Paper 42:text
10/16/09
12:29 PM
Page 271
VON KETELHODT, L. Viability of optical sorting of gold waste rock dumps. World Gold Conference 2009, The Southern African Institute of Mining and
Metallurgy, 2009.
Viability of optical sorting of gold waste rock dumps
L. VON KETELHODT
CommodasUltrasort, South Africa
During the period October 2003 to June 2004, a containerized optical sorting pilot plant was
operated at Kloof Gold Mine on a waste rock dump. In this paper, we describe the test work and
its results leading to the decision to operate this pilot plant. This paper presents the operational
data and financial figures achieved during this project, in particular, for the month of June 2004
when the plant was run at full capacity on three shifts, 24 hour operation. These results are also
extrapolated and presented to show the profitability of such an operation at today’s gold price
levels and cost structure.
Introduction
Mining activities around the Witwatersrand complex have
been going on for over 120 years. Waste rock from shaft
sinking and development operations has been dumped on
massive waste dumps throughout the area. Inherently, these
dumps contain a fair amount of gold which was misplaced
during the tramming and hauling operations.
In recent years, gold mines reclaimed some of this gold
by screening off the fines, 16 mm and blending it with the
run of mine as a feed to the processing plant. These fines
usually have higher 0.8 to 1.0 g/t gold content compared to
the dump’s overall grade which normally ranges between
0.2 to 0.9 g/t. Some mines even send the entire waste rock
dump to the plant to make sure that all the gold gets
recovered. This is typically done at Anglo Gold Ashanti’s
Vaal Reefs operations and Goldfield’s Driefontein Mine.
A number of mines have tried hand picking of gold ore,
for example, at Buffelsfontein Gold Mine and also at Kloof
Gold Mine. Usually, these operations failed due to the
inconsistencies and inefficiencies of such an operation.
Particularly in the finer size ranges of 50 mm hand picking
is labour intensive and difficult to produce adequate
tonnages.
In some Witwatersrand ore, a relationship between gold
and uranium, related to a particular reef has been exploited
by using radiometric sorting to selectively pre concentrate
coarse gold (and uranium) bearing rock from waste rock1.
In the 1970’s and 1980’s, radiometric sorters were used at
Buffelsfontein Gold Mine. Density separation techniques
have not worked for this application where all the rock
types have a similar specific gravity.
Optical sorting, using colour and brightness properties of
the different rock types, was seen as a potential
beneficiation technology for this waste rock sorting
application. Extensive test work was conducted at an
optical sorting test plant at Mintek before a pilot plant was
erected and operated at Kloof Gold Mine.
late 1940’s (Wills 1992)2. Although still a relatively small
industry, ore sorting equipment can be applied to a variety
of different applications. ‘Ore Sorting involves the appraisal
of individual particles and the rejection of those particles
that do not warrant further treatment’ (Wills 1992)2. Salter
and Wyatt (1991)3 discuss that the sorting process can be
divided into four interactive sub-processes:
• Particle presentation
• Particle examination
• Data analysis
• Particle separation.
Feed preparation is more critical for sorters due the
importance of surface characteristics and physical size of
the particles. Most sorters need a 3:1 or 2:1 ratio between
the largest and smallest particle to be efficient. Once the
particles have been properly prepared for sorting, they must
be presented to the sensor. To operate efficiently, the sensor
must be able to analyse each single particle. As a result,
feed rate and the materials handling methods are the critical
components, with this most commonly being done by a
conveyor belt or chute (Wotruba, 2006)4.
The critical stage of examining the particle and
determining whether material is valuable or barren, is done
by a combination of sensor and processing unit. Once the
decision of has been made as to accept or reject a given
particle, a mechanical device is required to physically make
the sort. High pressure jets of air, or water, and mechanical
arms or paddles are generally used to make this separation.
Of all the components in a sorter, it is the choice of sensor
that controls the design of a sorter (Weatherwax, 2007)5.
Sensor based sorting technology
Ore sorting itself is not a new concept, with hand sorting
being one of the first methods of minerals processing.
Electronic ore sorting equipment was first produced in the
Figure 1. Two types of sorting systems: Chute vs. Belt (Harbeck,
Kroog, 2008)6
VIABILITY OF OPTICAL SORTING OF GOLD WASTE ROCK DUMPS
271
Paper 42:text
10/16/09
12:29 PM
Page 272
Table I
The electromagnetic spectrum and the different sensors available for sensor based sorting in mineral processing (Harbeck, Kroog, 2008)6
Figure 2. Kloof Gold Mine: Gold-bearing rock types
A multitude of different sensors are available, and the
choice is generally driven by the mineralogy of a given ore.
Optical sensors are the most common sensor type, which
has been very successfully used in the industrial minerals
industry (Wotruba, 2006)4.
Table II shows the Au grades of the seven rock types.
Note that the dolomite, lava, green and grey quartzite were
assigned the same grade, since, these samples were
identified as waste and were submitted as a single sample.
Test work7
Table II
Au head grades of the various rock types
Preliminary test on small sample
The material supplied for the test work comprised of seven
samples of the most common rock types found in the dump
as well as a 1/2 ton randomly selected bulk sample. The
random sample was taken from the <75 mm >30 mm
stockpile using a front end loader. The various rock type
samples were used to formulate a sorting program as well
as emulate a sweetened synthetic feed.
272
Rock types
VCR
Cobble
Marginal
Dolomite
Lava
Green Quartz
Grey Quartz
Au (g/t)
14.50
3.70
0.55
<0.08
<0.08
<0.08
<0.08
WORLD GOLD CONFERENCE 2009
Paper 42:text
10/16/09
12:29 PM
Page 273
Prior to submitting the rock types for assaying, the
classified rocks were used to establish a sorting algorithm
and a synthetic feed for optimizing the optical sorter
Program set-up and design
Classification of the most common rock types:
• VCR
• Cobble
• Marginal reef
• Dolomite
• Lava
• Green quartzite
• Grey quartzite.
Line scan images were then taken of each rock type, after
which the colour and brightness were analysed, thereby
allowing distinct colour classes to be defined. The
dolomite, lava, green and grey quartz were defined as waste
and given the accept function in the programme.
The VCR, which was in the minority, was defined as the
product and given the reject function in the program. The
cobble and marginal reef were also included in the VCR
colour cloud. All particles defined as product in the feed
stream, would then be rejected by the sorter to the product
stream.
Figure 4 shows the classification of the different colour
classes in the designed program.
The following colour classes were defined:
1. BG (background):
‘Back’
(background colour)
2. RE (reject):
‘VCR’
(VCR, cobble & marginal reef)
3. RE (reject):
‘Brown’
(Oxidized reef)
4. AC (accept):
‘Lava’
(Lava and dolomite)
5. AC (accept):
‘Grey_qrtz’
(Grey quartz)
6. AC (accept):
‘Waste’
(Green quartz)
7. FGDC (foreground)
‘REST’
(All other undefined pixels)
The focus of the test work was to maximize VCR
recovery as well as maximizing waste rejection. One
distinct colour used in the sorting algorithm was rusty
brown which is related to the oxidized sulphide which is
associated with the gold reefs. The positive side effect of
sorting for reef also removes the sulphides which cause acid
drainage.
Method of sorting
Figure 5 shows the flow diagram for a single pass sorting
method. The aim of the single pass method was to obtain
maximum VCR recovery to the concentrate stream with
small to moderate dilution by waste particles.
The synthetic feed’s head grade was calculated to be 1.77
g/t and is higher than expected in the actual dump. The
discard stream comprised of 81% of the synthetic feed mass
to the sorter, which was calculated to be 0.59 g/t. The Au
lost to tailings amounts to 27% of the optical sorter feed.
The cobble reef contributed 21% to the loss of Au in the
tailings.
97%, 10% and 100% of the VCR, cobble and marginal
reef in the sort feed reported to the concentrate resulting in
a calculated grade of 6.95 g/t. Only 9% of the waste in the
feed to the sorter reported to the concentrate stream,
majority of which being the grey quartzite.
Bulk test work
After a successful test conducted with 500 kg of Kloof
waste, a 6 ton sample was tested to establish both the head
grade and the reproducibility of the first test.
Table IV shows the results of a single stage of sorting in
which both VCR and Cobble was recovered.
Recovering both VCR and Cobble resulted in a 13.5%
mass pull to concentrate, which assayed at 1.06 g/t Au. The
discard contained 0.12 g/t Au and this amounts to 40.1% of
the gold in the feed to the sorter. A representative sample of
the fines adhering to the coarse material was assayed at
0.98 g/t.
Figure 3. Kloof Gold Mine: Waste rock types
VIABILITY OF OPTICAL SORTING OF GOLD WASTE ROCK DUMPS
273
Paper 42:text
10/16/09
12:29 PM
Page 274
Figure 4. Programme with colour classes
Figure 5. Flow diagram for a single pass sorting method
Table III
Optical sorter results for the synthetic feed
Stream
number
1
2
3
Stream description
Mass (%)
Au (g/t)
Feed
Tails (accept class)
Concentrate (reject class)
100.0
81.4
18.6
1.77
0.59
6.95
% Au
recovery
100.0
27.0
73.0
Table IV
Optical sorter results recovering VCR and cobble reef
Stream description
Feed
Single stage concentrate
Single stage discard
Mass (%)
Au (g/t)
100.0
13.5
86.5
0.24
1.06
0.11
% Au
recovery
100.0
59.9
40.1
Pilot plant operation
The primary focus of this campaign was to evaluate the
economic viability of optically sorting the +16 mm size
fractions that were produced at Kloof’s rock dumps. This
material was generated by screening the waste dumps to
recover the +16 mm fraction, which has sufficient gold
274
content to be processed further by the mine. In order to
evaluate the sorting option, two issues were investigated:
• The range of variability of the head grade in the
+16 mm fractions
• The throughput and efficiency of the test sorter in order
to specify a large-scale plant performance.
A pilot optical sorting plant was brought onto site from
September 2003 until June 2004. This period can be divided
into 2 phases:
Phase 1: From September 2003 until February 2004 the
plant was operated in its initial design and construction.
After the first week of commissioning using one day shift
only, the plant was ramped up to be operational on a 3 shift
basis during the week and one shift on Saturdays.
Phase 2: After the completion of the first phase, a detailed
evaluation in terms of production performance, gold
recoveries and profitability of the operation was conducted.
After a period of repairs and maintenance and modifications
to the plant, it was decided to run the sorter for one more
month during June 2004 at a 3 shift full capacity.
Description of operation and plant:
• The feed to the sorting plant was contracted to Saldanha
plant hire (SPH) who were already operating the dump
screen at 16mm to produce a supplementary feed to the
mine’s process plant
• The plant includes a feed hopper with variable speed
feeder, conveyors, washing screen and water recirculation
sump, sorter, power generator and loading and hauling
equipment for feed, product and waste handling
• Site establishment took approximately one week,
including wet commissioning of all equipment
• The sorter feed was loaded directly into a feed hopper
equipped with an oversize grizzly and a variable speed
feeder
• The water circulation system around the sorter initially
consisted of a waste skip overflowing into a portable
metal sump. The degritted water was circulated to the
washing screen by means of a submersible pump.
Intermittently, the skip will be drained and the fines will
be collected and sampled for analysis
WORLD GOLD CONFERENCE 2009
Paper 42:text
10/16/09
12:29 PM
Page 275
Figure 6. Flow sheet of optical sorter pilot plant at Kloof Gold Mine
Figure 7. Optical sorter pilot plant at Kloof Gold Mine
Figure 8. Sorted product (left) discarded waste (right)
VIABILITY OF OPTICAL SORTING OF GOLD WASTE ROCK DUMPS
275
Paper 42:text
10/16/09
12:29 PM
Page 276
Table V
Operating data – plant tonnages and availability
Table VI
Operating data – grades and recovery
• During the first month of operation, it turned out that this
skip water treatment system was inadequate to handle the
large volumes of slimes that have been washes off the
feed material. A conical settling tank, including a
flocculation system, was installed. A sludge pump
underneath the cone transferred the slimes to a slimes
dam close to the sorter plant. Later, the dry slimes were
dug up and added to the sorter product
• 4 operators were employed and trained for this project.
Production was run at one operator per shift. The site was
managed by the subcontractor SPH, the Mine and the
general Manager of TollSort
• In order to accommodate the dual purpose of treating
sufficiently large bulk samples and to obtain performance
data, it is proposed that the 8hr day shifts be run on a
single sorting algorithm. This effectively fixed the mass
pull and throughput for that shift. As a rule, feed to the
plant should also come from a single source/area from the
dump
• Product and waste was sampled intermittently during the
shift, and then combined to form one composite product
sample (1 ton) and one composite tailing sample (1 ton)
per day
• During the course of the period, the product and tailings
samples were kept in 1 t bags and transported to Mintek
where they were crushed, sub sampled, pulverized and
analysed for gold content. The 1t samples were
transported weekly to Mintek to minimize cost of
transport and setup of crusher facilities at Mintek
• Kloof took control samples to confirm the assay results.
• The waste produced by the sorting process was loaded
and redumped onto the waste stockpile. The product from
all the sorter runs was stockpiled separately before taken
to the Kloof gold plant for further processing.
Production data and operational results
During the period of operation a total of close to 110 000
tons was processed through the optical sorter. The results
are shown in Tables V and VI:
276
Table VII
Operating costs for Kloof pilot plant
Cost structure
USD
Depreciation (60mnths)
Finance cost
Operating cost (SPH)
Accommodation and S&T
Kloof processing/transport
Salaries
Assays
Maintenance
Hire of ablution
Hire of office
Total
($/day)
($/day)
($/day)
($/day)
($/day)
($/day)
($/day)
($/day)
($/day)
($/day)
($/day)
846
402
1.633
70
415
204
114
379
3
30
4.095
Total cost per ton feed
Total costs per day
Total costs per month
($/t)
($/day)
($/month)
2,47
4.095
81.902
Table VIII
Profit/loss evaluation Kloof pilot plant
Gold price June 2004
$/ounce
380
R/$ exchange rate
R/$
6,6
Gold Price
R/g
81
Feed rate
(t/hr)
82
Mass percent to sorter concentrate
(%)
2,78
Head grade of feed
(g/t)
0,27
Sorter concentrate grade
(g/t)
5,7
Fines grade
(g/t)
1,5
Overall concentrate grade
(g/t)
3,9
Value of product per month
($/month)
72.474
Total costs per month
($/month)
81.902
Loss per month
($/month)
(9.428)
WORLD GOLD CONFERENCE 2009
Paper 42:text
10/16/09
12:29 PM
Page 277
Table IX
Profit/loss prediction at 2009 gold price and cost levels
Gold price June 2004
$/ounce
800
R/$ exchange rate
Gold price
Feed rate
Mass percent to sorter concentrate
Head grade of feed
Sorter concentrate grade
Fines grade
Overall concentrate grade
R/$
R/g
(t/hr)
(%)
(g/t)
(g/t)
(g/t)
(g/t)
8,0
206
82
2,78
0,27
5,7
1,5
3,9
($/month)
($/month)
($/month)
152.577
108.111
44.466
Value of product per month
Total costs per month
Loss per month
Operating results June 2004
Even though it was only a pilot plant, the participating
parties were expecting this operation to be a profitable
entity. Unfortunately, in 2004 the low Gold price $380 per
oz made this operation unprofitable. It was for this reason
that the project was stopped.
The overall operating cost per ton was 2.47 US$ per t.
The upgrade from the feed grade of 0.27 g/t to a sorter
product of 5.7 g/t shows a 20 fold improvement. The waste
stream had negligible gold losses.
Estimated operating results 2009
Over the last year, the price of Gold has improved
significantly. When predicting a scenario for 2009, we have
taken these 2004 figures as a basis and increased the costs
by 60% and used a Gold price of $800 per oz to get the
following estimated results Table IX):
This may be a very simplistic approach, but it highlights
the potential of optical sorting in a gold waste dump
recovery application.
Conclusion
Sensor based sorting is still not widely recognized and used
as a beneficiation process in mineral applications (other
than industrial minerals). The pilot work at Kloof Gold
Mine in 2004 has shown that misplaced gold reef can
effectively be separated from waste rock at low mass pull to
concentrate (5% to 10%) and at a gold recovery rate of
70%.Rapid advances in the development and improvement
of sensors open up more and more application fields.
Previously uneconomical resources, such as shown here
with waste rock gold dumps, can now be exploited at good
profit margins.
References
1. MARSDEN, J.O. and HOUSE, C.I. Published by
SME in 2006.
2. WILLS, B.A. Camborne School of Mines, Cornwall,
UK, ‘Mineral Processing Technology–An
Introduction to the Practical Aspects of Ore
Treatment and Mineral Recovery’ 5th Edition,
Pergamon Press. 1992.
3. SALTER, J.D. and WYATT, N.P.G. Sorting in the
Minerals Industry: Past, Present and Future, Mineral
Engineering, vol. 4, nos. 7–11, 1991. pp. 779–796,
Pergamon Press, Great Britain.
4. WOTRUBA, H. Sensor Sorting Technology – is the
minerals industry missing a chance?, XXIII
International Mineral Processing Congress, Istanbul,
Turkey, 2006. pp. 21–30.
5. WEATHERWAX, T.W. Integrated Mining and
Preconcentration Systems for Nickel Sulphide Ores,
The University of British Columbia. 2007.
6. HARBECK, H. and KROOG, H. New Developments
in Sensor Based Sorting, Montan University Loeben,
Austria, January 18, 2008.
7. BERGMANN, C. Kloof Test Report, 2003, Mintek,
TollSort.
Lütke von Ketelhodt
General Manager, Commodas (Pty) Ltd
Lütke was born and raised in Johannesburg, South Africa. He is a Mining Engineer with a
Higher National Diploma in Metalliferous Mining. For 4 years he worked in the Deep-Level
Gold Mines in the Witwatersrand area as well as Coal (Douglas Mine) and Chrome (Winterveld
Chrome Mine). He also holds a Bachelor of Commerce degree from the University of the
Witwatersrand.
He moved from South Africa to Germany and then Barcelona Spain, where he held various
positions in Financial Controlling and Financial management within the AEG Group.
He returned to South Africa where he got back into the mining game as managing director of
WIMICO, which was a company, specialized in underground Rock-Consolidation.
Further Financial management positions followed within the Mannesmann Demag Group before
he joined IMS Engineering as the General Manager for the MikroSort Division. This is where he
worked closely together with his current employer Commodas.
It is thus since 2002 that Lütke is involved in introducing Sensor Based Sorting equipment into the Southern African mining
and minerals market. Apart from sorting equipment he also managed the IMS sales department for crushing and screening
equipment.
July 2007 he joined Commodas as General Manager – North America. He is setting up and managing the newly formed
Commodas Inc which is based in Toronto, Canada to serve the North American mining market with the Commodas sorting
technology. In July 2009 Lütke returned to South Africa to manage Commodas (Pty) Ltd which have their office in Sandton.
VIABILITY OF OPTICAL SORTING OF GOLD WASTE ROCK DUMPS
277
Paper 42:text
278
10/16/09
12:29 PM
Page 278
WORLD GOLD CONFERENCE 2009