You are on page 1of 6

Minerals Engineering 24 (2011) 559–564

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Understanding the discrepancy between prediction and plant GRG recovery


for improving the gold gravity performance
Sunil Koppalkar a,⇑, Ahmed Bouajila a, Claude Gagnon a, Guillaume Noel b
a
COREM, 1180, rue de la Minéralogie, Québec, Canada G1N 1X7
b
IAMGOLD – Doyon Division, P.O. 970, Rouyn-Noranda, Québec, Canada J9X 5CT

a r t i c l e i n f o a b s t r a c t

Article history: Gravity recoverable gold (GRG) content in a sample is typically determined using the standard three-
Available online 20 October 2010 stage test developed at McGill University (Laplante et al., 2000). A comparison of GRG recovery in labo-
ratory with its actual recovery in plants shows that plants recover between one-third and two-thirds of
Keywords: the predicted GRG from laboratory test.
Gravity concentration Research conducted at COREM seeking the identification and the understanding of potential causes of
Gravity recoverable gold this discrepancy reveals that particle shape transformation during grinding is a major contributor among
Knelson concentrator
other causes such as bleed percent processed, quality of cyclone classification and gravity concentrator
Particle shape
Gold gravity recovery
operating conditions.
To analyze the gold particle shapes and their deportment, samples were collected from different
streams of a grinding–gravity (Knelson) circuit in a gold plant and compared to a concentrate sample
from a laboratory Knelson concentrator (KC MD3). A particle characterization based on gold morphology
was developed to diagnose typical shapes generated in the gold grinding–gravity circuit. The study
revealed that the GRG particles reporting to the plant cyclone overflow are the most flattened particles
among the samples analyzed and that on the other hand gold particles recovered in the laboratory Knel-
son concentrator are the least flattened, highlighting the potential cause of discrepancy between the two
recoveries.
In this paper, the GRG mapping in a conventional grinding–gravity circuit and the transformation of
GRG to non-GRG, as mainly explained by gold particle reshaping will be discussed.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction are typically installed on a bleed stream from the cyclone under-
flow, mill discharge or cyclone feed.
Knelson concentrators are the most commonly used semi-con- The potential of any gold ore for gravity recovery is estimated
tinuous centrifugal gravity concentrators for recovering gold and using the standard GRG test presented in Fig. 1. It is a three-stage
platinum group minerals (PGM) from grinding circuit streams. sequential liberation and recovery process using a 3-in. Knelson
Gravity concentration using centrifugal concentrators has gained concentrator (KC MD3) designed to assess the gravity recovery po-
increasing popularity since the invention of Knelson concentrator tential of a gold ore (Laplante et al., 2000). The test typically uses
in the 1980s. The benefits of gravity recovery are now well under- 50–100 kg of ore, depending on the gold grade and its distribution,
stood and the acceptance by the industry is overwhelming. The in order to provide reproducible results. The main purpose of the
presence of reasonable amounts of gravity recoverable gold standard GRG test is to predict the amenability of a given gold
(GRG) in an ore prompts gold mills to install gravity circuits. Thus, ore to gravity recovery. This test is not only applied for evaluating
the increased use of gravity concentration has necessitated a better greenfield projects but also for evaluating the performance of
understanding of the behavior of gold in grinding circuits and of existing gravity circuits. Thus, the test has become an industry
the factors affecting the performance of the gravity concentrator. standard for assessing the potential of a gold ore to gravity concen-
Due to its high density (19.3 g/cm3 when pure) and malleability tration. More than 200 gold ore samples from over the world have
properties, it tends to accumulate in circulating load streams mak- been processed at McGill University (Huang and Koppalkar, 2007).
ing them attractive for gravity recovery. The gravity recovery units In a typical Knelson concentrator-based gravity circuit, simula-
tions by Laplante (1996) indicate that about two thirds of the GRG
⇑ Corresponding author. present in the feed ore can be recovered. The remaining one-third
E-mail address: sunil.koppalkar@corem.qc.ca (S. Koppalkar). GRG not recovered is due to the fact that:

0892-6875/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2010.09.007
560 S. Koppalkar et al. / Minerals Engineering 24 (2011) 559–564

Sample (~ 100 kg)

Stage 1 tailings
Splitting
(saved)
Crushing

Screening Grinding

100% - 850 µm 50% - 75 µm


Stage 1 Stage 2

KC MD3 Tailings KC MD3 Tailings


Grinding

80% - 75 µm
Stage 3
Conc. Conc.
Tailings Stage 3 tailings
Screen Screen KC MD3 (saved)
Screen Screen

Conc.

Screen Screen

Fig. 1. Standard GRG test procedure (Xiao, 2001).

 Only part of the circulating load is processed for gravity 1.2. Circuit configuration related
recovery.
 Plant gravity units are not very efficient compared to the lab The process stream chosen for the gravity unit depends on the
units. amount of GRG and its size distribution. Bleed stream from the cy-
 Overgrinding of GRG can be minimized but cannot be entirely clone underflow in a ball mill circuit is typically processed on a
eliminated. Knelson due to its ease of operation and also, on the premise that
gold recirculation is high and has high chances to be recovered be-
A comparison of the gravity recoveries of plants with their pre- fore it is ground into non-GRG and exits through the cyclone over-
dicted recoveries are presented in Table 1. flow. Processing of low bleed % and stream type could also impact
It can be seen that plant gravity recoveries range from 34% to on the recovery. For example, processing ball mill discharge in-
85% of the predicted values from the standard GRG test, validating stead of cyclone underflow may improve gold recovery, since that
the assessment made by Laplante (1996). in such a configuration gold particle reshaping is low due to the
Optimized plant gravity operations like P4 and P5 achieved fact that the particles are not subjected to progressive flattening
recoveries approaching the predicted values. and are recovered immediately after discharge from the mill. Un-
Further data analysis of different plants, within the on-going re- der-optimized Knelson and cyclone operations are also potential
search work at COREM, suggests many causes for the discrepancy. factors for the discrepancy; under these conditions GRG losses
could be significant.
1.1. Machine related
1.3. Ore characteristics related
There is an inherent discrepancy between the laboratory-scale
and the full-scale machines owing to the processing methodology. Gold’s association with gangue minerals as well as its liberation
In a standard GRG test, the laboratory Knelson concentrator is size will impact recovery at plant-scale. Generally, the recovery of
operated under-optimized operating conditions to produce maxi- gold from low density gangue minerals like silicates will be higher
mum recovery. The laboratory unit is typically operated at low spe- compared to the recovery from high density gangue minerals like
cific feed rates (defined as the feed rate per concentrating bowl sulphides. If the gold particles liberation size is finer and associated
surface area, t/h/m2) and at a feed top size of 850 lm for optimal with the gangue minerals, fine grinding is required and that may
performance. Plant units are operated at high specific feed rates produce shapes unsuitable for gravity recovery. Chapman and
and the feed top size is typically maintained at 2.0 mm, yielding Houseley (1996) have studied the effect of particle shapes on grav-
lower recoveries than achieved at laboratory-scale. Moreover, the ity recovery and showed that mass rather than shape resulting
laboratory unit recovers gold that is progressively ground in three from grinding was responsible for the recovery of coarse
stages, which minimizes overgrinding and smearing of gold. (+300 lm) particles, whereas progressive flattening of gold grains
during grinding was attributed to the drop in recovery for the fine
particles ( 150 + 53 lm).
Table 1
Comparison of some reported predicted and plant gravity recoveries. COREM has undertaken a research program aiming at improv-
ing the gravity recovery of gold by addressing the causes of dis-
Plant P1 P2 P3 P4 P5
crepancy between GRG tests and plant results. In this research,
Predicted gravity recovery, % 44 55 65 90 86 new work aiming at the understanding gold reshaping and its ef-
Plant gravity recovery, % 15 30 45 70 73
fect on gravity recovery should help find ways to minimize the
Gravity recovery as % of predicted 34 55 69 78 85
generation of deleterious particle shapes.
S. Koppalkar et al. / Minerals Engineering 24 (2011) 559–564 561

In order to investigate and understand the discrepancy, a plant 3. Experimental


gravity circuit was studied in detail with a view of ‘‘tracking” gold
particle shapes in different grinding–gravity circuit streams. The 3.1. Sampling campaign
objectives were to:
Sampling of the grinding–gravity circuit at IAMGOLD-Doyon
 Measure the discrepancy between the lab test and plant results. mine was undertaken in order to find out how much of the GRG
 GRG mass balance around the grinding and gravity circuit to (gravity recoverable gold) available in the ore was recovered by
identify the source of discrepancy. the operating Knelson concentrator. Four sampling campaigns on
 Study the gold particle shapes in the different streams. the Mouska ore, three in 2008 and another in 2009 were performed
adopting the GRG test protocols (Koppalkar, 2010). The following
IAMGOLD-Doyon mine was used as a case study in this project. were collected throughout the Knelson recovery cycle period of
45 min:
2. Case study and plant description
 SAG feed samples to determine the GRG content in the ore.
The Doyon mine, one of the 21 member plants of COREM, be-  Knelson concentrate and tailings samples to assess the gravity
longs to the IAMGOLD Corporation, one of the 12 corporate mem- recovery.
bers of COREM. The Doyon Division is comprised of the Doyon and  Cyclone overflow and underflow samples to measure the GRG
Mouska underground gold mines, located approximately 40 km content in those streams.
east of Rouyn-Noranda and 80 km west of the city of Val-d’Or in
northwestern Québec, within the Cadillac-Bousquet gold belt in 3.2. GRG evaluation
the Abitibi Region, one of the most prolific gold mining regions
in Canada. In order to understand and explain the potential sources of the
Vein-type mineralization dominates at the Doyon Division. Vein discrepancy, GRG testing was performed as a first step on the sam-
assemblages typically comprise varying proportions of pyrite, ples collected from the plant. Two types of tests were performed:
quartz, chalcopyrite, carbonate and gold. The Mouska deposit is
classified as a lode-type deposit. Economic mineralization is asso-  Three-stage test using the standard GRG protocol for the SAG
ciated with laterally and vertically continuous narrow quartz veins feed samples.
and contains 5–15% pyrite–pyrrhotite and 5–10% chalcopyrite.  Single stage test using the McGill protocol for a stream sample
The Doyon gold mine includes a conventional mill equipped on:
with a SAG mill, gravity circuit, flotation circuit and CIP processing  The Knelson concentrator tailings samples.
facilities.  Cyclone underflow and cyclone overflow samples.
The 3500-tonne per day Doyon processing plant uses the car-
bon-in-leach (CIL) and carbon-in-pulp (CIP) processes to recover 3.3. GRG mass balance
gold and produce gold bullion from both Doyon and Mouska ores.
Flotation produces a gold bearing copper concentrate from Mouska As a second step towards understanding the discrepancy, GRG
ores that is then shipped to a smelter for final metal recovery. Cop- metallurgical balances of the plant Knelson concentrator were
per is a by-product and is credited against Mouska’s production determined using the size-by-size gold assays of the plant concen-
costs. trate and size-by-size gold assays from the tailings. GRG recovery
The simplified grinding and gravity circuit of the plant is of the Knelson concentrator was calculated assuming gold recov-
presented in Fig. 2. Gravity circuit is installed in the SAG grinding ered by the Knelson as 100% GRG and GRG in Knelson tailings pro-
circuit. A Knelson (KC XD30) concentrator recovers gold from a 12– portional to that recovered by the 3-in. laboratory Knelson
13% bleed stream from the primary cyclone screened at 2 mm as concentrator. GRG content of the SAG feed is determined from
shown in the figure. The concentrates produced from the Knelson the GRG test. From the data on GRG in the feed and the GRG recov-
concentrator are further upgraded in the gold room on a Gemeni ered by the gravity, GRG remaining, i.e., GRG in the cyclone over-
table to produce bullions. flow and GRG returned for grinding are calculated.

3.4. Shape analysis

It is suspected that the grinding operation contributes to the


production of particle shapes deleterious to gravity separation.
The transformation and reshaping of gold particles during grinding
could be pertinent in analyzing the discrepancy. Gold particle
shapes of the different grinding–gravity streams were analyzed
using two particle shape classification methods: Zingg classifica-
tion and the Corey shape factor (CSF).
Zingg classified particles in four classes: disks, blades, rods and
spheroids based on the ratio of the three main axes of a particle,
large (L), intermediate (I) and small (S) (Pettijohn, 1975) as shown
in Fig. 3.
Data generated from particle shape studies were also used to
determine Corey shape factor (CSF) for different products. It is de-
fined as the ratio of the smallest dimension of the particle to the
square root of the large and intermediate dimensions (Walsh and
Kelly, 1993). Lower CSF number indicates that the particles are flat-
Fig. 2. Grinding and gravity circuit of IAMGOLD-Doyon concentrator. ter. Chapman and Houseley (1996) reported using microscopes
562 S. Koppalkar et al. / Minerals Engineering 24 (2011) 559–564

I/L S/I Shape

>2/3 <2/3 Oblate (Discoidal)

>2/3 >2/3 Equiaxial (spheroidal)

<2/3 <2/3 Triaxial (bladed)

<2/3 >2/3 Prolate (rod-shaped)

Fig. 3. Zingg diagram for the classification of shapes (after Zingg (1935) in Pettijohn (1975)).

fitted with graticular eyepiece for measuring major dimensions. nificant portion, 37% of the gold, is recovered at stage 1 while fur-
The thickness of the grains up to 150 lm was measured using a ther liberation has contributed to 17% recovery at stage 2, and
micrometer. It is clear that the thickness measurement technique another 6% at stage 3 to an overall recovery of 60%.
needs better methodology for more accurate assessment. In the Fig. 5 presents a comparison of the GRG recoveries for the four
current research, measurement of the large and intermediate SAG feed samples. These curves represent the typical ‘‘average GRG
dimensions was performed using optical microscope fitted with a curve”, with a fairly coarse GRG, 25% coarser than 106 lm, ame-
camera and Image-Pro Plus software. nable for gravity recovery (Laplante et al., 2000). Table 2 shows the
GRG contents determined for the ore samples with their calculated
feed gold grades. GRG contents of the four tests, 56–62% are similar
4. Results and discussions

The GRG contents of the SAG feed sample from the Mouska ore
are presented in the graphical form as cumulative recovery as a Table 2
function of gold particle size in Fig. 4. Comparison of GRG results for the SAG feed samples.
Recoveries are cumulated from stage 1 to stage 3 with stage 3 Sample # Feed grade, g/t % GRG
showing the overall recovery for the test. It can be seen that a sig-
Total gold GRG
SAG 1 13.84 8.53 61.62
100 SAG 2 16.8 9.81 58.39
Stage 1 SAG 3 16.25 9.3 57.25
Stage 1+2
% Cum. recovery

SAG 4 13.33 7.4 55.52


80 Stage 1+2+3

60

40 Table 3
Comparison of predicted and plant GRG
recovery.
20
SAG feed, % Plant Knelson
0 GRG recovery
10 100 1000
61.62 27.23
Particle size, µm 58.39 15.82
57.25 18.18
Fig. 4. Graphical representation of GRG for SAG feed 1 sample.

10
GRG in cyclone Overflow (%)

9 Campaign 1
100 Campaign 2
8
SAG#1
7
Cum. recovery, %

80 SAG#2
SAG#3 6
60 SAG#4 5
4
40
3
2
20
1
0 0
10 100 1000 212 150 106 75 53 37 25 20 15
Particle size, µm Particle size, µm
Fig. 5. Comparison of GRG recoveries. Fig. 6. Size by size GRG content in cyclone overflow samples.
S. Koppalkar et al. / Minerals Engineering 24 (2011) 559–564 563

within the sampling and assaying errors. The results indicate that Table 7
head grade had no significant impact on GRG recovery. Zingg classification for Plant Knelson tailings.

Table 3 shows the discrepancy between the measured GRG in Size, lm Disk (%) Blade (%) Rod (%) Spheroid (%)
the ore and the gold actually recovered by the plant gravity con- 300 10 90 – –
centrator. It can be seen that the plant Knelson recovered 33% of 212 16 72 12 –
the GRG available in the ore. The reported monthly average gravity 150 8 75 17 –
recovery for the sampled month was 15%. 106 3 78 19 –
75 5 89 3 3
Fig. 6 shows GRG content in the cyclone overflow as a function
of gold particle size for the two cyclone overflow samples. In cam-
paign 1, the overall GRG recovery was up to 30%. For campaign 2,
the overall GRG recovery was 27%. Of that gold, coarser fractions Table 8
(from 212 lm to 75 lm) accounted for 4.25 to 6% of the total Zingg classification for cyclone overflow.
GRG. Normally, these gold grains should easily be recovered by Size, lm Disk (%) Blade (%) Rod (%) Spheroid (%)
the Knelson concentrator. In fact, Knelson concentrators can effec-
300 – 100 – –
tively recover gold particles as fine as 20 lm. One of the possible 150 14 86 – –
explanations for the loss of these gold particles could be the shape 106 27 73 – –
transformation during grinding. 75 – 100 – –
A detailed GRG mass balance was carried out from the GRG
measurement on the samples of the two campaigns and the results
are presented in Table 4 and 5. It was assumed that GRG not recov-
1.0
ered by gravity is equal to GRG exiting through the cyclone over-
0.9 KC Conc
flow plus the GRG circulated to the mill.
0.8 KC Tail
Data from the test work showed that the Knelson concentrator
and cyclone operations require optimization. A close look at the 0.7
CyOF
GRG balances of the first campaign shows that 24% of the GRG 0.6

CSF
Lab KC Conc
(106 g/h) is recovered by the gravity circuit and 44% of the GRG 0.5
(192 g/h) is lost through the cyclone overflow. The remaining 0.4
32% of the gold is ground in the mill to produce shapes and sizes 0.3
unsuitable for gravity separation. The GRG balances for the second 0.2
campaign (Table 5) have also produced results similar to those to 0.1
the first campaign. 0.0
Plant Knelson concentrate, tailings and cyclone overflow sam- 10 100 1000
ples were subjected to morphological studies under an optical Particle size, µm
microscope equipped with a camera and the Image-Pro Plus
Fig. 7. Comparison of CSF for different product streams.
4.5 software for measuring the length and breadth precisely. The
main problem encountered was the measurement of the smallest microscope-operator dependent, consistent and reliable data was
dimension, the thickness. Thickness was thus assigned as a fraction generated. However, for particles less than 75 lm, measuring the
of the largest dimension which was measured accurately using a thickness posed a serious challenge.
camera and software. Even though the thickness estimates were To investigate the shape transformation related issues, samples
from the different products (plant Knelson conc., tailings and cy-
clone overflow) were subjected to a new morphological study
Table 4 based on Zingg classification as shown in Fig. 3. From Table 6, it
GRG balance for grinding–gravity circuit (campaign 1). can be seen that the all particle types are recovered in the Knelson
concentrates; however, the proportion of disk type particles is rel-
GRG units in GRG recovered GRG lost in the cy. GRG converted
SAG feed, g/h by gravity, g/h overflow, g/h into non-GRG, g/h atively low compared to other particle types, since these particles
tend to deport to the tailings because of their shape.
437 106 192 139
Whereas from Table 7, it can be seen that the deportment of
spheroid particles to the tailings is almost negligible since these
types of particles are easily recovered in the concentrates and
Table 5 hence, do not report to the tailings stream. However, the recovery
GRG balance for grinding–gravity circuit (campaign 2). of blades and disks is relatively higher because of the flattened
GRG units in GRG recovered GRG lost in the cy. GRG converted shapes. It can be generalized that the proportion of disk and blade
SAG feed, g/h by gravity, g/h overflow, g/h into non-GRG, g/h particle shapes is lower for the concentrate stream, whereas it is
406 97 173 136
higher for the Knelson tailing stream.
Table 8 shows the proportion of particle types recovered in the
overflow. It is apparent that the proportion of blades and disks is
higher compared to the concentrate stream because of their flaky
Table 6 morphology.
Zingg classification for plant Knelson concentrate. Fig. 7 presents the CSF for the different plant samples and the
Size, lm Disk (%) Blade (%) Rod (%) Spheroid (%)
concentrate from the laboratory Knelson concentrator. Lower CSF
values are indicators of flattened particles. It is interesting to ob-
600 9 66 25 –
425 10 45 40 5
serve that the cyclone overflow particles are the flattest among
300 10 52 34 4 the particles analyzed, whereas the concentrate produced from
212 – 58 42 – the laboratory Knelson concentrator has the highest CSF number.
150 – 50 36 14 This is probably due to the fact that laboratory grinding generates
106 9 47 28 16
a lower amount of transformed particle shapes. It can also be seen
564 S. Koppalkar et al. / Minerals Engineering 24 (2011) 559–564

that CSF values for all products increase with decreasing gold par- 6. Future work
ticle size.
The shape relation from CSF can be shown as: Lab Knelson con- In order to further the maturity of this enabling technology
c. > Plant Knelson conc. > Plant Knelson Tailings > Cyclone COREM (2009), research is on-going, focusing on how to improve
overflow. shape measurement and model the shape effect on the classifica-
The grinding of samples for the GRG test is typically performed tion and gravity separation of the gold. Another attempt is being
as open grinding and thus, particle shape transformation is not se- made to reduce the generation of undesirable shapes. New gravity
vere as compared to industrial closed circuit grinding. Gold parti- schemes are being conceived to recover the coarse gold reshaped
cles lost to the Knelson tailings are relatively flatter, thus carried particles (i.e., flattened) lost in the overflow for improving gravity
to the tailings by the effect of Knelson fluidizing water, which circuit performance. Alternative grinding–gravity configurations
mainly helps to discard the low density gangue minerals concen- suitable for maximizing gravity recovery by minimizing the gener-
tration in the concentrating bowl of the Knelson. ation of deleterious shapes will also be explored.

5. Conclusions Acknowledgements

The conclusions drawn from the on-going research work are: The authors wish to thank IAMGOLD-Doyon for its support. The
authors also wish to acknowledge the efforts of Jacques Béland and
 The gap between prediction and plant GRG is significant and Jean Robitaille during testing and sampling campaigns. The authors
could be up to 58% vs. 15%. thank COREM for the permission to publish this paper.
 Some of the GRG is converted into non-GRG during grinding due
to gold’s malleable and high density properties. References
 GRG balance around the grinding–gravity circuit shows that a
significant portion of GRG (44%) exits through the cyclone Chapman, R.J., Houseley, K. 1996. The effect of particle geometry on the recovery of
gold grains by gravity concentration methods. In: Proceedings of the 28th
overflow, due among others, to the shape transformation of cer- Annual Conference of Canadian Mineral Processors, Ottawa, Paper 23, 1996, pp.
tain particle types. 335–357.
 Two particle shape characterization methods, the Zingg classifi- COREM Technology Register, Version December 2009 (Private).
Huang, L., Koppalkar, S., 2007. Evolution and optimization of the Gravity
cation and the CSF, were used to validate and quantify the find- recoverable gold test. In: Proceedings of the 38th Annual Conference of
ings from mass balance. Canadian Mineral Processors, Ottawa, Canada, Paper 19, 2007, pp. 317–330.
 Over 75% of the particles lost in the cyclone overflow and Knel- Koppalkar, Sunil., 2010. Effect of Operating Variables in Knelson Concentrators: A
Pilot Scale Study. Ph.D. Thesis.
son tailings are blade-type compared to 50% recovered in the Laplante, A.R., 1996. Characterization and behaviour of gold in grinding and gravity
Knelson concentrates. circuits, personal communication.
 Gold morphological studies reveal that the gold particles lost to Laplante, A.R., Woodcock, F., Huang, L., 2000. Laboratory procedure to characterize
gravity-recoverable gold. Transactions-Society for Mining, Metallurgy and
the cyclone overflow are highly flattened as indicated by CSF
Exploration 308, 53–59.
analysis. Pettijohn, F.J., 1975. Sedimentary Rocks, 3rd ed. Harper and Row Publishers, New
 Gold particle shape transformation seems to be one of the main York.
contributors to the discrepancy. Walsh, D., Kelly, E.G., 1993. Technical note nominal diameter of gold particles.
Minerals Engineering 6 (2), 193–198.
 COREM has developed a method for determining GRG distribu- Xiao, Z., 2001. Developing Simple Regressions for Predicting Gold Gravity Recovery
tion (mass, size and shape) throughout a grinding circuit by in Grinding Circuit. M. Eng. Thesis, McGill University.
using the test results of lab-scale Knelson concentrator as an Zingg, T., 1935. Beitrage zur Schotteranalyse: Schweizerische Mineralogische und
Petrologische Mitteilungen 15, 39–140.
enabling technology for increasing the productivity of gold
gravity separation circuits by optimizing their operating
variables.

You might also like