3 0.6399527 -0.0063615 0.6335911 0.0075767 1.20 0.0059033 0.93 0.61
4 1.1677094 -0.0102179 1.1574915 0.0127284 1.10 0.0062073 0.54 0.24
5 0.1522891 -0.0032505 0.1490386 0.0020067 1.35 0.0009669 0.65 0.23
7 mass 7.2794283 0.0016875 7.2811158 0.0425291 0.58 0.0274504 0.38 0.42
1 0.1043722 0.0008902 0.1052624 0.0012963 1.23 0.0008917 0.85 0.47
2 1.3783868 -0.0039229 1.3744639 0.0149447 1.09 0.0105225 0.77 0.50
3 5.6260314 0.0270241 5.6530556 0.0609982 1.08 0.0454820 0.80 0.56
4 0.1297311 -0.0000231 0.1297081 0.0010555 0.81 0.0008866 0.68 0.71
5 0.0045662 0.0000017 0.0045679 0.0000634 1.39 0.0000602 1.32 0.90
It is of interest to determine whether or not there is some is of interest. Recognising that the expressions involve matrices
general statement that can be made about the improvement in and using Equation 20, the quantity of interest is:
variance that can be achieved by introducing additional balance R = Σ −1 [ Σ − ΣG T (GΣG T )−1 GΣ ]
streams and nodes. The ratio that has been calculated is the ratio
of the variance of the adjusted component flow to its measured = I − G T (GΣG T )−1 GΣ
variance. In general terms, a ratio of the nature:
To simplify the structure of the uncertainties, let it be assumed
Σ that all measurement uncertainties are statistically independent.
R= a
Σ In such a case, the matrix Σ is diagonal. Then the ratio of the
Sampling and Blending Conference Sunshine Coast, QLD, 9 - 12 May 2005 133
G J LYMAN
TABLE 4
Results for least squares adjustment of component flows with variances using a single balance node. The WSSQ value is 2.4 (expected
value 6) and the average variance ratio is 0.70.
Stream Flows Measured Adjustment Adjusted SDM RSD (%) SDA RSD (%) Var ratio
1 mass 99.5172900 0.3239744 99.8412650 0.4542322 0.45 0.3664377 0.37 0.65
1 0.9975235 0.0066914 1.0042149 0.0120095 1.20 0.0057209 0.57 0.23
2 10.8411280 0.0755441 10.9166720 0.1142302 1.05 0.0644001 0.59 0.32
3 11.9086810 0.1305750 12.0392560 0.1393566 1.16 0.0820157 0.68 0.35
4 12.9694230 0.0128725 12.9822950 0.0943178 0.73 0.0698344 0.54 0.55
5 4.9862954 0.0135906 4.9998861 0.0517201 1.03 0.0374113 0.75 0.52
2 mass 83.6325490 -0.4331022 83.1994470 0.7634572 0.92 0.3609159 0.43 0.22
1 0.0208311 -0.0001119 0.0207191 0.0004256 2.05 0.0003919 1.89 0.85
2 2.2212457 -0.0138327 2.2074130 0.0320104 1.45 0.0265553 1.20 0.69
3 7.3236451 -0.0727133 7.2509318 0.1034831 1.43 0.0763607 1.05 0.54
4 11.7113400 -0.0366727 11.6746670 0.1258338 1.08 0.0698590 0.60 0.31
5 4.8689083 -0.0228845 4.8460237 0.0622700 1.28 0.0374251 0.77 0.36
3 mass 13.9745150 -0.0216839 13.9528310 0.0561305 0.40 0.0522713 0.37 0.87
1 0.9835833 -0.0022829 0.9813004 0.0065377 0.67 0.0057010 0.58 0.76
2 8.5839255 -0.0282493 8.5556762 0.0715574 0.84 0.0602064 0.70 0.71
3 4.1633542 -0.0148434 4.1485108 0.0422278 1.02 0.0404544 0.98 0.92
4 0.1399166 -0.0002047 0.1397119 0.0016089 1.15 0.0015961 1.14 0.98
5 0.0015772 -0.0000024 0.0015748 0.0000211 1.34 0.0000210 1.33 0.99
4 mass 2.6890580 -0.0000715 2.6889865 0.0250386 0.93 0.0249522 0.93 0.99
1 0.0021955 -0.0000001 0.0021953 0.0000523 2.38 0.0000523 2.38 1.00
2 0.1535941 -0.0000116 0.1535825 0.0020041 1.30 0.0020005 1.30 1.00
3 0.6399527 -0.0001391 0.6398136 0.0075767 1.18 0.0075567 1.18 0.99
4 1.1677094 0.0002074 1.1679167 0.0127284 1.09 0.0126607 1.08 0.99
5 0.1522891 -0.0000016 0.1522876 0.0020067 1.32 0.0020030 1.32 1.00
TABLE 5
Recovery of selected species with uncertainties in the recovery values: comparison of single and mutli-node balances. Stream numbers
relate to the three node balance configuration.
134 Sunshine Coast, QLD, 9 - 12 May 2005 Sampling and Blending Conference
SAMPLING, METALLURGICAL ACCOUNTING AND REDUCTION OF BALANCE ESTIMATION VARIANCE
Sampling and Blending Conference Sunshine Coast, QLD, 9 - 12 May 2005 135
136 Sunshine Coast, QLD, 9 - 12 May 2005 Sampling and Blending Conference
APPENDIX F – DETERMINATION OF THE SAMPLING
CONSTANT
INDIVIDUAL PARTICLE ANALYSIS (50-PIECE EXPERIMENT)
For an alternative version of this experiment as applied to gold ore sampling, see
Bartlett and Hawkins.
Fundamental Bias:
m(FE) = -Y[1/MS – 1/ML]
Fundamental Variance:
σ2(FE) = Z[1/MS – 1/ML]
When in the normal situation ML >> MS the equations can be written as:
m(FE) = -Y/MS
σ2(FE) = Z/MS
Z = {(0,25v)/(aQ2MQ)}{Σq(aq – aQ)2Mq2/Vq}
where:
Mq = Mass of individual pieces (g)
Vq = Volume of individual pieces (cm3)
aq = Analytical value of individual piece (e.g. % Cu)
MQ = Σq Mq
v = (Σq Vq)/q
q = Number of pieces
aQ = Σq aqMq/MQ
where:
W = (aq – aQ)*Mq2/Vq
X = (aq – aQ)2Mq2/Vq
by calculation:
MQ = 5371
aQ = 37.91
v = 37.67
Y = 1.8
Z = 5.9 (Sampling Constant)
using the relationship for Fundamental Variance, given previously:
σ2 = Z/MS
the following minimum sample masses (kg) can be calculated at the indicated
variances.
Variance 10% 5% 3% 1%
(Relative)
Mass 2.4 9.4 26.2 236
Abstract
In this paper it is shown how variance propagate through plant material balances. As
stockpiles of raw materials, products and inventory always fall within the battery limits
of a smelter plant they make up part of the working capital, a figure directly reflected
on the balance sheet of each smelter. The cumulative effect of uncertainty in the
measurement of stockpile volumes, moisture levels and bulk densities propagates
non-linearly through to the variance in average plant feed assays and estimated
recoveries. There is little published information on what precision can be ascribed to
the copper recovery from concentrate to anode or blister over a given metallurgical
accounting period.
Where the mine and the smelter are an integrated unit the overall recovery from mine
to metal is of primary importance and changes can be made either at the mill or the
smelter to maximize the overall recovery. Where the smelter recovery is extremely
important is when concentrates are purchased or tolled because of the contract
structures where the payable metal values mean that the smelter must achieve a
certain minimum recovery in order to break even.
This paper is an attempt to quantify the variance in the recovery from first principles
using plant data where the accounting principles are based on a “check in – check
out” system (Wills, 2005) which is often used for pay-metal accounting on a smelter.
However, the “check in-check out” system becomes impractical at smelters for
nonpay-elements and volatile elements such as sulphur (also elements such as
arsenic, antimony, zinc) where bulk quantities are not transported via a checkpoint
and continuous and variable losses occur via flue gases and fugitive emissions (such
as from Pierce Smith converters, and molten matte transport in ladles).
Various calculation strategies are now considered for a range of typical metal
accounting approaches. In the discussion below, bias (due to gross or systematic
measurement error) has been ignored. In other words, all listed plant measurements
are assumed to be unbiased. A number of methods exist for bias detection, such as
the “cusums method” (Bartlett, 2005) for a lumped, plant-wide test, or through
statistical tests (François-Bongarçon, 2005) on individual sampling points using
paired t-tests. Should biases be found to be present, they should be eliminated
through proper calibration, maintenance, or repair (such as leaks) as required.
Sometimes total redesign of the sampling system is required according to the
principles of representative sampling.
If the weights of monthly smelter feed, blister copper (product) and discard slag (tails)
are F , P and T respectively and their corresponding assays f ,p, and t then :
F =P+ T
(1)
i.e. : total material input = total material output, over the period of a month, and
F ⋅ f = P⋅p + T ⋅ t
(2)
The bar above the feed and tailings variables indicate representative quantity over
the monthly period, with inclusion of stocpiles for all feed or tailings material types.
It follows that:
F ⋅ f = P⋅p + ( F – P) ⋅ t
(3)
P (f − t)
=
F (p − t )
(4)
p⋅P
R= ⋅ 100%
f ⋅F or:
(5)
p⋅( f − t)
R= ⋅ 100%
f ⋅ (p − t )
(6)
The variance in any calculated amount (say recovery, or calculated stockpile mass,
or average concentration) can be found from the variances in the measured
quantities (such as assays, volumes, measured weights, etc.) through the summation
of the squared partial derivatives of the calculated variable with respect to the
independent measured quantities each multiplied with the respective variance in the
measured quantity. The mathematical proof can be found in many a good text on
mathematical statistics, such as Kreyszig (1970) and is based on the Taylor series
expansion of the error of a multivariable function. This relationship is usually referred
to as the “Propagation of Variance” or the “Law of Propagation of Error (LPE)”. This
Xiao and Vien (2003) has noted that the addition of the second Taylor series term
may be required in cases of large relative standard deviations of measurements.
They state that the law of propagation of error (variance) is only an approximation,
because second- and higher-order derivatives are neglected. For the non-linear
variance propagation case, the precision of the approximation increases with
decreasing relative variance, i.e. for small relative errors. Therefore, the truncation of
the Taylor series approximation after the first term becomes less accurate when
double digit relative standard deviations are measured on the plant, such as often
found for slags and dusts (Eksteen et al., 2004). Additionally, it is assumed that the
measurements are independent, which may not be the case in practice.
The variance in plant recovery, R, based on the derived two product formula
(Equation 6), can be found from it’s derivatives as follows:
2 2 2
⎛ ∂R ⎞ ⎛ ∂R ⎞ ⎛ ∂R ⎞
V R = ⎜⎜ ⎟⎟ V f + ⎜⎜ ⎟⎟ V p + ⎜ ⎟ Vt
⎝ ∂f ⎠ p ,t ⎝ ∂p ⎠ f ,t ⎝ ∂t ⎠ p , f
(7)
⎡ p2 ⋅ t2
V R = ⎢ 2 Vf +
( f − t )2 ⋅ t 2 V + ( p − f )2 ⋅ p 2 V ⎤⎥ ⋅ 1002
⎢⎣ f ( p − t )2 p ( p − t )2 t ⎥⎦ f 2 ( p − t )2
(8)
p⋅P
As copper recovery is defined by equation 5: R = ⋅ 100% where P, p, F , and
f ⋅F
f are defined as in Equations 1 and 2.
the variance of R can be found from the partial derivatives with respect to each
variable:
2 2 2 2
⎛ ∂R ⎞ ⎛ ∂R ⎞ ⎛ ∂R ⎞ ⎛ ∂R ⎞
V R = ⎜⎜ ⎟⎟ Vf + ⎜ ⎟ VF + ⎜⎜ ⎟⎟ Vp + ⎜ ⎟ VP
⎝ ∂f ⎠ F , p ,P ⎝ ∂F ⎠ f , p ,P ⎝ ∂p ⎠ f ,F ,P ⎝ ∂P ⎠ p , f ,F
(9)
⎡⎛ − p ⋅ P ⎞ 2 ⎛ − p⋅P ⎞
2
⎛ P ⎞
2
⎛ p ⎞ ⎤
2
V R = ⎢⎜⎜ ⎟ V f + ⎜⎜ 2
2 ⎟
⎟⎟ VF + ⎜⎜ ⎟⎟ V p + ⎜⎜ ⎟⎟ VP ⎥ ⋅ 1002
F⋅f ⎠ ⎝F ⋅f ⎠ ⎝F⋅ f ⎝F⋅ f
⎣⎢⎝ ⎠ ⎠ ⎦⎥
(10)
When inventory changes are to be incorporated, over and above new metal bearing
material, the copper recovery is defined as follows:
p⋅P
R=
f ⋅F + ∑ ∆m a
i
i i
(11)
The net change in in-process inventory, per material type i, between month end and
beginning is:
(12)
(13)
The sum of the variances for the total inventory change (all material types) is then the
sum of the variances for the individual items (as they are statistically independent),
i.e.:
V∆maT = ∑Vi
∆mi ai
(14)
Estimation of Variances in Feed Masses and Assays each per Material Type
The calculation for each material treated in the smelter is in two parts:
∆Fk = M dkr + M ko − M kc
(15)
The variance in feed mass is simply the sum of the opening and closing stockpile
variances and the receipts variance, as the variances in opening, closing, and receipt
masses are statistically independent (However, independence is questionable if there
is a bias in stock measurement):
V∆Fk = VM r + VM o + VM c
dk k k
(16)
where the superscripts o , c and r refer to the opening stocks, closing stocks and
receipts respectively.
M k = Volk BDk (1 − mf k )
(17)
The variance of the dry weight of stockpile Mk is defined by the following equation :
2 2 2
⎛ ∂M k ⎞ ⎛ ∂M k ⎞ ⎛ ∂M k ⎞
VM k = ⎜⎜ ⎟⎟ VVolk + ⎜⎜ ⎟⎟ VBDk + ⎜⎜ ⎟⎟ Vmf i
⎝ ∂Volk ⎠ BDk ,mf k ⎝ ∂BDk ⎠Volk ,mf k ⎝ ∂mf k ⎠ BDk ,Volk
(18)
(19)
M dkr = (1 − mf k ) ⋅ M wkr
(20)
where Mdkr is the dry weight of the shipment, Mwkr is the as-received (wet weight) as
measured on a weigh bridge and mfk is the moisture fraction as previously defined.
The corresponding variance of the dry weight of the receipt is defined by the Eq.19:
2 2
⎛ ∂M dkr ⎞ ⎛ ∂M dkr ⎞
VM r = ⎜⎜ ⎟⎟ VM r + ⎜⎜ ⎟⎟ Vmf k
⎝ ∂M wk ⎝ ∂mf k
r
⎠ mf k ⎠ M wkr
dk wk
(21)
VM r = (1 − mf k ) ⋅ VM r + (M wk ) ⋅Vmf i
2 r 2
dk wk
(22)
Average assay and associated variances for one material type over month
duration
The check in – check out system of accounting defines the average feed assay fk for
one material type as follows:
akr M dkr + ako M ko − akc M kc nett mass of copper fed of material type k during a month
fk = =
M dkr + M ko − M kc total mass of material type k during a month
(23)
2 2 2 2 2 2
⎛ ∂f ⎞ ⎛ ∂f ⎞ ⎛ ∂f ⎞ ⎛ ∂f ⎞ ⎛ ∂f ⎞ ⎛ ∂f ⎞
V f k = ⎜⎜ kr ⎟⎟ Va r + ⎜⎜ ko ⎟⎟ Va o + ⎜⎜ kc ⎟⎟ Va c + ⎜⎜ kr ⎟⎟ VM r + ⎜⎜ k o ⎟⎟ VM o + ⎜⎜ k c ⎟⎟ VM c
⎝ ∂a k ⎠ ⎝ ∂a k ⎠ ⎝ ∂a k ⎠ ⎝ ∂M dk ⎠ ⎝ ∂M k ⎠ ⎝ ∂M k ⎠
k k k dk k k
(24)
The partial differentials with respect to mass and assay are illustrated by the
following:
⎛ ∂f k ⎞ M dkr
⎜⎜ r ⎟⎟ = r
⎝ ∂ak ⎠ M dk + M k − M k
o c
(25)
A copper smelter usually treats more than one concentrate type and in some cases a
variety of mattes and purchased scrap as well. The monthly average assay, as fed to
the furnace (i.e. the equivalent of all stockpiles combined and fed via belt conveyor
into the furnace over a one-month period) can be defined as follows:
∑ f ∆F k k
f = k
∑ ∆F k
k
(27)
F= ∑ ∆F
k
k
(28)
2 2
⎛ ∂f ⎞ ⎛ ∂f ⎞
Vf = ∑ k
⎜⎜
⎝ ∂f k
⎟⎟ V f k +
⎠
∑
k
⎜⎜
⎝ ∂∆Fk
⎟⎟ V∆Fk
⎠
(29)
The partial differentials with respect to mass and assay are illustrated by the
following:
⎛ ∂f ⎞ ∆Fk
⎜⎜ ⎟⎟ =
⎝ ∂f k ⎠ ∑ ∆F k
k
(30)
n n
⎛ ∂f ⎞
fk ∑ ∆F − ∑ f ∆F k k k
⎜⎜ ⎟⎟ = k =1 k =1
⎝ ∂∆Fk
2
⎠ ⎛ n
⎞
⎜
⎜
⎝
∑
k =1
∆Fk ⎟
⎟
⎠
(31)
So that the variance in the average feed assay, including the effect of stockpiles is
given by:
2
⎛ n n ⎞
⎜
∑ ∑ ⎟
2
⎛ ⎞ ∆Fk − f k ∆Fk ⎟
⎜ ∆F ⎟ ⎜ fk
Vf = ∑∑ ⎜
⎜
k
∆Fk
⎟ Vf +
⎟ k ∑ ⎜
⎜
k =1
⎛ n
k =1
⎞
2
⎟ V∆F
⎟ k
⎜ ⎟
∑
⎜ ∆Fk ⎟
k k
⎝ k ⎠ ⎜ ⎟
⎜ ⎜ ⎟ ⎟
⎝ ⎝ k =1 ⎠ ⎠
(32)
In the case considered here the product blister is weighed on an assized weighbridge
with an accuracy of ±20kg in 100 tones. Any weight measurement on this bridge has
a relative standard deviation (RSD) is 0.01. Assuming a normal distribution of
measurement uncertainties, the standard error can be said to be ±1.96×σ, within a
95% confidence interval (Kreyszig, 1970).
∑ (m − m )
j =1
2
Where: σ =
n −1
(33)
The RSD was calculated by analysis of assay results from multiple spoon samples
taken from the same ladle during blister copper pouring. The RSD was determined
as 0.047.
Although the RSD for copper in discard slag was not directly measured, work done
by Eksteen et al. (2004) as well as within the AMIRA P754 project suggests that the
RSD is likely to be in the range 15 - 20% for metal/matte elements, where the
element/species can be entrained (e.g. Cu in this case). The RSD is much smaller for
dissolved metal losses than for entrained metal losses in slags. Due to the difficulty
of achieving equiprobable sampling for melts, there is a significant risk of introducing
bias due to the sampling process. This bias is also difficult to assess as finding the
“true” value is not economically viable. There is therefore a very good reason for
using feed and product assays only, rather than including waste materials (such as
slags and dusts). However, the inclusion of waste does aid in material balancing
(using data reconciliation), but only if the composition can be determined without
significant bias.
Worked Example
The base data for the following cases is July 2005 actual data for a Southern African
copper smelter A. For confidentiality reasons the sources of the materials treated are
not given.
Recovery and Variance Calculation from Assays using the 2-product formula:
R = 97.32%
and therefore the σR = 0.55, and the standard error with 95% confidence limits =
±1.1%
From this calculation the recovery at the 95% confidence level will be 97.3% ± 1.1%.
In this case the recovery variance is dominated by the variance in copper in the
discard slag with a small contribution coming from the variance in the average copper
in feed mix.
Then by substitution:
R = 100.1%
From this calculation the recovery at 95% confidence limits is: R = 100.1% ± 5.2%
Input Data
∑ ∆m a
i
i i = copper in inventory change: 80 mt
Then by substitution :
R = 97.2%
and σR = 2.6
For this case the inventory variance has some significance but the feed assay
variance is still dominant.
Discussion of Results
This work was undertaken to try and understand why the recovery numbers and
unaccounted losses in smelter A varied by up to 5% in any month accounting period
when the process streams and recycles were accounted for and the flowsheet was
closed except for discard slag.
In the case of the mass estimates the variances in stock pile volumes, bulk densities
and moisture levels can be established in the field. In the case of assays the
sampling and analytical errors are reasonably well known. The major contributor to
the feed assay variance is the assay variability between lots and within lots which
comes back to the performance of the concentrator at the mine. The initial
assumption in this analysis was that the variance in copper assay between lots of
nominally 400 tones would be representative of the variance within a lot. This was
checked using 10 sequential lots of the individual concentrates and 10 wagons within
a lot. The results are shown in Table 1.
Base % % % %
Comp Case Cont Case 1 Cont Case 2 Cont Case 3 Cont Case 4 %
Vf
4.72 72 1.90 50 4.32 93 1.80 85 5.63 68
V∑ ∆mi ai
i 0.27 4 0.27 7 0.27 6 0.27 13 0.27 3
• The Base Case is the actual case as observed for a given month for smelter
A.
• Case 1 assumes that the copper assay variance for concentrate A was
applicable for all concentrates
• Case 2 assumes that all concentrates received during the month were treated
during the month
• Case 3 is the “best” case where there are no opening and closing stocks and
the assay variance for concentrate A applies to all concentrates
In all cases the contributions of variance of both the product weight (blister copper)
and the product assay were insignificant, due to the high degree of precision.
Conclusions
The following conclusions could be drawn from the analyses presented above.
• A variance propagation model has been formulated for a copper toll treatment
smelter. This variance propagation model took into account changes in
stockpiles, as well as changes in inventories.
• It was determined that, despite the 2-product formula being more precise (i.e.
smaller variance in recovery estimate), it is inherently less accurate, due to
challenges associated with the sampling of melts, dusts and gases in cases
when these become the “tailings”, or waste of a smelter.
• The assay variance in feedstocks was always the largest contributor to
variance in recovery, and was always greater than the contribution due to
variances in net masses fed (which includes stockpile and inventory
changes).
• The large observed variances in recoveries could be explained through the
variance propagation model.
• The actual observed case (Base Case) showed that the standard deviation in
the recovery estimate was 2.6%. The standard error in the monthly recovery
estimate is therefore +/- 1.96 × 2.6% = ±5.1% with 95% confidence.
• There is therefore a clear business case to improve sampling practice and
instrumentation on this specific toll treatment smelter. However, the most
effective way to reduce the recovery variance is to reduce the assay
variability between stockpiles. Subsequent to the reduction in assay
variability, is the reduction in the variability in stockpile masses due to
variability in volumes, bulk densities and moisture fractions. Good control over
blending might go a look way in achieving this goal, i.e. variation between
material type assays can be reduced through proper blending based on
known assays of all material types being blended. However, doing this well
usually requires a major capital investment.
• Bulk density variation can be reduced through the use of a narrow particle
size distribution, volume variances can be reduced through improved
surveying techniques and the variance in moisture fractions can be reduced
by, for example, shielding the stockpiles from environmental changes in
moisture (e.g. using a shed).
The AMIRA P754 program is aimed at delivering a Code, a set of Guidelines and a
set of Toolboxes for Metallurgical Accounting and Reconciliation with the aims of:
Eksteen J.J., Frank S.J., Reuter M.A., Distributed compositional and temperature
nature of melts in submerged and open arc furnaces in high carbon ferrochrome
production, Proceedings of the 10th International Ferroalloys Congress (INFACON X),
1-4 February, Cape Town, South Africa, pp. 83 -93, 2004
Narasimhan, S., and Jordache, C. Data Reconciliation and Gross Error Detection: An
Intelligent use of Process Data, Gulf Publishing, Houston, Texas, 406 pp. 2000
Romagnoli, J.A., and Sánchez, M.C., Data Processing and Reconciliation for
Chemical Process Operations, Process Systems Engineering, Vol. 2, Academic
Press, San Diego, California, 270 pp., 2000
Seinfeld, J.H., and Lapidus, L., Mathematical Methods in Chemical Engineering, Vol.
3 (Process Modeling, Estimation and Identification), Prentice-Hall, Englewood Cliffs,
New Jersey, 545 pp, 1974.
Xiao, Z., Vien, A., Limitations of variance analysis using propagation of variance,
Minerals Engineering, Vol. 16, pp. 455–462, 2003
AOAC – www.aoac.org
APLAC – www.ianz.govt.nz/aplac
ASME – www.store.asme.org
ASTM – www.astm.org
CITAC – www.citac.org
DIN – www.normas.com/DIN/indexes/
EURACHEM – www.eurachem.org
IEC – www.iec.org
ILAC – www.ilac.org
ISA – www.isa.org
ISO – www.iso.ch
JIS – www.jis.org
NATA – www.nata.asn.au
NIST – www.nist.gov
OIML – www.oiml.org
SABS – www.stansa.co.za
SANAS – www.sanas.co.za
UKAS – www.ukas.com
VAM – www.vam.org.uk