You are on page 1of 5

SME Annual Meeting

Feb. 23-25, Denver, Colorado

Preprint 04-12
PROCESS MINERALOGY – A NEW GENERATION FOR ORE CHARACTERIZATION AND PLANT
OPTIMIZATION

W. Baum
Phelps Dodge Mining Co.
Safford, AZ

N. O. Lotter and P. J. Whittaker


Noranda/Falconbridge Ltd., Sudbury, ON, Canada

SUMMARY Also, these investigations often aim at improving the quality of the
flotation concentrate(s) in terms of reduced gangue dilution, because
The last twenty years of development in Process Mineralogy are the downstream process is often a pyrometallurgical one. These
reviewed. Since the milestone review by Henley in 1983, which pro- activities, when successfully performed, enhance the business case
jected the concept and suggested closer working relations between of the production concentrator to a significant extent [Henley, 1983].
mineral processing and mineral science, the second generation First Generation Process Mineralogy: 1983
(1983-2003) has made significant progress. Examples are to be Henley reviewed the relatively young discipline of Process
found in improved sampling and high-confidence flotation testing, Mineralogy in a milestone paper in 1983. This approach involves the
and in automated quantitative mineral measurement. The recent merging of the two disciplines, i.e. mineral processing and mineralo-
addition of end-member studies has provided a clearer view of the gy (or mineral science). In this review, a flowsheet from ore body
overall ore mixture milled from its constituent geological components. exploration to optimal plant operation was proposed. This flowsheet
In all, better predictive abilities have resulted from these improve- had the objective of developing a predictive process mineralogy
ments to Process Mineralogy. These improvements have led to more model for the ore body and concentrator in question. The early inputs
significant business impacts in mineral resource characterization, from mineral science, describing the bulk mineral assemblage, pos-
and in optimization of existing concentrator operations. sible liberation size ranges, and identifying problematic minerals,
were shared with the mineral processing engineer. The latter would
INTRODUCTION develop a flowsheet partly based on this information, and partly on
laboratory and pilot plant testing. Following commissioning of the
Milling and flotation, sometimes accompanied by gravity sepa- plant, an optimization program was pursued by a second (and sub-
ration, are well-established unit processes, which have traditionally sequent iterations) of this study. These iterations especially extend-
been used to treat ore that has been mined. This is because these ed beyond new samples of drill-core, into samples taken from the
processes liberate, then separate, the valuable minerals, which con- operating plant, which were studied by the two disciplines to identify
tain paymetals, from host gangue. This is often done by a combina- flow sheet limitations. Various mineral science technologies were
tion of crushing and grinding processes to liberate the paymetal min- suggested as suitable means for developing the information. These
erals, followed by gravity separation and/or flotation, to separate included X-ray diffraction, microprobe analysis, scanning electron
these valuable minerals into a concentrate, or several concentrates. microscopy, classical optical microscopy with point-counting and
These concentrate(s) have a lower bulk than the parent ore from image analysis techniques. The image analysis approach for miner-
which they are derived. Further, they contain a high percentage of al quantification was pursued by Petruk (2000), starting in the early
the valuable minerals. Accordingly, they are more economically 1970’s. This approach made use of an electron-probe microanalyser
treated in subsequent beneficiation processes than the ore from to generate backscatter electron images of minerals for measure-
which they were produced [Lotter,2002]. ment by an early image analysis system. The results of this work by
The choice of this concentration process invariably results in a Petruk, at what is now CANMET, were aimed specifically at the min-
variable paymetal loss to the concentrator tailing. This is discarded eral processing industry. The first commercially available electron-
and disposed of in a tailings dam, and has no immediate business probe microanalyser became available in the late 1950’s to early
use. This paymetal loss forms the largest loss of paymetal in the 1960’s and marked the beginning of quantitative mineral analysis.
entire flowsheet by which refined, or saleable, metal(s) are produced Such an instrument was used for locating and identifying minerals in
[Cramer, 2001]. an automated manner by Jones & Gavrilovic (1966) at Imperial
From the foregoing, the sampling and analysis of these produc- College, London. This technology provided the foundation for future
tion concentrators towards improved metallurgical performance, has development of the x-ray microanalysis approach to quantitative min-
logically been the object of technical effort. A history of these eralogy used by QEM*SEM. At the time of writing in 1983, Henley
endeavours exists [Cameron et. al., 1971, Restarick, 1976, Hartley noted the beginnings of the QEM*SEM development and suggested
et. al., 1977; Weller and Stearns, 1984;] These sampling exercises, a size-by-size structure for all of these analyses, in keeping with
sometimes called diagnostic metallurgy, or surveys, have aimed at Trahar’s models [Trahar, 1981]. Some discussion was offered on the
identifying and diagnosing segments, or areas, of the concentrator relative errors of point-counting. Henley’s review, however, did not
flowsheet which have potential for improved metallurgical perform- address the issue of representative sampling or desired confidence
ance. Implementation of improved processing in these flowsheet levels for the overall models.
areas reduces the paymetal loss to flotation tailings, and increases
the recovery of the paymetal(s) to gravity or flotation concentrate(s).

1 Copyright © 2004 by SME


SME Annual Meeting
Feb. 23-25, Denver, Colorado

Second Generation Process Mineralogy: 1983-2003 • In surveying an operating mill, the level of operator
Since 1983, some twenty years of development have seen sig- awareness of the survey tends to drive the concentra-
nificant advances in Process Mineralogy, in terms of sampling, flota- tor performance upwards; thus the benchmark survey
tion testing, mineral measurement techniques, and especially model- result probably represents less of the short-term
ling of the data. Baum et al. (1989) showed how continuous use of process variability than would normally be the case.
process mineralogy contributed to multi-million dollar plant improve- This sets the benchmark above what is more typical of
ments at the El Indio operation. the operation. The aim of the benchmark survey is to
The purpose of this paper is to update the progress made to measure plant performance whilst processing normal
date with the second generation, and to communicate to the mineral run-of-mine ore.
processing discipline the value of this approach.
FLOTATION TESTING
KEY PROBLEM AREAS
A problem equivalent to that of the concentrator survey sam-
• In order for meaningful diagnosis of whatever nature to pling used to exist at the laboratory scale flotation test level. Before
be made from the samples extracted from a concentra- one could meaningfully conduct a laboratory scale flotation test, it
tor survey, these samples have to be representative of was first necessary to prove that the sample of ore presented to the
the flowsheet streams or drill core from which they were flotation test was representative. Further, it was necessary to
taken. demonstrate that the flotation test results were reproducible. Earlier
• In order to better understand the actual behaviour of the work in South Africa addressed these problems. The work showed
minerals in the concentrator flowsheet, a quantitative that a fundamental approach to the ore sampling and sample prepa-
mineral measurement system is needed. Using chem- ration protocols prior to the tests reduced the fundamental variance
ical analysis to determine the paymetal content of the of the ore sample to 5%, with quantitative proofs of trueness [Lotter,
samples does not necessarily describe the behaviour of 1995a]. A series of replicated flotation tests with quantitative diagnos-
the minerals. It is the mineral content, size and tex- tics for laboratory scale flotation testing of platinum-bearing ores
tures, not the assays, which determine the mineral sep- improved the platinum accountability, and thus the reproducibility of
aration behaviour. results, across the flotation test [Lotter, 1995b]. Apart from defining
primary sampling protocols, this work measured and minimized the
The second of the above two problems has already been evaluation bias errors of PGE in specialised sample processing
solved. The advent of modern quantitative mineral measurement ahead of lead collection fire-assay [Lotter et.al., 2000].
technologies in the 1990’s has enabled accurate, quantitative meas- There are several interests in the type of information desired
urement of minerals in samples taken from production concentrators. from representative sample material. These samples are grouped
There are two known commercially available technologies. These into concentrator survey samples, and drill-core samples.
are Quantitative Evaluation of Minerals by Scanning Electron
Microscopy (QEM*SEM – and later, Qem*SCAN with light element CONCENTRATOR SURVEY SAMPLES
detection capability), manufactured by the Commonwealth Scientific
and Industrial Research Organisation, Australia, and the Mineral For the concentrator, these interests extend beyond the routine
Liberation Analyser (MLA), manufactured by the Julius Krutschnitt metal accounting of battery limit streams such as final concentrate(s),
Research Centre, Australia. In these two technologies, overall miner- final flotation tailings and flotation feed. A well-formulated ‘plant sur-
al assemblage and liberation of minerals are described at a sized vey’ wherein detailed sampling of internally recirculating or interme-
fraction level. Representative sampling is critical to obtaining mean- diate streams is often highly informative, requires far more than a
ingful QemSCAN or MLA data. one-off ‘snapshot’ sampling approach. In the case where the type of
In current research, the first of the above problems will be information desired will be produced from advanced technologies
addressed [Lotter, 2002-). A fundamental approach will be taken to such as QemSCAN or the MLA, which are expensive units, the chal-
improve the representativity, or trueness, of sample material taken lenge of sample representativity at the polished section level is criti-
from a production concentrator. This goes beyond routine metal cal. It is therefore very important to have quality control diagnostics
accounting, rather a coordinated sampling campaign across the at hand. These diagnostics certify the representativity of the plant
entire flow sheet is desired. This will be with the specific purpose of survey material prior to any advanced analysis being performed.
producing a suite of representative, or true, samples taken from an High-quality information from such technologies places the investiga-
overall flotation flow sheet. This new process will be called statistical tor in a position to practice Process Mineralogy on flowsheet optimi-
benchmark surveying, and will relate to the generic features of the sation, or in the case of exploration, to model the behaviour of the ore
ore being mined. Two of the critical issues are: body in a variety of flow sheet options. These are predictive models.

• The ore presented to the mill for processing has been DRILL-CORE
mined from an ore reserve that typically has log-nor-
mally distributed metal values. This lognormal distribu- The other key problem to be solved is that of drill-core sampling
tion is associated with large variances. The blasting, and testing ahead of a decision to proceed with a new mine, or with
underground blending, and surface crushing processes a replacement mineral resource for an older operation. It is key to the
tend to reduce the log variance of the pay metal values success of the project that accurate, reliable characterisation of drill-
in the time dimension as compared to the spatial core is performed, in both quantitative mineral terms and in high-con-
dimension. It is still common to observe reduced, but fidence flotation tests. This approach was recently reported [Lotter
visible, remnants of this variance, expressed as time- et. al. 2003].
correlated effects in sampling the mill feed and flotation In this work, it was shown that by approaching the drill core sam-
circuit products. The sampling arrangements have pling and testing from an end-member structure, individual character-
therefore to take this variance into account. istics and flotation behaviours could be identified and treated. By
treating the drill-core on an overall composite basis (such as would

2 Copyright © 2004 by SME


SME Annual Meeting
Feb. 23-25, Denver, Colorado

be the case in a production mill feed), a problematic feature may be A common argument presented to a QEM*SEM/QemSCAN
identified but would not necessarily be as effectively treated. user is that the polished section displays the minerals in only two
This approach is not limited to assessment of future operations. dimensions, thereby causing QemSCAN to overestimate the degree
The end-member study is as effective when used ahead of a first of liberation in composite particles.
concentrator survey.
The typical structure of an end-member study includes: In a key study to describe and quantify stereological bias in pol-
ished sections of sized fractions which are measured by QEM*SEM
• Listing of end members in the orebody or mineral or QemSCAN, Latti and Adair concluded in 2001 that, for the sample
resource, from a geological perspective. material studied, stereological biases were minimal. In this study, a
• Sampling each end-member from either drill core or hard-rock titanium deposit was sampled. The size fraction
working stopes. –850+425 micrometres was chosen for the polished section because
• Mineralogical characterisation of each end-member, this size corresponded with the processing conditions of the operat-
using a diverse array of appropriate compositional, tex- ing circuit.
tural and grain size measurement technologies. The –850 + 425 micrometre size fraction was prepared into a
• Chemical analysis of each end-member. polished section, then measured by QemSCAN in this instance.
• High-confidence flotation tests on each end-member, After measurement, the section was taken back to the polished sec-
characterising the metallurgical performance. tion preparation laboratory and was cut down by 25 micrometres. At
• High-confidence flotation tests on an ore mixture of end- this point, the section was measured again by QemSCAN. This
members to simulate the likely production mill feed, in process was repeated several times.
keeping with the mining plan. A 3-dimensional picture of the section was then constructed,
and compared to the two-dimensional images. The results showed
When all of this information is gathered and reviewed, a more that there was only a minimal difference between the two models.
fundamental understanding of the basic causes of certain flotation
circuit problems develops [Lotter et. al., 2003]. This often leads to a TWO CASE STUDIES
list of expected performance issues ahead of performing a survey on
the operating concentrator, or in the case of a mineral resource being Raglan Concentrator: 1998-2000
studied from drill core, a list of issues that the greenfield flowsheet will The Raglan concentrator was commissioned in January 1997,
have to address. and formed a major replacement source for nickel arising out of
Characterisation of the Mexican San Nicolas deposit was Sudbury for Falconbridge Ltd. Two surveys were conducted on this
improved using QEM*SEM [Winkers, 2002]. This fine-grained poly- concentrator, one in June 1997, and the other in February 2000. In
metallic VMS deposit required the more fundamental mineral-based both cases, the difference of nickel recovery between rougher and
description; conventional flotation tests failed on their own to prove final concentrate was 5-7%. It was therefore important to understand
head grade/recovery models. In this study, Winkers [2002] showed what was leading to this difference. The loss was present in scav-
that the Concentration Efficiency Index (CEI) in the flotation test cor- enger tailings, a final discard material arising from the cleaner circuit
related with the Phase Specific Surface Area (PSSA). [Lotter et. al., 2002].
QemSCAN analysis of these scavenger tailings produced a dis-
The CEI was defined as tribution of liberation across the 106 micrometre to 3 micrometre size
range as shown in Figure 1.
CEI = (C-F)*R/(100-F)

Where

C = concentration of mineral in concentrate 16

F = concentration of mineral in feed 14


R = percent mineral recovery.
12
%TotalPentlandite

PSSA is defined from line scan measurements using linear 10

intercept lengths. This means that the surface area per unit volume 8

of a mineral, or phase specific surface area (PSSA), can be reliably 6


Liberated
(90-100)
estimated [Sutherland, Pers. Comm., November, 2002; Sutherland et 4
2 3 Middling
(30-90) ParticleType
al, 1991]. PSSA is in mm / mm , or 1/length. For sizes in microme- 2

tres, the function “6000/PSSA” is used. This is based on an equant 0 Locked(0-30)

particle approximating a cube (6 sided). Since this is an inverse rela-


-3
+3 +7 +15 +25
+53 +106
tionship, as PSSA becomes larger, grain sizes become smaller. SizeFraction(micrometers)

QEMSEM produces a PSSA value directly.


In this model, larger PSSA values (smaller grain sizes) correlat- Figure 1 - Raglan Scavenger Tailings : Pentlandite Distribution
ed with lower CEI, indicating that, for this ore, the flotation perform- by Size and Liberation Class
ance was liberation-controlled. By transposing the CEI equation to
predict recovery, it was possible via the correlation between CEI and
PSSA to predict recovery from the PSSA. The overall distribution of pentlandite in this stream was found
to be: Liberated : 26.6%; Middling : 28.6 %; and Locked: 44.7 %.
IMPROVEMENTS IN QEM*SEM OR QEMSCAN Locked and middling particles thus accounted for 54% of all pent-
OPERATION landite in this stream. This clearly indicated a need for regrinding the
cleaner tailing prior to scavenger flotation. The regrind mill was com-
missioned into this duty in September 2000, and resulted in a sus-

3 Copyright © 2004 by SME


SME Annual Meeting
Feb. 23-25, Denver, Colorado

tainable improvement to nickel, copper, platinum and palladium


recoveries.

Candelaria Concentrator 2000


The use of automated mineral analysis via QemSCAN technol-
ogy was implemented in 2000 at Phelps Dodge Mining Company. It

% Chalcopyrite
is now extensively used within Phelps Dodge for ore characterisation,
concentrator surveys and hydrometallurgical applications [Baum et
al. 2002, 2003a] and Kendrick et al [2003b]. A QemSCAN survey was
performed on the Candelaria concentrator with the objective to a)
quantitatively characterise the process mineralogy of key ore types,

[CS7]

[CS6]

[CS4-5]
b) assess options for gold recovery improvement and c) establish a

[CS1-3]

[-38/+20]

[-75/+38]

[-106/+75]

[-150/+106]

[-300/+150]
size-by-size mineralogy of the circuit streams.

[+300]
Pertinent findings of the QemSCAN analysis can be sum-
marised as follows:
Figure 2 Size-by-size Chalcopyrite Liberation in Rougher Feed
• The liberation characteristics of the chalcopyrite exhibit
variances between the different ore types. Also, textur-
al differences impact the liberation in the comminution Further improvements in Cu recovery, therefore, need to target
circuit and govern middlings generation, i.e. insolubles the 4.2% lost from the rougher circuit. In particular, the form of the
content in the final concentrate and losses of ultrafine chalcopyrite lost to the +106 micrometre fractions of the rougher tail-
chalcopyrite to the rougher tails. ings need to be investigated because these fractions account for over
• Ore types with higher quartz and magnetite contents 50% of the lost copper. Figure 3 shows the size-by-size liberation
reduce the liberation efficiency of the chalcopyrite. distribution of the chalcopyrite lost to the rougher tailings.
• Coarse grained chalcopyrite textures (80-120 microme- The distribution resolves into two very discrete populations of
tre) such as semi-massive and veinlet occurrences lib- particles with the following characteristics:
erate well.
• Silicate gangue (such as chlorite, amphibole, epidote,
biotite, clay) and carbonates occur in the fine to ultra-
fine size range and represent the major causes for slim-
ing during primary grinding and re-grinding.
• Lower recoveries (<95%) are most often associated
with a combination of high Bond WI and fine chalcopy-
rite sizes.
• The monitoring of ore feeds for particle size distribution
on a mineral-by-mineral basis and control of hard
gangue and sliming minerals is beneficial to optimal
chalcopyrite liberation and good mill operation.
[CS7]

[CS6]

[CS4-5]

[CS1-3]

[-38/+20]

[-75/+38]

[-106/+75]

The Candelaria concentrator exhibited typical metallurgy during


[-150/+106]

[-300/+150]

[+300]
the survey period and averaged 95.3% Cu recovery. Losses to tail-
ings averaged 4.7%, of which 0.5% was lost from the scavenger cir-
cuit and 4.2% from the rougher circuit.
It is clear that the excellent liberation characteristics identified in Figure 3: Size x size Chalcopyrite Liberation in the Rougher
the –75-micrometre fractions of the feed contributed to the high Tailings
recoveries in the same size fractions of the concentrate. It is also
apparent that the cleaner/scavenger circuits were performing excep-
tionally well with losses of only 0.5% Cu relative to the feed. • The most dominant population consists of locked chal-
The size-by-size chalcopyrite liberation and locking characteris- copyrite in coarse particles, typically larger than 75
tics in the rougher feed are shown in Figure 2. The degree of libera- micrometres. Some of these particles are also present
tion of both the chalcopyrite and the gangue provides quantitative in the intermediate fractions and consist of disseminat-
evidence that there are no major textural barriers to further upgrad- ed chalcopyrite typically enclosed within silicate gangue
ing this stream to final concentrate grade. minerals such as feldspar, chlorite and amphibole/epi-
dote. The average grain size of this disseminated chal-
copyrite is 25 to 30 micrometres and will never be fully
recovered without additional fine grinding. Locked par-
ticles account for 58% of the copper losses to rougher
tailings and represent a recovery target of 2.5%.
• The second population consists of liberated particles in
the fine –20-micrometre fractions. Some liberated par-
ticles are also lost in the +20 and +38 micrometre frac-
tions. Liberated particles account for 31% of the cop-
per lost to the rougher tailings and represent a recovery
target of 1.3% copper.

4 Copyright © 2004 by SME


SME Annual Meeting
Feb. 23-25, Denver, Colorado

The remaining 11% of the chalcopyrite occurs as middling par- Cramer, 2001, The Extractive Metallurgy of South Africa’s
ticles in intermediate and coarser size fractions (38 to 150 microme- Platinum Ores, Journal of Metallurgy, October 2001, pp. 14-18.
tres). This loss represents a recovery target of 0.4% Cu. It is clear Hartley, DG, Trahar, WJ, and Weller, KR, 1977, Analysis of
from this analysis that the fine size fractions offer the best target for Concentrator Operations at Mount Lyell, Conference Series,
improved recovery because these fractions are clearly dominated by Australasian Institute of Mining and Metallurgy, 6, pp. 155-169.
liberated chalcopyrite losses. Since poorer flotation kinetics are Henley, KJ, 1983, Ore Dressing Mineralogy – A Review of
responsible for the losses of liberated chalcopyrite, mineral process- Techniques, Applications and Recent Developments, Spec. Publ.
ing solutions need to address over-grinding in the primary grinding Geol. Soc. S. Afr., 7, (1983) pp. 175-200.
circuit, and improved fine particle flotation in the rougher flotation cir- Jones, MP, and Gavrilovic, J, 1968. Automatic searching unit
cuits. for the location of rare phases by electron-probe X-ray microanalysis,
In the rougher feed, 84% of the silicates are in liberated particles Institution of Mining and Metallurgy, (77) pp. B137-B143.
(containing more than 90% silicates). Furthermore, significant pro- Kendrick, M., Marsden, J., Baum, W., Wilkie, G., Gottlieb, P.,
portions of liberated silicates are found in the coarser fractions as and Thompson, P., 2003b, The Use of the QemSCAN Automated
well as the finer –20 micrometre fractions. It can, therefore, be Mineral Analyzer in the Candelaria Concentrator, Copper 03,
expected that the majority of the feldspar and biotite particles are Santiago/Chile, Proceedings, paper in preparation.
either liberated or they occur in association with other gangue miner- Latti, D, and Adair, BJI, 2001, An Assessment of Stereological
als in this stream. Adjustment Procedures, Minerals Engineering, 14, no,. 12, pp. 1579-
1587, 2001.
CONCLUSIONS Lotter, NO, 1995a, Review of Evaluation Models for the
Representative Sampling of Ore, J. S. Afr. Inst. Min. Metall., 1995,
Both the Raglan and Candelaria QemSCAN plant surveys (4), pp. 149-155.
(associated with metallurgical testing) represent highly successful Lotter, NO, 1995b, A Quality Control Model for the Development
case histories for size-by-size circuit audits as essential tools for plant of High-Confidence Flotation Test Data, M.Sc.(Eng.)(Chem.) Thesis,
optimization. University of Cape Town, June 1995.
At Candelaria, the QemSCAN survey and the subsequent Lotter, NO, O’Connor, CT, and Clark, I, 2000, The Evaluation
implementation of circuit changes and associated metallurgical tests Bias Errors of the Platinum-Group Elements (PGE) Associated with
resulted in increased copper recovery and reagent savings. Gravimetric Fire-Assaying Practice in Rustenburg Section of the
Further, this work has resulted in significant gold recovery improve- Bushveld Complex, South Africa, SME Preprint, Annual Meeting and
ments since 2001. Exhibit, Salt Lake City, February 2000.
Several revenue-generating optimization targets were identified Lotter, NO, 2002-, Statistical Benchmark Surveying of
by the QemSCAN work. The QemSCAN surveys, even at well oper- Production Concentrators, PhD Thesis in Preparation, McGill
ated/high recovery plants, can achieve substantial benefits from cir- University, 2002-.
cuit audits using quantitative size-by-size mineralogy. Lotter, NO, Whittaker, PJ, Kormos, L, Stickling, JS, and Wilkie,
It is concluded that the use of automated mineral analysis rep- GJ, 2002, The Development of Process Mineralogy at Falconbridge
resents a powerful and highly economic process mineralogy tool for Ltd., and Application to the Raglan Mill, CIM Bulletin, vol. 95, no.
ore type characterisation, control and optimisation of grinding and 1066, November/December 2002, pp. 85-92.
flotation circuits. Lotter, NO, Kowal, DL, Whittaker, PJ, Tuzun, MA, and Kormos,
L, Sampling and Flotation Testing of Sudbury Basin Drill Core for
ACKNOWLEDGEMENTS Process Mineralogy Modelling, SME Annual Meeting and Exhibit,
Cincinnati, February 2003, Preprint # 03-23.
The authors would like to thank the management of Noranda- Petruk, W, 2000. Applied Mineralogy in the Mining Industry.
Falconbridge and Phelps-Dodge Mining Company for permission to Published by Elsevier, New York, pp.268.
publish this paper. Prof. AR Laplante, of McGill University, and Ms. Restarick, CJ, 1976, Pulp Sampling Techniques for Steady
L Kormos, of Noranda/Falconbridge, kindly provided valuable editing State Assessment of Mineral Concentrators, Sampling Symposium,
comment during the preparation of this paper. Aus. Inst. Min. Metall., Melbourne, 1976, pp. 161-168.
Sutherland, D.N., Gottlieb, P., Wilkie G. and Johnson, C.R.,
REFERENCES 1991, Assessment of ore processing characteristics using automated
mineralogy, XVIIth International Mineral Processing Congress (
Baum, W., Marsden, J., Vanderbeek, J, and Wilkie, G., 2003a, Dresden ), pp. 353 – 361.
Mineralogical Audits of Grinding and Flotation Circuits with the Sutherland, DN, 2002, Estimation of Grain Sizes using
QemSCAN Mineral Analyzer, Applied Mineralogy Conference 03, QemSCAN (pers. comm., November 2002.
Helsinki, 3 p. Trahar, WJ, 1981, A Rational Interpretation of the Role of
Baum, W., Fennel, M., Vanderbeek, J., Wilkie, G. , and Gottlieb, Particle Size in Flotation, Int. J. Min. Proc., 8, (1981), pp. 289-327.
P., 2002, The use of the QemSCAN Automated Mineral Analyzer for Weller, KR, and Sterns, UJ, 1984, The Use of Mineralogical
Ore Characterization and Concentrator Operation, abst. SME Data in Computer-Aided Optimisation of a Tin Concentrator, AIME
Arizona Conference, Tucson, December 2002. Symp. Applied Mineralogy, Los Angeles, 22-25 February 1984, pp.
Baum, W., Sanhueza, J., Smith, E., and Tufar, W., 1989, The 313-330.
Use of Process Mineralogy for Plant Optimization at the El Indio Winkers, AH, 2002, Metallurgical Mapping of the San Nicholas
Gold-Silver-Copper Operation (Chile), Erzmetall, 42, Nr. 9, pp. 373 Deposit, Canadian Mineral Processors Conference, Ottawa, January
– 378. 2002, paper no. 3, pp. 35-55.
Cameron, AW, Kelsall, DF, Restarick, CJ, and Stewart, PSB,
1971, A Detailed Assessment of Concentrator Performance at
Broken Hill South Limited, Proc. Aust. Min. Met. 240, Dec. 1971, pp.
53-67.

5 Copyright © 2004 by SME

You might also like