You are on page 1of 8

Selection and Design Of Blasting Techniques for Selective

Open Pit Gold Mining using Knowledge Based Systems Technology


J JIANG1 and T N L1TTLE

ABSTRACT
The Western Australian School of Mines is continuing to investigate the
blasting techniques and mining strategies that are used for ore/waste
separation (or high grade/low grade ore separation) or to prevent ore and
waste mixing.
The aim of all the mining strategies is to produce an optimum product.
The optimum prpduct is one which contains no diluent and undergoes no
mixing of the various identified ore and waste blocks and is fully
recovered at the minimum overall operating cost.
The current study includes the use of computerised Knowledge Based
Systems (KBS) technology to assist the user with the blast technique
selection and blast design process. These aspects are highly dependent on
the geology, (geometry; rock mass properties, grade distribution),
explosive performance and the expectations of the mining company in
terms of production rates and equipment utilisation.
Most of these factors are not well defined and nor are they easy to
quantify. However, a large amount of knowledge does exist and is
available in the following forms: rules of thumb, empirical formulae,
engineering judgement and past experience. KBS techniques are well
adapted to handle such'forms of knowledge.
The primary objective of this paper is to illustrate an approach for the use
of Knowledge Based Systems during the analysis and interpretation
stages of the blast design process. In particular, the components and
methodology of the developed system aFtl described for selective open pit
gold mining. The Knowledge Based System developed was written using
the declarative programming language PROLOG.

INTRODUCTION
In many open pit gold mining operations the delineation of
blocks of differing grade and the ability to mine those blocks
separately is of prime importance to the profitability of the
operation.
This type of mining is termed selective mining and the
approaches used for the control of the grade are called grade
control techniques.
Grade control techniques are used to define ore blocks, predict
recoverable reserves and to enable reconciliation of grade and
tonnages to be made between calculated and actual values.
In order to recover the defined ore blocks with minimum
dilution and ore loss certain limitations need to be placed on the
blasting and loading operations.
The two major limitations imposed on the blasting are to
prevent ore and waste (or high grade and low grade) mixing and
to obtain the required fragmentation. In order to realize the above
we need to select the appropriate blasting strategy and refme the
design until optimum performance and product is achieved.
The current research is an extension to some initial research
work that has been done at the Western Australia School of Mines
(Little and van Rooyen 1987, Humphryson 1989, Federici
1989).
1 Visiting Research Scholar, Western Australian School of Mines
2 Research Supervisor and Acting Head of Department ME&MS WASM.

FRAGBLAST '90

This paper starts with a brief overview of the selective and bulk
blasting strategies used in gold mines in Australia. This is
followed by a brief discussion on the applicatiop of KBS and
Expert Systems technology to the field of blasting.
The main part of the presentation then focuses on the
objectives and' implementatioJ!. of the KBS currently under
development
Finally the current KBS is critically reviewed and the paper
concludes with a discussion of our future research plans and
direction.
.

GOLD MINING STRATEGIES


Blasting strategies
Various innovative .mining strategies and blasting techniques
have been devised to cope with the complex geometry and grade
distribution exhibited by many Australian gold deposits. The
blasting strategies can be categorised according to whether a
horizontal free face (paddock blasting) or a roughly vertical free
face (bench blasting) is utilised. A further distinction can be
made between blasting techniques which collectively blast
various ore and waste blocks (bulk blasting) and those which
blast only single blocks (high grade, medium grade, low grade or
,waste blocks) at anyone point in time and space.
Table 1 summarises the four basic strategies.

TABLE

Categories ofGold Mining Strategies


Blasting
Strategy

Orientation
Free Face

Blasting
Blocks

Digging
Requirements

Bulk Paddock Horizontal

All blocks
in area shot

Mark out then


Selectively

Selective
Paddock

Horizontal

Single grade
blocks shot

Clean up then
Non-Selective

Bulk Bench

Sub-Vert

Full bench

Mark out then


Selectively

Selective
Bench

Sub-Vert

Single grade
blocks shot

Clean up then
Non-Selective

The above blasting strategies represent four clearly definable


methods. In practic~ mine operators may use anyone technique
exclusively or may use a combination of two or more hybrid
techniques to suit the situation at hand. For example some
operations which use bench blasting (sub-vertical free face) also

Brisbane August 26 - 31

391

JJIANGand TNUITLE

use in excess of ten rows and design the blast for minimum
forward movement are in effect using many of the principles of
paddock blasting.
Other modifications to the basic four methods are routinely
undertaken when the impact of the mixing of adjacent blocks is
low, for example when:
both high grade and ROM blocks are adjacent
low and high grade blocks are dug selectively and the
boundary material stockpiled for later re-sampling.

Recovery and dilution


For gold mining maximum ore recovery is of paramount
importance for all'types of orebod,ies. Limiting the amount of
diluent material in the mill feed becomes increasingly important
for low grade. marginal deposits. The actual mill head reserve
(grade and tonnage) is influenced by both geological and mining
considerations. The estimated recoverable reserve depends on
the quality of the initial in-pit reserve estimation work and the
quality of the mine planning (estimates of ore loss and dilution)
and grade control activ.ities. The actual difference between the
recoverable reserve estimate and the mill head reserve is the
actual amount of ore losses and dilution introduced during the
blasting and loading activities, assuming no transport losses
occur.
During the mining operation four phenomena contribute to ore
losses and dilution gains. These are mixing at boundaries due to
blasting, incorrectly marking out ore boundaries (shifted an
unknown direction and amount), loading operation inaccuracies
and fly rock losses and gains.

KNOWLEDGE BASED SYSTEMS AND BLASTING


Research into blasting is hampered by the extremely hostile
environment close to the detonating biasthole and the enormous
variability of geological materials. Hence many areas of open pit
blast design are empirical in nature. That is to say, that a large
amount of knowledge and empirical formulae are available.
Experienced persons and researchers do have some judgement
and heuristic rules to enable then to und~ake such blast designs.
Knowledge Based Systems are sophisticated, interactive
computer programs which use all fornls of knowledge in some
narrow problem domain to solve a complex problem in that
domain. KBS have also been referred to as expert systems,
intelligent assistant and analysis sYstem.
In the last decade, some expert system tools, or shells, have
been developed for applications for use ill some special areas. As
expert system shells are easier to use than developing an expert
system from scratch, shells are widely accepted.
As initial investigations of a few expert system shells revealed
considerable limitations, it was decided to write a KBS using the
declarative programming language PROLOG (specifically Turbo
Prolog 2.0 from Borland) which can run on an IBM/AT or
compatible computer.
Knowledge Based Systems have been applied to civil
engineering for some time and the ASCE sponsored conference
titled, "First Symposium on Expert Systems in Civil
Engineering", was held in Seattle, Washington in 1986. The
proceedings of this symposium indicate that a wide range of
applications have been found. Some researchers (Ramani and
Prasad 1987, Bodkin 1988), have identified domains in J11ining
engineering where the application of KBS could be beneficial,

392

and include mine ventilation and ground control in underground


mining.
A group of researchers at Ohio University, (Scheck, Chatterjee
and Wei Sun 1987, Scheck 1988), have developed a surface mine
blast design and consultant system to help with blast design and
blast vibration analysis. The current system under development at
WASM differs considerably in concept and form from the Ohio
University system and was developed totally independently of
that system. The next section of this paper outlines the current
WASM system development in terms of objective and
developmental status.

SYSTEM DEVELOPMENT
Objectives
The developers of the current KBS anticipate two distin<;:t
functions and user groups for the system:
The overall objective of the KBS, then, is to provide advice on
blasting during both the pre-production and operating stages of a
project. Thus the KBS can be used as planning tool during the
feasibility stage of a project, and provide valuable information for
equipment selection (drilling, loading and crushing).
Alternatively the KBS can provide a friendly and seemingly
intelligent adviser to assist a shotflrer or blasting engineer to
obtain a blasting strategy and blast design which achieves the
required product (size and recovery) and at the lowest cost.

Developmental stage
To develop a KBS a number of development stages need to be
undertaken~ The development process can be summarized by the
following nine steps (Finn and Reinschmidt 1986):
1.

Preliminary functional specification

2.

Selection of prototype tools

3.

Prototype development

Evaluation of prototype

5.

Revised performance specification

6.

Selection of operational tool

7.

Deve~opment of operational product

8.

Feedback from users

9.

Enhancements (go back to step 5.)

The prototype KBS under development is currently undergoing.


evaluation and revisions to the specification are being formulated
i.e. steps 4 and 5. The next section deals with the KBS design and
following that section a discussion of the limitations and revised
performance specifications are given.

Source .of information


Our knowledge base is derived from three sources:

1.

On-going research, including blast vibrations, airblast and


explosive performance monitoring and practical shotfiring
work by the researchers and others at WASM.

2.

Published
technical
specifications.

Brisbane August 26 - 31

literature

and

manufacturer

FRAGBLAST '90

BLASTING TECHNIQUES FOR OPEN PIT GOLD M!N.ING

3.

A detailed survey (Humphryson 1989) involying twenty


parameters for 23 gold mines in the Eastern Goldfields of
Western Australia.

SYSTEM DESIGN
The system can perform common arithmetic operations and
accommodate natural language descriptors as input for
information that is either hard to quantify or is uncertain.
Weighting factors and confidence indicators are used in the
system for decision making.
Three kinds of user input are employed. These are:

1.
2.

3.

The current avail~ble explosive information is stored in the


data base, the information includes: explosive type; product
name; water resistance properties; relative weight strength;
velocity of detonation (VOD); explosive density; unit price and
other properties. The data base can be updated at any time.
The rule base consists of six sub-sections as schematically
represented in Figure 2.
SELECTION OF
BLASTING STRATEGY
BLASTING
DESIGN

Boolean inputs are used for simple decisions which require


the user to give yes or no answers.

FRAGMENTATION
ESTIMATION

Numerical inputs are used for quantifiable data concerning


particular parameters, for example annual production rate - 2
Mtpa. Only realistic parameter values are accepted.

1.

Blasting strategy

2.

Blast design

3.

Prediction ofrock fragment size distribution

4.

Prediction of ore loss and dilution

5.

Unit cost

The system can be used for parameter sensitivity studies and


interactive optimisation can thus be performed.

System structure
The system consists of three major components; a user interface,
an inference engine and a knowledge base. The knowledge base
can be further divided into a database and a rule base, as shown
in Figure 1.

USER
INTERFACE

FIG. 1. Anatomy of system

FRAGBLAST '90

Us.r~ Decision FinAl

ORE LOSS AND DILUTION


ESTIMATION

N aturallanguage input for subjective and qualitative ~wer~


to system queries, for example orebody outline - very weak.

For some user input more than one form of input can be
accommodated. For example to calculate a blasting index the
uniaxial cornpressive strength and Young's modulus is prompted
for, if this information is not available to the user the system
further prompts for a hardness descriptor.
Another system attribute is its user friendliness including the
"why" and "how" question capabilities. The system can provide
explanations, input suggestions and answer ranges for particular
input parameters.
The KBS has been developed so that the user makes the final
decision as to which strategies and design parameters are further
processed. For some decisions, the col)fidence level is indicated,
this provides the user additional information if he decides to
over-rule the advice given by the system.
Five categories of output are used, these are: .

U5.r~ D~cision FinAl

COST

ESTIMATION
SENSITIVITY
STUDIES (OPTIMISATION)

FIG. 2. Structure of Rule Base

The first sub-section is concerned with the selection of the


mining strategy (blasting and loading). Four c~tegories are
recognised, see Table 1. All subsequent design and.,prediction is
influence by the selected mining strategies. TIle user can
over-rule the KBS advice at this stage if required, for ~xample to
accommodate an existing practice.
'!.'he second sub-section deals with blast design. Based on the
mining strategy selected above, a method specific design is
interactively formulated. The design output includes: explosive
type; powder factor; burden; spacing; sub-drill depth and
stemming column.
The next sub-section predicts fragmentation. Based on the
above blasting method and design, a particle size distribution is
estimated. The user can then judge whether the required degree of
fragmentation is achieved. If this is not the case the user can
interactively modify the blast design until satisfactory results are
obtained. During the modification process the system gives useful
advice.
The fourth part deals with mining recovery. The ore losses and
dilution that occur during mining are mainly dependant on the
blasting method, blast design and loading practices. If the ore loss
and dilution percentages are not satisfactory, 'the user can modify
the blast design to achieve the desired result.
The fifth sub-section deals with cost information. The cost
includes: drilling; charging; explosives and blasting accessories;
secondary blasting and loading.
The final sub-section can be used to undertake sensitivity
analyses, with the aim of optimisation. In this block, the user can
change one or more blast design parameters, or some other input
such as oversize limit, bucket capacity and some cost
information. The new result includes fragmentation, ore loss,
dilution and cost. Comparing the different results, the user can
find the proper decision and design information.
More detailed information concerning the system developed is
presented in the following sections.

Brisbane August 26 - 31

393

J JIANG and TN LITTLE

Input requirements
The information regarding explosive prqperties is buil,~. into the
system as a data base. The system can retrieve data from the data
base when explosive information is needed.
Five other categories of input are interactively solicited from
the user:
1.

Mine requirement

2.

Orebody geology and geometry

3.

Rock mass properties

4.

Equipmentinformation

5.

method. For example, the membership function for production


rate (P) on selective blasting (SB) methods can be expressed

as:

{ 1 P<O.05
{l-(P-0.05)j(2-0.05) 2>10.05

SB(P)

{012
A weighting system is used to express the relative importance
of each attribute, eg production rate, to the selection process.
Thus the index for a particular blasting method is estimated using
an equation of the form:
n

(Blast Method Index)i

Cost information.

= 2, Xij . Wj
j=l

Selection of blasting strategy


Selective mining methods aim to prevent the mixing of ore and
waste. (or high grade and low grade). Thus the products of
selective mining blasting (paddock or bench) are less tonnes of
ore at higher grade and higher tonnages of waste, compared with
bulk blasting products. However bulk (or non-selective) blasting,
can achieve higher production rates than selective blasting.
Furthermore, for low bench heights, paddock blasting can
achieve higher production rates than bench blasting.
The selection of the appropriate mining strategies is based on the
mine production rate, number of ore bodies, grade, orebody
outline and the bench height. The bench height is selected by the

KBS.
Some general rules used for the selection of the mining strategy
are as follows:
IF

grade high

THEN

minimum ore loss required,


preferred

IF

production rate high

THEN

bulk blasting method is preferred

IF

production rate low

THEN

selective blasting can be used

selective blasting

where:
n
is the number of attributes,
Xij
is the membership function of attribute j for method i,
Wj
is the importance (weighting) of attribute j to method i.
The method with the highest Blast Method Index is
recommended as the appropriate blasting strategy. The magnitude
of the respective Blast Method Indices is used to express the
reliability of the decision.

Blast design
Blasting design includes the selection of explosive type, the
quantity of explosive, the burden and spacing dimensions, the
depth of subdrill, and the length of stemming column.
To reflect the difference between bench and paddock blasting,
design modifications to the base case (bench blasting) have been
incorporated in the system, see Table 2.

TABLE

Design Parameter Modification

Bench Blasting

Paddock Blasting

Burden

stemming +( 0.5 x
charge column)

Pattern

BxS

SxS

Sub-drill

Sub

KsxSub

PF

KpxPF

IF

production rate greater than 2 Mtpa

THEN

bulk blasting is recommended

IF

number of orebodies is high

THEN

paddock blasting is preferred

Powder
Factor

IF

strong ore outline

The values adopted for Ks and Kp are between 1.1 and 1.3.

THEN

the easier it is to blast to that outline and hence


selective blasting is preferred

IF

bench height is high

THEN

bench blasting is preferred

These general rules are formulated into membership functions


in order to enable the selection of the most appropriate blasting
394

Fragmentation
In order to concentrate on system development eX1S1;lIlg
blastability index and fragmentation models have been
incorporated.
The Kuz-Ram model (Cunningham 1983) is employed to
estimate rock fragment size distribution due to blasting. The
most sensitive parameter for the Kuz-Ram model is the rock
factor. To determine the rock factor a blastability index (BI) based

Brisbane August 26 - 31

FRAGBLAST '90

BLASTING TECHNIQUES FOR OPEN PIT GOLD MINING

on the work of Lilly (1986), has been modified to reflect values


normally encountered when gold mining. The modified rock
factor calculation follows:
A = K(RMD+JPS+1PO+SOI+b(UCS*y)1/1)

If UCS and/or Y are not known the KBS prompts for a rock
hardness (H).

Ore loss and dilution


The developed KBS only considers muung related factors
contributing to ore losses and dilution gains. For ore loss and
dilution due to blast design the following general rules have been
included:

IF

selective blasting stralegy is employed

THEN

the integrity of ore/waste interfaces can be retained


leading to lower ore losses and dilution gains

IF

pre-splitting control blast is used

THEN

horizontal movement can be restricted to a certain


extent leading to lower ore losses and dilution gains

IF

small burdens and spacings are employed

THEN

less mixing of ore and waste leading to lower ore


losses and dilution gains

TABLE 3

Definition ofthe Symbols Used

RMD

Rock Mass Description. 9 levels from extremely


powdery to extremely massive, value range 10-50.

JPS

Joint Plane Spacing. Spacing from close ( <0.05 m)


to wide (> 0.5 m), value range 10-50.

JPO

Joint Plane Orientation. In degrees from 0 to 180


degree, value range 10-40,see Figure 3 and 4.

SOl

Specific Gravity Influence, SOI=45xSG-45.

UCS

Uniaxial Compressing Strength, MPa.

IF

loading bucket is small

Young's modulus, GPa.

THEN

modification coefficient

more accurate control on loading can be achieved


leading to lower ore losses and dilution gains

rock hardness scale, trom 1-10.

IF

stemming column short

It is
modification coefficient for rock type.
dependent on rock structure, properties and the rock
response to blast loading. The value 0.05 is used for
goldrnines in Western Australian.

THEN

lead to higher fly rock ore losses and dilution gains

Dip out of FAC. 2tI

I
\

Sub-V.rUCAl
Fr.. FAc.

HorizontAl 1.

I
\:1
1:\

Dip into FAC. 4tI

38
PArAll.l to FAc.

Floor

FIG. 3. Relationship of Joint Orientation and JPO for Bench


Blasting

Horizontal F'r ..e


\

The minimum mining zone, or mining accuracy is dependant


on the bucket capacity and width, and the skill of the operator.
Hence for loading the follow general rules have been used:

IF

clean mining strategy adopted

THEN

average ore loss equals 0.5 times mining accuracy


and dilution due to loading is zero

IF

mining along the ore/waste interface

THEN

average ore loss equals 0.2 times mining accuracy


and dilution gain equals 0.2 times mining accuracy

IF

mining exclusively outside the orebody

THEN

average dilution equals 0.5 times mining accuracy


and ore loss due to loading is zero

Brisbane August 26 - 31

395

::-.c:e

1
\

1
\

30

1
\:1
1:\

Parallel to
Free Fac.

40

With regard to loading, three strategies are available in respect


of mining near ore/waste interfaces. The first stralegy is to mine
exclusively within the orebody, see Figure 5(a), this can be
termed clean mining and is not generally appropriate for gold
mining. The second strategy involves excavating to a pre-defined
boundary, see Figure 5(b), this method is commonly used at gold
mines and gives rise to approximalely equal quantities of ore loss
and dilution. The final stralegy involves mining exclusively on
the outside of the ore/waste interface, see Figure 5(c), and aims at
maximising ore recovery at the expense of dilution.

:10

\ 40

FIG. 4. Relationship of Joint Orientation and JPO for Paddock


Blasting

FRAGBLAST '90

J llANG and T N UITLE

WAst.

+ (Area4 x Dilution4 + Area5 x DilutionS)] x Dw

1111111111111111111111111111111111
1111111111111111111111111111111111

..
Dilution (Tw)
Percent DilutIOn = Ore in Blast (To) x 100%

.1019 Accuracy

or.

Mill Head Reserve (Tm

(A)

Gm>

=Total Reserve in Blast (To x Go)


- Ore Loss (T01 X Go)

+ Dilution (Tw x G w)
Mill Head Tonnage (Tm) = To - To! + Tw

(b)

Mill Head Grade (Gm> = To x Go - Tol


-----'_-L.-----JL--L.-----JL--L----JL-....J

or.

.in1no accuracy

Tm

Go.+ T w x G w

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

where Do and D w are the density of ore and waste respectively;


Go and G w aie the grade of ore and waste respectively.

(c)

FIG. 5. Loading Strategy and Mining Accuracy

DISCUSSION AND CONCLUSION

Four individual ore loss and dilution prediction models


corresponding to the four blasting strategies have been
developed. An example of the predictive model for selective
paddock blasting is illustrated as Figure 6.
The calculation of percentage ore loss and dilution and mill
head grade follows:
Ore Loss (To!) =
[(Length1 x Loss 1 + Length2 x Loss2) x BenchHeight
+ Area3 x Loss3 + Area4 x Loss4 + AreaS x Loss5] x Do
Ore Loss (T01)
Percent Ore Loss = Ore in Blast (To) x 100%
Dilution (Tw) =
[(Length1 x Dilutionl + Length2 x Dilution2) x BenchHeight

The KBS approach to blasting method selection and blast design


has many advantages over more manual, less structured
approaches.
The KBS approach:

is easy to use
is very quick
allows "what if' questions
is easily incrementally extended

is able to reason with uncertain data


is able to explain its train of reasoning
allows developers to easily add new rules
allows developers to easily modify existing rules.
The current system under development is ambitious in that it
attempts to select the appropriate blasting strategy and predict
fragmentation, recovery and cost. Hence the KBS can be used as
planning tool during the feasibility stage of a project, and provide

SELECTIVE
PADDOCK
BLASTING

I
Pre-split
blastin.g
length
Length!

I
Loading
method
Loading
equipment
Pattern
Loss!
Dilution!

Standard
blasting
length
Length2

I
Loading
method
Loading
equipment
Pattern
Loss2
Dilution2

Free
face
area
Area3

I
Hanging
wall
area
Area4

I
Pattern
Stemming
Loss3

I
Foot
wall
area
AreaS

Loading
method
Loading
equipment
Pattern
Loss4
Dilution4

Loading
method
Loading
equipment
Pattern
LossS
DilutionS

FIG. 6. Information Flow for Ore Loss and Dilution

396

Brisbane August 26 - 31

FRAGBLAST '90

BLASTING TECHNIQUES FOR OPEN PIT GOLD MINING

valuable information for equipment selection (drilling, loading


and crushing). Alternatively the KBS can be used as mining
engineers adviser for production blasting.
The limitations of the current prototype KBS at this stage of
development are as follow:
1.

The initiation sequence influence on fragmentation and


recovery is not included.

2.

The knowledge base in some areas is inadequate and


requires further research effort.

3.

The grade control aspects such as incorrect marking out of


ore blocks and its influence on recovery has not been
included.

4.

The system inputs equipment information. A more flexible


approach is to allow equipment requirements to be either an
input or an output (planning tool).

The only conclusion to be drawn from the current work is that


this KBS approach to grade control blasting has great potential
and research work is continuing.

REFERENCE
Bodkin, K.E. 1988. Expert Systems for Colliery Problems, Colliery
Guardian, July, p 216-219.

FRAGBLAST '90

Cunningham,. C. 1983. The Kuz-Ram Model for Prediction of


Fragmentation from Blasting, First International Symposium on Rock
Fragmentation by Blasting, Lulea, Sweden, August, p 439-453.
Federici, P. 1989. Blast Design and Expert Systems, Jpdustrial Project,
WASM.
Finn, G.A. and Reinschmidt, K.F. 1986. EXP,!lrt System in an
Engineering-Construction Firm, Proceedings of Symp, Expert Systems
in Civil Engineering, ASCE, New York, p 40-54
Humphryson,R. 1989. Open Pit Blasting and Grade Control Practices for
Gold Orebodies, Industrial Project, WASM.
Lilly, P.A. 1986. An Empirical Method of Assessing Rock Mass
Blastability, Large Open Pit Mining Conference, AusIMMJIEAust
Newman Combined Group, October, p 89-92.
Little, T.N. and van Rooyen, F. 1988. The Current State of the Art of
Grade Control Blasting in the Eastern Goldfields, Explosive in Mining
Workshop, AusIMM, Melbourne Victoria, November, p 87-95.
Scheck, D.E., Chatterjee, P.K. and Wei Sun, M.S. 1987. Surface Mine
Blast Design and Consultant System, Proc of the Twentieth Int Symp
on the Application of Computers and Mathematics in the Mineral
Industries. Volume 1: Mining. Johannesburg, SAIMM, p 181-188.
Scheck, D.E. 1988. Expanded Expert System for Blast Design, Mine
Planning and Equipment Selection, Singhal(ed.), Balkema, Rotterdam,
p361-367.
Ramani, R.V. and Prasad, K.V.K. 1987. Application of Knowledge Based
Systems in Mining Engineering, Proc of the Twentieth Int. Symp. the
Application of Computers and Mathematics in the Mineral Industries.
Volume 1: Mining. JohannesbUIg, SAIMM, p 167-180.

Brisbane August 26 - 31

397

398

Brisbane AugUSI26 - 31

FRAGBLAST '90

You might also like