You are on page 1of 11

International Journal of Mineral Processing 146 (2016) 54–64

Contents lists available at ScienceDirect

International Journal of Mineral Processing

journal homepage: www.elsevier.com/locate/ijminpro

CFD simulation on influence of suspended solid particles on bubbles'


coalescence rate in flotation cell
A.R. Sarhan a,b, J. Naser a,⁎, G. Brooks a
a
Department of Mechanical and Product Design Engineering, Swinburne University of Technology, Hawthorn, Victoria 3122, Australia
b
Department of Mechanical Engineering University of Anbar, Ramadi, Anbar 31001, Iraq

a r t i c l e i n f o a b s t r a c t

Article history: A computational fluid dynamics (CFD) model was used to investigate the influence of solid concentration on
Received 29 May 2015 bubbles' coalescence rate in flotation cell using Eulerian–Eulerian approach. CFD simulations were performed
Received in revised form 19 November 2015 with AVL-FIRE 2009.2, and the existing flow field was modelled for two-phase (gas–liquid) and three-phase
Accepted 30 November 2015
(gas–liquid–solids). The liquid phase was treated as a continuum and the gas phase (bubbles) and solid particles
Available online 2 December 2015
were considered as dispersed phases. The population balance equation for bubble break-up and bubble coales-
Keywords:
cence rate and the interfacial exchange of mass and momentum as well as bubble–particle attachment and
CFD detachment have been included in the CFD code by writing subroutines in FORTRAN. This investigation focused
Flotation on studying the effect of solid particle on bubble break-up and bubble coalescence rate in the flotation cell at
Solid concentration different superficial gas velocity values. The results predict that the presence of solid particles reduced the gas
Gas holdup holdup in a flotation column. With the increase of the superficial gas velocity the size of gas bubble that were
Population balance generated inside the cell decreased, leading to increased gas holdup. The result also shows that the Sauter
mean diameter of bubbles decreases with the increase of solid concentration. Reasonably good agreement was
obtained between simulation and experimental results for the effect of solid concentration on gas hold-up and
axial pressure profile. In the current study, the froth zone was neglected, only the pulp zone was simulated.
This is a deficiency of the present model, as the pulp is only one part of the flotation process, and it is physically
linked to the froth. However, the model is a step towards gaining a complete view to describing the processes
within a flotation cell through inspect the impact of presence of solid particles on the bubble coalescence rate
under the different operation conditions.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction bubbles and commence their journey to the top of the pulp zone due
to difference in density between the solid particle–gas bubble aggregate
The interaction of solid-coated bubbles is important for a wide range and water report to the froth product. Whereas, part of these particles
of catalytic reactions used in many industrial applications such as min- remain entrained in the froth which can be drained off or mechanically
eral flotation, deinking of recycled paper, water treatment and oil sepa- skimmed away thereby effecting separation (Verrelli et al., 2014; Zhang
ration. Specifically, the process has crucial role in froth flotation because and Finch, 2014).
the ultimate recovery of particles relies on the survival chance of the There are many studies in open literature that describe the mixing of
bubbles during the interaction (Grevskott et al., 1996; Tao, 2005; Ata, two phases systems, byt hardly any on three phase systems (Grevskott
2008, 2009). Froth flotation is a complex physico-chemical process et al., 1996; Eggers et al., 1999; Hagesaether et al., 2002b; Sattar et al.,
which is extensively used in mineral industry to separate hydrophobic 2013b, 2014). The phenomena that govern flotation cells are even
minerals (valuable minerals) from hydrophilic ones (gangue). The more complicated since the scales of the two dispersed phases (solid,
importance of this process lies on its ability to treat low and complex gas) are different. There is a big difference in sizes of gas bubbles and
grade ores where the mineral particle size is too small for other separa- solid particles, where the diameters of gas bubbles are in the range of
tion techniques to be efficient. The flotation process utilises the differ- 1 mm while the particle size range between approximately 50 to
ences in the ability of air bubbles to selectively adhere to valuable 600 μm which are much smaller than this value (Sobhy and Tao,
mineral surfaces in order to separate it from the gangue component. 2013). The gas and solid phases (dispersed phases) are contained only
In froth flotation, solid particles were fed at the top of collection zone. in the liquid phase (continuous phase). Furthermore, the measurement
Only hydrophobic particles have the opportunity to attach to the of system parameters such as gas holdup, bubble coalescence rate or
even the bubble size is complicated due to the presence of particles in
⁎ Corresponding author. the system, where the particles play a crucial role in bubble coalescence
E-mail address: jnaser@swin.edu.au (J. Naser). and froth stability.

http://dx.doi.org/10.1016/j.minpro.2015.11.014
0301-7516/© 2015 Elsevier B.V. All rights reserved.
A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64 55

Coalescence of bubbles occurs when two gas bubbles collide, trap- 2009). This code is based on the finite volume technique which
ping liquid between them whose thickness decreases gradually to a crit- rests on integral conservation statements applied to a general con-
ical thickness due to the drainage of the liquid before the film ruptures trol volume. The conservation of mass, momentum, and energy of
(Prince and Blanch, 1990). The models that describe this process are each phase in each cell was calculated by using the two-fluid model
based on two phase systems (gas, liquid). However, the presence of or Eulerian–Eulerian description which is applied for both the two-
solid in the inter film may inhibit liquid drainage by creating a steric ob- phase gas–liquid flow and the three-phase gas–liquid–solid. In
structions. The effect of this obstruction depends on solid concentration, Eulerian–Eulerian multiphase flow approach, gas, liquid and solid
particle size, shape, and contact angle and the orientation of the parti- phases interact with each other and there is significant exchange of
cles at the interface (Ata, 2008). Ata et al. (2003) studied the bubble mass and momentum between phases. User-defined subroutines
size as a function of height above the froth–pulp interface. They found (UDS) were employed to predict the interfacial mass and momen-
that the amount of entrained solids in the froth considerably reduces tum between phases in each cell. The interfacial exchange terms in
bubble coalescence. This is because it increases the slurry viscosity be- the conservation equations dealing with the exchanges of mass, mo-
tween the bubble films, and thereby reducing the drainage rate of liq- mentum, and energy were incorporated. In current model, both the
uid. Bao et al. (2008) conducted an experimental investigation to dispersed (gas and solid) and the continuous (liquid) phases are
determine the impact of solids concentration and gas flow rate on the simulated in the Eulerian frame of reference as interpenetrating con-
gas holdup, just-suspension agitation speed, and power consumption tinua. The turbulent viscosity in the continuous phase is calculated
in a multi-impeller hot three-phase system with settling particles. using the standard k - ɛ turbulence model, which is a modification
Slurry temperature has significant impact on the physical properties of of the original k - ɛ model proposed by Launder and Spalding
the system. The bubble size and the gas holdup are likely to influence (1974). In the k - ɛ model, the turbulent kinetic energy k and the
by changes in the liquid viscosity, interfacial tension, and gas density. dissipation rate ɛ are calculated as follows:
The results have shown that the presence of solid particles only has a Turbulence energy:
small influence on gas dispersion in the system.   !
Ata (2008) studied the interaction between two solid-coated bubbles ∂ðρkÞ ∂ ρU j k ∂U j ∂ μ t ∂k
þ ¼ ρu j ui þ μþ  ρε: ð1Þ
using a high speed camera. The fractional coverage of the surface by par- ∂t ∂X j ∂X j ∂X j σ k ∂X j
ticles was used to determine the time required for two bubbles of equal
volumes to coalesce. His study shows that the increase of the bubble sur- Energy dissipation:
face coverage leads to increased coalescence. Although this work was
  !
carried out to study coalescence behaviour in the presence of solid parti- ∂ðρε Þ ∂ ρU j ε ∂U j ε ∂ μ t ∂ε ε2
cles, it was only applied to one bubble size generated under laboratory þ ¼ C ε1 ρu j ui þ μþ  C ε2 ρ : ð2Þ
∂t ∂X j ∂X j k ∂X j σ ε ∂X j k
conditions. There are different bubble classes within flotation cell
which require more extensive investigations. Gallegos-Acevedo et al.
where Cε1, Cε2, σk, σε and Cμ are empirical constants in the standard
(2010) used high speed photograph to study the effect of the presence
k - ɛ turbulence model, and their values are 1.44, 1.92, 1.0, 1.3, and 0.09
of solid particles on the bubble coalescence phenomena. Their results re-
respectively.
veals that the presence of solid particles has a great influence on bubble
coalescence phenomena within collecting zone since the gas holdup
2.1. Governing equations
value is εg b 0.15, while it has less effect in froth zone where εg N 0.6. How-
ever, all these studies were conducted under laboratory conditions,
The governing equations in Eulerian–Eulerian approach that is used
where bubbles were controlled to behave in desired condition.
to describe the motion of gas, liquid and solid in the cell are given as
Zhang and Finch (2014) conducted an experimental studied to ex-
follows (Manual, 2009):
amine the influence of solid particles on pulp hydrodynamics, bubble
coalescence intensity within froth phase and water overflow rate with
2.1.1. Mass conservation equation
solids present. They considered two types of solids; hydrophilic
The mass conversation for each phase in each cell was calculated
particles (silica) and hydrophobic particles (talc). Often, hydrophobic
using the following equation (Manual, 2009):
particles are expected to accumulate at the bubble liquid–air interface,
while hydrophilic particles will concentrate in the interstitial water
∂ðα m ρm Þ
between bubbles (within the froth). þ ∇:α m ρm vm ¼ 0; m ¼ 1…M ð3Þ
∂t
The aim of this work is to obtain a better understanding of the im-
pact of the presence of solid particle in flotation column under different where αm, vm and ρm are the volume fraction, average velocity and
conditions, through predicting of the changes of the bubble classes' the density of phase m. The compatibility condition must be observed
distribution in the cell due to bubble break up and coalescence. Special as:
attention has been given to gas holdup because it is considered to be
one of the most important variables in the collection zone which can af- M
∑m¼1 α m ¼ 1: ð4Þ
fect significantly flotation performance (Tavera et al., 2001). The param-
eters of interest are gas flow rate and the concentration of solid
particles. This paper presents a 3D CFD model similar to the experimen-
2.1.2. Momentum transfer equation
tal model of Shukla et al. (2010). The developed a CFD model is used to
The momentum conservation equation used for the present model is
predict the number density of different bubble class. The source term in
given by:
the CFD simulations, accounting for the death and birth of different
bubble classes due to coalescence and break-up was obtained with ∂ðα m ρm vm Þ  
the model derived by Hagesaether et al. (2002b) and modified by þ ∇:ðα m ρm vm vm Þ ¼ α m ∇p þ ∇:α m τ m þ T tm þ α m ρm f
∂t
Sattar et al. (2013c). þMm ; m ¼ 1…M ð5Þ

2. The hydrodynamic model where f is the body force vector which comprises of gravity g and the in-
ertial force in rotational frame; Mm represents the momentum interfacial
The commercial CFD software package AVL FIRE 2009.2 was used interaction between phases (m = g , l , s), and p is pressure which is as-
for modelling the hydrodynamics of a flotation column (Manual, sumed to identical for all phases (i.e. p =pm , m= 1 … M). Shear stress
56 A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64

for each phase can be calculated as: τ m ¼ μ m ½ð∇vm þ ∇vTm Þ  23 ∇:vm , where BBi, BCi, BCi, and DCi are birth and death due to break-up and
and Reynolds stress is determined using the following equation: coalescence respectively. The source term of break-up and coalescence
of bubbles in Eq. (8) that was developed by Hagesaether et al. (2002a)
 
  2 2 and corrected by Sattar et al. (2013c) was used in the present model
Ttm ¼ μ tm ∇vm þ ∇vTm  ∇:vm I  ρm K m I ð6Þ
3 3 (Sattar et al., 2013c):

N   i  
where I is a unit tensor usually known as Kronecker delta (Sattar et al.,
BBi ¼ ∑ Ω B φ j ; ϑφi þ ∑ X iþ1;k Ω B φiþ1 ; φ j ; i ¼ 1; ……; N
2013c). The turbulent viscosity μtm is computed by using the following j¼iþ1;i≠N j¼1;i≠N
equation (Manual, 2009): ð15Þ

K 2m i1   i  
μ tm ¼ C μ ρm ð7Þ DBi ¼ ∑ ΩB φ j ; φi þ ∑ Xi;k ΩB φ j ; φi i ¼ 2; ……; N ð16Þ
εm
j¼1 j¼1;i≠N

where K and ε are turbulent kinetic energy and its dissipation rate   i1 
i1   
respectively. BC i ¼ ∑ X i; j Ω C φi ; φ j þ ∑ 1  X i1; j Ω C φi1 ; φ j ; i ¼ 2; ……; N
The following equations (Manual, 2009) was used to calculate the j¼1;i≠N j¼1

momentum interfacial exchange by considering the drag force: ð17Þ

1 N1   i  
Mp ¼ C D ρp Ai jV r jV r ¼ Mq ð8Þ DC i ¼ ∑ Ω C φi ; φ j  ∑ X i; j Ω C φi ; φ j ; i ¼ 1; ……; N  1 ð18Þ
8 j¼1 j¼1

where Ai is interfacial area density, and Vr is the drift velocity which is where ΩB(φj, φi) is the break-up rate of bubble class j with volume vj
defined as the velocity of a secondary phase (q) relative to the velocity that goes into bubble class i with volume vi and the complementery
of primary phase (p): fraction can be introduced as (Hagesaether et al., 2002b):
V r ¼ V q  V p: ð9Þ vk ¼ v j  vi : ð19Þ

The drag coefficient CD is taken from (Schlichting, 1979): It was assumed that there are ten bubble classes in the cell, and the
8 volume of upper bubble class is twice the volume of the lower class. In
< 24  
the current model, it was assumed that there is no breakage death of
1 þ 0:15Reb 0:687 Reb ≤1000
CD ¼ Re ð10Þ the lowest class and no breakage birth of the highest class, respectively.
: b
0:438 Reb N1000:
On the other hand, it was assumed that there is no coalescence death of
the highest class and no coalescence birth of the lowest class, respec-
The Reynolds number Reb is a function of the Sauter mean diameter tively. The following equation was used to calculate the complementary
which can be calculated as (Schiller and Naumann, 1935): fraction in case that the break-up volume fraction vk falls between
classes (Hagesaether et al., 2002b):
V r d32
Reb ¼ : ð11Þ  
vl vk ¼ x j;i v j1 þ 1  x j;i v j ð20Þ

In present study, the Sauter mean diameter is used as the represen- x j;i ¼ 21þi j ; i b j ð21Þ
tative size for the bubble size distribution in the cell. The Sauter mean
bubble diameter (d32) can be defined as the volume-to-surface mean where xj , i is the fraction of bubble distributed among classes. In
bubble diameter (Yianatos et al., 1986): Eqs. (17) and (18) the term ΩC(ϑi, ϑj) is representing the coalescence
rate of bubble i and j, where the product of coalescence could be
N 3
∑i¼1 ni di distributed among the classes as follows (Hagesaether et al., 2002b):
d32 ¼ N 2
ð12Þ
∑i¼1 ni di  
vi þ v j ¼ xi; j vi þ 1  xi; j viþ1 ð22Þ
where ni and di are the number and diameter of each bubble class
respectively. The Sauter mean bubble diameter is commonly used in flo- xi; j ¼ 1  2 ji ; i ≥ j: ð23Þ
tation studies because the surface areas of the bubbles have significant
effect on the rate of flotation (Yianatos et al., 1986). The volume fraction ϕi of bubble class i in each cell relies on the
The population balance equation (PBE) was used to predict the volume fraction of gas ϕi = αi/αg. The total volume αi of bubble class i
number density of different bubble classes. This approach has been can be obtained by using the following equation (Bannari et al., 2008):
successfully used in modelling of slag forming behaviour (Sattar et al.,
α i ¼ ni V i ð24Þ
2013b). Population balance equation was introduced as scalar transport
equation in AVL-Fire. The volume fraction of each bubble classes were where ni and Vi are the total number of bubble class i and the volume of
calculated by using the scalar transport equation. The population single bubble of that class respectively.
balance equation can be written as:
2.1.3. Bubble break-up model
∂ 
α g ρg ϕi þ ∇:α g ρg vg ϕi ¼ ∇:α g ρg Dgi ∇ϕi þ Si ð13Þ The phenomenon of bubble break-up is a result to the collision of
∂t
bubble with turbulent eddy. The energy and size of the eddy is the
where (Si) is the source term of bubble class i due to break-up and main parameters that inducing bubble break-up (Huh et al., 2006).
coalescence which is given by Bannari et al. (2008): Due to collision of bubble with eddies that are smaller or equal to its
size will cause bubble break-up, where either eddies that are larger
  than the bubble or eddies which are much smaller than the bubble do
Si ¼ ρg BBi  DBi þ BC i  DC i ð14Þ not cause the bubble to break. While the very small eddies do not
A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64 57

have energy to break the bubble (Schlichting, 1979), the larger eddies summation of the turbulent collision rate, buoyancy-driven collision
merely carry the bubble. Commonly, there are different mechanisms rate and laminar shear collision rate, however in the present study
of bubble break-up into its lower counterpart namely; turbulent fluctu- only the turbulence and laminar shear collision rate was considered.
ation mechanism, viscous shears stress mechanism, shearing-off mech- The turbulent collision rate can be calculated by Prince and Blanch
anism and interfacial instability mechanism. The binary break-up model (1990) equation as follows:
of bubble based on turbulent fluctuation mechanism which has been
developed by Luo and Svendsen (1996) is adopted in the present π 2 0:5
θTnm ¼ nn nm ðdn þ dm Þ u2tn þ u2tm : ð31Þ
study for its simplicity. In this model, predict break-up rate for bubbles 16
does not require a predefined distribution for daughter bubble size and
thus helps to avoid some of the problems of the original model. At the  
same time this model contains no adjustable parameters and assumes utn and utm are the average turbulent fluctuation velocities of bubble
that parent bubbles break-up into only two daughter bubbles (Sattar class n and m respectively. The laminar shear collision rate can be deter-
et al., 2013a, 2013b, 2014). The rate of bubble break-up ΩB is a function mined as (Prince and Blanch, 1990):
of the probability of bubble break-up PB and the collision frequency ωB.
The following equation was used to calculate the break up rate of bubble 3 U l; max
θLS
nm ¼ 1:7667nn nm ðdn þ dm Þ ð32Þ
class n into m colliding with the eddy size λi (Luo and Svendsen, 1996; Dc
Sattar et al., 2013c):
where Dc is the diameter of the column. The centreline liquid circulation
Ω B ðϑ n ; ϑm Þ ¼ ωB ðdn ; λi Þ  P B ðdn ; λi ; dm Þ: ð25Þ velocity can be written as follows:
 
Here ΩB(ϑn, ϑm) is the rate of break-up of bubble of volume vn into 1  0:75α g α g gD2c
U l; max ¼ : ð33Þ
bubble of volume vm, ωB(dn, λi) is the collision frequency between 1  αg 48υt
bubble classes n with eddy size λi, and PB(dn, λi, dm) is the probability
of breakage. The turbulent collision is written as (Luo and Svendsen, υt is the turbulent kinetic viscosity which can be calculated from the
1996; Sattar et al., 2013c): following equation:

π 2
ωB ðdn ; λi Þ ¼ ðdn þ λi Þ  uλ  nλ  ndn ð26Þ D1:77
4 υt ¼ 0:0536 C
: ð34Þ
ρl

where the average turbulent velocity of eddy uλ¼ β1=2 ðελi Þ1=3, the num-
4
ber density of eddy nλ ¼ 0:822ð1  α d Þ=λi , and the number of bubble Collision efficiency is a function to the contact time between bubbles
in each cell ndn = αn/Vn. The probability of break up can be calculated and the time required for bubbles to coalesce (Prince and Blanch, 1990).
as (Luo and Svendsen, 1996): The coalescence efficiency that was developed by Coulaloglou and
Tavlarides (1977) is used in present study:
P B ðdn ; λi ; dm Þ ¼ expðχ c Þ ð27Þ
P C ðr n ; rm Þ ¼ exp  ðt nm =τnm Þ ð35Þ
where χc is the critical dimensionless energy which can be computed
from the following equation (Luo and Svendsen, 1996; Sattar et al.,
where tnm is the coalescence time and τnm is the contact time. In the
2013c):
present model, the presence of solid particles was not included into
12C f σ the bubble coalescence rate equation.
χc ¼ 5=3 11=3
: ð28Þ
βρc ε2=3 dn ξ
3. Kinetic of flotation process

Here, the increase of coefficient of surface energy Cf = f 2/3 BV +


The efficient capture of hydrophobic particles by air bubbles is the
(1 - fBV)2/3 - 1, the size ratio between eddy and bubble ξ = λ/d, and the
fundamental condition to successful particle separation in a flotation
volume fraction of parent bubble of volume vn that break-up into
cell, which can be divided into three distinct processes: collision, adhe-
bubble of volume vm can be obtained from the following equation:
sion, and detachment, as depicted in Fig. 1 (Tao, 2005). In order to get a
vm reliable model of flotation cell these sub-processes must be identified
f BV ¼ ð29Þ and used to determine the cause and effect relationships between the
vn
system variables (Polat, 2000; Tao, 2005; Koh and Schwarz, 2007,
2008).
2.1.4. Bubble coalescence model There are various kinetic models of mineral flotation available in the
The collision of bubbles in turbulent flows causes bubble coalescence open literature utilising different expressions of the attachment and de-
which occurs due to a variety of mechanisms such as the turbulence of tachment process (Sarhan et al., 2014). Flotation is usually represented
liquid, laminar shear and buoyancy. The collision of two bubbles creates as a first-order rate process with respect to the concentration of parti-
thin liquid film between them. The liquid film reaches to critical thick- cles and bubbles in both stages the attachment and the detachment pro-
ness due to the liquid drains and a film rupture occurs resulting in cess (Koh et al., 2000). The influence of flotation kinetics is modelled
coalescence. The coalescence model of Prince and Blanch (1990) due through applying source terms in the transport equation for concentra-
to turbulent induce collision has been adopted in this study. The coales- tion of particle as follows:
cence rate of bubbles ΩC can be expressed as follows (Prince and Blanch,
1990): ∂
ðα k nk ϕi Þ þ ∇:ðα k nk uk ϕi Þ ¼ Sa þ Sd ð36Þ
∂t
Ω C ¼ θnm  P C ðdn ; dm Þ: ð30Þ
where Sa is the source term of attachment rate, and Sd is the source term
θnm is the collision rate of two bubbles and PC is coalescence efficien- of detachment rate. The rate of removal of solids in a given volume relies
cy. In this model, collision frequency is calculated by determining the on the rates of two simultaneously running processes of attachment and
58 A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64

Fig. 1. Illustration of three sub-processes in flotation system (Tao, 2005).

detachment. For the Lagrangian framework written in terms of time, the where uf is the mean fluctuating fluid velocity. The collision efficiency
kinetic equation can be written as (Koh and Schwarz, 2006): ( Z1) for fine particles and bubbles confined within eddies in the low
turbulent dissipation regions can be calculated as follows:
dN p1
¼ k1 Np1 NbT ð1  βÞ þ k2 NbT β ð37Þ sffiffiffiffiffiffiffiffiffiffiffi 3
dt 8πε dp þ db
Z1 ¼ : ð43Þ
15v f 2
where k1 and k2 are kinetic constants of attachment and detachment
processes, Np1 and NbT are the number concentrations of free particles
(m-3) and the total number of bubbles per unit volume, β is the loading The relative velocity between the particle–bubble aggregate and the
parameter, respectively. The kinetic constants of attachment (k1) and surrounding fluid affects on the detachment frequency (Z2) which has
detachment are computed as (Koh and Schwarz, 2006): been computed as follows (Bloom and Heindel, 2003):

k1 ¼ Z 1 P c P a P s ð38Þ  2=3
Z 2 ¼ ðC 1 Þ0:5 ε1=3 dp þ db : ð44Þ

k2 ¼ Z 2 ð1  P s Þ: ð39Þ
The constant C1 is an empirical constant with value 2 as suggested by
Bloom and Heindel (2003).
Here, Pc, Pa and Ps are the probabilities of particle–bubble collision,
The tendency of particles to follow the fluid streamlines around the
adhesion and stabilisation against external forces, respectively. The
bubble and avoid actual contact is accounted in the collision rate
particle–bubble collision frequency (Z1) depends on several variables
equation by the collision efficiency (Pc), which is more significant for
such as the size of bubbles and particles, hydrodynamics of the
particles that are much smaller than the bubble. In the current study,
collecting zone. Bloom and Heindel (2003) applied the following
the simple model of Yoon and Luttrell (1989), which is valid for inter-
equation to calculate the particle–bubble collision frequency in turbu-
mediate bubble size and particles smaller than 100 μm, was used:
lent flows:
" #
4Re0:72 Dp 2
 2  2 2 0:5 P c ¼ 1:5 þ : ð45Þ
Z1 ¼ 1:25 dp þ db up þ ub ð40Þ 15 Db

where up and ub are the turbulent (rms) fluctuating velocities of parti- Reynolds number can be calculated: Reb = ubdb/vf (Schiller and
cles and bubbles relative to the fluid. In flotation cell, these variables are Naumann, 1935).
functions of the local turbulent dissipation rate (ε) (Liepe and Möckel, The probability of adhesion (Pc) dependent on the particle and bub-
1976): ble sizes, the bubble Reynolds number and the induction time which
can be expressed as follows (Yoon and Luttrell, 1989):
"

#
0:4ε 4=9 dp;b
7=9

ρ  ρ
2=3
p;b f
2 0   13
Up;b ¼ ð41Þ  45 þ 8Re0:72
b ub;rise t ind
1=3
vf ρf Pa ¼ sin 24
2 tan 1
exp@   A5 ð46Þ
15db db =dp þ 1

where ρp , b represent densities of solid particles or gas bubbles, dp , b is


the diameter of solid particles or gas bubbles, and vf, ρf are the kinematic where Ub,rise is the bubble rise velocity, and tind the induction time. The
viscosity and the density of fluid, respectively. The above model is valid above equation has been calculated under assumption that particle–
only when the diameter of the particle or bubble is greater than the bubble collision occurs uniformly over the entire upper half of the
critical diameter (dcr), given by the following equations: bubble surface, which might be not entirely correct (Dai et al., 1999).
The induction time is a function of the solids characteristics such as par-
ticle size and hydrophobicity which can be experimentally obtain. If the
15v f ρ f u2f
2 2
dp;b N dcr ¼ ð42Þ induction time shorter than the sliding time, the particle will have long
ρp;b ε enough contact time to attach to the bubble (Yoon and Luttrell, 1989).
A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64 59

The following expression was used by Koh and Schwarz (2006) to Table 1
calculate the induction time (tind): tind = (75/θ)d0.6 Bubble classes used in the simulation.
p .
Based on work of Schulze (1993); Bloom and Heindel (2003) applied Bubble class 1 2 3 4 5 6 7 8 9 10
the following equation to calculate the probability of stability (Ps): Bubble diameter (×10−3 m) 0.5 0.63 0.8 1 1.26 1.58 2 2.5 3.17 4
  
1
P s ¼ 1  exp 0:5 1   ð47Þ
Bo mesh consisting 160,343 numerical meshes was chosen for further
simulations.
where Bo⁎ is the modified Bond number which can be expressed as In the present study, ten bubble classes were considered where the
(Bloom and Heindel, 2003): volume of the lower bubble class was one half the volume of the
upper bubble class vn + 1 = 2vn. Table 1 presents the diameters of the
  1=3   
2 d 4σ θ bubble classes that were tracked in this simulation. It was assumed
dp Δρp g þ 1:9ρp ε2=3 2p þ d2b þ 1:5dp  db ρ f g sin2 π 
d 2 that the bubble of class 1 (the lowest class) enter the calculation domain
Bo ¼   b
θ θ through the inlet at the bottom of the column. The mathematical model
6σ sin π  sin π þ
2 2 used for both the two (gas and liquid) and three (gas, liquid and solid)
ð48Þ phase systems were in an Eulerian–Eulerian frame of reference. In the
following calculations, solid particles were considered with a size of
where Δρp is the density difference between particle and fluid (Δρp = 63 μm and a density of 1400 kg m− 3. The influence of the gas flow
ρp -ρb), g is the acceleration due to gravity, and σ is the surface tension. rate and solid concentration on the gas holdup and bubble coalescence
rate was studied through measure the effect of changing these variables
4. Model description on Sauter mean diameter.
The following assumptions were adopted for modelling:
In this study, three-dimensional flow of a flotation column similar to
the experimental model of Shukla et al. (2010) has been simulated with • Particles and bubbles have a spherical shape.
AVL-FIRE 2009.2 CFD software. Fig. 2 illustrates the geometrical aspects • Bubble loading parameter is equal to zero for clean bubbles, and equal
of the geometry and generated computational grid of the model. The to one for full loaded bubbles.
flotation column is 1.68 m high and 0.1 m in diameter. Gas bubbles • The particle contact angle θ = 38o.
were injected through a ceramic sparger at the bottom of column. In • There is no liquid inlet in the system.
the current study, two different cell sizes with increasing number of el-
ements (200,127 and 160,343) were constructed to ensure indepen- These assumptions are consistent with previous studies (Koh and
dence of obtained results on mesh density. In order to evaluate the Schwarz, 2003, 2006; Ghaffari and Karimi, 2012).
impact of the grid on the results, simulations were carried out on both
grid using gas flow rate of 1.68, 2.47, 3.29, 4.14 and 5.01 cm s− 1 and
solids concentration 50 kg m−3. It has been verified that the percentage 5. Results and discussion
of change of the solution in both cases is small. Also, it was found that
there were no significant differences in the results. Therefore, the Results of two-phase and three-phase systems were obtained for a
flotation column at different superficial gas velocity values. In the cur-
rent study 10 bubbles classes were considered (see Table 1). Gas was in-
troduced through the inlet at the center of the bottom of the column,
where the superficial gas velocity and bubble diameters are specified.
In the present study, it was assumed that the bubbles enter the calcula-
tion domain with the initial diameter measured at H = 0 plane in the ex-
perimental model (Shukla et al., 2010). In the current study the bubble
break-up and coalescence model corrected by Sattar et al. (2013c) for
the population balance equation was used. The simulations were carried

Fig. 2. Grid generated for the numerical analysis. Fig. 3. Effect of gas flow rate on gas holdup in the absence of frother at Cs = 0 kg m−3.
60 A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64

Fig. 4. Effect of gas flow rate on gas holdup in the absence of frother at Cs = 50 kg m−3. Fig. 6. Effect of superficial gas velocity on axial pressure profile at Cs = 50 kg m−3.

out for superficial gas velocity of Ug = 0.94, 1.30, 1.68, 2.07, 2.47, 2.88, heterogeneous flow regime which can be attributed to coalescence of
3.29, 3.71, 4.14, 4.57 and 5.01 cm s−1. the small bubbles leading to relatively lower gas holdup (Krishna
In order to verify the reliability of the CFD model and accuracy of the et al., 1997). This explains the kink that can be observed it in the gas
numerical method, results obtained with the CFD model are compared holdup curve at 0.5 m/s. Generally, bubble size decreases with the
with the experimental data that are available in the literature for gas– increase of gas flow rate due to promoting bubble breakup due to
liquid and gas–liquid–solid system (Shahbazi et al., 2010). increasing bubble-eddies collision rate (Banisi, 1995a, 1995b).
There are several process variables that affect the gas holdup in flo- Although, there were several studies that have been conducted to in-
tation column such as gas flow rate, feed flow rate, solid concentration, vestigate the effect of solid concentration on the gas holdup, the results
particle size, frother and collector concentration, bubble size and slurry are often contradictory where both increasing and decreasing effects are
viscosity. The effect of gas flow rate on gas holdup for different solid reported (Banisi, 1995a). However, in the current study it was clear that
concentration is presented in Figs. 3 and 4. The gas holdup is defined the presence of solid particles decreased the gas holdup in flotation cell.
as the volumetric fraction of gas in the considered zone which can be To study the axial pressure characteristics inside the flotation cell,
calculated as follows (Schwarz and Alexander, 2006): the simulation model was used to predict the pressure differences at dif-
ferent locations of the cell. The effect of solids concentration on
Hg  H column's axial pressure profile is plotted in Fig. 5. The points in Fig. 5
εg ¼ ð49Þ
Hg represent data obtained from present simulation. The lines are obtained
from curve fitting through the data points. The plot show that the
where Hg is the height of the liquid after aeration. increase in solid concentration leads to a slight increase in pressure dif-
It can be seen that the gas hold-up increases linearly with increasing ference per unit length inside the column. The effect of solids concentra-
gas superficial velocity. This is in line with the other relevant studies tion on axial pressure profile has been attributed to the increase of solid
(Frijlink et al., 1990; Zhou et al., 1993; Banisi et al., 1995a, 1995b; loading in the system. Figs. 6–9 demonstrate the effect of gas flow rate
Grevskott et al., 1996; Zhang and Finch, 2014). Gas holdup increases on the axial pressure profile. It was observed in both gas–liquid
with gas flow rate because there is more gas in the system. It also can (not shown here due to brevity) and gas–liquid–solid systems that
be observed that at higher gas velocities the column moved towards there is a slight decrease in the pressure difference in each zone as the

Fig. 5. Axial pressure profile with different solid concentrations. Fig. 7. Effect of superficial gas velocity on axial pressure profile at Cs = 100 kg m−3.
A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64 61

Fig. 8. Effect of superficial gas velocity on axial pressure profile at Cs = 150 kg m−3. Fig. 10. Sauter mean diameter with solid concentration at H = 400 mm predicted by the
CFD Model.

gas flow rate is increased. The increase in gas flow rate leads to an
increase in gas holdup causing this pressure reduction.
The effect of solid concentration on the Sauter mean diameter d32 in-
side flotation cell for different superficial gas velocity is shown in
Figs. 10–12. The Sauter mean diameter (d32) of the bubbles was investi-
gated at different locations of the flotation cell (H = 400, 1000 and
1400 mm). It is evident from Figs. 10, 11 and 12 that the Sauter mean
bubble diameter d32 decreases with increasing solid concentration
for all gas flow rates. However, the major decrease in Sauter mean
diameter d32 occurs at heights far away from the pulp-froth interface re-
gion. The bubbles close to the bottom are subjected to higher turbulence
from the inlet region. Whereas, the bubbles close to froth zone
experience higher rate of coalescence leading to bigger Sauter mean
diameter d32. The Sauter mean diameter decreased with the increase
of gas velocity in almost all the locations and solid particle concentra-
tion. This is because the bubble breakup increases with the increase of
gas velocity due to increased bubble-eddies collision rate (Bukur et al.,
1990; Saxena et al., 1992; Li and Prakash, 1997; Kumar et al., 2012).
Figs. 13 to 17 present the impact of gas flow rate on the Sauter mean
diameter at different heights of flotation cell for different solid concen-
tration. The predicted Sauter mean diameter increased with height ap- Fig. 11. Sauter mean diameter with solid concentration at H = 1000 mm predicted by the
proaching a constant value half way towards the top of the pulp zone. CFD Model.
This is because bubbles experience high breakage rate at the bottom
of the flotation cell compared to the top. The increase of superficial

Fig. 12. Sauter mean diameter with solid concentration at H = 1400 mm predicted by the
Fig. 9. Effect of superficial gas velocity on axial pressure profile at Cs = 200 kg m−3. CFD Model.
62 A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64

Fig. 13. Sauter mean diameter with Cs =0 kg m−3 predicted by the CFD Model. Fig. 16. Sauter mean diameter with Cs = 150 kg m−3 predicted by the CFD Model.

0 kg m− 3 and Cs = 200 kg m−3 are shown in Figs. 18 and 19. The


model that is used in the current study to calculate bubble break-up
and bubble coalescence was developed by Prince and Blanch (1990)
only for gas–liquid systems, however it has successfully employed by
Sawyerr et al. (1998) for gas–liquid–solid systems as well. The results
presented in Figs. 18 and 19 show that the bubbles of different classes
changed along the flotation column due to bubble coalescence and
breakup. The figures depict that the number densities of higher bubble
classes are less than its lower counterparts. This is because in the current
study the bubble class 1 (the lowest class considered in this study) was
introduced into the cell through the inlet at the center of the bottom,
hence its number density is more at the inlet. Furthermore, it can be
seen from the figures that the number density of higher bubble class
increases as the bubbles move upward and the bubble size grows due
to coalescence of lower bubble classes.

6. Conclusions

Fig. 14. Sauter mean diameter with Cs =50 kg m−3 predicted by the CFD Model. CFD simulations were conducted for both gas–liquid column and
gas–liquid–solid flotation columns to provide useful insights into the
gas velocity leads to produce lower bubble sizes because of the bubble effect of presence of solid particles on coalescence and break-up of bub-
break-up is more likely to happen than bubble coalescence. bles. User subroutine was written using FORTRAN to incorporate the
The volume fraction of different bubble classes inside flotation cells scalar transport equation source term for bubble break up and coales-
for superficial gas velocity of 5.01 cm s−1, solid concentration Cs = cence. The liquid phase was treated as a continuum and the gas phase

Fig. 15. Sauter mean diameter with Cs = 100 kg m−3 predicted by the CFD Model. Fig. 17. Sauter mean diameter with Cs = 200 kg m−3 predicted by the CFD Model.
A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64 63

Fig. 18. Model predictions for distribution of different bubble classes with solid concentration Cs =0 kg m−3.

(bubbles) and solid particles were considered as a dispersed phase. The solid particles within flotation cell have a great influence on bubble
population balance equation for bubble break-up and bubble coales- coalescence, where the hydrophobic particles act like a physical bar-
cence rate and the interfacial exchange of mass and momentum as rier between the bubbles. The study showed that the increase of solid
well as bubble–particle attachment and detachment were included in concentration leads to decrease the Sauter mean bubble size. How-
the CFD code. The predictions of the model are in reasonably good ever, the major decrease in bubble sizes occurs at heights far away
agreement with the experimental data of Shukla et al. (2010). The pulp-froth from interface region. The bubbles close to the bottom
main findings of the present study are: the presence of solid particles are subjected to higher turbulence from the inlet region. Whereas,
reduced the Sauter mean diameter of the bubbles in the flotation cell. the bubbles close to froth zone experience higher rate of coalescence
This is because the bubble breakup increases with the increase of leading to bigger Sauter mean diameter d32. The CFD approach used
bubble-eddy collision rate at higher gas velocity. With the increase here shows reasonably good agreement between simulation and exper-
of gas flow rate, the bubble size generated decreased leading to in- imental results for the effect of solid concentration on gas hold-up and
creased gas holdup. It also can be concluded that the presence of axial pressure profile.

Fig. 19. Model predictions for distribution of different bubble classes with solid concentration Cs = 200 kg m−3.
64 A.R. Sarhan et al. / International Journal of Mineral Processing 146 (2016) 54–64

Acknowledgements Krishna, R., deSwart, J.W.A., Ellenberger, J., Martina, G.B., Maretto, C., 1997. Gas holdup in
slurry bubble columns: effect of column diameter and slurry concentrations. AICHE J.
43, 311–316.
We gratefully acknowledge the Higher Committee for Education Kumar, S., Kumar, R., Munshi, P., Khanna, A., 2012. Gas hold-up in three phase co-current
Development in Iraq (HCED) for their financial support. bubble columns. Procedia Eng. 42, 782–794.
Launder, B.E., Spalding, D.B., 1974. The numerical computation of turbulent flows.
Comput. Methods Appl. Mech. Eng. 3, 269–289.
References Li, H., Prakash, A., 1997. Heat transfer and hydrodynamics in a three-phase slurry bubble
column. Ind. Eng. Chem. Res. 36, 4688–4694.
Ata, S., 2008. Coalescence of bubbles covered by particles. Langmuir 24, 6085–6091. Liepe, F., Möckel, H.-O., 1976. Untersuchungen zum Stoffvereinigen in flüssiger phase.
Ata, S., 2009. The detachment of particles from coalescing bubble pairs. J. Colloid Interface Chem. Technol. 30, 205–209.
Sci. 338, 558–565. Luo, H., Svendsen, H.F., 1996. Theoretical model for drop and bubble breakup in turbulent
Ata, S., Ahmed, N., Jameson, G.J., 2003. A study of bubble coalescence in flotation froths. dispersions. AICHE J. 42, 1225–1233.
Int. J. Miner. Process. 72, 255–266. Manual, A.F., 2009. AVL List GmbH. Graz, Austria.
Banisi, S., 1995a. Effect of solid particles on gas holdup in flotation columns–i. Measure- Polat, H.C.S., 2000. First-order flotation kinetics models and methods for estimation
ment. Chem. Eng. Sci. 50, 2329–2334. of the true distribution of flotation rate constants. Int. J. Miner. Process. 58,
Banisi, S., 1995b. Effect of solid particles on gas holdup in flotation columns—II. Investiga- 145–166.
tion of mechanisms of gas holdup reduction in esence of solids. Chem. Eng. Sci. 50, Prince, M.J., Blanch, H.W., 1990. Bubble coalescence and break-up in air-sparged bubble-
2335–2342. columns. AICHE J. 36, 1485–1499.
Banisi, S., Finch, J.A., Laplante, A.R., Weber, M.E., 1995a. Effect of solid particles on gas Sarhan, A.R., Naser, J., Brooks, G., 2014. A Review of CFD Modelling of Flotation Cells. 19th
holdup in flotation columns—I. Measurement. Chem. Eng. Sci. 50, 2329–2334. Australasian Fluid Mechanics Conference, Melbourne, Australia.
Banisi, S., Finch, J.A., Laplante, A.R., Weber, M.E., 1995b. Effect of solid particles on gas Sattar, M.A., Naser, J., Brooks, G., 2013a. Numerical simulation of creaming and foam for-
holdup in flotation columns—II. Investigation of mechanisms of gas holdup reduction mation in aerated liquid with population balance modeling. Chem. Eng. Sci. 94,
in presence of solids. Chem. Eng. Sci. 50, 2335–2342. 69–78.
Bannari, R., Kerdouss, F., Selma, B., Bannari, A., Proulx, P., 2008. Three-dimensional math- Sattar, M.A., Naser, J., Brooks, G., 2013b. Numerical simulation of two-phase flow with
ematical modeling of dispersed two-phase flow using class method of population bubble break-up and coalescence coupled with population balance modeling.
balance in bubble columns. Comput. Chem. Eng. 32, 3224–3237. Chem. Eng. Process. Process Intensif. 70, 66–76.
Bao, Y., Zhang, X., Gao, Z., Chen, L., Chen, J., Smith, J.M.A., Kirkby, N.F., 2008. Gas dispersion Sattar, M.A., Naser, J., Brooks, G., 2013c. Numerical simulation of two-phase flow with
and solid suspension in a hot sparged multi-impeller stirred tank. Ind. Eng. Chem. bubble break-up and coalescence coupled with population balance modeling.
Res. 47, 2049–2055. Chem. Eng. Process. 70, 66–76.
Bloom, F., Heindel, T.J., 2003. Modeling flotation separation in a semi-batch process. Sattar, M.A., Naser, J., Brooks, G., 2014. Numerical simulation of slag foaming on bath
Chem. Eng. Sci. 58, 353–365. smelting slag (CaO–SiO2–Al2O3–FeO) with population balance modeling. Chem.
Bukur, D.B., Patel, S.A., Daly, J.G., 1990. Gas holdup and solids dispersion in a three-phase Eng. Sci. 107, 165–180.
slurry bubble column. AICHE J. 36, 1731–1735. Sawyerr, F., Deglon, D.A., O'Connor, C.T., 1998. Prediction of bubble size distribution in
Coulaloglou, C.A., Tavlarides, L.L., 1977. Description of interaction processes in agitated mechanical flotation cells. J. South. Afr. Inst. Min. Metall. 98, 179–185.
liquid–liquid dispersions. Chem. Eng. Sci. 32, 1289–1297. Saxena, S.C., Rao, N.S., Thimmapuram, P.R., 1992. Gas phase holdup in slurry bubble
Dai, Z., Fornasiero, D., Ralston, J., 1999. Particle–bubble attachment in mineral flotation. columns for two- and three-phase systems. Chem. Eng. J. 49, 151–159.
J. Colloid Interface Sci. 217, 70–76. Schiller, L., Naumann, A., 1935. A drag coefficient correlation. Vdi Ztg. 77, 51.
Eggers, J., Lister, J.R., Stone, H.A., 1999. Coalescence of liquid drops. J. Fluid Mech. 401, Schlichting, H., 1979. Boundary-Layer Theory, New York.
293–310. Schulze, H.J., 1993. Flotation as a heterocoagulation process: possibilities of calculating
Frijlink, J.J., Bakker, A., Smith, J.M., 1990. Suspension of solid particles with gassed impel- the probability of flotation. Coagulation and Flocculation: Theory and Applications
lers. Chem. Eng. Sci. 45, 1703–1718. 126, p. 321.
Gallegos-Acevedo, P.M., Espinoza-Cuadra, J., Pérez-Garibay, R., Pecina-Treviño, E.T., 2010. Schwarz, S., Alexander, D., 2006. Gas dispersion measurements in industrial flotation
Bubbles coalescence: hydrofobic particles effect. J. Min. Sci. 46. cells. Miner. Eng. 19, 554–560.
Ghaffari, A., Karimi, M., 2012. Numerical investigation on multiphase flow simulation in a Shahbazi, B., Rezai, B., Javad Koleini, S.M., 2010. Bubble–particle collision and attachment
centrifugal flotation cell. Int. J. Coal Prep. Util. 32, 120–129. probability on fine particles flotation. Chem. Eng. Process. Process Intensif. 49,
Grevskott, S., Sannaes, B.H., Dudukovic, M.P., Hjarbo, K.W., Svendsen, H.F., 1996. Liquid 622–627.
circulation, bubble size distributions, and solids movement in two- and three-phase Shukla, S.C., Kundu, G., Mukherjee, D., 2010. Study of gas holdup and pressure character-
bubble columns. Chem. Eng. Sci. 51, 1703–1713. istics in a column flotation cell using coal. Miner. Eng. 23, 636–642.
Hagesaether, L., Jakobsen, H.A., Svendsen, H.F., 2002a. A model for turbulent binary break- Sobhy, A., Tao, D., 2013. Nanobubble column flotation of fine coal particles and associated
up of dispersed fluid particles. Chem. Eng. Sci. 57, 3251–3267. fundamentals. Int. J. Miner. Process. 124, 109–116.
Hagesaether, L., Jakobsen, H.A., Svendsen, H.F., 2002b. Modeling of the dispersed-phase Tao, D., 2005. Role of bubble size in flotation of coarse and fine particles—a review. Sep.
size distribution in bubble columns. Ind. Eng. Chem. Res. 41, 2560–2570. Sci. Technol. 39, 741–760.
Huh, B.G., Euh, D.J., Yoon, H.Y., Yun, B.J., Song, C.H., Chung, C.H., 2006. Mechanistic study Tavera, F.J., Escudero, R., Finch, J.A., 2001. Gas holdup in flotation columns: laboratory
for the interfacial area transport phenomena in an air/water flow condition by measurements. Int. J. Miner. Process. 61, 23–40.
using fine-size bubble group model. Int. J. Heat Mass Transf. 49, 4033–4042. Verrelli, D.I., Bruckard, W.J., Koh, P.T.L., Schwarz, M.P., Follink, B., 2014. Particle shape ef-
Koh, P.T.L., Schwarz, M.P., 2003. CFD modelling of bubble–particle collision rates and fects in flotation. Part 1: microscale experimental observations. Miner. Eng. 58,
efficiencies in a flotation cell. Miner. Eng. 16, 1055–1059. 80–89.
Koh, P.T.L., Schwarz, M.P., 2006. CFD modelling of bubble–particle attachments in Yianatos, J.B., Finch, J.A., Laplante, A.R., 1986. Holdup profile and bubble-size distribution
flotation cells. Miner. Eng. 19, 619–626. of flotation column froths. Can. Metall. Q. 25, 23–29.
Koh, P.T.L., Schwarz, M.P., 2007. CFD model of a self-aerating flotation cell. Int. J. Miner. Yoon, R.-H., Luttrell, G.H., 1989. The effect of bubble size on fine particle flotation. Miner.
Process. 85, 16–24. Process. Extr. Metall. Rev. 5, 101–122.
Koh, P.T.L., Schwarz, M.P., 2008. Modelling attachment rates of multi-sized bubbles with Zhang, W., Finch, J.A., 2014. Effect of solids on pulp and froth properties in flotation.
particles in a flotation cell. Miner. Eng. 21, 989–993. J. Cent. South Univ. 21, 1461–1469.
Koh, P.T.L., Manickam, M., Schwarz, M.P., 2000. Cfd simulation of bubble–particle Zhou, Z.A., Plitt, L.R., Egiebor, N.O., 1993. The effects of solids and reagents on the
collisions in mineral flotation cells. Miner. Eng. 13, 1455–1463. characteristics of coal flotation in columns. Miner. Eng. 6, 291–306.

You might also like