Seventh International Conference on CFD in the Minerals and Process Industries

CSIRO, Melbourne, Australia

9-11 December 2009

CFD MODELS OF MICROCEL AND JAMESON FLOTATION CELLS

P.T.L. KOH1 and M.P. SCHWARZ1

1
CSIRO Minerals, Clayton, Victoria 3169, AUSTRALIA

overflow launder. The remaining slurry is discharged from

ABSTRACT the cell.
This paper describes an investigation using computational
Computational fluid dynamics (CFD) models of the
fluid dynamics (CFD) to increase understanding of the
Microcel column and Jameson cell have been developed
hydrodynamics in coal flotation and to identify any
for coal flotation. The models have been used to increase
opportunities to improve the design and operation of both
understanding of the cell hydrodynamics and to identify
the Microcel column and Jameson cell. The flotation
any opportunities to improve design and operation. The
effect is included as three sub-processes involving
models incorporate micro-processes including particle-
collision, attachment and detachment. A turbulent
bubble attachments and detachments. From CFD
collision model has been used (Koh et al., 2000) to
simulations, liquid velocities, turbulent dissipation rates,
estimate the rate of particle-bubble encounters. The
gas hold-up, particle-bubble attachment rates and
probability of collision is due to the streamline effect of
detachment rates are obtained. The attachment and
fine particles moving around a larger bubble. The CSIRO
detachment rates per particle have been used for
CFD model incorporates micro-processes from cell
comparing cell performance.
hydrodynamics that affect particle-bubble attachments and
detachments.
NOMENCLATURE
As constant = 0.5 The issues facing coal flotation are two-fold: firstly, a
B buoyancy reduction force (N m-3) drive to increase capacity of existing installations, and
Bo* bond number secondly, a need to increase the recovery of coarse coal.
C1 constant = 2 These issues are influenced by the flow conditions inside
d bubble or particle diameter (m) the cell. The conditions that affect coal recovery can be
Di diffusivity of phase I (m2 s-1) identified through an analysis of the CFD results in the
F drag force (N m-3) tank. The effect of composite particles in coal flotation is
g gravity vector (m s-2) considered by the spread of specific gravities and particle
k turbulent kinetic energy (m2 s-2) sizes. The application of CFD modelling can provide
k1 rate constant (s-1) knowledge of the hydrodynamics involved in enhancing
n particle number concentration (m-3) particle-bubble attachments while minimizing particle
P pressure or probability (N m-2) detachment from bubbles.
Re Reynolds number
Si mass source or sink (kg m-3 s-1) MODEL DESCRIPTION
S surface ratio
t time (s) Transport Equations
U velocity (m s-1) Fundamental equations for the conservation of mass,
Z collision frequency (m-3 s-1) momentum and turbulence quantities are solved. Using the
α volume fraction Eulerian-Eulerian multi-fluid approach, the gas and pulp
β bubble loading parameter phases are treated as interpenetrating continua. The
γ shear rate (s-1) variables solved include Cartesian velocity components,
ε turbulent eddy dissipation (m2 s-3) pressure and volumetric fractions.
θ contact angle (degree)
μ dynamic viscosity (Pa s) ∂ (α i ρ i ) r
+ ∇ ⋅ (α i ρ iU i − ρ i Di ∇α i ) = S i (1)
ν kinematic viscosity (m2 s-1) ∂t
ρ density (kg m-3) r
σ surface tension (N m-1) ∂ (α i ρ iU i ) r r
+ ∇ ⋅ (α i ρ iU i ⊗U i )
∂t
( )
INTRODUCTION r r
= ∇ . α i ( μ L ,i + μ T ,i )(∇U i + (∇U i ) T
Flotation is used by the minerals industry to separate r
valuable minerals from the ore in air-sparged flotation + α i (Bi − ∇Pi ) + Fi + S iU i
cells. Separation is achieved when particles of the (2)
valuable minerals preferentially attach to air bubbles and
are floated to the surface, forming a froth layer. The The transport equations were solved within the
particles are removed from the froth by means of an computational fluid dynamics code CFX-4.4 (2001). The
transport of solids within the pulp phase is modelled using
the algebraic slip model. The source terms include the

Copyright © 2009 CSIRO Australia 1

pressure sources, the mass source or sink of gas entering Turbulent kinetic energy k and turbulent dissipation rate ε
and leaving the vessel, the generalised inter-phase force are solved for the continuous pulp phase. The turbulent
including drag and buoyancy forces. viscosity in the pulp phase is calculated using the standard
k-ε turbulence model (Launder and Spalding, 1974).
dn p1
= −k1 n p1 nbT (1 − β ) + k2 nbT β
Net attachment
The flotation effect is simulated as three sub-processes
rate dt involving collision, attachment and detachment. In
simulations with flotation kinetics, the transfer of particles
Attachment rate
k 1 = Z 1 Pc Pa Ps between the pulp and bubbles is achieved by applying
constant source terms for particle number concentration ni in the
transport equation as follows:
Detachment rate
k2 = Z 2 (1 − Ps ) ∂ (α i ni ) r
constant
+ ∇ ⋅ (α i niU i ) = −φ a + φ d (3)
∂t
np2
β= where Smax = 0.5S
S max nbT where φa and φd are sources or sinks specifying
Bubble loading 2
attachment and detachment rates respectively. The local
and S = 4 ⎜ db ⎟
⎛ ⎞ particle-bubble detachment rates are from the flow field
⎜d ⎟ using equations shown in
⎝ p⎠ Table 1 (Koh and Schwarz, 2006).
Collision ⎛ d p + db
2
⎞ ⎛ '2 1/ 2

Z 1 = 5.0 ⎜⎜ ⎟⎟ ⎜U p + U b' ⎞⎟ RESULTS

2
frequency
across eddies ⎝ 2 ⎠ ⎝ ⎠ Multi-phase flow simulations (for liquid, air and solids)
have been performed for the Microcel column and
Critical Jameson cell. Results of the simulations for both cells
15μ f U 'f
2
diameter of consist of velocity vectors, gas hold-up, turbulent
d i2 > d crit
2
=
particle or ρiε dissipation rates and the particle-bubble attachment and
bubble detachment rates.

⎛ d p + db ⎞ ⎛ ε ⎞
Collision 3
8π
1/ 2

frequency Z1 = ⎜⎜ ⎟⎟ ⎜ ⎟ Microcel Column

within eddies 15 ⎝ 2 ⎠ ⎝ν ⎠ The Microcel flotation column of 4.9m diameter and
10.7m height operating with 12 spargers is manufactured
Turbulent by Eriez Pty Ltd. The CFD mesh of the column is shown
2/3
fluctuating
2 0.4 ε 4 / 9 d i7 / 9 ⎛⎜ ρ i − ρ f ⎞
⎟
in Figure 1. The hydrodynamics of the column operating
velocity of U i' = with 250 μm particles at 12% w/w solids and solid density
ν 1/ 3 ⎜ ρ ⎟
particle or ⎝ f ⎠ of 1520 kg m-3 are plotted in Figure 2 (velocity vectors)
bubble and Figure 3 (void fraction). An average bubble size of
1 1.0 mm is used in the simulation for the Microcel column
Detachment C1 ε 3
Z2 = (Filippov et al, 2000; Pyecha et al, 2006).
frequency (d p + db ) 3
2

In Figure 2, there is a general circulating flow that rises in

2 the centre with downward flow near the wall. An average
Probability of ⎛ 4 ⎞ d
Pc = ⎜1.5 + Re b0.72 ⎟ 2
p
circulating velocity of 1 m s-1for the pulp can be compared
collision ⎝ 15 ⎠ db with a velocity of 5.4 m s-1 at the sparger exits. Feed inlet
velocity of 0.53 m s-1 has some effect on the circulating
Bubble 2 flow creating complex flow patterns at the top of the
Reynolds d b U b'
Re b = column. The effect will be greater with increasing
number ν throughput and potentially can affect the pulp-froth
interface.
⎛ ⎡ ⎤⎞
⎜ − (45 + 8 Reb0.72 ) U b' tind ⎥ ⎟
2
Probability of Pa = sin 2 ⎜ 2 arctan exp ⎢
⎢ 15d b (d b / d p + 1) ⎥⎟ In Figure 3, the contours of gas hold-up show a fairly
adhesion ⎜ ⎢⎣ ⎥⎦ ⎟⎠
⎝
uniform distribution of gas in the column with slightly
75 higher hold-up near the wall due to the downward flow.
Induction time tind = d p0.6 An average void fraction of 0.176 for the whole column
θ compares favourably with a value of 0.23 observed at the
⎡ ⎛ 1 ⎞⎤ plant for the top part of the column, and a value of 0.35
Probability of
Ps = 1 − exp ⎢ As ⎜1 − ⎟⎥ predicted at the sparger exits.
stabilisation ⎣ ⎝ Bo * ⎠⎦
⎡
In Figure 4, the turbulent dissipation rates are greatest
d ⎞ 3⎤
−1
2 ⎛ dp ⎛ 4σ ⎞ ⎛ θ⎞
d p2 ⎢Δρ p g + 1.9 ρ pε 3 ⎜⎜ + b ⎟⎟ ⎥ + 1.5d p ⎜⎜ − db ρ f g ⎟⎟ sin 2 ⎜ π − ⎟ near the sparger exits and near the two discharges. The
⎢ ⎝ 2 2⎠ ⎥ ⎝ db ⎠ ⎝ 2⎠
Bond number Bo* = ⎣ ⎦
⎛ θ⎞ ⎛ θ⎞
6σ sin ⎜ π − ⎟ sin ⎜ π + ⎟
turbulent dissipation rates near the pulp-froth interface are
⎝ 2⎠ ⎝ 2⎠
higher than values in the bulk because of the feed inlet
flows and the presence of feed pipes and deflectors.
Table 1: Equations for particle-bubble interaction.

Copyright © 2009 CSIRO Australia 2

The distribution of attached particles is shown in Figure 5 CONCLUSION
where the higher concentrations are found in regions with
higher gas hold-up. In Figure 6, the distribution of the CFD simulations for both the Microcel column and
unattached particles shows higher concentration at the Jameson cell have produced data that provide useful
base of the column, indicating some settling of unattached insights in relating cell hydrodynamics to the particle-
particles. bubble attachment and detachment rates for coal flotation.
It has been found that the local value of the turbulent
The attachment and detachment rates are plotted in Figure dissipation rate has a direct influence on the local particle-
7 and Figure 8 respectively. The maximum attachment bubble detachment rate. It is therefore important to
rates occur at the centre of the column as well as near the identify the locations of these regions with high turbulent
feed and sparger exits, while the maximum detachment dissipation rates.
rates are found mainly near the sparger exits where the
turbulent dissipation rates are highest. The local value of Under the conditions investigated, the Microcel column is
the turbulent dissipation rate has direct influence on the well designed with fairly good performance throughout
local particle-bubble detachment rate. It is important to the column, but further improvements are still possible. In
identify the locations of these regions with high turbulent the Jameson cell, the turbulent dissipation rates were
dissipation rates. found to be greatest near the inner launder where the flow
is influenced by the air coming from beneath the sloping
Jameson Cell launder wall. Because of this effect, the turbulent
The standard B6000-20 Jameson cell, manufactured by dissipation rates at the pulp-froth interface are also higher
Xstrata Technology, has a diameter of 6m operating with near the inner launder. Cell performance can be improved
20 downcomers. The CFD mesh is shown in Figure 9 with further investigations.
consisting of the pulp and froth zones with a bubble
disperser plate. This work has demonstrated that CFD modelling is a cost
effective means of developing an understanding of
CFD simulation has been performed for the case with particle-bubble attachments and detachments, and can be
7.5% w/w solids, 250 μm size and solid density 1520 used to identify and test potential cell or process
kg m-3. In Figure 10, the general pattern of flow from the modifications.
downcomer exit to the pulp-froth interface is around the
bubble disperser with a small portion of flow through the ACKNOWLEDGEMENT
perforated plate. The pulp-air mixture has a velocity of The authors are grateful to the Australian Coal
0.49 m s-1 with a void fraction of 0.43 at the downcomer Association Research Program (ACARP) for financial
exits. The velocities at the inner launder wall tend to be support and permission to publish. Thanks are also
higher than those at the outer launder wall. The direction presented to Xstrata Technology Pty. Ltd. and Eriez
of velocities at the inner launder wall is downward at the Magnetics Pty. Ltd. for suppling drawings of the flotation
deeper end, but it is upward at the shallow end. This is due cells.
to the high gas hold-up beneath the inner launder, as
shown in Figure 11. Because of the sloping wall, the air
REFERENCES
escapes towards the shallow end of the launder. The
average void fraction in the cell is 0.072, while in the CFX User Guide, Release 4.4, (2001), Computational
upper part of the cell, the value is about 0.16. Fluid Dynamics Services, AEA Industrial Technology,
Harwell Laboratory, Oxfordshire, UK.
In Figure 12, the turbulent dissipation rates are highest FILIPPOV, L.O., JOUSSEMET, R. and HOUOT, R.,
near the inner launder wall where the flow is influenced (2000) “Bubble spargers in column flotation: Adaption to
by the air coming out beneath the sloping launder wall. precipitate flotation”, Minerals Engineering, 13, 37-51.
Because of this effect, the turbulent dissipation rates at the HARBORT, G.J., COWBURN, J.A. and MANLAPIG,
pulp-froth interface are also higher towards the inner E.V., (2005) “The effect of residence time and aeration on
launder. coal recovery within the high intensity zone of a flotation
machine”, Centenary of Flotation Symposium, Brisbane,
In Figure 13, the distribution of the unattached particles p. 449-456.
shows higher concentration at the base of the cell, KOH, P.T.L., MANICKAM M. and SCHWARZ, M.P.,
indicating some settling of unattached particles. The (2000) “CFD Simulation of Particle-bubble Collisions in
distribution of attached particles is shown in Figure 14 Mineral Flotation Cells”, Minerals Engineering, 13, 1455-
where the higher concentrations are found in regions with 1463.
higher gas hold-up. In the simulation, 80% of the particles KOH, P.T.L. and SCHWARZ, M.P., (2006) “CFD
were assumed to be attached to the bubbles at the exit of Modelling of Particle-bubble Attachments in Flotation
the downcomers based on data by Harbort et al (2006). Cells”, Minerals Engineering, 19, No. 6-9, 619-626.
LAUNDER, B.E. and SPALDING, D.B., (1974) The
The attachment and detachment rates are plotted in Figure numerical computation of turbulent flows. Comp. Meths.
15 and Figure 16 respectively. The maximum attachment Appl. Mech. Engng., 3, 269-289.
rates occur near the downcomer exits while the maximum PYECHA, J., LACOUTURE, B., SIMS, S., HOPE, G.
detachment rates are found near the wall of the inner and STRADLING A., (2006) “Evaluation of a Microcel
launder where the turbulent dissipation rates are highest. sparger in the Red Dog column flotation cells”, Minerals
Thus, it is important to minimise these regions with high Engineering, 19, 748-757.
turbulent dissipation rates to minimise detachment.

Copyright © 2009 CSIRO Australia 3

Figure 1: CFD mesh of Microcel column. Figure 3: Void fraction in Microcel column.

Figure 2: Liquid velocity (m s-1) in Microcel column. Figure 4: Turbulent dissipation rate (W kg-1) in Microcel
column.

Copyright © 2009 CSIRO Australia 4

Figure 5: Attached particle-volume fraction in Microcel Figure 7: Attachment rate (m-3 s-1) in Microcel column.
column.

Figure 6: Unattached particle-volume fraction in Figure 8: Detachment rate (m-3 s-1) in Microcel column.
Microcel column.

Copyright © 2009 CSIRO Australia 5

 Figure 13: Attached particle-volume fraction in Jameson
Figure 9: CFD mesh of Jameson cell with twenty cell.
downcomers.

Figure 14: Unattached particle-volume fraction in

Figure 10: Liquid velocity (m s-1) in Jameson cell.
Jameson cell.

Figure 15: Attachment rate (m-3 s-1) in Jameson cell.

Figure 11: Void fraction in Jameson cell.

Figure 12: Turbulent dissipation rate (W kg-1) in Jameson

Figure 16: Detachment rate (m-3 s-1) in Jameson cell.
cell.

Copyright © 2009 CSIRO Australia 6