Designing otation circuits for high nes recovery

J.D. Pease
a,
*
, D.C. Curry
a
, M.F. Young
b
a
Xstrata Technology, Level 2, 87 Wickham Terrace, Brisbane, QLD 4000, Australia
b
Xstrata Zinc, Mount Isa, QLD 4825, Australia
Received 30 June 2005; accepted 30 September 2005
Available online 15 November 2005
Abstract
Many otation operations need to improve liberation to increase recovery, yet operators are often concerned that ner grinding will
produce slimes with low otation recovery. The development of large scale stirred mills like the IsaMill has now enabled simple circuits
with excellent nes recovery. The ability to grind in a small footprint with high energy eciency and sharp size distribution supports a
staged grinding and otation approach, applying grinding power only where liberation is needed. Most importantly, the inert grinding
environment has profound benets for otation compared with conventional steel grinding. Flotation rates and recoveries are improved
and reagent needs are lowered. This paper describes design principles for nes otation and challenges common myths about nes and
otation cells. It considers several operations where the design principles have been successfully applied, and describes performance on a
size-by-size mineralogical basis.
It is argued that the substantial benets of high eciency, distributed, inert grinding will have much wider implications for circuit
design beyond ultrane grained ores.
2005 Elsevier Ltd. All rights reserved.
Keywords: Classication; Fine particle processing; Froth otation; Grinding; Liberation
1. Introduction: Fines do oat!
The conventional view of otation of dierent size frac-
tions is shown in Fig. 1. Sampling of most otation plants
will produce a similar curve. Not surprisingly this leads to
the view that nes do not oat, and operators are careful
to avoid overgrinding or sliming of feed.
However for ne-grained ores there is no choiceclean
concentrates can only be made from liberated particles. In
some cases achieving liberation means grinding to sizes
below 10 lm. For example, at Xstratas McArthur River
Mine (MRM), grinding to P80 of 7 lm is essential to pro-
duce a saleable concentrate. At Mt Isa, grinding streams to
P80 of 12 lm and 7 lm is essential to get adequate recov-
ery. In these plants, creating slimes is absolutely essential
for good otation recovery. Between them, these opera-
tions produce around 1 Mt/y of concentrates by otation
of particles mostly ner than 10 lm, at over 80% recovery.
In fact, the best otation recovery is in the slimes. At
MRM, 96% of the individual particles recovered are ner
than 2.5 lm. So ne particles do oatand they oat very
well, as shown by Fig. 2, the recoverysize curve for sphal-
erite from rougher concentrate at Mt Isa.
1.1. Are ne particles dierent?
There is much discussion about the dierent behaviour
of nes in otation, and the need for special attention
high energy otation machines, special impellors, small
bubble sizes, attritioning, dierent reagents, entrainment.
In our experience, after making over 10 million tonnes of
ne concentrate in the last decade, there is nothing special
about nes, they just respond dierently because
0892-6875/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2005.09.056
*
Corresponding author. Tel.: +61 7 3833 8533; fax: +61 7 3833 8555.
E-mail address: [email protected] (J.D. Pease).
This article is also available online at:
www.elsevier.com/locate/mineng
Minerals Engineering 19 (2006) 831840
They have higher surface area per unit mass, so need rel-
atively higher reagents.
They have less momentum, so tend to follow water more
easily than coarse particlesless energy for bubble
attachment, more tendency for entrainment.
As a result, otation rates will be slower, and lower
cleaning densities or froth washing may be needed to
counter the extra entrainment.
They tend to be more aected by surface coatingsper-
haps because the high surface area ratio makes them
more reactive, or perhaps because their low momentum
means it less likely that loose surface deposits are
abraded o by other particles.
They tend to be more aected by water chemistry and
ions in solution.
The high surface area to volume means that froths can
be tenacious and thickening and ltering is more
dicult.
Their otation kinetics can be slower and may be similar
rate to coarse composite particles. They can still oat
with big bubbles, but smaller bubbles increase their o-
tation rate.
Though these eects become stronger as particles
become ner, there is no mysterious sharp distinction
between coarse and ne particles, just a steady grada-
tion as particles get ner.
1.2. Bursting the bubble myth on otation cell design
We believe that eorts to design otation impellors and
cell energy input specically to suit nes are misdirected.
Mt Isa and MRM successfully oat more nes than any
other sites, and we use a range of dierent otation equip-
mentotation columns, Jameson cells, several dierent
makes of conventional cell. Our selection criteria for ota-
tion cells has been on the basis of
Ensure adequate residence time.
Ensure adequate lip length.
Apply froth washing if benecial.
Choose the cheapest cell available that meets these
criteria.
If you nd this controversial, look again at Fig. 2. Flota-
tion at Mt Isa and MRM is done in ve dierent types of
cells, usually chosen because they were cheap (often because
we already had them in the plant or in the scrap yard). We
put a lot of science into achieving the results in Fig. 2, but
none of it was about otation mechanisms, it was about
achieving the correct level of liberation and the right surface
chemistry. Once this is achieved otation is easy in any
device. Put simply, otation impellors do not grind compos-
ites and they do not make particles hydrophobic. They just
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
C7 C6 C5/C4 C3-C1 38/53
75
Size fraction
Z
i
n
c

R
e
c
o
v
e
r
y

i
n

s
i
z
e

f
r
a
c
t
i
o
n

%
0.0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
0-4m 4-8m 8-16m 16-38m 38-75m
Recovery
Size Distribution
Fig. 2. Mt Isa zinc recovery from rougher concentrate by size.
Size (microns)
Recovery
5 10 30 50 200 80 100 150
Liberated Fines:
High surface area
Need high collector
and low depressant
Intermediate : fast floating
lower collector need,
composites need depressant
Coarse Particles
Low liberation
Fig. 1. The conventional viewnes do not oat.
832 J.D. Pease et al. / Minerals Engineering 19 (2006) 831840
make bubbles. The choice between cells is purely one of what
capital is needed to achieve the required residence time, and
to reduce entrainment (either by dilution cleaning or by froth
washing). In Mt Isas case, new Jameson cells would have
been a great technical solution, using ne bubbles to quickly
recover nes in a small space that makes froth washing easy.
While this would have been a low capital solution, it could
not compete with using the existing unfashionable old cells.
Of course, smaller bubbles have much higher surface
area, so nes that are already hydrophobic will be collected
in a smaller volume. In an existing plant with xed otation
capacity, this faster otation rate converts into higher
recovery in the xed equipment. This is probably the
source of the misunderstanding that nes need small bub-
bles. Big bubbles will still oat hydrophobic nes, just more
slowly so more cells are needed. So a device that creates
smaller bubbles like the Jameson cell or other pneumatic
devices will have a smaller footprint and lower capital to
achieve the same recovery, and the ability to froth wash
a small volume can further reduce the amount of cells
needed to reduce entrainment. Low footprint and low cap-
ital and froth washing are good, but they should not be
confused with the task of making particles hydropho-
bicthe right reagents on clean surfaces do this.
2. Design principles for good nes otation
The reason that nes do not oat well in many circuits is
because they are mixed with coarse particles and with com-
posites. The nes need more collector and more otation
timebut the otation conditions usually have to be set
to suit the coarser fractions. Designing dierent otation
impellors distracts from the real problemnes and coarse
will never oat well in the same cell, since the reagents can-
not be optimised for both. This is particularly relevant
when composites have to be rejected to maintain concen-
trate grade. For example, a 40 lm composite that includes
a 15 lm grain of sphalerite has to be rejectedbut the con-
ditions to do so will probably also reject a 15 lm liberated
sphalerite particle. Texts as old as Taggart (1927) described
the benets of sand/slimes splits into separate circuits.
This simple concept has been largely overlooked in the
push for circuit simplication and larger otation cells.
Surface analysis of nes lost to tailings almost invariably
shows that they are there because they are not hydrophobic
enougheither they have a hydrophilic surface coating, or
there simply is not enough collector on the surface (Grano
et al., 2004). Adding extra collector will oat the latter par-
ticles, but at the cost of oating other diluentsi.e. the
operator cannot aord to oat the nes because he will lose
selectivity. Fig. 3 shows a simple conceptual solutiontai-
lor the otation conditions to dierent size distributions to
achieve high recovery across many size ranges.
The Mount Isa circuit developed into an excellent bal-
ance of the needs of dierent minerals, relying on several
stages of grinding and otation. The design principals are
Achieve the correct mineral liberation you need to make
target grade and recovery.
Apply the most ecient grinding method, in the place
that needs least grinding power to achieve target
liberation.
Make clean surfaces in grinding, and oat as soon as pos-
sible before surfaces are oxidised again.
Float minerals in narrow size distributions.
Minimise circulating loadsgrind cleaner feed rather
than cleaner tail, and open circuit where possible.
Design launder and pumping systems for the more tena-
cious froths made by ne particles.
Design thickening and ltering for the ner particles.
These principles are explained below.
2.1. Achieving the correct liberation, in the
most ecient way
Selection of the right grind size is invariably determined
by economics. If a high grade concentrate is required, or if
a contaminant (e.g. silica) must be eliminated from concen-
trate, then the grinding must achieve high liberation of the
target mineral. For high value products where recovery is
much more important than concentrate grade (e.g. plati-
num), liberation only needs to be enough to expose enough
valuable mineral surface for otation recovery (though
more liberation to increase concentrate grade for smelting
may still be economic). Since required grinding power
increases exponentially as target size reduces, it is impor-
tant to limit the tonnage sent to ner grinding stages.
For Mt Isa and MRM the best technical solution would
be to grind all the ore to 7 lm. But these are relatively
low value orebodies, so the economic compromise is to
only grind rougher feed ne enough to give most values a
chance to report to rougher concentrate, then apply ne
grinding to smaller tonnage streams that have a concentra-
tion of composites.
Size (microns)
Recovery
5 10 30 50 200 80 100 150
Intermediate and coarse behaviour
Fines performance when
treated by themselves
Fines peformance when
treated with coarse particles
Fig. 3. Conceptual staged grind-oat circuit performance.
J.D. Pease et al. / Minerals Engineering 19 (2006) 831840 833
2.2. Use the most ecient grinding methodhigh
intensity stirred mills versus tower mills
Fine grinding is capital intensive and energy intensive,
so it is crucial to get good power eciency at the full plant
scale. As Fig. 4 shows, this broadly means tumbling mills
to around 40 lm, and stirred milling for ner sizes. Tower
Mills are more ecient than ball mills for sizes below about
40 lm, mainly because they typically use smaller balls and
operate at slower speeds that favour attrition grinding over
impact breakage. However Tower Mills struggle to grind
below about 25 lm, and are less ecient than the new high
intensity stirred mills like IsaMill (Fig. 5) or detritors.
These stirred mills are more ecient, and can grind to
much ner sizes, e.g. below 7 lm, because they can operate
with very ne media (e.g. 12 mm for the IsaMill), and with
very high intensity, both of which greatly accelerate the
attrition mechanism. Table 1 (Gao and Weller, 1993)
shows the very high intensity in a stirred mill, assisting
grinding and cleaning surfaces.
2.3. Classication is crucial
High grinding eciency requires good classication; and
good classication also produces narrow size distributions
ideal for otation. Too often ne grinding circuits are con-
strained by poor classication, causing higher energy con-
sumption, unnecessary production of ultranes, and poor
control of top-size. To classify sharply below 20 lm needs
small cyclones, theoretically 2-in. diameter. However this is
expensive, and in our experience they are virtually inoper-
able in a minerals plant. So most operators compromise
and use bigger cyclonesbut the atter size distribution
compromises grinding eciency and otation performance
(and especially leaching performance if this is the down-
stream process). The IsaMill addresses this by classifying
within the mill using the product separatoreectively
an internal centrifuge with a tip speed of 20 m/s, giving
sharper separation than even 2 in. cyclones. This allows
the mill to produce a very sharp size distribution in open
circuit conguration without cyclones.
2.4. Make clean surfaces in grinding and oat them
quickly
With their high surface area ne particles are highly
reactive, and are sensitive to oxidation and surface precip-
itates. Grinding with steel media can be very harmful. This
has been well reported by Frew et al., 1994 who examined
ne grinding with steel media at six sites. The grinding
achieved signicant liberation, but this benet was largely
lost due to the negative impact on surface chemistry. Sites
like Helyer were successful in applying high intensity con-
ditioning (Holder, 1994) to clean surfaces after Tower Mill-
ing, but at signicant additional capital cost, and operating
cost of power and maintenance. Greet and Steinier (2004)
reported X-ray photon spectroscopy (XPS) work on grind-
ing product with three dierent media. Forged steel balls
produced 16.6% Fe concentration on galena surfaces (pres-
ent as oxy-hydroxide species), reducing to 10% for high
chrome media, and to below the detection limit (less than
0.02%) for inert ceramic media. Using the same otation
conditions, this resulted in galena recoveries of 48%,
70%, increasing to 92% for inert media. Of course in prac-
Fig. 4. Grinding energy vs product size for a pyrite concentrate.
Table 1
Comparison of various grinding technologies independent laboratory data
Feature IsaMill Tower mills Vertical
pin mills
Grinding intensity (kW/L) 0.54 0.005 0.150.18
Residence time to 15 lm (min) 0.6 154 79
Power usage to 15 lm (kWh/t) 17.4 59.6 37.539.0
Media material Various Steel Steel
Media size (mm) 0.81.6 912 68
834 J.D. Pease et al. / Minerals Engineering 19 (2006) 831840
tice higher recovery is achieved after steel milling by
increasing collector additionthe problem is that this also
increases the recovery of unwanted species (i.e. loss of
selectivity), as well as the higher cost of reagents and slower
otation times.
Fine grinding creates a lot of new surfacethis is an
opportunity to create clean surfaces ideal for otation
rather than altered surfaces. The motto is make em clean
and oat em quick. The IsaMill makes clean surfaces
very quickly; if water chemistry and reagents are con-
trolled to keep the surface clean long enough to get collec-
tor coverage, then the resultant minerals can be very fast
oating. An excellent example of is the lead circuit at Mt
Isa, where a small (1.7 m by 3.2 m) Jameson cell at the
discharge of the lead IsaMills produces 3040% of the
concentrate at higher than concentrate grade target. It
takes advantage of the high otation rates of the freshly
produced minus 10 lm particles and small bubble size
(0.3 mm) produced in the Jameson cell, and the small sur-
face area is easy for froth washing to maintain high
grades. This was a low cost way to extend the capacity
of the existing otation cells.
2.5. Float minerals in narrow size distributions and
minimise circulating loads
Fig. 3 demonstrates that nes should not be mixed with
coarse particles in otation, yet this is what happens when
cleaner tailings are reground and sent back to roughing. If
the regrind is in a conventional mill then it often does not
eectively liberate ne composites anyway, it just worsens
the surface chemistry before adding back to the fresh
stream. This helps sustain the myth that nes do not oat.
The answer is a staged grind and oat approach as shown
in Fig. 6 for Mt Isa. Note that the rule not to recirculate
tails does not mean that particles only get one chance. Con-
sider the potential path of a sphalerite particle that has
been oated in roughing in Fig. 6. It has three chances to
make concentrate before we give up on it:
It can oat directly from roughers to columns to concen-
trate in the 37 lm circuitif it is liberated and fast
oating it will take this route. If it does not, then it is
slower oating either because it is composite or because
its surface is altered.
If so it is sent to the 12 lm regrind and cleaning circuit.
If it does not oat here, then it is still composite or has
developed new surface coatings.
If so, it is sent to the 7 lm regrind and cleaning circuit.
If after these three stages of grinding and otation the
particle still has not been recovered we give up on itwe
have run out of ideas (and grinding power). Though the
particle has three chances to be recovered, we would never
send it back to the head of the circuitwhy would we put
recalcitrant 7 lm particles back with fresh new feed? What
else can the circuit do with it? (Note that we do circulate
tailings within cleaner stagesclosed circuit cleaning is a
valid way to increase grade, the crucial point is that all
the particles within cleaning are in a similar size range
and liberation class; we do not send a ne cleaner tailings
to join a coarse rougher feed).
Another outcome is that it is better to regrind cleaner
feed rather than cleaner tailing. In fact, Fig. 6 shows that
you may do boththe cleaner tails of one stage may be
reground and become the feed of another stage of cleaning.
Conventional logic is to only regrind cleaner tailing why
regrind minerals that can already make a decent concen-
trate? The answer to this question is a better question:
why wait until you have made most of your concentrate
Fig. 5. M3000 IsaMill as used for lead zinc regrinding at Mt Isa.
J.D. Pease et al. / Minerals Engineering 19 (2006) 831840 835
before applying your most powerful tool for improving o-
tation? Certainly fast oating liberated minerals should be
recovered before grindinge.g. a quick oat in a cell or
column or Jameson cell. For Mt Isa about 30% of the
sphalerite is recovered this way, but the remainder is sent
to IsaMilling to improve liberation and clean particle sur-
faces, before being oated in the next cleaning circuit.
2.6. Build simpler circuits with less otation capacity, not
more
The above design principles make circuits look more
complex on paper. In reality the circuit is simpler, since
each otation stage can be tailored to suit its feed, and is
not disrupted by circulating loads. Theoretically nes oat
more slowly, but in reality the staged grind oat does not
need more otation capacityin our experience it may
need less capacity, since
With fresh clean surfaces and tailored reagents, the nes
oat faster than they did when mixed with coarse
particles.
Removing circulating loads can more than halve the vol-
umes to roughing and cleaning.
2.7. Designing froth handling and dewatering for ner
particles
As every operator knows, a technology is only as good
as the materials handling constraints. As discussed before,
operating 2-in. cyclones to close circuit ne grinding is
technically sound, but an operating nightmare, so ne
grinding technology can be seriously compromised by poor
classication. The advantages of ne grinding can also be
rendered useless by inadequate attention to froth launders,
froth pumps, pumboxes, thickeners and lters. Usually the
design features needed are not complex or expensive, but
they are crucial to get right. At MRM we thought we
had paid adequate attention to froth handling and dewa-
tering for 7 lm P80 concentrates. In fact we did not do
enough, and this seriously reduced tonnage and recovery
and plant hygiene for almost 4 years. The solutions are
very simple in hindsight, but very easy to get wrong with-
out experience. The solutions include
The overall circuit designincluding eliminating circu-
lating loads.
Setting reagent addition rates and reagent addition
points correctly. This sounds simple when you know
how, but is easy to get wrong. Starting with the wrong
reagents often locks in a high reagent position, where
too much collector and too much depressant compete
with each other, causing high loads and spillage. The
clean surfaces produced in a IsaMill require a clean
sheet approach to reagents.
Adequate sprays on launders, and enough room in laun-
ders for de-aeration. This is not as hard as it sounds,
since proper circuit design reduces the load on launders
by eliminating circulating loads. Also pipework design
must ensure the launders do not back-up with froth
and overow.
7um Regrind / Float
70um Primary Grind / Float 37um Secondary Grind / Float
12um Regrind / Float
12um Regrind / Float
Rod & Ball Milling
Tailings
Prefloat Pb Ro
Ball Milling
Pb Ro / Scav Zn Ro Zn Ro / Scav
Tailings
Zn Conc
Zn Columns
3 x 1.1MW
IsaMills
Pb Conc
Jameson
Cell
Pb Cleaners
1 x 0.52MW Tower Mill
2 x 1.1MW
IsaMills
Zn Cleaners
3 x 1.1MW
IsaMills
Zn Retreatment Cl
Zn Retreatment Ro
Zn Conc
Zn Conc
Tailings
Pb Conc
34% Pb Rec
46% Pb Rec
37% Zn Rec
39% Zn Rec
6% Zn Rec
Fig. 6. Mount Isa Pb/Zn concentrator ow sheet.
836 J.D. Pease et al. / Minerals Engineering 19 (2006) 831840
Pumpboxes with enough surface area for de-aeration,
and adequate sprays.
Water to spray lines should be strained, and the sprays
and the strainer must be easily accessed and cleaned
on-line by operators.
For true ne grinding applications, choose pumps to run
at a slow speed, and with a wide open impellor (e.g. the
Woody pump applied at Mt Isa, available from War-
man as their horizontal froth pump).
Control of pipe entry into pumpboxes, and control of
pumpbox level, to avoid air entrainment and the milk-
shake eect.
Adequate de-aeration of froths prior to pumping to the
thickener, so air is not entrained in the ocs, causing
froth on the thickener.
Adequate thickener residence time and bed height to
compress the concentrate bed to achieve the desired
thickener underow density.
Pumping and pipework systems that do not entrain air
into the thickened concentrate, to avoid lowering the
feed density to the ltering systems.
Filtering systems designed to handle the range of low
feed densities that may occur. That is, capacity to handle
increase cycle times, if the feed density is low.
3. Case study 1Mount Isa lead zinc concentrator
The changes to the Mount Isa circuit as part of the
George Fisher Project are detailed elsewhere (Young
and Gao, 2000; Young et al., 2000). In summary, the pro-
ject involved adding a further six IsaMills, to regrind lead
rougher concentrate to P
80
of 12 lm, most zinc rougher
concentrate to 12 lm, and a zinc regrind to P
80
of 7 lm.
Lead performance increased by 5% concentrate grade
and 5% recovery (equivalent to 10% increase in lead recov-
ery at the same grade). Zinc recovery increased by 10%, in
two steps, and zinc concentrate grade by 2% (equivalent to
16% increase in zinc recovery at the same grade). The story
of zinc metallurgy is told in Figs. 79.
The project predicted 5% higher zinc recovery (and no
extra concentrate grade) due to extra liberation. Fig. 7
shows this was achieved instantly. The surprise was the
second wave of a further 5% zinc recovery increase
and the 2% increase in zinc concentrate grade. This was
because nes otation improved after grinding ner.
It took about 6 months to discover how much better the
nes could perform because we were so used to otation
after conventional grinding rather than after IsaMilling.
Our three biggest mistakes were
Expecting to need a lot more reagents after IsaMilling
due to the huge new surface area created. Some reagent
additions were forecast to triple.
Not taking the depressant (lime to pH 11) o the zinc
cleaners.
Pulling otation harder because we thought otation
rates of the nes would be slower.
To our surprise, even though we introduced 6 MW of
extra grinding power, operating cost per tonne of feed
did not increase. This was because of
the lower reagent additions,
elimination of circulation loads between roughing and
cleaninga lot of power (and otation capacity) is
wasted in conventional circuits by pumping circulating
loads of 100300%.
Virtual elimination of spillagedue to new designs for
pumpboxes and pumps, the lower reagents, and espe-
cially the reduction in circulating loads.
The reduction in reagents deserves comment. We
expected the additional surface area would need more
62.5
65.0
67.5
70.0
72.5
75.0
77.5
80.0
82.5
Apr May
1999
Jun Jul Aug Sep Oct Nov
1999
Dec Jan Feb
2000
Mar Apr May Jun Jul Aug Sep Oct
2000
Nov
Baseline
Reduced grinding & flotation capacity, due to
equipment relocation during construction.
+2%
Conc Grade
(not shown)
2nd Wave +5%
Zinc Recovery
1st Wave +5%
Zinc Recovery
IsaMills
Commissioned
Zn
Rec %
Fig. 7. Zinc recovery increase from IsaMilling.
J.D. Pease et al. / Minerals Engineering 19 (2006) 831840 837
reagentsso we increased reagents when we commissioned
the mills; in fact we should have reduced it. Why? Certainly
additional surface needs more collector coverage. But our
mistake was to assume that more collector on the surface
meant more collector added to the pulp. Our zinc circuit
had never seen clean mineral surfaces beforethey had
always been masked by iron hydroxides from grinding, or
gypsum precipitation from pulp, or simply oxidation after
long otation times in the lead circuit. High collector addi-
tions to pulp were needed to diuse through surface layers
to get adequate mineral surface coverage. But the IsaMills
produced very clean mineral surfaces very quicklywith
the right water chemistry and adding reagents in the right
place, surface collector coverage was achieved with much
lower than expected collector concentrations in the pulp.
Another advantage of the lack of circulating loads is
that cleaner densities remain low, thereby assisting dilution
cleaning.
3.1. Size by size performance in the zinc circuit
Fig. 9 demonstrates the power of the staged grind-oat
circuit at Mt Isa. Fast oating liberated particles have a
chance to be recovered in the column (37 lm) circuitcol-
umn recovery is not high, and columns are not ideal for the
jobbut they were in the circuit so they were free. The
intent is simply to scalp out fast oating particles, then
send the remaining particles to the 12 lm circuit. In this cir-
cuit otation conditions are tailored to 12 lm particles, so
not surprisingly the highest otation recovery (90%) is in
this size fraction. Particles not recovered here are probably
still composite, so they are sent to the 7 lm grinding and
cleaning circuit. Flotation here is tailored to 7 lm particles
(a simple task using the old existing Agitair cellsotation
is easy since there is only one size range and few composites
to depress, and density is low because most of the solids
have already been recovered). Not surprisingly, the highest
recovery (above 95%) in this circuit occurs in the sub 7 lm
size fraction. Overall circuit recovery from rougher concen-
trate is determined by adding the products from all three
cleaning stages, resulting the at size-recovery curve.
Above 95% recovery from rougher concentrate is achieved
in all size fractions from 1 lm to 25 lmrecovery drops in
the coarser fractions since these particles are frequently
compositethe circuit is designed to reject them and send
them to an IsaMill to grind into the high-recovery nes
fractions.
4. Case study 2: South African platinum
industryconcentration of UG2 ores
Platinum otation operations also derive signicant
benets from the combination of liberation and improved
surface chemistry from high intensity inert grinding (Curry,
2002). These benets have been extensively demonstrated
on the UG2 ore body, part of the Bushveld Igneous Com-
plex located to the north of Johannesburg in South Africa.
UG2 ore is characterised by repetitive layers of dark, high-
density minerals (pyroxene, chromite) and light, less dense
feldspars. Typically, the highest grades of platinum group
metals (PGMs) occur at the lower contact zone of the lay-
ers, where a strong association with base metal sulphides
occurs (specically nickel, copper and iron). The bimodal
distribution of mineral densities is a signicant mineral
processing challenge to produce high grade PGM rich con-
centrates. The PGM grain size is ultra-ne (sub 10 lm) and
associated with the silicate and sulphide species, rather
than the large (200 lm), dense chromite minerals. Dicult
middling streams have a median mineral grain size of
20 lm and a median PGM grain size of 3 lm. The PGMs
0
20
40
60
80
100
1 10 100
SIZE (microns)
Whole circuit recovery, roughing and cleaning, including losses to lead concentrate
R
E
C
O
V
E
R
Y

(
%
)
Apr-99
Oct-00
Moved particles from low recovery and low grade fraction
to higher recovery and higher grade fines fraction
Increased recovery in fine sizes
Net result- 10% recovery increase and 2% concentrate grade increase
Fig. 8. Zn recovery vs sizebefore and after IsaMilling.
838 J.D. Pease et al. / Minerals Engineering 19 (2006) 831840
are most commonly locked in silicates, or on the grain
boundaries of silicates and base metal sulphides. As well
as the diculty associated with the ne grain size, many
of the sulphides coated in talc slimes that impede otation
recovery.
UG2 concentrators typically use staged grinding and
otation to rst recover fast oating valuable particles as
soon as possible. Primary and secondary circuits are most
common (MF1 and MF2), however tertiary mill/oat
options (MF3) are more frequently being considered. Sur-
prisingly, the use of regrinding within each MF circuit is
not common. A key part of circuit design is to reject chro-
mite particles, which are hard and dense. Various methods
are used to avoid overgrinding of chromite and reject it
while it is still coarse, e.g. gravity separation within grind-
ing circuits, or open circuit milling.
C7 C6 C5/C4 C3/C2 C1/38 53 75
Size fraction
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
0-4m 4-8m 8-16m 16-30m 30-53m
Recovery
Size Distribution
C7 C6 C5/C4 C3/C2 C1/38 53 75
Size fraction
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
0-4m 4-8m 8-16m 16-30m 30-53m
Recovery
Size
37 m zinc circuit
12 m zinc
7m Regrind / Float
70m Primary Grind/Float
37m Secondary Grind / Float
12m Regrind / Float
12m Regrind / Float
Rod & Ball Milling
Tailings
Prefloat PbRo
Ball Milling
Pb Ro/Scav Zn Ro Zn Ro/ Scav
Tailings
Zn Conc
Zn Columns
3 x 1.1MW
Pb Conc
Jameson Cell
Pb Cleaners *
1 x0.52MW TowerMill
2 x 1.1MW
ISaMills
Zn Cleaners *
Zn Retreatment Cl
Zn Retreat
Zn Conc
33%
Pb rec
46% Pb Rec
34% Zn rec, 50% Zn
6% Zn rec,47% Zn
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
C7 C6 C5/C4 C3-C1 38/53 75
Size fraction
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
0-4m 4-8m 8-16m 16-38m 38-75m
7m zinc circuit
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Size
Recovery
* 3 stages of closed circuit
conventional cleaning
3 x1.1MW IsaMills
Zn Conc
Tailings
Zn
Recovery
Zn
Recovery
Zn
Recovery
37%Zn Rec, 54% Zn
Combine 37 m circuit with 12 m circuit with 7 m circuit for overall zinc cleaning performance graph
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
C7 C6 C5/C4 C3-C1 38/53 75
Size fraction
Z
i
n
c


R
e
c
o
v
e
r
y


i
n

s
i
z
e


f
r
a
c
t
i
o
n

%
0.0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
0-4m 4-8m 8-16m 16-38m 38-75m
Recovery
Size Distribution
Overall Zinc Circuit Recovery by Size

Fig. 9. Zinc recovery in cleaning stages and for overall cleaning circuit.
J.D. Pease et al. / Minerals Engineering 19 (2006) 831840 839
Because primary and secondary grind sizes are much
coarser than the PGM grain size, cleaning of rougher con-
centrates leads to high circulating loads. Extra liberation is
needed to increase concentrate grade. Traditionally this is
achieved by depressing composites from cleaning then
regrinding cleaner tailing. However low intensity regrind-
ing with steel media further harms PGM otation. The
result is high circulating loads, slow otation kinetics, poor
nes otation, and poor grade/recovery performance. High
intensity inert grinding is an ideal solution to provide both
the liberation and improved surface chemistry to achieve a
step change in circuit design and performance. Typically
UG2 applications target regrind product size of 15
18 lm. At this size about 90% of the PGMs associated
with base metal sulphides are liberated and can easily be
recovered with xanthate collectorif the surfaces are clean.
Two major platinum producers have applied these
advances by installing IsaMills in their circuits with great
success. Consistent with experience at Mt Isa and MRM,
platinum producers are increasingly realising that cleaner
feed is the best target stream. Inert grinding of cleaner feed
increases the otation rate and performance by removing
surface layers of talc, oxidation and iron hydroxides. By
improving liberation and otation kinetics immediately,
particles in rougher concentrate have an opportunity to
report quickly to a high grade nal concentrate, rather than
being forced to recirculate to get to a grinding mill and then
being sent back to roughing. The practicality of achieving
this has been greatly enhanced by the 2.6 MW M10,000
IsaMill installed by Anglo platinum. All rougher concen-
trate is sent directly to a single IsaMill before cleaning.
No cyclones are needed before or after the mill, since the
internal product separator allows the mill to handle a wide
variety of feed densities, and produces a very sharp product
size distribution. Cleaner tailings can be open-circuiteda
novelty in platinum otation, but with immense benets
for the roughing circuit. Up to 50% of platinum roughing
capacity is often consumed by circulating loads, and fre-
quently the roughers are operated to suit the troublesome
recirculated particles rather than the fresh feed. By apply-
ing the right grinding technology in the right place, circuits
can be designed with considerably less otation capacity in
both roughing and cleaning. This is the opposite of the
common view that ner grinding will require more
otation.
A more conservative, but still successful, application is
to regrind cleaner tailings, then to oat mill product in a
separate small cleaning circuit tailored for the narrow, ne
size distribution produced by the IsaMill. Because libera-
tion and kinetics are good, ne cleaner tailing can report
to nal tailing. This practice is used by Lonmin.
Plant data is not made available for publication. How-
ever testwork on a wide range of platinum ores and a wide
range of operators shows that a single IsaMill regrinding
stage increases overall PGM recovery by 415%, at the
same or higher concentrate grade.
5. Conclusions
Contrary to common perception, nes otation is quite
simple and can achieve very high recoveries. It does not
require special otation cells or exotic reagents or long res-
idence times. It requires rstly an understanding of the size
by size liberation characteristics of the ore, then designing a
staged grinding and otation circuit to suit. Attention to
surface chemistry, water chemistry, classication and mate-
rials handling is important.
The ability to design cheap eective circuits for nes
recovery has been enabled by the introduction of high
intensity stirred mills using inert media. The clean mineral
surfaces can transform otation performance of nes
compared with conventional grinding with steel media.
The development of large scale units like the 2.6 MW Isa-
Mill, eciently producing liberated particles with clean
surfaces in narrow size distributions without the need
for cycloning has provided a powerful new tool for circuit
design.
References
Curry, D., 2002. The impact of IsaMill technology on modern concen-
trator design, paper presented to MMMA New Technologies Confer-
ence, Carltonville, RSA.
Frew, J.A., Davey, K.J., Glen, R.M., 1994. Eects of ne grinding on
otation performance: distinguishing size from other eects, paper
presented to AusIMM Fifth Mill Operators Conference, Roxby
Downs, Australia, 1620 October, pp. 263270.
Frew, J.A., Smart, R., Manlapig, E.V., 1994. Eects of ne grinding on
otation performance: generic statements, paper presented to
AusIMM Fifth Mill Operators Conference, Roxby Downs, Australia,
1620 October, pp. 245250.
Gao, M., Weller, K.R., 1993. Review of alternative technologies for ne
grinding. AMIRA Project P336, Report P336/20, November.
Grano, S., Weedon, D., Akroyd, T., Wiseman, D., 2004. Application of a
property based otation model in circuit, paper presented to Metal-
lurgical Plant Design and Operating Strategies AusIMM, Perth 2004.
Greet, C.J., Steinier, P., 2004. Grindingthe primary conditioner paper
presented to metallurgical plant design and operating strategies,
AusIMM, Perth 2004.
Holder, R.K., 1994. Improvements in copper and silver otation at
Hellyer using high energy conditioning, paper presented to AusIMM
Fifth Mill Operators Conference, Roxby Downs, Australia, 1620
October.
Taggart, A.F., 1927. Handbook of Mineral Dressing. John Wiley and
Sons, New York, pp. 1292 to 1297.
Young, M.F., Gao, M., 2000. Performance of the IsaMills in the George
Fisher owsheet, paper presented to AusIMM Seventh Mill Operators
Conference, Kalgoorlie, Australia, 1214 October.
Young, M.F., Pease, J.D., Fisher, K.S., 2000. The installation of the
George Fisher owsheet in the Mount Isa lead/zinc concentrator,
paper presented to AusIMM Seventh Mill Operators Conference,
Kalgoorlie, Australia, 1214 October.
840 J.D. Pease et al. / Minerals Engineering 19 (2006) 831840