Available online at www.sciencedirect.

com

Int. J. Miner. Process. 85 (2007) 16 – 24

www.elsevier.com/locate/ijminpro

CFD model of a self-aerating flotation cell

P.T.L. Koh ⁎, M.P. Schwarz
CSIRO Minerals, Clayton, Victoria 3169, Australia
Received 11 April 2007; received in revised form 26 July 2007; accepted 26 August 2007
Available online 5 September 2007

Abstract

The effect of impeller speed on the air flow in a self-aerated Denver laboratory flotation cell was investigated using
computational modelling. Air is induced into the slurry by the impeller's rotating action. The rate of air flow is determined by the
suction pressure created by the impeller, the hydrostatic head of the slurry and the frictional losses along the delivery shaft from the
inlet valve to the impeller. From two-phase simulations of the flotation cell at varying impeller speeds, the predicted air flow rates
have been found to compare favourably against measured values reported in the literature. The effect of increasing impeller speed is
to increase the air flow rate and gas holdup in the cell. Simulations with flotation kinetics showed that the gravitational force acting
on the attached particles is significant. The effect is a decrease in the bubble rise velocity which in turn affects the flotation rate as
predicted by the model. The effects of the local turbulence level on the local attachment rate and bubble loading have been
discussed and quantified.
Crown Copyright © 2007 Published by Elsevier B.V. All rights reserved.

Keywords: Flotation modelling; Flotation kinetics; Flotation machine; Cell hydrodynamics

1. Introduction for flotation is generated by a rotor–stator mechanism

Flotation is an important process in the minerals
industry and most flotation is carried out using
mechanically agitated cells ranging in volume up to
300 m3. Separation of the particles of valuable minerals
from the ore is achieved when particles of the valuable
minerals preferentially attach to air bubbles and are
floated to the surface of the slurry via a froth layer. From
the froth layer, the particles are removed by means of an
overflow launder. The remainder of the slurry exits the
cell through an outlet in the lower part of the cell. A class
of the commonly used cells has air pumped into the slurry
via a central shaft. The dispersion of air as fine bubbles
⁎ Corresponding author. Fax: +61 3 9562 8919.
E-mail address: peter.koh@csiro.au (P.T.L. Koh).
0301-7516/$ - see front matter. Crown Copyright © 2007 Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.minpro.2007.08.006
which also serves to mix the slurry and air bubbles.
Various proprietary mechanisms are used in the minerals
industry.
Self-aerating flotation cells, such as the Denver and
WEMCO series, are also used in the minerals industry. In
this case, air is induced into the slurry by the impeller's
pumping action. The rate of air flow is controlled by the
suction pressure created by the impeller, the hydrostatic
head and the frictional losses along the delivery shaft
from the inlet valve to the impeller. In this situation, the
air flow rate is directly related to the impeller speed. A
schematic diagram of the Denver cell is shown in Fig. 1.
Metallurgical performance is measured in terms of the
recovery of the valuable mineral in the concentrate, the
level of grade of the concentrate and the level of valuable
mineral left in the tailings. Performance is affected by
design features such as impeller and stator design, and
 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24
Fig. 1. Schematic diagram of flotation cell.
tank geometry and internals. Flotation is also affected by
operating variables such as slurry concentration, aeration
rate and impeller speed. In self-aerating cells, the aeration
rates and the impeller speeds are not independent.
The aeration behaviour of the Denver cell and several
other self-aerating flotation machines has been extensively
investigated by Arbiter et al. (1976), and by Harris and
Mensah-Biney (1977) in both laboratory and full-size
machines. Some of the key findings with respect to induced
air flow rates are summarised as follows. The induced air
flow rate increases with impeller speed, it increases as pulp
level is decreased, it increases as pulp density is reduced,
and it decreases with frother addition. These dependencies
are explained as follows: increasing impeller speed
increases low pressure zones behind the impeller blades,
which increases suction; decreasing pulp level and pulp
density reduce back pressure near the impeller; and
addition of frother decreases the size of the bubbles but
increases the number of fine bubbles which then move
away more slowly from the impeller and this build-up of
fine bubbles reduces fluid density and hence reduces power
consumption and air flow rate.
For application in the chemical industry, there have
been extensive investigations on gas-inducing impellers
in agitated reactors (eg. Sawant et al., 1981; Joshi et al.,
1982; Rielly et al., 1992; Forrester and Rielly, 1994;
Sardeing et al., 2006). Some of the reactors are in fact
quite similar to Denver and Wemco machines. There is
continuing interest in the search for new improved gas-
inducing impellers for chemical processing.
17
CFD technology for the numerical modelling of com-
plex unsteady, multi-phase flows typically found in agi-
tated tanks of this type has developed to the point where it
now has the potential to identify design and operational
modifications to the process which can lead to process
improvements. CSIRO Minerals has successfully applied
CFD technology to mineral processes with complex
unsteady multi-phase flows. For example, in the AMIRA
P780A project, flotation cells up to 150 m3 from
Outokumpu and Metso Minerals have been investigated
to improve flotation performance (Koh et al., 2003).
In the present paper, the effect of impeller speed on the
air flow in a self-aerated Denver laboratory flotation cell
was investigated using CFD modelling (Koh et al., 2000;
Koh and Schwarz, 2003, 2006). The air flow rates
predicted by CFD are compared against measured values
reported in the literature for this cell (Girgin et al., 2006).
2. Model description
The Denver flotation cell modelled here has a volume
of 5.3 litres fitted with an eight-bladed impeller of diameter
72 mm. The standpipe has a diameter of 42 mm and a
rotating shaft of diameter 16 mm. The cell is operated
under semi-batch conditions (batch water, continuous air).
The CFD mesh used for the flotation cell and impeller has
103,000 grid points as illustrated in Fig. 2.
In the model, the multi-phase flow equations for the
conservation of mass, momentum and turbulence
quantities are solved using an Eulerian–Eulerian ap-
proach in which the phases are treated as interpenetrating
continua. The multi-phase equations for unsteady flow
can be written as follows (where i = 1 for liquid, i = 2 for
gas):
Aðai qi Þ Y
þ jd ðai qi U i  qi Di jai Þ ¼ Si ð1Þ
At
Y
Aðai qi U i Þ Y Y
þ jd ðai qi U i  U i Þ
At
Y Y
¼ jd ðai ðlL;i þ lT;i Þðj U i þ ðj U i ÞT Þ
Y
þ ai ðBi  jPi Þ þ Fi þ Si U i : ð2Þ
The variables solved include velocity components
Ui, pressure Pi and volumetric fractions αi for each of
the phases. The mass source terms, Si, include sources
or sinks of gas entering and leaving the vessel. The inter-
phase force Fi is given by the standard drag force with
modifications for fluid turbulence. The body forces Bi
include the buoyancy force, and centrifugal and Coriolis
forces in the rotating frame of reference. In the equation,
18 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24
Fig. 2. CFD mesh of flotation cell.
ρi is the density, μL,i in the laminar viscosity and Di is
the turbulent diffusivity.
The turbulent kinetic energy k and turbulent
dissipation rate ε are for the continuous liquid phase.
The turbulent viscosity μT,i in the liquid phase is
calculated using the standard k–ε turbulence model
(Launder and Spalding, 1974).
The transport equations are solved using the compu-
tational fluid dynamics code CFX-4.4 (2001) and the
multiple frames of reference technique has been used for
modelling the impeller rotation in the stationary tank.
The flow of air in the standpipe is not modelled. The
inlet and outlet boundaries for the air flow are shown in
Fig. 3 where a degassing boundary condition has been
applied at the fluid surface. A complete CFD model could
include flow through the complex air valve drawing air
from the atmosphere. An alternative approach which is
used in this work involves using an overall pressure drop
for delivering air from the atmosphere to the impeller. In
this method, the total pressure drop is obtained by apply-
ing Fanning friction factor (f) for air flow through the
standpipe, together with another factor (K) in equivalent
length to diameter ratio to account for pipe roughness and
pressure drop through the air valve and bends. These
factors are included in the pressure drop equation as
follows:


L
DP ¼ 2f qg V 2 þK
De
where ΔP is the pressure drop, ρg the air density, V the
mean flow velocity and L pipe length. The equivalent
pipe diameter (De) is taken as four times the flow area
divided by the wetted perimeter, De = (Do − Di) where
Do = 26 mm and Di = 16 mm respectively are the outer
and inner diameters of the annular flow in the standpipe.
The friction factor f is calculated using the equation by
Blasius which is expressed as a function of the pipe
Reynolds number (Re) as follows:
0:0791
f ¼ ð4Þ
Re0:25
De V qg
Re ¼ : ð5Þ
l
A pressure balance across the inlet boundary above the
impeller shown in Fig. 3 involves the suction pressure
created by the impeller, the hydrostatic head of the liquid
and the pressure drop of the air flow from the atmosphere.
To achieve a particular air flow the suction pressure must
be sufficient to overcome both the hydrostatic head and
the frictional pressure loss. In the CFD model, the air flow
rate as indicated by the velocity (V) in Eq. (3) is
determined by iteration until the pressure drop (ΔP)
matches the pressure at the inlet boundary as obtained by
CFD. The pressure across the inlet boundary has to be
averaged because it varies with radius due to the swirling
motion. The air flow rate is adjusted between iterations if
the pressure difference is greater than 10 Pa until a
converged solution is obtained. The average pressure
across the inlet boundary was observed to fluctuate
between iterations as a result of changes to the air holdup
above the inlet. This fluctuation caused unstable air flow
rate into the cell and a relaxation technique was
implemented to minimise the fluctuation.
The froth layer is not included in the computational
domain; only the pulp zone is simulated. At the froth–
ð3Þ
Fig. 3. CFD setup of semi-batch flotation cell showing inlet and outlet
boundaries for air.
 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24 19

pulp interface, gas bubbles with attached particles are
transferred from the pulp zone to the froth layer at the rate
bubbles arrive at the interface. An average bubble diame-
ter of 1 mm is applied in the model based on measure-
ments by Girgin et al. (2006) in the air–water system at a
frother concentration of 50 ppm. At this relatively high
frother concentration, the bubble sizes remain fairly
constant over the range of impeller speeds reported
because coalescence is reduced.
For simulations with flotation kinetics, the transfer of
particles between the pulp and bubbles is achieved by
applying source terms for particle number concentration
ni in the transport equation as follows:
transient simulation uses variable time steps such that
the mass error is less than 0.1% for each time step. The
equations used in the model for collision frequency and
attachment probability reported previously (Koh and
Schwarz, 2006) are shown in Table 1.
3. Results
Results of two-phase simulations have been obtained for
the Denver cell at different impeller speeds, and a typical mid-
plane plot of the liquid velocity vectors at impeller speed of
1200 rpm is shown in Fig. 4. CFD predictions indicate a
complex flow field within the flotation cell and large fluid
velocities observed in the impeller region.

Aðai ni Þ Y 3.1. Comparing air flow

þ jd ðai ni U i Þ ¼ /a þ /d
At
where ϕa and ϕd are sources or sinks specifying at-
tachment and detachment rates respectively. The
ð6Þ
From simulations in the air–water system at various impeller
speeds, the predicted air flow rates have been obtained and are
compared against literature values by Girgin et al. (2006). Based

Table 1
Bubble–particle equations used in the CFD model as reported previously (Koh and Schwarz, 2006)
Net attachment rate dnp1
¼ k1 np1 nbT ð1  bÞ þ k2 nbT b
dt
Attachment rate constant k1 ¼ Z1 Pc Pa Ps
Detachment rate constant k2 ¼ Z2 ð1  Ps Þ
Bubble loading
2
np2 db
b¼ where Smax ¼ 0:5S and S ¼ 4
Smax nbT dp
Collision frequency across eddies

dp þ db 2 2 1=2
Z1 ¼ 5:0 ðUp þ Ub2 Þ
2
Critical diameter of particle or bubble 15lf Uf2
di2 Ndcrit
2
¼
qi e
rffiffiffiffiffiffi

Collision frequency within eddies 8k dp þ db 3 e1=2
Z1 ¼
15 2 m
Turbulent fluctuating velocity of particle or bubble 7=9

0:4e4=9 di qi  qf 2=3
Ui ¼
m 1=3 qf
Detachment frequency pffiffiffiffiffiffiffiffiffiffiffiffiffi
C1 e 1=3
Z2 ¼
ðdp þ db Þ2=3
Probability of collision
 2
4 dp
Pc ¼ 1:5 þ Re0:72 b
15 db2
Bubble Reynolds number db Ub
Reb ¼
m
Probability of adhesion
 
ð45 þ 8Reb0:72 ÞUb tind
Pa ¼ sin 2 arctan exp
2
15db ðdb =dp þ 1Þ
Induction time 75 0:6
tind ¼ dp
h
Probability of stabilisation 

1
Ps ¼ 1  exp As 1  T
Bo
Bond number "
 #


2=3 dp db 1=3 4r h
dp Dqp g þ 1:9qp e
2
þ þ 1:5dp  db qf g sin2 k 
2 2 db 2
BoT ¼


h h
j6rsin k  sin k þ j
2 2
20 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24
Fig. 4. CFD predictions of velocity vectors (m/s) at impeller speed of
1200 rpm.
on the reported flow rate at 1800 rpm, a value for the factor K of
342 was needed in the model to match the predicted and reported
flow rates. Simulations at other impeller speeds have been
performed using the same value of K. The gas rates obtained by
CFD at various impeller speeds are shown to compare
favourably against measured data at frother concentrations of
50 ppm in Fig. 5. The differences in the measured gas rates at
frother concentrations of 20 ppm and 50 ppm are small because
at high frother concentrations the bubble sizes are fairly constant.
CFD results compare very favourably against either set of data.
At low frother concentrations, the equivalent length factor K is
likely to decrease as the size of the produced bubbles increases.
However, in flotation cells with frother addition operating with
discrete bubble formation, the approach of using K is reasonable.
It can be seen from measured results that the air flow rate
increases with decreasing frother concentration. The effect of
decreasing frother concentration on the air flow rate is likely
explained by the production of fewer and larger bubbles with
increased buoyancy. This means a reduced voidage especially in
Fig. 5. Comparison of CFD predictions against measured air flow rate
at different frother concentrations plotted against impeller speed.
the impeller region with an increased local density and pumping
power resulting in higher gas rate as observed in measured
results. The intercept on the x-axis represents the minimum
speed required just to maintain air in the standpipe. By
extrapolation, the CFD predicted intercept appears to be greater
than in the experiment. A reason for this difference is that at low
impeller speeds the air flow becomes laminar and the Blasius
friction factor no longer applies. Further simulations with
modified friction factors are needed to resolve the situation.
Increasing impeller speed increases the air flow rate because
the suction pressure increases. Values of gauge pressure at the
pipe tip of 1721, 2301 and 2965 Pa have been obtained at
impeller speeds of 1200, 1800 and 2300 rpm respectively.
At low frother concentrations, Girgin et al. (2006) found that
bubble size increases with increasing impeller speed in the Denver
cell. This observation is different to the case in a forced air
machine where an increase in impeller speed would produce finer
bubbles through the additional shear. The explanation is that in a
self-aerated Denver cell the resulting increase in air flow rate with
increasing impeller speed causes an increase in bubble size that
more than offsets any effect of increasing shear. However, the
effect at high frother concentrations, say 50 ppm, (which is the
condition used in our simulation) is different where the bubble size
is almost constant due to a decrease in bubble coalescence. In this
situation, it is reasonable for the air flow in the cell to be modelled
with a constant bubble size of 1 mm.
The other effect of increasing impeller speed is an increase in
the gas holdup. With increasing impeller speed, the gas has a
greater retention time in the cell because drag on the bubbles and
pulp circulation both increase with increasing turbulence
intensity which are accounted for in the model. A profile of
gas holdup in the flotation cell has been reported previously
(Koh et al., 2000).
3.2. Flotation kinetics
Using the model previously reported by Koh and Schwarz
(2006), simulations with flotation kinetics in the Denver cell
have been performed with pulps of 10% by weight of monosized
particles with a density of 2600 kg/m3.
Fig. 6. Aggregate density of a 1-mm bubble with fractional surface
coverage of particles plotted against particle diameter.
 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24
Fig. 7. Rise velocity of bubble–particle aggregate with fractional
surface coverage on a 1-mm bubble plotted against particle diameter.
The gravitational force acting on particles attached to bubbles
decreases the rise velocity of the aggregate especially with larger
particles at higher loadings. The additional body force on the
attached particles is incorporated in the present model by a source
(Bp) in the gas phase as follows:

2
db
Bp ¼ 4fc ðqp  ql Þgnb ð7Þ
dp
where fc is the fractional surface coverage, nb the local number
concentration (m− 3) of bubbles, ρp the density of particles, and db
and dp are the diameters of bubble and particles respectively. The
effects of the particle diameter on both the rise velocity and the
density of bubble–particle aggregates are illustrated in Figs. 6
and 7 where values have been calculated using a spreadsheet.
The aggregate density increases with particle diameter
because the total mass of attached particles and bubble increases
more than the total volume of particles and bubble for larger
particles. The aggregate rise velocity approaches zero for particles
of 250 μm size with a surface coverage of 0.5 because the
effective aggregate density is close to that of water as shown in
Fig. 6. The rise velocity at various loadings in Fig. 7 is obtained
Fig. 8. CFD prediction of the fraction of free and attached particles
remaining in the cell plotted against time for different particle
diameters (μm).
21
Fig. 9. CFD predictions of rate constant plotted against particle
diameter, with and without the effect of buoyancy reduction due to
attached particles, and comparison with literature data for quartz
(Ahmed and Jameson, 1985).
for bubbles in a quiescent fluid. One can expect a similar effect in
turbulent flow where the bubble rise velocity is reduced by the
attached particles. The bubble–particle aggregates can sink if the
particles are large and the bubble load also is high. The
calculations assume that the aggregate shapes are spherical and
the effect is greater on the rise velocity if the bubble–particle
aggregates are markedly different from spherical.
From CFD simulations with flotation in the Denver cell, the
total number of particles remaining in the cell is determined by
summing particles in all the finite volumes. The results for
different particle diameters are shown in Fig. 8 where the sum
of free and attached particles remaining in the cell operating at
1200 rpm is plotted as a function of time. The rate constants
are obtained from this diagram assuming first order kinetics
and plotted against particle diameter in Fig. 9.
Fig. 10. CFD predicted net attachment rates (m− 3 s− 1) after flotation
time of 129 s for particles of 15 μm size. Negative values indicate net
detachment.
22 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24
Fig. 11. CFD predicted net attachment rates (m− 3 s− 1) after flotation
time of 129 s for particles of 60 μm size. Negative values indicate net
detachment.
The effect of incorporating Eq. (7) into the model is shown
in Fig. 9 where flotation rates with and without body force at
different particle sizes are compared. The effect is greater at
intermediate particle size range of 100 to 300 μm where the
particle mass and particle loading on the bubble contribute
significantly to reduce the buoyancy. For very large particles of
about 400 μm, the net attachment rates are smaller and the
particle loadings on the bubble are also smaller, resulting in
lower reductions in the buoyancy. Hence, the flotation rate for
large particles remains almost the same with or without this
body force. The reduction in rise velocity due to reduced
Fig. 12. CFD predicted net attachment rates (m− 3 s− 1) after flotation
time of 129 s for particles of 120 μm size. Negative values indicate net
detachment.
Fig. 13. CFD predicted bubble loading after flotation time of 129 s for
particles of 15 μm size.
buoyancy confirms indirectly that transport rate of attached
particles to reach the froth–pulp interface is an important
contributor to the overall flotation rate.
In the present simulations, the maximum loading on a bubble
has been assumed to be equal to a half mono-layer coverage
(fc = 0.5) in the absence of experimental data. Flotation rates of
quartz particles in a Rushton-turbine stirred tank reported in the
literature (Ahmed and Jameson, 1985) are also plotted in Fig. 9
for comparison. The CFD predicted rates are higher because
power consumption is greater in the Denver cell (impeller speed
Fig. 14. CFD predicted bubble loading after flotation time of 129 s for
particles of 60 μm size.
 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24
1200 rpm) than in the stirred tank (impeller speed 600 rpm).
Besides, there is no froth in the model: the reported rates include
recovery through the froth layer and would therefore be lower
than the predicted rates which account only for flotation to the
pulp–froth interface. There are other possible reasons for the
high predictions including the assumed maximum loading of
half monolayer which was used as a first approximation. It is also
possible that detachment rates in the model may be too low. With
higher rates of detachment, the maximum flotation rate constant
could occur at less than 100 μm as normally observed in flotation
tests. Work on this aspect of the model is still in progress.
The distributions of net attachment rates in the Denver cell
operating at 1200 rpm after a flotation time of 129 s for three
particle diameters of 15, 60 and 120 μm are plotted in Figs. 10,
11 and 12 respectively. The negative values represent detach-
ment. For small particles of 15 μm, the net attachment rates are
highest near the impeller tip and in regions of highest shear.
There is little detachment for 15 μm particles. For larger particles
of 60 and 120 μm, the locations of highest net attachment rates
are away from the impeller. In fact, for 120 μm particles, the
highest rates are outside the shroud. The high detachment rates
for 60 μm particles are near the impeller tip, while the high
detachment rates for 120 μm particles are around the impeller
and outside the shroud.
The results indicate that the smaller particles need higher
impeller speeds for flotation while larger particles float best at
lower speeds. Thus, in plant operations, flotation cells should
operate with impeller speeds that perform best for a given
particle size. This agrees with practice in many mineral plants
where the feed streams are sized into coarse, fines and slimes and
treated separately in customised flotation cells where the speeds
are adjusted to suit the particle size.
The distributions of bubble loadings in the Denver cell
operating at 1200 rpm after a flotation time of 129 s for three
Fig. 15. CFD predicted bubble loading after flotation time of 129 s for
particles of 120 μm size.
23
particle diameters of 15, 60 and 120 μm are plotted in Figs. 13, 14
and 15 respectively. Bubble loading is defined as a fraction of
the maximum coverage available on a bubble; the maximum
coverage is assumed to be equal to a half mono-layer (fc = 0.5) in
the present simulations. The effect of particle size on bubble
loading is significant. For smaller particles of 15 and 60 μm, the
bubble loading is close to the maximum coverage everywhere in
the flotation cell indicating there is very little detachment of
particles from the rising bubbles. For larger particles of 120 μm,
the loading varies more across the flotation cell and is generally
lower in value in comparison to smaller particles. The dis-
tribution of bubble loading in Fig. 15 indicates that the loading
reaches a maximum in the region around the shroud and
decreases as the bubbles rise towards the surface. In this
situation, detachment of particles from the rising bubbles is
occurring. The effect is greater for larger particles than it is for
smaller particles.
The variation in bubble loading within the flotation cell is
caused by the different bubble–particle attachment and
detachment rates in different parts of the cell. The local
turbulence level affects both the attachment and detachment
processes occurring in each region and there is a balance
between the two rates. For a given particle size, the local
attachment rate can be greater than the local detachment rate in a
particular region, while the reverse may be true in other regions
resulting in negative net attachment rates. Thus, the bubble
loading can vary between regions in the flotation cell and is
dependent on whether the local attachment or detachment rates
predominate.
4. Conclusions
The hydrodynamics and flotation kinetics in the
Denver laboratory cell have been studied using CFD
modelling. From two-phase simulations at different
impeller speeds, the predicted air flow rates have been
found to compare favourably against measured values
reported in the literature. The model provided detailed
hydrodynamics of the Denver cell that are very useful for
understanding batch flotation test results and for inves-
tigations in the design and operation of larger flotation
cells.
Simulations with flotation kinetics showed that the
gravitational force acting on the attached particles is
significant. The effect is a decrease in the bubble rise
velocity which in turn causes the flotation rate to decrease.
Model predictions showed that net bubble–particle
attachment for coarse particles occurs mainly outside the
shroud. Near the impeller, both attachment and detach-
ment rates are high with the net effect being detachment.
The coarse particles have been shown to detach from
rising bubbles. For fine particles, the attachments occur
predominately inside the shroud with high loading on the
bubbles carrying particles to the surface.
24 P.T.L. Koh, M.P. Schwarz / Int. J. Miner. Process. 85 (2007) 16–24

The results confirm current practice in many mineral References

plants where the feed streams are sized into coarse, fines
and slimes and treated separately in customised flotation Ahmed, N., Jameson, G.J., 1985. The effect of bubble size on the rate
cells where the speeds are adjusted to suit the particle size. of flotation of fine particles. International Journal of Mineral
Processing 14, 195–215.
For practical application, the CFD model can be used to Arbiter, N., Harris, C.C., Yap, R.F., 1976. The air flow number
investigate performance of a flotation cell. The design and in flotation machine scale-up. International Journal of Mineral
operating conditions can be identified for operations with Processing 3, 257–280.
close to maximum bubble loading. Flotation cell design CFX User Guide, Release 4.4, 2001. Computational Fluid Dynamics
should ideally minimise the regions of net detachment Services, AEA Industrial Technology. Harwell Laboratory,
Oxfordshire, UK.
while maximizing the attachment rates available in the Forrester, S.E., Rielly, C.D., 1994. Modelling the increased gas
cell. capacity of self-inducing impellers. Chemical Engineering Science
49, 5709–5718.
Nomenclature Girgin, E.H., Do, S., Gomez, C.O., Finch, J.A., 2006. Bubble size as a
As constant = 0.5 function of impeller speed in a self-aeration laboratory flotation
cell. Minerals Engineering 19, 201–203.
B body force Harris, C.C., Mensah-Biney, R.K., 1977. Aeration characteristics of
Bo⁎ bond number laboratory flotation machine impellers. International Journal of
C1 constant = 2 Mineral Processing 4, 51–67.
d bubble or particle diameter Joshi, J.B., Pandit, A.B., Sharma, M.M., 1982. Mechanically agitated
D pipe diameter gas–liquid reactors. Chemical Engineering Science 37, 813–844.
Koh, P.T.L., Schwarz, M.P., 2003. CFD modelling of bubble–particle
Di diffusivity of phase i collision rates and efficiencies in a flotation cell. Minerals
f friction factor Engineering 16, 1055–1059.
fc surface coverage Koh, P.T.L., Schwarz, M.P., 2006. CFD modelling of bubble–particle
F drag force attachments in flotation cells. Minerals Engineering 19, 619–626.
Koh, P.T.L., Manickam, M., Schwarz, M.P., 2000. CFD simulation of
g gravity vector
bubble–particle collisions in mineral flotation cells. Minerals
k turbulent kinetic energy or rate constant Engineering 13, 1455–1463.
K resistance factor Koh, P.T.L., Schwarz, M.P., Zhu, Y., Bourke, P., Peaker, R., Franzidis,
n particle number concentration J.P., 2003. Development of CFD models of mineral flotation cells.
P pressure or probability Third International Conference on Computational Fluid Dynamics
Re Reynolds number in the Minerals and Process Industries, Melbourne, Australia,
December, pp. 171–175.
S mass source or sink, or surface ratio Launder, B.E., Spalding, D.B., 1974. The numerical computation of
t time turbulent flows. Computer Methods in Applied Mechanics and
U velocity Engineering 3, 269–289.
Z collision frequency Rielly, C.D., Evans, G.M., Davidson, J.F., Carpenter, K.J., 1992.
α volume fraction Effect of vessel scaleup on the hydrodynamics of a self-aerating
concave blade impeller. Chemical Engineering Science 47,
β bubble loading 3395–3402.
ε turbulent eddy dissipation Sardeing, R.F., Poux, M., Xuereb, C., 2006. Development of a new
θ contact angle gas-inducing turbine family: the partially shrouded turbine.
μ dynamic viscosity Industrial and Engineering Chemistry Research 45, 4791–4804.
Sawant, S.B., Joshi, J.B., Pangarkar, V.G., Mhaskar, R.D., 1981. Mass
ν kinematic viscosity
transfer and hydrodynamic characteristics of the Denver type of
ρ density flotation cells. The Chemical Engineering Journal 21, 11–19.
σ surface tension

Subscripts
1 free particle or attachment
2 attached particle or detachment
b bubble
f fluid
i phase index
p particle
T total