Advanced Powder Technology 23 (2012) 445–453

Contents lists available at SciVerse ScienceDirect

Advanced Powder Technology

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

Original Research Paper

CFD simulation of solid–liquid stirred tanks

Divyamaan Wadnerkar, Ranjeet P. Utikar ⇑,1, Moses O. Tade, Vishnu K. Pareek
Department of Chemical Engineering, Curtin University, Perth 6102, Australia

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

Article history: Solid liquid stirred tanks are commonly used in the minerals industry for operations like concentration,
Received 31 October 2011 leaching, adsorption, effluent treatment, etc. Computational Fluid Dynamics (CFD) is increasingly being
Received in revised form 6 March 2012 used to predict the hydrodynamics and performance of these systems. Accounting for the solid–liquid
Accepted 17 March 2012
interaction is critical for accurate predictions of these systems. Therefore, a careful selection of models
Available online 25 April 2012
for turbulence and drag is required. In this study, the effect of drag model was studied. The Eulerian–Eule-
rian multiphase model is used to simulate the solid suspension in stirred tanks. Multiple reference frame
Keywords:
(MRF) approach is used to simulate the impeller rotation in a fully baffled tank. Simulations are con-
Solid–liquid suspension
Stirred tanks
ducted using commercial CFD solver ANSYS Fluent 12.1. The CFD simulations are conducted for concen-
Hydrodynamic study tration 1% and 7% v/v and the impeller speeds above the ‘‘just suspension speed’’. It is observed that high
Drag models turbulence can increase the drag coefficient as high as forty times when compared with a still fluid. The
Homogeneity drag force was modified to account for the increase in drag at high turbulent intensities. The modified
Cloud height drag is a function of particle diameter to Kolmogorov length scale ratio, which, on a volume averaged
CFD basis, was found to be around 13 in the cases simulated. The modified drag law was found to be useful
to simulate the low solids holdup in stirred tanks. The predictions in terms of velocity profiles and the
solids distribution are found to be in reasonable agreement with the literature experimental data. Turbu-
lent kinetic energy, homogeneity and cloud height in the stirred tanks are studied and discussed in the
paper. The presence of solids resulted in dampening of turbulence and the maximum deviation was
observed in the impeller plane. The cloud height and homogeneity were found to increase with an
increase in impeller speed. The work provides an insight into the solid liquid flow in stirred tanks.
Ó 2012 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder
Technology Japan. All rights reserved.

1. Introduction behaviour of turbulence kinetic energy, suspension quality and

cloud height are extensively discussed.
Solid–liquid mixing systems are amongst the common opera-
tions used in the field of chemical and mineral industry. The main 2. Literature review
purpose of mixing is the contact between the solid and liquid
phase for facilitating mass transfer. In industrial processes effective Micale et al. [5] used Settling Velocity Model (SVM) and Multi
mixing is necessary at both micro and macro level for adequate fluid Model (MFM) approaches to analyse the particle distribution
performance. At the micro level, micromixing governs the chemical in stirred tanks. In SVM, it is assumed that the particles are trans-
and mass transfer reactions. Micromixing is facilitated by mixing at ported as a passive scalar or molecular species but with a superim-
macro level. Numerous factors such as the just suspension speed, posed sedimentation flow, whereas in MFM, momentum balances
critical suspension speed, solids distribution, etc. dictate the mix- are solved for both phases. Computationally intensive MFM was
ing performance. CFD has proved to be a useful tool in analyzing found to be better than SVM, but for both the models it was neces-
the impact of these factors on the flow characteristics of such sys- sary to take into account the increase in drag with the increasing
tems [1–7]. Proper evaluation of interphase drag is essential for turbulence. Micale et al. [4] simulated the solids suspension of
accurate predictions using the CFD model. In this study four differ- 9.6% and 20% volume fractions using the MFM approach and slid-
ent drag models are analysed and their validity is checked by com- ing grid (SG) approach using the Schillar Nauman drag model.
paring the results of CFD simulations at low concentrations of solid Schillar Nauman is applicable on spherical particles in an infinite
with the experimental data available in the literature [8]. The stagnant fluid and accounts for the inertial effect on the drag force
acting on it. It provided satisfactory results at low impeller speed.
⇑ Corresponding author. Tel.: +61 892669837; fax: +61 892662681. Derksen [9] conducted Eulerian–Lagrangian simulations to
E-mail address: r.utikar@curtin.edu.au (R.P. Utikar). study the velocity field, turbulence, solid distribution and
1
Present address. particle-impeller and particle–particle collision and frequencies

0921-8831/$ - see front matter Ó 2012 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
http://dx.doi.org/10.1016/j.apt.2012.03.007
446 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453

in a stirred tank system. Inertial, gravitational and drag forces were It is quite clear from the review that solids suspension and dis-
included, but Saffman, Magnus and stress forces were kept condi- tribution is highly dependent on the turbulence and interphase
tional and their effect were studied. The effect of these forces was drag in the tank. At low impeller speeds, turbulent fluctuations
found to be negligible due to the high magnitude of drag, inertial are less and hence do not affect the predictions much. However,
and gravitational force. at higher impeller speeds, the drag and turbulence become increas-
Ochieng and Lewis [7] simulated nickel solids loading of ingly important. Moreover, there is no consensus on the appropri-
1–20% w/w with impeller speeds between 200 and 700 rpm using ate drag for liquid–solid stirred tanks. Therefore, in this study, the
both steady and transient simulations and found out that transient impact of drag model on the flow distribution and the velocity
simulations, although time consuming, are better for stirred tank fields is investigated. Different drag models are assessed to provide
simulations. The initial flow field was generated using the multiple a clear understanding of the selection criterion of drag in a partic-
reference frame (MRF) approach and then the simulations were ular case.
carried out using SG. The Gidaspow model was used for the drag
factor, which is a combination of the Wen and Yu model and the
3. CFD model
Ergun equation [10]. Wen and Yu drag is appropriate for dilute
systems and Ergun is used for dense packing. For the study of just
3.1. Model equations
suspended of solids using solids at the bottom of the tank as an
initial condition, it provided satisfactory results.
The hydrodynamic study is simulated using Eulerian–Eulerian
The suspension can also be modelled as a continuous phase
multiphase model. The phases, in this model, are treated as inter-
using a viscosity law and the shear induced migration phenomenon
penetrating continua represented by a volume fraction at each
generated by gradients in shear rates or concentration gradients
point of the system. The Reynolds averaged mass and momentum
can be captured at a macroscopic scale. For the prediction of
balance equations are solved for each of the phases. The governing
shear-induced particle migration, the Shear Induced Migration
equations are given below:
Model (SIMM) was used, which states that, in a viscous concen-
Continuity equation:
trated suspension, small but non-Brownian particles migrate from
@  
regions of high shear rate to regions of low shear rate, and from re- ðaq qq Þ þ r: aq qq~
uq ¼ 0
gions of high concentrations to regions of low concentrations in @t
addition to which settling by gravity is added. In the case of a mix- Momentum equation:
ing process, owing to the action of shear and inertia, the particles
may segregate and demix, thereby generating concentration gradi- @    
aq qq~
uq þ r: aq qq~ uq ¼ aq rP þ r:sq þ aq qq~
uq~ g
ents in the vessel. This shear-induced migration phenomenon can @t
 
be simulated at the macroscopic scale, where the suspension is þ ~F td þ ~
Fq þ ~
F lift;q þ ~
F v m;q þ ~
F 12
modelled as one continuous phase through a viscosity law [1].
However, this model shows potentially erratic behaviour in where q is 1 or 2 for primary or secondary phase, respectively, a is
close-to-zero shear rate and high concentration zones. volume fraction, q is density, ~ u is the velocity vector, P is pressure
The dependency of the drag on the turbulence was numerically and is shared by both the phases, s is the stress tensor because of
investigated by Khopkar et al. [3] by conducting experiments using viscosity and velocity fluctuations, g is gravity, ~ F td is force due to
single phase flow through regularly arranged cylindrical objects. A turbulent dissipation, ~
F q is external force, ~
F lift;q is lift force, ~
F v m;q is
relationship between the drag, particle diameter and Kolmogorov virtual mass force and ~F 12 is interphase interaction force.
length scale was fit into the expression given by Brucato et al. The stress–strain tensor is due to viscosity and Reynolds stres-
[11]. They found that the drag predicted by the original Brucato ses that include the effect of turbulent fluctuations. Using the
drag model needs to be reduced by a factor of 10. This modified Boussinesq’s eddy viscosity hypothesis the closure can be given
Brucato model was then used for the simulation of liquid flow field to the above momentum transfer equation. The equation can be gi-
in stirred tanks [2]. It was able to capture the key features of liquid ven as:
phase mixing process.    
2
Panneerselvam et al. [12] used the Brucato drag law to simulate sq ¼ aq lq r~ uTq þ aq kq  lq r:~
uq þ r~ uq I
3
7% v/v solids in liquid. MRF approach was used with Eulerian–Eule-
rian model. There was mismatch in the radial and tangential com- where l is the shear viscosity, k is bulk viscosity and I is the unit
ponents of velocity at impeller plane. This discrepancy was stress tensor.
attributed to the turbulent fluctuations that dominate the impeller
region, which the model was not able to capture successfully. 3.2. Equations for turbulence
Guha [13] conducted numerical simulations and assessed dif-
ferent approaches viz. LES and Eulerian–Eulerian (using Schiller– k-e mixture turbulence and k-e dispersed turbulence models are
Nauman drag model) to simulate turbulent solid–liquid flow in a used in the present study. The mixture turbulence model assumes
low solid loading (1% by volume) stirred tank by comparing with the domain as a mixture and solves for k and e values which are
results from the CARPT experiment. Either of the simulation ap- common for both the phases. In the dispersed turbulence model,
proach was not able to predict a stronger lower circulation loop the modified k-e equations are solved for the continuous phase
as is observed in the experiments. The stronger loop is because and the turbulence quantities of dispersed phase are calculated
of the high magnitude axial velocities striking the walls, which is using Tchen-theory correlations. It also takes the fluctuations due
not the case in the region above impeller. Comparisons were done to turbulence by solving for the interphase turbulent momentum
with respect to the solids velocities, turbulent kinetic energy, and transfer. For the sake of brevity, only the equations of mixture
solids sojourn times at various parts of the tank. LES predicted model for turbulence are given below. Other equations can be
radial velocities better than the Euler–Euler in the impeller plane. found in the Fluent user guide [14].
Elsewhere, both the predictions were comparable. It was because 
of the high turbulent fluctuations in this region that was not taken @ lt;m 
ðq kÞ þ r:ðqm~
um kÞ ¼ r: rk þ Gk;m  qm e
into account by the drag formulation used. @t m rk
 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453 447

 
@ lt;m e  are available in the literature [17,18]. For the conditions studied
ðq eÞ þ r:ðqm~
um eÞ ¼ r: re þ C 1e Gk;m  C 2e qm e
@t m re k in this paper, the drag force calculated using these models was
similar to the modified Brucato drag model. Therefore, Brucato
C1e and C2e are constants, rk and re are turbulent Prandtl numbers.
and modified Brucato models were used for further study.
The mixture density, qm and velocity, ~um are computed from the
In this paper, the Gidaspow, Wen and Yu, Brucato and modified
equations below:
Brucato Drag models are assessed. The implementation of Brucato
X
N drag model uses Kolmogorov length scale calculated for each cell,
qm ¼ ai qi rather than using a value from the mean power dissipated in the
i¼1
system and applying it all over the domain as is commonly
practiced.
X
N X
N
~
um ¼ ai qi~
ui = ai qi
i¼1 i¼1
4. Methodology and boundary conditions
Turbulent viscosity, lt;m and turbulence kinetic energy, Gk;m are
computed from equations below: 4.1. Vessel geometry
2
k In the current study, a flat bottomed cylindrical tank was simu-
lt;m ¼ qm C l
e lated. The dimensions used are tank diameter, T = 0.2 m and tank
height, H = T. The tank has four baffles mounted on the wall of
Gk;m ¼ lt;m ðr~ uTm Þ : r~
um þ r~ um width T/10. The shaft of the impeller (of diameter = 0.01 m) was
concentric with the axis of the tank. A six-bladed Rushton turbine
was used as an impeller. The Rushton turbine has a diameter,
3.3. Turbulent dispersion force
D = T/3. For each blade, the length = T/12 and the height = T/15.
The impeller off-bottom clearance was (C = T/3) measured from
In the simulation of solid suspension in stirred tanks, the turbu-
the level of the impeller disc. The fluid for the system was water
lent dispersion force is significant when the size of turbulence ed-
(q = 1000 kg/m3, l = 0.001 Pa.s) and the solids were small glass
dies is larger than the particle size [2]. Its significance is also
particles of density 2550 kg/m3 and diameter of 0.3 mm.
highlighted in some previous studies [15]. The role of this force
is also analysed in this study. It is incorporated along with the
momentum equation and is given as follows: 4.2. Numerical simulations
~
F t;d ¼ K pq~
udr Fig. 1 shows half of the computational domain with baffles and
~ is given by,
where drift velocity, dr stirrer. Owing to the rotationally periodic nature, half of the tank
  was simulated. Multiple reference frame (MRF) approach was
Dp D used. A reference moving zone with dimensions r = 0.06 m and
~
udr ¼  rap  q raq
rpq ap rpq aq 0.03995 < z < 0.09325 was created (where z is the axial distance
from the bottom). The impeller rod outside this zone was consid-
Dp and Dq are diffusivities and rpq is dispersion Prandtl number.
ered as a moving wall. The top of the tank was open, so it was de-
fined as a wall of zero shear. The specularity coefficient is 0 for
3.4. Interphase drag force
smooth walls and is 1 for rough walls. The walls of stirred tank
were assumed to be smooth and a very small specularity coeffi-
The drag force represents interphase momentum transfer due
cient of 0.008 was given to all other walls. In the initial condition
to the disturbance created by each phase. For dilute systems and
of the simulation, a uniform average concentration (0.01 v/v or
low Reynolds number, particle drag is given by Stokes law and
0.07 v/v) glass particles was taken in the tank. The rotation speed
for high Reynolds number, the Schillar Nauman drag model can
of the impeller was 1000 rpm that was above the speed of just-sus-
be used. In the literature review other drag models such as Gidas-
pension of glass particles in the liquid. For modelling the turbu-
pow model [10] and Wen and Yu model [16] have also been dis-
lence, a standard k-e mixture model was used. The model
cussed. But for stirred tank systems, there should be a model
parameters were Cl: 0.09, C1: 1.44, C2: 1.92, rk: 1.0 and re = 1.3.
that takes turbulence into account as with increasing Reynolds
In few cases the standard k-e dispersion model was also used with
number and with the increase in the eddy sizes, the impact of tur-
the turbulence Schmidt number, r, taken as equal to 0.8. The stea-
bulence on the drag increases. Considering this Brucato et al. [11]
dy state numerical solution of the system was obtained by using
proposed a new drag model making drag coefficient as a function
the commercial CFD solver ANSYS 12.1 FLUENT. In the present
of ratio of particle diameter and Kolmogorov length scales. So, with
work, simple pressure–velocity coupling scheme was used along
the change in the turbulence at some local point in the system, the
with the standard pressure discretization scheme. The grid inde-
drag will also change. The drag coefficient proposed by Brucato
pendency of the geometry was checked by conducting single phase
et al. is given below:
flow simulations on total number computational grid of 175460,
   3
C D  C Do dp 224280 and 428760 cells. The grid of 224280 predicted a correct
¼K
C Do k
where, K is constant with value of 8.76  104, dP is particle diam- Table 1
Details of cases simulated.
eter and k is Kolmogorov length scale.
Khopkar et al. [3] performed DNS simulations for conditions Case Impeller speed Reynolds Power
closer to those in stirred tanks. Based on these simulations they ob- (rpm) number number

tained a modified version of Brucato drag that is more appropriate Single phase 1000 73926 4.972403
for stirred tanks. This modified drag has a constant value of flow
1% Solids 1000 75071.85 4.950141
8.76  105. Few more drag correlations that also take the depen-
7% Solids 1000 81946.97 4.63385
dency of drag on volume fraction and density into consideration
448 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453

value of power number that did not change with further refine-
ment of the grid. The details of cases simulated and discussed in
the paper are given in Table 1.

5. Results and discussion

5.1. Preliminary numerical simulations

Initial simulations were conducted to assess the effect of turbu-

lence dispersion force. The flow field was analysed and it was
found that there was negligible effect of this force in this particular
case of 0.01 volume fraction. The turbulence dispersion force has
higher influence at higher concentration of solids where its magni-
tude will be high enough to be comparable with the other forces
being exerted on the secondary phase [15].
In order to verify that the simulations have converged, the
residuals as well as additional parameters namely turbulence dis-
sipation over the volume and torque on the shaft were monitored.
Once the residuals and additional parameters were constant, a Fig. 2. Solid velocity vectors on plane between the baffles for solid volume fraction
simulation was deemed to be converged. of 0.01 and 1000 rpm.

5.2. Flow field exerted on the flow due to the presence of bottom wall and ab-
sence of top wall. In the simulations, different boundary conditions
Fig. 2 shows the velocity vectors on a centre plane. For the imposed on the vessel on the top wall (free slip) and bottom wall
Rushton turbine, an outward jet stream is formed due to the out- (no slip) result in this angular inclination. This effect was also
ward thrust of the impeller. This high velocity jet approaches to- shown in the simulation of Sbrizzai et al. [19]. The velocity vectors
wards the wall of the stirred tank and strikes it. The jet splits near the top surface of the stirred tank show a very weak flow field
into two streams. One stream moves in axially upward and another in this region.
in axially downward direction. It creates an anticlockwise velocity
field in the region above the impeller and a clockwise velocity field
in the region below the impeller. The velocity near walls for the re- 5.3. Analysis of drag models
gion above impeller is upwards and below the impeller is down-
wards. It is opposite when the velocity field is observed near the 5.3.1. Slip velocity
centre. The intensity of the recirculation in the region below the The slip velocity in the stirred tanks was analysed using Gidas-
impeller is stronger than that above the impeller. Converged solu- pow, Brucato and modified Brucato drag models. A significant dif-
tion showed similar flow field (velocity field vectors) as compared ference in the magnitude of slip velocity was observed between
to that available in the literature [8]. All the flow characteristics Gidaspow and Brucato drag model in the impeller zone (Fig. 3).
discussed above were captured by the CFD simulation and of the The reason behind the disparity in the predictions of drag models
flow are clearly visible in Fig. 2. The simulations were also able is the basis. Brucato drag correlation is entirely dependent on the
to capture the upward inclination of the jet and its asymmetry turbulence and is calculated from the ratio of diameter to Kol-
(Fig. 2). This inclination is the result of the imbalance in the forces mogorov length scale. Gidaspow drag model is valid when the

2.5
Outer Stationary Domain

2
Slip Velocity [m/s]

1.5

Inner Rotating Domain 0.5

0
0 0.02 0.04 0.06 0.08 0.1
Radius [m]

Fig. 3. Slip velocity on impeller plane for solid volume fraction of 0.01 and
1000 rpm. - - - - Gidaspow model, –  – Brucato drag model, — Modified Brucato
Fig. 1. Computational domain and grid distribution in stirred tank. drag model.
 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453 449

internal forces are negligible which means that the viscous forces duced in the flow with the ability to mimic the motion in the phase
dominate the flow behaviour. The flow in the stirred tank is mainly of interest. Considering the limitation of CARPT technique which
turbulent. Although k-e turbulence model is not able to simulate has a spatial resolution of the data is 7 mm [26], the CFD results
turbulence effects at the Kolmogorov length scale, this effect has are reported on ensemble average basis in a 7 mm zone around
been taken into account in the modified Brucato drag model. the centreline of the measurement point.
Khopkar et al. [3], while simulating the array of cylinders, have The radial velocity of the solid particles at impeller plane is
incorporated a source term in order to keep the lambda of the or- shown in Fig. 4. On x axis, r is radial position, Ri is impeller radius
der as is observed in a stirred tank. As a result, the underprediction and R is stirred tank radius. Out of the four drag models wide
of slip velocities in the impeller zone is addressed by using the Bru- disparity with experimental data was observed when using the
cato and modified Brucato drag model. The magnitude of the drag Wen and Yu and the Gidaspow model. These two models predicted
increases with turbulence and hence, influences the slip velocity the highest radial velocities. Guha et al. [27] observed similar
between primary and secondary phase. As seen from Fig. 3, the overprediction of radial velocities while using the Schiller–Nauman
Brucato drag modifies the Wen–Yu drag to a greater extent than drag model. The Brucato drag model slightly overpredicted the
the modified Brucato drag, but still the value of slip velocity it is radial velocity, whereas the predictions from the modified Brucato
predicting is below that predicted by modified Brucato drag model. drag were in reasonable agreement with experimental data. The
This is because of the impact of these models on the turbulence. solid velocities are higher at the impeller tip. As the solids approach
Brucato drag, due to its magnitude, has a higher influence on the towards wall, the velocity gradually decreases. Due to no slip
turbulence. The simulations show a high value of turbulent kinetic condition on the wall, the velocity should gradually reach zero
energy with the modified Brucato drag rather than the Brucato value at wall. But, quantitatively, there is an over-prediction of
drag model. And the dissipation noticed in the former case was the velocities in simulations in near wall region. The disparity can
smaller than the latter case. It has a negative impact on the drag be attributed to lesser number of data points available for averaging
calculated and hence the drag is underpredicted due to the low in experiments. The experiments clearly show a zero ensemble
turbulence calculated in the stirred tank. In the cases simulated, averaged value even at (r  Ri)/(R  Ri) = 0.8, which is not
the maximum observed difference in dampening of turbulence reasonable.
was 10%. The impact of turbulence on drag is maximum visible At low solid concentrations, Gidaspow drag model acts like Wen
in the impeller plane where turbulence is dominant. In this region, and Yu model and at higher concentrations it takes the form of the
the ratio dp/k is greater than 10 and for dp/k > 10, the interaction Ergun equation. Therefore, both Wen and Yu model and Gidaspow
between energy dissipating eddies and particles become important models predict the same result. The modified Brucato drag model
for the solids concentration distribution [20]. Derksen et al. [9] accounts for the effect of solid phase on the turbulence. At higher
have compared the slip velocities in terms of linear and rotational impeller speed, the role of turbulence in calculation of drag is vital
Reynolds number and pointed out the dominance of high slip factor, hence, the modified Brucato drag model predicts better re-
velocity in this impeller zone. All the compared drag models were sults as compared to the other drag models.
able to capture the high slip velocities in this region. But, only Fig. 5 shows the comparison between the simulations results and
modified Brucato drag model predicted reasonable values of slip experimental data for radial velocity at axial plane r/R = 0.5. A po-
velocities. When compared with the linear slip velocity values for sitive radial velocity is expected in the upper zone of the impeller. As
vertical plane midway between two baffles, predictions by the compared to the negative velocity in the region higher than the
Gidaspow and Brucato drag model were below the respective impeller zone where the magnitude of negative velocity is merely
values from Derksen et al. [9]. The maximum slip velocity values 2% of the maximum velocity attained by the solids, the negative ra-
obtained were 0.58, 0.63 and 0.72 for Gidaspow, Brucato and mod- dial velocity in the bottom of the tank is 10% of the maximum veloc-
ified Brucato drag models, respectively in the plane as compared to ity. This corresponds to a relatively stronger flow towards the centre
0.75 in case of Derksen et al. [9]. in the bottom of the tank. It indicates the presence of stronger clock-
Another zone of higher slip velocities is the region in which the wise currents. All the drag models could qualitatively capture this
direction of axial velocities is upwards. In this zone, the velocity of flow behaviour. For the experimental data, the highest tangential
the particles is against the force of gravity and increases the differ- velocity is observed at z/T = 0.36 ± 0.04. This compares well the sim-
ence in the velocity of continuous phase and the dispersed phase. ulation result of z/T = 0.34. In the lower region of the stirred tank,
The observed phenomenon is exactly opposite in the regions with
axial velocity vectors pointing downwards [15,20–24]. The slip 0.60
velocities for 1% volume fraction and 7% volume fraction at
1000 rpm were compared. All the parameters of the study were 0.50
kept the same for the two cases except the solid volume fraction.
Since, the velocity in the case of 1% volume fraction case was
0.40
greater than the just suspension speed and that in the case of 7%
Ur/Utip

volume fraction was below the speed of just suspension, a greater

non-uniformity was observed in the solid concentration in the 0.30
stirred tank. It resulted in an increased local value of slip velocity
in regions with low solid concentration in the cases of higher 0.20
average concentration of solids. Although the increase in concen-
tration should result in the decrease in slip velocity, but for cases
0.10
with impeller speed below and above the just suspension speed,
local regions with higher slip velocities can be noticed. This
behaviour was also observed by Sardeshpande et al. [25]. 0.00
0.00 0.20 0.40 0.60 0.80 1.00
(r-Ri)/(R-Ri)
5.3.2. Velocity components
The simulations were run using different drag models and the Fig. 4. Radial velocity at impeller plane for 0.01 solid volume fraction and
results were then compared with the experimental data. Guha 1000 rpm.  Guha et al. (2007), – – Wen and Yu model,  Gidaspow Model,
et al. [8] used CARPT technique, in which a single particle is intro- –  – Brucato drag model, — Modified Brucato drag model.
450 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453

0.50 0.50

0.45 0.45

0.40 0.40

0.35 0.35
z/T

z/T
0.30 0.30

0.25 0.25

0.20 0.20

0.15 0.15
-0.10 0.10 0.30 0.50 -0.20 -0.10 0.00 0.10 0.20
Ur/Utip Uax/Utip

Fig. 5. Radial velocity at axial plane r/R  0.5 for 0.01 solid volume fraction and Fig. 7. Axial velocity at axial plane r/R  0.5 for 0.01 solid volume fraction and
1000 rpm.  Guha et al. (2007), – – Wen and Yu model,  Gidaspow Model, –  – 1000 rpm.  Guha et al. (2007), – – Wen and Yu model,  Gidaspow Model, –  –
Brucato drag model, — Modified Brucato drag model. Brucato drag model, — Modified Brucato drag model.

where the effect of turbulence is not as prominent as the upper re- ible in Fig. 2, the axial velocities shown in Fig. 6 fail to predict it. It
gion, the predictions from all the drag models compared well with is because of the bigger circular loop in the lower region of the
experimental data. In the upper region, discrepancy was observed. impeller clearly seen in Fig. 2, which also affects the ensemble
Around the impeller zone, where, the turbulence and velocity fluc- averaging of values in this particular zone. At the impeller plane,
tuations are higher, the Wen and Yu and Gidaspow drag models the axial velocity is zero as it is distributed as the other two com-
show large overprediction compared to the experimental data. On ponents of velocities.
the other hand, the Brucato and modified Brucato drag show reason- Modified Brucato drag model was found to be the most appro-
able agreement. Fig. 6 shows the comparison between the simula- priate drag model and, therefore, the remaining analysis showed in
tions results and experimental data for tangential velocity at axial the paper is based on the simulations conducted using this drag
plane r/R = 0.5. Similar trend to that observed in radial velocity model.
is observed.
The axial velocity profile is shown in Fig. 7. The reversal of flow 5.3.3. Turbulent kinetic energy (TKE)
can be clearly seen. Above the impeller, the axial velocities are neg- The flow in a stirred tank is turbulent flow that results in the
ative that means the flow is in downward direction. It reverses in fluctuating components of velocity due to formation of eddies.
the region below impeller. At the impeller, the axial velocity is zero The k-e model used in the RANS simulation assumes isotropic tur-
and is distributed as the other two components of velocities viz. ra- bulence and uses the Boussinesq hypothesis to relate the Reynolds
dial and tangential. All the drag models were able to capture the stresses to the mean velocity gradients. The TKE represents the
flow reversal qualitatively. Moreover, the predictions of all the magnitude of turbulence present in the system. The presence of
drag models were comparable. The experiments show higher axial particles dampens turbulence. In order to access the impact of par-
velocity in the lower region compared to the upper region. ticles, the TKE for single phase flow is compared with the TKE at
Whereas, the simulations predicted similar velocities in the lower solid loadings of 1% and 7%.
and upper region of the impeller. Although this phenomenon is vis- For a single phase system, the liquid is agitated by the impeller.
The high velocity and trailing vortices result in large velocity fluc-
tuations in the impeller plane. For this reason TKE was found to be
0.50 the maximum at the blade tip (Fig. 8(a)). As the velocity decrease
radially in this plane, the TKE also decreases. The magnitude of
0.45 TKE in the other parts of the tank is approximately 103 times lower
than those in the impeller plane. Michelleti et al. [28] used LDA
0.40 technique to measure dissipation rates at various points in the stir-
red tank and found the variation by more than 2 orders of magni-
0.35 tude between impeller region and the bulk. Dissipation rate
z/T

follows the same trend as TKE not only in the impeller region
0.30 but in the other regions of the stirred tanks as well. The TKE, sim-
ilar to the trend observed by Michelleti et al. [28] for turbulence
0.25 dissipation rate, is comparatively high near the walls and near
the axially centre line where the axial velocity field is dominant.
0.20 Similar behaviour is observed in presence of particles (Fig. 8(b)).
However, the magnitude of TKE is much lower.
0.15 Fig. 9 shows the comparison between the TKE profile at impel-
0.00 0.10 0.20 0.30 0.40
ler plane for 0%, 1% and 7% solids volume fraction. Within the
Ut/Utip
impeller radius, the TKE increases due to the increase in turbu-
Fig. 6. Tangential velocity at axial plane r/R  0.5 for 0.01 solid volume fraction and
lence. Initially, rate of increase is low as the impeller disc offers
1000 rpm.  Guha et al. (2007), – – Wen and Yu model,  Gidaspow Model, –  – resistance. After the disc, the TKE increases steeply and reaches
Brucato drag model, — Modified Brucato drag model. a maximum slightly beyond impeller radius. This behaviour is
 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453 451

1.112

0.281

0.071
0.018

0.005

0.001
J kg-1

(a) Single Phase Flow (b) Solid- Liquid System (1% v/v)

Fig. 8. Turbulent kinetic energy contours in stirred tanks at 1000 rpm.

1.2 values between 50 and 70%. The decrease in turbulence with

Impeller Radius
increase in solid concentration was also observed by Barresi and
Baldi [30], Micheletti et al. [29,31] and Ayazi Shamlou and
Turbulent Kinetic Energy [ J / Kg ]

1
Koutsakos [32]. Micheletti et al. [29] conducted experiments to
study velocity characteristics in stirred solid liquid suspension.
0.8 The flow field measurement in the presence of solids revealed
significant influence of their presence. The maximum difference
was observed in the impeller plane that diminished with increas-
0.6 Single Phase
ing radial distance. These points support the findings in this paper
1%
where the turbulence is the maximum in the single phase flow
0.4 7%
and corresponding lower values of turbulence is observed for
higher solid concentration. The difference in turbulence also
0.2 decreases with the increase in the radial distance. In practical
conditions, due to the increase in solids concentration, the
dissipation of energy will be higher due to the high frequency
0 of particle–particle, particle-wall and particle blades collision.
0 0.02 0.04 0.06 0.08 0.1
The turbulence dampens in the presence of solids and the magni-
Radial Location [ m ]
tude of vortices leaving the impeller decreases. For the same
Fig. 9. Turbulent kinetic energy profiles at impeller plane at 1000 rpm. reason, a shift in the peak of TKE is observed with increase in
solid concentration.

attributed the vortices leaving the impeller blade that result in 5.3.4. Cloud height and homogeneity
high magnitude fluctuating velocities. After this point, the TKE The velocities of the jet in the region above the impeller begin
gradually decreases along the radius due to decrease in velocities. to decay after reaching a certain height. Negative buoyant jet
As the velocity jet hits the vessel wall, it creates eddies resulting behaviour is observed at this point and it results in a sudden con-
in fluctuating velocities. As a result a small peak in the TKE is ob- centration gradient and is termed as ‘cloud height’. Beyond this
served near the wall. height the velocity is not able to drag the solids. The cloud height
The kinetic energy in the liquid is imparted to solids resulting was calculated for the stirred tanks at volume fractions 0.01 and
in the solids following the jet. It is also the reason of maximum 0.07. Apart from just suspension speed, the cloud height is an
energy dissipation in this zone. The comparison shows 50% and important parameter for the representation of homogeneity. The
65% decrease in the kinetic energy observed for 1% and 7% vol- cloud height between 0.45 and 0.55T shows just suspension.
ume fraction of solids, respectively in the impeller plane. The The cloud height below this height shows poor homogeneity.
kinetic energy of the liquid is dissipated in the suspension and And a cloud height that reaches 0.9T or above shows the highest
dispersion of solid particles. This results in the decrease in the quality of homogeneity. The cloud height in the stirred tank is
level of turbulence and is visible as lower levels of TKE. Nouri shown as the iso-contours of the average volume fraction in the
and Whitelaw [21] measured and analysed liquid and solid phase stirred tank (Fig. 10).
velocities in stirred vessels with solid concentration up to 0.02%. Fluctuations in the velocity and the solid concentration below
The effect of presence of particles, particle concentration and the cloud height was observed to be negligible. Due to the sudden
density is studied on the slip velocities and turbulence and the change in the flow field and concentration, a zone of high turbulent
turbulence was found to decrease by up to 25%. Specifically, they fluctuations and macroinstabilities forms at the cloud height. The
found the dampening in the turbulence in impeller zone. In the velocity changes frequently and the cloud height fluctuate around
impeller zone, both the TKE and particle concentration are maxi- a constant value.
mum. As a result, the dissipation of energy is the maximum in At low solid concentration, the influence of solids on the liquid
this region and leads to the maximum decrease in turbulence flow field is less. At the same time the kinetic energy dissipation
as particle concentration increases. The dampening of turbulence of the continuous phase for the suspension of the dispersed phase
found by Derksen et al. [9] was around 15%. This value is far low- is lesser as compared to the high loading systems. As a result, the
er than as observed in this paper. Similar observations were made magnitude of the axial velocity is not altered to a great extent.
by Michelleti et al. [29] that presented the turbulence dampening The homogeneity in the stirred tank is, therefore, achievable at
452 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453

Fig. 10. The cloud height in stirred tanks. (a) 0.01 volume fraction 1000 rpm (b) 0.07 volume fraction 600 rpm (c) 0.07 volume fraction 800 rpm (d) 0.07 volume fraction
1000 rpm.

low impeller speeds. The cloud height in this case is observed to be more number of particles and high frequency of particle–particle,
higher for the low solid loading system than the high solid loading particle-wall and particle-blade collisions.
system (Fig. 10).
Homogeneity is a measure of quality of suspension. The qual-
ity of suspension increases with the impeller speed. The increase References
in the impeller speed results in a higher kinetic energy of the con-
[1] L. Fradette, P.A. Tanguy, F.O. Bertrand, F. Thibault, J.-B.T. Ritz, E. Giraud, CFD
tinuous phase which is available for disposal to the solids. A phenomenological model of solid-liquid mixing in stirred vessels, Computers
strong velocity field for continuous phase is present at high and Chemical Engineering 31 (2007) 334–345.
impeller speeds in various zones of a stirred tank. The velocity [2] G.R. Kasat, A.R. Khopkar, V.V. Ranade, A.B. Pandit, CFD simulation of liquid-
phase mixing in solid-liquid stirred reactor, Chemical Engineering Science 63
of the jet near the top surface is also larger. The magnitude of up-
(2008) 3877–3885.
ward axial velocity near the walls is found to be higher when [3] A.R. Khopkar, G.R. Kasat, A.B. Pandit, V.V. Ranade, Computational fluid
compared with those at lower impeller speeds. This result in dynamics simulation of the solid suspension in a stirred slurry reactor,
Industrial and Engineering Chemistry Research 45 (2006) 4416–4428.
the attainment of a higher cloud height at higher impeller speeds
[4] G. Micale, F. Grisafi, L. Rizzuti, A. Brucato, CFD simulation of particle
as is evident from Fig. 10. It is in accordance with the results ob- suspension height in stirred vessels, Chemical Engineering Research and
tained by Sardeshpande et al. [33] for impeller speeds above Design 82 (2004) 1204–1213.
‘speed of just suspension’. The non-monotonic behaviour in the [5] G. Micale, G. Montante, F. Grisafi, A. Brucato, J. Godfrey, CFD simulation of
particle distribution in stirred vessels, Chemical Engineering Research and
cloud height is observed at very low impeller speeds [33]. This ef- Design 78 (2000) 435–444.
fect is due to the presence of pseudo-bottom formed because of [6] G. Montante, G. Micale, F. Magelli, A. Brucato, Experiments and CFD predictions
the accumulation of undispersed solids at the bottom causing of solid particle distribution in a vessel agitated with four pitched blade
turbines, Chemical Engineering Research and Design 79 (2001) 1005–1010.
‘false-bottom’ effect. As a result the off-bottom clearance and [7] A. Ochieng, A.E. Lewis, CFD simulation of solids off-bottom suspension and
the amount of suspended solids increases with increasing the cloud height, Hydrometallurgy 82 (2006) 1–12.
impeller speed and hence, the cloud height decreases. Since, the [8] D. Guha, P.A. Ramachandran, M.P. Dudukovic, Flow field of suspended solids in
a stirred tank reactor by Lagrangian tracking, Chemical Engineering Science 62
impeller speed studied were not very low that can cause ‘false (2007) 6143–6154.
bottom effect’, this non-monotonic behaviour was not observed. [9] J.J. Derksen, Numerical simulation of solids suspension in a stirred tank,
The height of the interface between the clear liquid layer and American Institute of Chemical Engineers, AIChE Journal 49 (2003) 2700.
[10] J. Ding, D. Gidaspow, A bubbling fluidization model using kinetic theory of
the solid suspension layer is lower in the region above the impel-
granular flow, AIChE Journal 36 (1990) 523–538.
ler as the recirculating jet forces solids downwards near the axis. [11] A. Brucato, F. Grisafi, G. Montante, Particle drag coefficients in turbulent fluids,
This phenomenon is prominent at low impeller speeds due to the Chemical Engineering Science 53 (1998) 3295–3314.
[12] R. Panneerselvam, S. Savithri, G.D. Surender, CFD modeling of gas–liquid–solid
weak upward axial flow and the dominance of the drawing action
mechanically agitated contactor, Chemical Engineering Research and Design
right above the impeller. 86 (2008) 1331–1344.
[13] D. Guha, Evaluation of large Eddy simulation and Euler–Euler CFD models for
solids flow dynamics in a stirred tank reactor, American Institute of Chemical
Engineers, AIChE Journal 54 (2008) 766–778.
6. Conclusion [14] ANSYS, Fluent User Guide, in: ANSYS Inc., Canonsburg, PA, http://
www.fluent.com, 2009.
CFD simulations of solid suspension in stirred tank were per- [15] M. Ljungqvist, A. Rasmuson, Numerical simulation of the two-phase flow in an
axially stirred vessel, Chemical Engineering Research and Design 79 (2001)
formed. The predictions of four different drag models were com- 533–546.
pared. It was observed that turbulence dispersion force had [16] C.Y. Wen, Y.H. Yu, Mechanics of fluidization, in: Chemical Engineering Progress
negligible effect due to a low volume fraction of solids. Axial, radial Symposium Series, 1966, pp. 100–111.
[17] G.L. Lane, M.P. Schwarz, G.M. Evans, Numerical modelling of gas and liquid
and tangential velocities were compared at different axial loca- flow in stirred tanks, Chemical Engineering Science 60 (2005) 2203–2214.
tions. It was observed that all four models could qualitatively cap- [18] D. Fajner, D. Pinelli, R.S. Ghadge, G. Montante, A. Paglianti, F. Magelli, Solids
ture the flow in stirred tank. The Wen and Yu and Gidaspow model distribution and rising velocity of buoyant solid particles in a vessel stirred
with multiple impellers, Chemical Engineering Science 63 (2008) 5876–5882.
showed biggest deviation from the experimental data while results [19] F. Sbrizzai, V. Lavezzo, R. Verzicco, M. Campolo, A. Soldati, Direct numerical
from the modified Brucato drag model were in reasonable agree- simulation of turbulent particle dispersion in an unbaffled stirred-tank reactor,
ment for the liquid flow fields. Chemical Engineering Science 61 (2006) 2843–2851.
[20] A. Ochieng, M.S. Onyango, Drag models, solids concentration and velocity
Maximum turbulence kinetic energy was found in the impeller
distribution in a stirred tank, Powder Technology 181 (2008) 1–8.
zone. The turbulence dampens with the increase in the solids con- [21] J.M. Nouri, J.H. Whitelaw, Particle velocity characteristics of dilute to
centration and this effect was the most significant in this zone. For moderately dense suspension flows in stirred reactors, International Journal
achieving homogeneity at low loading stirred tanks, a low impeller of Multiphase Flow 18 (1992) 21–33.
[22] P. Guiraud, J. Costes, J. Bertrand, Local measurements of fluid and particle
speed is adequate. However, high impeller speed is needed for high velocities in a stirred suspension, Chemical Engineering Journal 68 (1997) 75–
solid loading systems, as the energy dissipation is significant due to 86.
 D. Wadnerkar et al. / Advanced Powder Technology 23 (2012) 445–453 453

[23] M. Ljungqvist, A. Rasmuson, The two-phase flow in an axially stirred vessel stirred reactors, Chemical Engineering Research and Design 82 (2004) 1188–
investigated using phase-doppler anemometry, The Canadian Journal of 1198. http://www.sciencedirect.com/science/article/pii/S0263876204726064.
Chemical Engineering 82 (2004) 275–288. [29] M. Micheletti, M. Yianneskis, Study of fluid velocity characteristics in stirred
[24] P. Pianko-Oprych, A.W. Nienow, M. Barigou, Positron emission particle solid-liquid suspensions with a refractive index matching technique,
tracking (PEPT) compared to particle image velocimetry (PIV) for studying Proceedings of the Institution of Mechanical Engineers, Part E: Journal of
the flow generated by a pitched-blade turbine in single phase and multi-phase Process Mechanical Engineering 218 (2004) 191–204.
systems, Chemical Engineering Science 64 (2009) 4955–4968. [30] A. Barresi, G. Baldi, Solid dispersion in an agitated vessel, Chemical Engineering
[25] M.V. Sardeshpande, V.A. Juvekar, V.V. Ranade, Solid suspension in stirred Science 42 (1987) 2949–2956.
tanks: UVP measurements and CFD simulations, The Canadian Journal of [31] M. Micheletti, Particle concentration and mixing characteristics of moderate-
Chemical Engineering 89 (2011) 1112–1121. to-dense solid–liquid suspensions, Industrial and Engineering Chemistry
[26] J. Chaouki, F. Larachi, M.P. Dudukovic, Noninvasive tomographic and Research 42 (2003). 6236-6236.
velocimetric monitoring of multiphase flows, Industrial and engineering [32] P. Ayazi Shamlou, E. Koutsakos, Solids suspension and distribution in liquids
chemistry research 36 (1997) 4476–4503. under turbulent agitation, Chemical Engineering Science 44 (1989)
[27] D. Guha, Evaluation of large Eddy simulation and Euler–Euler CFD models for 529–542.
solids flow dynamics in a stirred tank reactor, American Institute of Chemical [33] M.V. Sardeshpande, A.R. Sagi, V.A. Juvekar, V.V. Ranade, Solid suspension and
Engineers, AIChE Journal 54 (2008). 766-766. liquid phase mixing in solid–liquid stirred tanks, Industrial and Engineering
[28] M. Micheletti, S. Baldi, S.L. Yeoh, A. Ducci, G. Papadakis, K.C. Lee, M. Yianneskis, Chemistry Research 48 (2009) 9713.
On spatial and temporal variations and estimates of energy dissipation in