Minerals Engineering 18 (2005) 785–790

This article is also available online at:

www.elsevier.com/locate/mineng

Influence of open circuit regrind milling on UG-2 ore

composition and mineralogy at ImpalaÕs UG-2 concentrator
a,*,1
E. Nel , M. Valenta b, N. Naude c

a
Impala Platinum Limited, P.O. Box 9784, Rustenburg 0300, South Africa
b
Hatch Africa (Pty) Ltd., Private Bag X20, Gallo Manor 2052, South Africa
c
Tshwane University of Technology, Private Bag X680, Pretoria 001, South Africa

Received 13 December 2004; accepted 20 January 2005

Available online 31 May 2005

Abstract

Impala Platinum Limited is the worldÕs second largest platinum producer, producing more than 1.9 million ounces of the pre-
cious metal last year. As part of ImpalaÕs ongoing expansion programme, tonnage through their principal UG-2 processing facility,
near Rustenburg, was increased by 25% in 2001.
The expansion project achieved the increased throughput target but the anticipated recovery increase was not realised. Test work
as part of an optimisation project indicated that open circuit regrind milling would significantly increase the overall recovery of the
plant. Subsequently, the high grade regrind mill was converted from closed circuit to open circuit configuration, during September
2003, and a 3.7% recovery increase was realised.
This paper attempts to explain the recovery increase achieved by examining and comparing metallurgical as well as mineralogical
changes in mill product from before and after the open circuit conversion. Other initiatives implemented during the structured opti-
misation project have already been published by Martin2 et al. [Martin, C.J., Nel, E., Raabe, H., 2004. PGM ore processing at Impa-
laÕs UG-2 concentrator in Rustenburg, South Africa. In: Proceedings of 36th Annual Canadian Mineral Processors Conference,
Ottawa] and will thus not be discussed any further.

2005 Elsevier Ltd. All rights reserved.

Keywords: Precious metal ores; Grinding; Liberation; Ore mineralogy; Particle size

1. Introduction chromite (Cr2O3) with the remainder being made up of

1.1. UG-2 ore mineralogy and processing
Extensive mineralogical studies done by Martin et al.
(2004) on the Impala UG-2 ore have characterised the
mineralogical composition of the UG-2 ore processed.
The ore normally contains between 20% and 30%
*
Corresponding author. Tel.: +27 14 569 8035; fax: +27 14 569 8002.
E-mail address: eugen.nel@implats.co.za (E. Nel).
1
Paper is in partial fulfilment of M.Tech. (Metallurgical Engineer-
ing) degree.
2
SGS Lakefield, Canada.
aluminium silicates (primary magnesium alumino-sili-
cates such as feldspars, pyroxenes and chlorite, and
hydro-thermally altered silicates such as amphiboles
and talc).
It was also shown that sulphide mineralisation is
sparse compared to Merensky ores. The feed to ImpalaÕs
UG-2 plant contains 0.1–0.2% sulphides; these are com-
posed mainly of Pyrrhotite (roughly 50% of all sulp-
hides), Pentlandite (roughly 35%) and Chalcopyrite
(roughly 10%).
The mix of PGE in UG-2 constitutes about 45% plat-
inum, 25% palladium, 10% rhodium and 15% ruthenium.
UG-2 also contains minor quantities of copper and

0892-6875/$ - see front matter
2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2005.01.031
786 E. Nel et al. / Minerals Engineering 18 (2005) 785–790
nickel sulphides. PGE sulphides comprise 70% of the
platinum group minerals (PGM) in the UG-2 plant
feed, the remainder largely being alloys of iron, lead, ar-
senic and antimony, or tellurides. As is typical of such
ores, UG-2 platinum group mineralisation is highly
complex.
From the metallurgical perspective, however, PGM
speciation is a minor issue as it tends not to dictate flo-
atability. For example, QemSCAN studies of the Impala
UG-2 plant feed have shown that the composition of
fast-floating PGM is similar to that of the slow floating
PGM, and the unfloated PGM, indicating no clear
‘‘hierarchy’’ in mineral floatability. This is consistent
with similar studies on the Lac des Iles circuit in Canada
(Martin et al., 2003).
To a large extent the PGMs are finely disseminated,
the average grain size being less than 10 lm, so that
grain size, liberation and association tend to dictate min-
eral floatability. Platinum group mineral grain size and
association can be split into four categories:
• liberated PGM,
• PGM associated with base metal and iron sulphides,
• PGM occurring on host mineral grain boundaries
(mainly silicates),
• PGM locked in silicates.
Floatability of these four categories decreases from
the top to bottom of the list with locked PGMs being al-
most impossible to float.
Low Grade Rougher
UG-2 Plant
Feed
Crushing
plant
Low Grade Mill
LG Final Concentrate
High Grade
Regrind Mill
Primary High Grade Rougher
Primary High Grade Cleaner
Primary High Grade Recleaner
Fig. 1. Ore separation circuit.
Processing of the UG-2 ore has always been problem-
atic due to the high chromite content of the ore feed and
even though chromite does not respond to floatation,
ultra fine chromite particles will report to the floatation
product largely through entrainment.
Concentrates with high chrome values present signif-
icant complications to the downstream smelting process
and thus Impala limits UG-2 chrome in concentrate val-
ues to less than 3%. This is normally achieved through
reducing the generation of chrome slimes through over
grinding in the secondary milling step and by restricting
water recovery.
1.2. Ore separation circuit
UG-2 ore has been processed through a designated
plant at Impala since the early 1990s with a number of
circuit changes being made through the years. Originally
the ore was processed through a conventional mill–float
circuit that was changed to a MF2 (mill–float–mill–
float) circuit in the mid-90s. During the late 1990s the
need was identified to increase UG-2 processing as the
ore profile from the mine was changing to include more
UG-2 and less Merensky ore.
Pilot plant test work by Hinde and Naik (1999a,b)
from Mintek established a unique ore separation flow
sheet that would allow for a 25% increase in mill
throughput at increased PGM recovery.
The main features of the ore separation circuit as
shown in Fig. 1 include:
Final
Tails
Low Grade Cleaner
Low Grade Recleaner
Secondary High Grade Rougher
Secondary High Grade
Cleaner and Recleaner
HG Final Concentrate
 E. Nel et al. / Minerals Engineering 18 (2005) 785–790
1. conversion of the closed circuited primary SAG mill
to a fully autogenous open circuit mill with classifica-
tion done by screens instead of cyclones;
2. the screen undersize is chrome and PGM rich and is
treated through a primary float followed by regrind
milling and secondary floatation (mill–float–mill–
float circuit);
3. the screen oversize is relatively low in chromite and
PGMs and is treated separately through a secondary
milling and floatation circuit (mill–mill–float circuit).
The initial test work included open circuit high grade
regrind milling but the final design was based on con-
ventional closed circuit milling with primary float tails
combining with mill discharge to form cyclone feed. Cy-
clone overflow was fed directly to the secondary float
circuit.
The circuit, as shown in Fig. 1, was commissioned in
2001 and after a 12-month commissioning period the de-
sign throughput was easily achieved. Unfortunately the
expected recovery increases were never completely rea-
lised which led to an extensive optimisation campaign
being initiated in October 2002.
2. Bench and pilot scale open circuit tests
A Mintek modeling and simulation study (Hinde,
2002) done at the Impala MF2 plant (closed circuit re-
grind similar to the UG-2 plant and treating the same
UG-2 ore) indicated the following:
1. There was a significant bypass of fine material to
cyclone underflow and some bypassing of coarse
material to cyclone overflow. Up to 50% of the fines
and water was bypassing the classification process
and reported directly to the underflow.
2. More significant was the difference in cut-size seen
from the Cr2O3 and SiO2 partition curves (Fig. 2)
indicating a cut-size for the silicates being almost
double that of the chromite. The difference in cut-size
was not only due to differences in specific gravity but
also due to dense media effects. This results in over
grinding of the chromite fraction resulting in
increased chromite recovery downstream.
The sampling results were used in computer simula-
tion studies by Mintek to determine the effect of improv-
ing cyclone efficiencies on the partition curves for SiO2
and Cr2O3. It was shown that an increase in cyclone effi-
ciency will only have a minimal effect on reducing the
+106 lm fraction of silicates (less than 2% reduction)
while the 38 lm chrome fraction increased by 7%.
Improving cyclone efficiencies will thus only lead to an
increase in the chrome-sliming problem through
entrainment.
787
Fractional recovery to
cyclone underflow
1.2
1
Solids
Cr2O3
0.8 SiO2
0.6
0.4
0.2
0
10 100 1000
Particle mesh size
Fig. 2. Partition plot for chrome and silicates at MF2 plant.
Based on the campaign results and as part of the ex-
tended UG-2 optimisation programme, the influence of
open circuit high grade regrind milling at the UG-2
plant itself was investigated by Hinde (2002). Evaluation
data of the circuit obtained during mass balancing cam-
paigns was used in simulation models and it was deter-
mined that
1. open circuit milling would result in a major coarsen-
ing of the chromite size distribution with only minor
changes to the silicate size distribution;
2. the power draw of the high grade mill was signifi-
cantly lower than theoretically predicted by MorrelÕs
power equations. This poor milling performance was
mainly due to the low ball filling as well as an exces-
sive volume of pulp in the mill;
3. if the ball filling and milling conditions could be
improved silicate grinds of up to 70% 75 lm could
be achieved;
4. if these improved grinds are achieved and optimal flo-
tation conditions exist (such as in a lab environment),
then a recovery increase of 2–3% could be realised.
Accordingly, the secondary high grade regrind circuit
was modified into an open circuit configuration as
shown in Fig. 3.
This circuit was commissioned in September 2003
with an immediate increase in plant recovery being
observed.
Primary float rougher tail
Secondary float feed
Fig. 3. Open circuit regrind milling.
788 E. Nel et al. / Minerals Engineering 18 (2005) 785–790

3. Effects of open circuiting draw was also achieved although the change in the %
75 lm fraction was never realised. This questions
The change in plant performance for the 12 months whether using ‘‘% 75 lm’’ as a measure in the study
preceding open circuiting and the 12 months following of the effect of particle size in the flotation response is
open circuiting are shown in Table 1. appropriate.
There was a significant increase in circuit recovery at
very much the same grind (the +150 lm particle fraction 3.1. PGM distribution
was not reduced at all). This would indicate that grind
alone is not very good indicator of floatability with The PGM size distributions, as shown in Fig. 4, show
other factors also affecting flotation performance. a marked move towards the finer fraction. After open
Also interesting to note is the significant increase in circuit milling almost 60% of the PGMs are contained
power draw with a small apparent impact on product in the 38 lm fraction. This indicates a marked
size. Under the closed circuit milling conditions the cir- improvement in PGM liberation and improvement in
culating loads made operating at relatively high ball recovery potential for the milled product. However,
loads (greater than 25%) virtually impossible as the flow there is some doubt as to the capability of the existing
through the mill would flush the balls out. Under open flotation cells to recover the ultra-fine particles which
circuit conditions the feed to the mill inlet is reduced could lead to an increase in the losses in this fraction.
by up to 60% and the feed density increased to 2.2 t/ Of benefit is also the reduction in PGMs in the coarse
m3. This allowed for operating at ball loads in excess +150 lm fraction from 14% to 7% as this size is nor-
of 30% and improved solids residence times at the higher mally very difficult to float.
feed densities.
The reduction in chrome in concentrate was as ex- 3.2. Chrome distribution
pected from the studies while the increase in recovery
confirmed the findings in the tests. The increased power The chrome distribution (Fig. 5) shows a change in
the chrome contained in the coarse fraction. This indi-
cates that the over grinding of chromite particles (slim-
Table 1
ing) during closed circuit milling has been eliminated.
Plant performance data changes
The percentage chrome contained in the fine fraction
Closed circuit Open circuit
( 38 lm) has also been reduced which should lead to
Tons milled 448,000 t/m 462,000 t/m reduced entrainment and thus lower chrome in concen-
Cr2O3 in conc. 2.88% 2.39%
trate values.
High grade circuit recovery Increase by 3.2%
Overall plant recovery Increase by 3.7%
Conc. grade 133 g/t 139 g/t 3.3. Mineralogy
High grade mill % 75 lm 59% 62%
% +150 lm 13.8% 14.3% Samples were taken of the regrind mill product before
kW h/t milled 7.42 11.93
and after conversion to open circuit milling and sent for

Fig. 4. PGM distribution by size.

 E. Nel et al. / Minerals Engineering 18 (2005) 785–790 789

Fig. 5. Cr2O3 distribution by size.

Liberated
13%

Locked in silicates
Associated with BMS 33%
9%

Associated with BMS/Silicates

11%

Attached to silicates
34%

Fig. 6. PGM deportment during closed circuit milling.

QemScan analysis at SGS Lakefield, Canada. The min-
eralogical distribution can be seen in the graphs (Figs. 6
and 7).
The liberated PGM fraction increased significantly
from 13% to 24% mainly at the expense of the PGMs
attached to silicates. Overall the fraction of PGMs
locked in silicates (non-floating fraction) remained
about the same at between 33% and 35%. This would
indicate that the amount of exposed PGMs did not
change significantly but rather that the float rank or ease
at which the particles will float increased.
It is surmised that the reason for this is that the con-
ditions in the regrind mill in open circuit mode is much
more conducive to efficient attritioning, partly due to the
effect of the high density chrome particles in the slurry as
well as improved mill densities being maintained. This
mode of grinding should preferentially ‘‘shear’’ the
PGM/BMS particles off the hard silicate particles at
the granular interface zone as this is the point of weakest
bond. Of concern was a slight decrease in the fraction of
liberated PGM particles as well as total PGM particle
size observed as the smaller sized particles will be more
790 E. Nel et al. / Minerals Engineering 18 (2005) 785–790
Liberated
24%
Associated with BMS
19%
Associated with BMS/Silicates
11%
Fig. 7. PGM deportment during open circuit milling.
difficult to float. Further studies are however, being
done to confirm this.
4. Conclusions
It was predicted that open circuit regrind milling at
the UG-2 plant would result in
1. less circulation of higher density chromite particles
resulting in a coarser chrome particle size distribution
for the circuit. This in turn should lead to reduced
Cr2O3 content in the final concentrate due to reduced
entrainment of chromite slimes;
2. reduction in the amount of coarse silicates which
would lead to an increase in floatability of the ore;
3. increased fineness of grind associated with improved
mill power draw and efficiencies as well as the
improved silicate grinding.
All three of these outcomes were achieved to varying
degrees with the change in fineness of grind not being as
substantial as predicted. However, it was also observed
that
1. fraction of PGMs locked in silicates were not reduced
at all;
2. the mode of breakage favours shearing rather than
impact breakage;
3. overall circuit size distribution expressed as %
75 lm did not change significantly;
4. because the PGM particles themselves form such a
small fraction of the overall feed to the mill (4 g/t
equates to 0.0004% by mass of total feed) an increase
in liberated PGMs cannot always be correlated with
an increased grind. It is thus possible to improve lib-
eration and floatability in a circuit without changing
Locked in silicates
35%
Attached to silicates
11%
overall grind but rather by changing the mineralogi-
cal characteristics of the mill product. One has to
consider the distribution of valuable mineral particle
size rather than the absolute figure i.e., in terms of %
75 lm.
A number of challenges been identified, the most sig-
nificant being the identification of a flotation process
suited to the recovery of the ultra fine liberated PGM
particles (less than 10 lm).
Acknowledgements
I would like to thank the management of Impala for
the support for this paper as well as the personnel at the
UG-2 plant for their efforts in ensuring the success of the
open circuit conversion.
References
Hinde, A.L., 2002. Modeling and simulation study to guide the
optimization of the MF2 mill circuit at ImpalaÕs Main Plant.
Confidential Mintek Report.
Hinde, A.L., Naik, S.M., 1999a. Piloting test work to provide design
parameters for upgrading ImpalaÕs UG-2 plant. Confidential
Mintek Report C2901M.
Hinde, A.L., Naik, S.M., 1999b. Optimisation of secondary high-grade
mill circuit at ImpalaÕs UG-2 plant. Confidential Mintek Report
C3542M.
Martin, C.J., Ohrling, T., Olsen, T., Taggart, P., 2003. Addressing
palladium floatation start up problems at the new Lac des Iles
Concentrator, Thunder Bay, Ontario. In: Proceedings of 35th
Annual Canadian Mineral Processors Conference, Ottawa.
Martin, C.J., Nel, E., Raabe, H., 2004. PGM ore processing at
ImpalaÕs UG-2 concentrator in Rustenburg, South Africa. In:
Proceedings of 36th Annual Canadian Mineral Processors Confer-
ence, Ottawa.