Implementation of An Alternative Sampling Protocol at a Typical UG2

Concentrator in South Africa

N. Sukha1, C. Kruger2 & W. Slabbert3

1MetallurgicalTechnical Engineer, Anglo American Platinum, Johannesburg, South Africa

2Principal Sampling & Evaluation, Anglo American, Johannesburg, South Africa
3 Process Manager, Multotec Process Equipment Pty (Ltd), Johannesburg, South Africa

ABSTRACT

A UG2 Concentrator Plant in South Africa was historically under-accounting in terms

of Platinum Group Metals. The key reason being the biased sampling of the main feed
to the plant resulting from sub-sampling of the primary increment with a secondary
vezin sampler system. Conventionally, a primary vezin or linear cross stream sampler
is used to take a minimum number of primary increments per sampling campaign. For
larger increments the flow through the secondary vezin, could be restricted by means of
a conical hopper with regulated compressed air at the reduced hopper outlet to agitate
the primary slurry increment until sub-sampling is complete. The air agitation alone did
not keep all particles, of varying size and density in “equal” suspension throughout the
sub-sampling duration of the primary increment. The finer, high grade particles were
suspended for much longer compared to the coarser and heavier, low grade particles
resulting in a final sample that was biased low in coarse, low grade particles. This was
established through a series of vezin credibility tests. To combat the resulting
segregation error, a re-design of the intermediate hopper system was considered and
this included (i) retrofitting the design of the discharge nozzle on the original hopper (ii)
use of a new hopper design that mechanically agitates the slurry with an original
discharge nozzle design (iii) use of a new hopper design with the new discharge nozzle
design. Replica vezin credibility and chronological sub-sampling tests were done using
all the intermediate hopper configurations to measure the evident bias. The results
obtained using the original hopper and new nozzle design showed the best
improvement in the bias related to particle size distribution in the sub-sample and reject,
indicating that particle segregation had been significantly reduced to give more credible
sampling results. A notable improvement in plant accountability was realized with an
unbiased relationship between build-up (calculated) and measured feed grade.

Keywords

Secondary vezin sampling, intermediate hopper, mechanical agitation, particle segregation, bias.
INTRODUCTION

The importance of sampling in the mining industry cannot be over-emphasized, whether in exploration, in
mining or in mineral processing (Bartlett, 2005). The conventional wisdom suggests that when the rules and
procedures for representative sampling are well defined and followed and the sampling equipment is in
good order, unbiased samples will be obtained. In sampling, there are two major areas where bias can exist,
namely, sampling and sample preparation. Sampling bias generally occur when (i) increments coincide
with cyclic events (ii) when only a portion of the stream is being sampled (iii) where cutter specifications
are not being adhered to (iv) when sample containers are overfilled (Kruger & Millar, 2002). Therefore, the
best defense against any sampling bias is the correct sampling protocol, correct mechanical design of the
sampling rig and adequate control and maintenance during its operation (Bartlett, 2005; Kruger & van
Tonder, 2014).

The 4T (individual elemental analysis comprising of elements: Platinum, Palladium, Rhodium and Gold)
accountability for all Concentrator Plants within Anglo American Platinum is calculated, trended and is
used as a risk management tool to rapidly determine metal content discrepancies between input and output
streams. A UG2 Concentrator Plant in South Africa was historically under-accounting in terms of Platinum
Group Metals (Platinum, Palladium, Rhodium and Gold). The current practice at this Concentrator Plant is
that crushed run-of-mine UG2 ore is milled in a semi-autogenous (SAG) mill and the mill product is
classified using a screen to produce undersize and oversize material streams respectively. The oversize
classification screen material is sent back to the SAG mill for further grinding. The undersize material is
gravity fed to a surge tank and this material is then pumped to the primary rougher flotation circuit. Prior
to being fed to the primary rougher flotation circuit, the material is sampled by an automatic, two-stage,
vezin-vezin sampler.

The debate on the source of under-accounting pointed towards possible non-representative sampling or
biased sampling occurring in the Concentrator Plant feed sampling system. There were numerous
indications that pointed in the direction of the feed sampling system being the problem. Process related
challenges such as poor classification screening efficiency and missing classification screen panels led to
unnecessary chokes in the feed to the vezin-vezin sampling system. These chokes only heightened the poor
accountability trend further. Internal and external audits conducted indicated that all the other parameters
contributing to the determination of the accountability value were not to be questioned. There were also no
obvious or noted changes in ore blend ratios and primary mill grind. A sensitivity analysis done using the
plant accountability model also indicated that the likely contributor to the poor accountability was the feed
grade as opposed to the other parameters. It was hypothesized that the main reason for the consistent
under-accounting may be due to over sub-sampling of finer material into the official samples and
consequently under sub-sampling of the coarser material present in the feed slurry streams. UG2 feed
material size by assay analyses indicate that higher platinum and palladium grades are associated with the
sub 75µm size fractions as opposed to the coarser size fractions above 75µm (Ntlhabane, 2014). Indications
are that because of the under sub-sampling of coarse material, the head grade of the feed into the plant is
overstated leading to an under accountability of metal content.

VEZIN-VEZIN SAMPLER ARRANGEMENT

A vezin sampler is a multipurpose device that collects samples from materials that are either free-falling
from pipes, chutes and hoppers or being pumped vertically down through a pipe. The schematic of the
vezin-vezin sampler combination used at the Concentrator Plant and in which tests were performed is
shown in Figure 1 overleaf.
 Figure 1. Schematic of dual primary vezin samplers.

This device operates by one or more cutters revolving on a central shaft, passing through the sample stream
and collecting a fixed percentage of the total material (Trottier & Dhodapkar, 2012). This sampling system
comprises of a primary sampler and a secondary sampler. The bulk stream that is sampled is generally large
and thus the primary increment that results is often too large to be further processed or prepared. A
secondary sampler is then incorporated as part of the overall sampling system to reduce the primary sample
into a more manageable sub-sample size. A secondary sampler is generally in the form of a vezin sampler
or rotary splitter. Kruger & van Tonder (2014) explain in detail the mechanical design of a typical vezin
sampler and its mode of operation.

EXPERIMENTAL METHODOLOGY

Dual primary vezin sampling system as depicted in Figure 1 was available on site. Under normal operating
conditions, only one primary vezin sampler is meant to operate at a time. The primary sampler has a single
vezin cutter arrangement and the secondary sampler has a four cutter vezin arrangement. The primary
sampler typically produces a primary slurry sample which is then discharged from the primary cutter outlet
through a flexible pipe (“1” in Figure 1) into a Y-feed pipe (“2” in Figure 1) that enables the secondary vezin
to collect primary increments from either primary unit and then into the intermediate hopper (“3” in Figure
1) of 20L volume capacity. The primary sample increment is then discharged from the intermediate hopper
through a nozzle via gravity, at which point, sub-sampling begins. Sub-sampling of the primary increment
results in the generation of an official and reject slurry sample.

Principal bias testing relating to segregation of particles in the intermediate hopper was conducted using
the Vezin Credibility technique (Kruger & Millar, 2002). The primary and secondary sampling stages were
used to collect samples.

The experimental work reported in this paper was conducted in 4 stages. Stage 1 involved experimental
baseline test work on the existing sampling equipment i.e. original hopper and original nozzle. Stage 2
involved test work using the original hopper and a re-designed nozzle. Stage 3 involved test work using
the new hopper design (Mechanical Agitated Hopper - MAH) and the original nozzle. Stage 4 involved test
work using the new hopper and new nozzle designs respectively. All test work conducted was performed
under controlled conditions and supervision. The results from the test work therefore provided a snapshot
of particle size distributions under such conditions.

Twin stream analyses were done for all samples sent to the analytical laboratory to determine the analytical
variance. The analytical laboratory that conducted the assaying is ISO 17025 accredited. A Certified
Reference Material (CRM) matching the samples was used for quality control purposes. These were
randomly placed in each batch of samples that were analysed. All samples were analysed in triplicates and
the relative standard deviation were used to eliminate outliers. If there were no outliers, the average value
of the three results was then reported. Back-up samples were reserved (where possible) for repeat analysis.

HOPPER AND NOZZLE DESIGNS USED IN THE EXPERIMENTS

Original hopper design: The intermediate hopper is a 20L vessel with a conical bottom and fitted with a
pipe close to the bottom exit point that delivers regulated compressed air supply (~2-4 bar pressure) to keep
the slurry particles well agitated and mixed before exiting the hopper through the discharge nozzle.

New hopper design: As per Figure 2 below, the new hopper (MAH) design is equipped with three impellers
or agitators throughout its height to keep the slurry completely agitated for its entire residence time. The
agitators can be adjusted along the length of the agitator shaft and can be rotated 180°.The makeup of the
impellors, in this instance, includes a down thrust impellor positioned as the last impellor on the agitator
shaft with the two up thrust impellors being positioned above the down thrust impellor. The down thrust
impellor is meant to lift the slurry whereas the two up thrust impellors are meant to mix and circulate the
slurry. The mechanically agitated hopper is a 60L vessel with a hemispherical bottom that can hold more
than one primary increment at a time.

Figure 2. Hopper with mechanical agitation (Courtesy: Multotec Process Equipment Pty (Ltd)).

Original nozzle design: The nozzle is 154mm long and has an internal diameter of 49mm at the point of
connection to the hopper. This diameter is reduced gradually from 49mm to 20mm after a straight length
of 105mm. The nozzle discharges into the secondary vezin sampler which is used for sub-sampling until
the intermediate hopper is emptied out.
 Figure 3. Engineering schematic of original nozzle design (Courtesy: Multotec Process Equipment Pty (Ltd)).

New nozzle design: The outline dimensions of the new nozzle are the same as the original nozzle design
to fit into the same installation slot. However, the tapering of the nozzle internal diameter happens
immediately at the top of the nozzle opening.

Figure 4. Engineering schematic of new nozzle design (Courtesy: Multotec Process Equipment Pty (Ltd)).

EXPERIMENTAL PROGRAMME

The first objective was to investigate whether particle segregation occurs in the intermediate hopper of the
UG2 feed sampling system in its original configuration leading to biased sub-sampling by the secondary
vezin sampler. In order to test the hypothesis that particle segregation is present in the intermediate hopper,
two different tests were performed, namely, a vezin credibility test (test 1) and a chronological sub-sampling
test (test 2). The purpose of a vezin credibility test is to verify whether the vezin sampler produces sound,
repeatable and unbiased results. If a vezin sampler is credible then the characteristic/analyte under
consideration should almost be identical for both the official sample and reject sample despite different
mass splits to each stream. For the vezin credibility tests, a total of five test runs for Stage 1 and Stage 2, and
a total of three test runs for Stage 3 and Stage 4 respectively were performed for repeatability purposes. For
each test run of Stage 1 and Stage 2, five individual samples were generated i.e. sample A – reject sample,
sample B - official sample, sample C - primary/feed sample, sample D – reject sample, sample E – official
sample. Similarly, primary, reject and official samples were generated for Stage 3 and 4.

A schematic diagram of the sampling protocol and sample preparation methodology for the vezin
credibility test work is shown in Figure 5.

Figure 5. Sampling and sample preparation methodology for Stage 1, test 1.

The chronological sub-sampling test work was also conducted and these tests were designed to determine
the presence of particle segregation during the intermediate hopper discharge. The intermediate hopper
discharge was sampled intermittently (for example, t0-6sec, t1-12sec, t2-18sec, t3-24sec, t4-until the hopper
emptied out) over a predetermined period of time in order to examine the constitution of the samples with
respect to particle size. In the case that coarse particles settle faster in the intermediate hopper, as the Stokes
Law would suggests (McCabe, Smith & Harriott, 1993), they should exit the hopper earlier and more quickly
during sub-sampling and are therefore not sub-sampled for the entire duration of the sub-sampling
campaign.

A schematic diagram of the sampling protocol and sample preparation methodology for the chronological
sub-sampling test work is shown in Figure 6 overleaf.
 Figure 6. Sampling and sample preparation methodology for Stage 1, test 2.

RESULTS AND DISCUSSION

Stage 1 - Base line test work – original hopper and original nozzle

Figure 7 overleaf shows the cumulative particle size distribution of the primary, official and reject samples
for all the 5 test runs performed. It is clear from Figure 7 that the reject sample is consistently coarser than
both the official and primary “feed” sample and that the primary sample particle size distribution (PSD)
lies between the official sample and reject sample PSD’s.
 Figure 7. Stage 1: Cumulative % passing comparison for primary, official and reject samples.

These results demonstrate that the original sampling system tends to sub-sample less of the coarser particles
than the fine fractions. Figure 8 overleaf is a re-plot of the cumulative mass passing data in Figure 7 and it
shows clearly that a consistent bias exists between the reject and official samples, with the official samples
consistently having a higher cumulative % passing than that of the reject samples across all screen sizes.
Had no bias existed, all the data points for both official and reject samples on the parity chart would be
lying on the 45° line.
 Figure 8. Stage 1: Cumulative % passing comparison (official & reject samples).

The % bias between the reject and official samples were calculated per size fraction and is shown in Figure
9.

Figure 9. Stage 1: % Bias between reject and official samples passing each sieve size.

Figure 9 shows the extent of the bias present between official and reject sample across the spectrum of
particle sizes. The official samples are consistently finer than the reject samples and the bias increases as the
particles size decreases. The largest % bias of 18.5 % is observed for the -38 μm fraction. Considering the
previous suggestion that sub 75 μm size range particles normally have higher PGM grades, such biases
would therefore have an effect on the overall grade of the primary, reject and official secondary sample
resulting in the declaration of an incorrect feed grade for metal accounting purposes.

A size by assay analysis was performed and is shown in Figure 10 overleaf. It can be observed that there is
almost an exponential increase of the Platinum Group Metals (PGM) grade with the decrease in particle
size.
 *Assay data has been factorized for confidentiality purposes

Figure 10. PGM grade association with particle size.

These assay results emphasize that if sampling has a significant bias towards the coarser fraction which is
much lower in PGM grade then the overall grade of the official sample will be much higher than that of the
primary or reject samples. From Figure 11, the official samples (4B and 4E) clearly have the highest PGM
grade values followed by the primary sample (4C) and reject samples (4A and 4D).

*Assay data has been factorized for confidentiality purposes

Figure 11. Stage 1: PGM assays for run 4.

Figure 12 overleaf shows the results of the chronological sub-sampling of the official sample. From the
results, it can be see that for the initial 12 seconds of secondary sampling (T2t0 to T2t1) the 4T grade remains
constant. The 4T grade thereafter increases as the % +38 μm fraction decreases. This proves the point that
coarser particles have a tendency to exit the intermediate hopper faster than finer particles. The base metal
(BM) composition remains fairly constant during the entire sampling period suggesting that deportment of
base metals does not change much with change in particle size compared to the deportment of PGMs.
Historical mineralogy investigations of this ore material have shown that BM deportment is less dependent
of particle size at those PSD ranges (Ntlhabane, 2014). The trend for Cr2O3 deportment on the other hand is
opposite to that of PGM’s as is decreases with particle size.

% +38um 4T g/t

99.99 9.00

99.98 8.00

7.00
99.97
6.00
99.96
% +38μm

4T (g/t)
5.00
99.95
4.00
99.94
3.00
99.93
2.00

99.92 1.00

99.91 0.00
T2t0 T2t1 T2t2 T2t3 T2t4

Time (seconds)

% +38um Cr2O3 avg BM (Cu, Ni, Co, Fe) Avg

99.99 18.00

99.98 16.00

14.00
99.97

BM & Cr2 O3 Avg (%)

12.00
99.96
% +38μm

10.00
99.95
8.00
99.94
6.00
99.93
4.00

99.92 2.00

99.91 0.00
T2t0 T2t1 T2t2 T2t3 T2t4

Time (seconds)

*Assay data has been factorized for confidentiality purposes

Figure 12. Stage 1: Variation of grade with % +38µm retained at different time intervals.

With the clear presence of particle segregation causing consistent bias on sub-sampling in the original
sampling system configuration it was logical to assume that;

(i) In the case of the compressed air supply to the base of the original hopper achieving adequate mixing of
the slurry particles as they leave the intermediate hopper then the source of the segregation would be the
geometry of the discharge nozzle. The original nozzle remains wide for most of its length and tapers off at
the exit end. Segregation could possibly occur in the wide straight section of the nozzle where the coarser
particles can preferentially settle out faster than finer ones. The idea of a new nozzle design was then
suggested and the new nozzle tapers off much earlier and has a long narrower discharge length.
(ii) In the case of the compressed air supply to the base of the original hopper not achieving adequate mixing
of the slurry particles as they leave the intermediate hopper then segregation happens in the hopper itself.
To counter this possibility, a new hopper with hemispherical bottom to combat wall effects possible in the
conical design of the original hopper as well as the addition of internal mechanical agitation was designed.

Replica test work for vezin credibility and chronological sub-sampling as explained in Stage 1 were then
conducted using different combinations of the original hopper, original nozzle and new hopper and new
nozzle designs. The experimental methodology remained essentially the same with minor adjustments with
the incorporation of the mechanical hopper design in Stage 3 and 4.

Stage 2 - Original hopper and new nozzle

Figure 13 shows the individual and combined cumulative particle size distribution of the primary, official
and reject samples for all the 5 test runs performed. There is generally a closer agreement between reject,
official and primary samples as seen on the combined cumulative % passing curve.

Figure 13. Stage 2: Cumulative % passing comparison for primary, official and reject samples.

The parity chart shown in Figure 14 overleaf shows that the reject and official samples are not necessarily
identical at each respective size as the data is generally scattered around the 45° line however an averaged
cumulative % passing comparison shows no net segregation. This indicates that over an entire sampling
campaign, the random bias observed for individual increments would mostly likely average out and not
result in a consistent bias in terms of particle size and hence overall grade of the reject and official samples.
 Figure 14. Stage 2: Cumulative % passing comparison (official and reject Samples).

The % bias between the reject and official samples obtained with the original hopper-new nozzle
arrangement were calculated per size fraction and plotted in Figure 15 which also compares the new bias
values with those obtained from the test work with original hopper-original nozzle. These results show that
no consistent bias is seen in any particular direction as compared with that observed in Stage 1, Test 1
(original hopper-original nozzle). The quantum of the bias per size class is also much less than those
recorded with the original hopper-original nozzle experiments.
 Figure 15. Stage 1 & 2: % Bias between reject and official samples passing each sieve size.

The chronological sub-sampling results for the tests with the new nozzle are shown in Figure 16 and they
indicate that for the initial 18 seconds of secondary sampling (T1t0 to T1t2) the 4T grade remains fairly
constant. The 4T grade thereafter increases as the % +38 μm fraction decreases though to a lesser degree
than what was observed with the original nozzle. The BM composition remains fairly constant as time
progresses with the Cr2O3 composition again following the % +38 μm trend.

*Assay data has been factorized for confidentiality purposes

Figure 16. Stage 2: Variation of grade with % +38µm retained at different time intervals.

Stage 3 – New hopper and original nozzle

Figure 17 overleaf shows the individual and combined cumulative particle size distribution of the primary,
official and reject samples for test runs performed on the new hopper-original nozzle combination.
 Figure 17. Stage 3: Cumulative % passing comparison for primary, official and reject samples.

The reject and official samples are identical at sieve sizes 425μm, 300μm, 212μm and 150μm. The PSD’s at
<150μm deviate with the official sample PSD being finer than the reject sample PSD. The standard deviation
at the 95% confidence levels for the cumulative % passing the various sieve sizes are also shown in Figure
17 by means of error bars.

Figure 18 indicates the comparison of the cumulative % passing the particular sieve sizes for the official and
reject samples for all three runs.

Figure 18. Stage 3: Cumulative % passing comparison (official & reject samples),
From Figure 18, it is clear that the reject and official samples are not identical at each of the respective size
fractions. Two of the three test runs, runs 1 and 3, conducted however indicate a consistent bias with the
official samples being finer than the reject samples. The combined graph indicated a very subtle bias with
the official samples being bias low towards the coarse fraction (i.e. finer). The significance of this bias would
need to be further investigated with more test runs being done to obtain enough data points to perform
sound statistical analysis.

Stage 4 – New hopper and new nozzle

Figure 19 shows the individual and combined cumulative particle size distribution of the primary, official
and reject samples for test runs performed on the new hopper-new nozzle combination.

Figure 19. Stage 4: Cumulative % passing comparison for primary, official and reject samples.

Figure 20 overleaf is a re-plot of the cumulative % passing data in Figure 19 for the official and reject samples
for all three runs.
 Figure 20. Stage 4: Cumulative % passing comparison (official & reject samples).

From Figure 20, it is clear that the reject and official samples are close to each other at the respective size
fractions with the exception of run 12. Run 12 indicates a consistent bias with the official samples being finer
than the reject samples. The combined graph indicates a very slight bias with the official samples being
slightly bias low towards the coarse fraction meaning that segregation may have been reduced with this
configuration (new nozzle design and mechanical agitation). In general, minimal bias is observed.

Figure 21. Stage 1, 2, 3 & 4: % Bias between reject and official samples passing each sieve size

The extent of the bias between official and reject samples measured using all hopper-nozzle combinations
is summarized in Figure 21. In this graph, it is evident that the original hopper-original nozzle configuration
results in greatest bias towards the finer particles compared to other configurations and this bias is
consistent and becomes progressively stronger as particle sizes decrease. The configuration that resulted in
the least bias is the original hopper-new nozzle combination. The new hopper-original nozzle and new
hopper-new nozzle configurations also substantially improve the bias in comparison to the original system
of the original hopper-original nozzle combination though not to the extent recorded with the original
hopper-new nozzle arrangement. This suggests that the replacement of the original system with the
original hopper-new nozzle combination will give improved sampling outcomes that will reduce the
perennial problem of under accounting of 4T that the concentrator has been experiencing.

The paired t-test method was applied to confirm whether there is a consistent bias between the reject and
official samples that are correlated (i.e. it is expected that these measures would change with the change in
the feed conditions) and are significantly different from 0. The differences in % mass retained for each pair
of reject and official sample arising from an independent feed condition was compared in the statistical
analysis. Table 1 indicates the confidence levels for the significance in the bias between reject and official
samples. Red, yellow and green cells indicate greater than 95 % confidence, between 90 % and 94.9 %
confidence and less than 89.9 % confidence respectively.

Table 1. Statistical Confidences from Paired t-tests between reject and official samples.
Summary of Statistical Confidence in the Differences of % Retained between Reject and Official Samples
Stage 1 Stage 2 Stage 3 Stage 4

(Reject Sample A - (Reject Sample D - (Reject Sample A - (Reject Sample D - (Reject Sample F - (Reject Sample F -
Official Sample B) Official Sample E) Official Sample B) Official Sample E) Official Sample E) Official Sample E)

+425μm 99.8 42.7 58.9 32.8 30.02 10.19

+300μm 96.3 95.7 85.0 97.1 73.95 2.75
+212μm 98.0 100.0 51.6 24.5 51.39 80.03
+150μm 93.3 98.7 1.8 57.5 30.22 85.29
+106μm 99.1 99.9 51.8 69.3 75.18 4.11
+75μm 53.7 67.7 82.6 94.5 75.50 97.18
+53μm 29.4 99.2 80.0 12.1 92.06 96.41
+38μm 80.1 90.2 26.2 36.0 24.97 83.37
-38μm 99.8 99.8 8.4 31.4 37.84 92.62

For Stage 1 test work, there is generally a greater than 95% confidence that a bias exists and that the bias is
real and significant between the reject and official sample. The statistical confidence level for Stage 2 test
work indicates that bias is not significant and highlights that modifying the hopper discharge nozzle design
was sufficient in reducing particle segregation in the intermediate hopper of the sampling system. Stage 3
confidence in the difference is comparable to Stage 2 with only 3 test runs being conducted. Stage 3 and
Stage 4 configurations still produced results that were deemed not as statistically biased as Stage 1
configuration. This inherently means that the addition of the agitator does assist in producing positive
results (comparing specifically Stage 1 to Stage 3 where the only change is the hopper design with the nozzle
design being constant).

CONCLUSIONS

The hypothesis that particle segregation is present in the intermediate hopper of the UG2 feed sampling
system was confirmed in the vezin credibility and chronological sub sampling test-work on the original
sampling system. A consistent bias was observed between the reject and official samples with the official
samples being depleted of coarse particles and higher in 4T grade than the reject samples. By means of a
paired t-test, the calculated bias for % mass retained was deemed significant at the 95% confidence level.
This outcome together with the size by assay analysis performed indicated that an under-accounting
scenario would result.
Further work was done to improve the sampling system in the backdrop of the confirmed persistent bias
caused by the original system. A new design of the intermediate hopper and discharge nozzle were done
and retrofitted into the existing sampling system to create various combinations of hopper-nozzle
arrangements which were used in the experiments in an attempt to improve on the sampling bias observed
with the original system. This further work involved test work using (i) the original hopper and a re-
designed nozzle (ii) the new hopper design and the original nozzle and (iii) the new hopper and new nozzle.

The results from the tests conducted using these permutations of the hopper-nozzle arrangements showed
that the original hopper-original nozzle configuration results in greatest bias towards the finer particles
compared to other configurations and this bias is consistent and becomes progressively stronger as particle
sizes decrease. The configuration that resulted in the least bias is the original hopper-new nozzle
combination. The new hopper-original nozzle and new hopper-new nozzle configurations also
substantially improve the bias in comparison to the original system of the original hopper-original nozzle
combination though not to the extent recorded with the original hopper-new nozzle arrangement. The
statistical confidence level for test work with original hopper-new nozzle tests indicates that bias is least
significant and highlights that modifying the hopper discharge nozzle design was sufficient in reducing
particle segregation in the intermediate hopper of the sampling system. This suggests that the replacement
of the original system with the original hopper-new nozzle combination will give improved sampling
outcomes that will reduce the perennial problem of 4T under accounting that the concentrator has been
experiencing.

WAY FORWARD

The nozzle re-design although effective in significantly reducing the bias between the official sub-sample
and reject sample, was still limiting in that only a certain number of primary increments could be taken per
shift or per sampling campaign. The nozzle re-design did not cater for any process variability. By producing
a variogram, the time interval that a sample needs to be taken to overcome process variability can be
determined. The MAH design of 60L capacity together with the nozzle re-design allows for superior
benefits including additional flexibility to increase the number of primary increments taken per campaign,
agitating and circulating the collected increments and sub-sampling thereafter to produce a more
representative sample. This will be applicable in all sampling scenarios but more specifically in scenarios
where the process variability in the feed (such as mill grind) is high and increments need to be taken over
shorter but regular time intervals. It is envisaged that the MAH will replace all conventional intermediate
hoppers in the Platinum division as time progresses.

ACKNOWLEDGMENTS

The authors would like to thank:

 Anglo American Platinum for the opportunity to conduct test work at one of their Concentrator
Plants and providing various resources throughout the test work period;
 Multotec Process Equipment Pty (Ltd), the original equipment manufacturer of the sampling
equipment for assisting with the re-designs and providing graphics included in this paper;
 The Technical Services department of the Concentrator Plant where the experiments were
sanctioned and carried out.
REFERENCES

Bartlett, H.E. 2005. Confidences in Metallurgical Balances Estimated from the Errors in Mass Measurement,
Sampling and Analytical Determinations. Proceedings of the Conference on Exploration, Mine, Met and
Environmental Sampling. 03-04 November 2005. The South African Institute of Mining and Metallurgy,
Geological Society of South Africa, Geostatistical Association of Southern Africa, Eskom Conference
Centre, South Africa.

Kruger, C.J. & Millar, N.C. 2002. Evaluation Manual. Group Evaluation Metal Accounting, Anglo American
Platinum, South Africa.

Kruger, C.J. and van Tonder, E. 2014. Pitfalls in Vezin Sampling for Finely Crushed Materials, The Journal
of The Southern African Institute of Mining and Metallurgy, Vol 114, pp. 47 – 52.

McCabe, W.L., Smith, J.C. & Harriott, P. 1993. Unit Operations of Chemical Engineering. 5th Edition.
Singapore: McGraw-Hill.

Ntlhabane, S. 2014. Benchmark Mineralogical Investigation of UG-2 Concentrator. (Research report, no.
MPR/14/183). Johannesburg, South Africa: Anglo Research, Crown Mines Campus.

Trottier, R. & Dhodapkar, S. 2012. Sampling Particulate Materials the Right Way. ProQuest SciTech Collection.