You are on page 1of 13

Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2, pp.

282294 (2013)

SOLID SUSPENSION IN STIRRED TANK EQUIPPED WITH MULTI-


SIDE-ENTERING AGITATORS
J. Fang *, X. Ling and Z.F. Sang

College of Mechanical and Power Engineering, Nanjing University of Technology, Nanjing, 210009,
China
* E-Mail: fjian04@sina.com (Corresponding Author)

ABSTRACT: The flow field and solid distribution in a stirred tank equipped with four side-entering agitators were
investigated using both experimental measurements and CFD simulations. Experiments were carried out to
determine the regularity of the solid sediment distribution on the tank bottom and the critical impeller speed for off-
bottom suspension. The CFD simulation was performed using the Eulerian-Eulerian two-phase approach, the
standard k- turbulence model and the multiple reference frame (MRF) approach. The predicted critical impeller
speeds were compared with the experimental data to validate the CFD model. The effects of the solid loading and
some installation parameters including inclined angles of the impeller, plunging length and mounting height of the
shaft on the critical impeller speed were investigated. The predicted results showed reasonably good agreement with
the experimental data for both the distribution of solid accumulation and critical impeller speeds. The solid loading
has a greater influence on the critical impeller speed in the side-entering stirred tank than that in the top-entering
tank. On the basis of the CFD simulation determined critical impeller speeds under different conditions, the above
installation parameters of the agitator were optimized. The results and discussion here will have useful implications
for design and optimization of the solid-liquid suspension in side-entering stirred tanks.
Keywords: solid suspension, CFD, critical impeller speed, side-entering agitator

1. INTRODUCTION obtain detailed solid distribution in the whole


vessel through experimental measurements.
Solid-liquid stirred reactors are widely used in Computational fluid dynamics (CFD) is a
many process industries such as fine chemicals, powerful tool in the stirring area as it could
biotechnology, cosmetics, consumer products, provide detailed information on the flow field for
food, and mineral processing. The primary single and multi-phase systems in stirred tanks.
objective of solid-liquid mixing is to avoid solid However, the reliability of CFD predictions is
accumulation in the stirred tank or to ensure that critically dependent on many factors such as the
the solid particles are distributed uniformly grids of impeller/tank configurations, the
throughout the tank. Many factors critically affect multiphase model and numerical discretization.
the suspension of solid particles in the stirred Montante et al. (2001) investigated the solid
vessels, including impeller speed, solid and liquid suspension in a tall tank agitated with four pitched
properties and stirred tank construction. A good blade turbines by the CFD simulation using a
understanding of the distribution of solid particles Sliding-Grid approach coupled with the Eulerian-
in the tank is significant for the design and scale- Eulerian model. The predicted axial concentration
up of solid-liquid suspension systems. Many profiles of solid particles were in good agreement
experimental investigations on mixing and solid with the experimental data. Ochieng and Onyango
suspension in vessels of different geometrical (2008) studied the effects of drag models on the
configurations agitated by various types of simulation of solid suspension in a vessel stirred
impellers have been performed in the past (Hicks by a hydrofoil propeller. Based on the consistency
et al., 1997; Wu et al., 2001; Sharma and Shaikh, between the CFD and experimental results, drag
2003; Al-Qaessi and Abu-Farah, 2009; models suitable for various solid loadings were
Sardeshpande et al., 2009; Jirout and Rieger, proposed. Khopkar et al. (2006) used the Euler-
2011). However, the main objective of these Euler model together with the standard k-
investigations was the determination of some turbulence model to predict the solid-liquid flow
macro-parameters such as the critical impeller in a stirred vessel. Different criteria for
speed for just off-bottom suspension and the determining the critical impeller speed were
cloud height in the vessel. It is quite difficult to

Received: 26 Dec. 2011; Revised: 26 Apr. 2012; Accepted: 21 May 2012

282
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

discussed and the predicted speed values showed of the shaft on the critical impeller speed were
good agreement with the experimental data. investigated to obtain the influence of these
Stirred tanks with side-entering agitators, which parameters on the solid suspension in the side-
are usually used in many industries such as pulp entering stirred tank.
and paper manufacturing, flue gas
desulphurization system and petroleum, have 2. EXPERIMENTS
received limited attention. The side-entering
agitator has such features as small relative Experiments were conducted in a scaled-down,
diameter ratio, low power consumption and high flat-bottomed, cylindrical tank, of diameter
power efficiency in comparison to the top- T=0.58 m and equipped with four side-entering
entering agitator. The majority of researches on agitators, as shown in Fig. 1. The vessel is made
the stirred tank are based on the top-entering of perspex and immersed in a glass rectangular
agitators, while the researches concerning side- tank filled with water, so as to minimize the
entering stirrers are quite limited. Previous studies visual fault owing to the refraction of the water
on the side-entering stirred tanks were mainly and curved wall of the inner vessel. Four agitators
focused on the flow structure and mixing are equally installed around the lower part of the
characteristics (Wesselingh, 1975; Dskhel and vessel, each with a vertical inclination angles
Rahimi, 2004; Ford et al., 2006; Saeed et al., and a horizontal angle . Based on the design
2007; Gomez et al., 2010; Fang et al., 2011). Wu experience of side-entering stirred tanks used in
(2012) studied the power and flow number, the industries, the inclination angle and in this
mixing characterization in large-scale circular and test equipment were initially set as 10 and 5,
rectangular stirred tanks with single and multi- respectively. The diameter of the impeller is
side-entering impellers using CFD technique. The D=0.1 m and its configuration is shown in Fig. 1c.
effect of inclined angles and some installation It is one type of axial-flow impeller with four 35
parameters of the agitators on mixing were pitched blades. The blade was with variable
discussed and optimized. Kipke (1984) sections and the maximum ratio between blade
investigated the effect of the impeller position, height and impeller diameter is 0.15. One arc
impeller types and diameter ratio on the plate was set at the end of each blade to enhance
suspension characteristics of the side-entering the axial pumping capacity. The agitators were
impellers using various scale tests based on a tank installed 0.12 m (h=1.2D) above the bottom of the
in desulphurization industries. As far as we know, vessel and the shaft length stretched into the
no research has been reported in the open vessel l=D, with a shaft diameter of 0.015 m. The
literature on a combined experimental and CFD working liquid was tap water (the viscosity
study of solid suspension in stirred tanks with l=0.001 Pas and density l=998 kg/m3) and the
multi-side-entering agitators. Therefore the liquid height H in the vessel was 0.8T. Glass bead
present study was carried out to make further particles (s=2500 kg/m3) of mean diameter
analysis of the solid-liquid flow and suspension to dp=100 m were used as the solid phase and
get the inherent regularity of solid-liquid mixing various solid loading by volume s viz 1-5% v/v
in agitated vessels of this kind. were carried out in the experiments. The solid
In the present work, based on the validated single suspension state in the tank was studied visually
phase flow field using PIV measurement data by through the transparent tank wall and tank bottom
Fang et al. (2011), both the experimental and with the aid of a mirror placed directly underneath
CFD (Fluent 6.3) methods were used to it. Critical impeller speed for off-bottom
investigate the characteristics of the solid-liquid suspension Njs was determined visually according
flow field in a laboratory-scale cylindrical stirred to the definition of Zwietering (1958)- no
tank equipped with four axially side-entering particles were observed to remain at rest on the
agitators. The flow characteristics of the liquid tank bottom for more than 1-2 s. Each
phase and solid distribution in the solid-liquid measurement was performed after 10 min. of
system were discussed. The critical impeller stirring as the flow and solid particle suspension
speed for just off-bottom suspension obtained by reached the fully stabilized state. To minimize the
both experimental measurement and CFD observation error in Njs, each of Njs was obtained
simulation were compared so as to validate the from three different runs by different workers and
accuracy of the CFD model. The effects of then averaged.
different operating and installation parameters
including solid loading, inclined angles of the
impeller, the plunging length and mounting height

283
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)


( q qU q ) + ( q qU q U q )
t (2)
= q p ( q q ,ij ) + q q gi + FB + FI

Here, q denotes the continuous phase (liquid) or


the suspended phase (solid), and i is the direction.
q and q are the density and volume fraction of
phase q, respectively. Uq is the time-averaged
value of the velocity and p is the pressure.
Further, q ,ij is the stress tensor of the phase q.
q q gi is the external body force on the phase q.
(a) vertical plane FB denotes the centrifugal and Coriolis forces. FI
represents the interfacial force, which includes
drag force (FD), non-drag forces (FND) and the
turbulent dispersion force (FTD). The turbulent
dispersion force (FTD) is considered to be
important only when the sizes of the turbulent
eddies are larger than the particle sizes. In the
solid-liquid stirred vessel, the ratio of the largest
turbulent eddy to the particle size was found to be
about 10 (Khopkar et al, 2006). So the
contribution of FTD is significant. Ljungqvist and
Rasmuson (2001) and Ochieng and Lewis (2006)
highlighted the importance of the effect of the
turbulence dispersion force while simulating
solid-liquid mixing in stirred tanks. Therefore, the
(b) horizontal plane turbulence dispersion force was considered in this
work, which is given by
CTD q kq q
FTD = (3)

with coefficient CTD taken as 0.1 (Alajbegovic et


al., 1999).
The non-drag forces (FND) contain the lift force
(FL) and virtual mass force (FVM), and others.
Ljungqvist and Rasmuson (2001) studied the
influence of these forces on the hydrodynamics
simulation in the stirred vessel and found that the
drag force has a major role in the solid-liquid
flow while the effect of virtual mass and lift
(c) impeller forces can be neglected. Therefore, only the drag
force was taken into consideration in this paper.
Fig. 1 Configuration of stirred tank and impeller.
The expression of drag force in the i direction is
written as
3. CFD SIMULATION
FD ,i = Fl ,i = Fs ,i
3.1 Mathematical model
3 l s l CD ( (U s ,i U l ,i ) 2 )0.5 (U s ,i U l ,i ) (4)
In the present study, the mathematical model is =
4d p
formulated based on the Eulerian-Eulerian multi-
fluid model. The Reynolds averaged mass and In the solid-liquid stirred tank, the drag
momentum balance equations in the turbulent coefficient (CD) is a complicated function of the
flow regime, without considering the mass and the solid phase hold-up, the drag coefficient in
energy transports, can be written as stagnant liquid (CD0) and bulk turbulence.
Ochieng and Onyango (2008) tried different drag
( q q ) + ( q qU q ) = 0 (1) models including those of Schiller-Naumann,
t

284
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

Brucato et al. (1998) and Gidaspow to simulate (LES) method to simulate the solid suspension in
solid-liquid flow in a stirred vessel and found that stirred tanks. However, this method requires a
the model of Brucato et al. (1998) gave the best great amount of computing time, but does not
prediction. Based on their studies, this model for make a dramatically improvement over the
estimating the drag coefficient CD was used and standard k- model. Therefore, the RANS based
the correlation is defined by mixture k- turbulence model is used here to
simulate the solid-liquid flow in the stirred tank.
dp
CD = CD 0 1 + 8.76 104 ( )3 (5) The governing equations for turbulent kinetic
energy k and turbulent energy dissipation rate
are taken as:
where

24 ( m k ) + ( mU m k ) = tm k + Gkm m (9)
CD 0= (1 + 0.15 Re p 0.687 ) (6) t m
Re p

Here, is the Kolmogorov length scale, dp the ( m ) + ( mU m )
particle diameter and Re p the particle Reynolds t (10)

number, which is defined as = tm + (C1 Gkm C2 m )
k
d p U s ,i U l ,i l
Re p = (7) where Um is mean velocity of the mixture, which
l is calculated from
The Reynolds number of a single impeller for N

solid-liquid flow in the side-entering stirred tank U q q q


is defined as Um = n =1
N
(11)

D 2 N m q q
Re = (8) n =1
av
Further, Gkm is the generation of turbulent kinetic
Here, m is the density of the mixture in solid- energy due to the mean velocity gradients, which
m l l + s s . av is the
liquid stirred tank, = is defined by
average viscosity of the mixture and 1
1.8 =
G tm (U m + (U m )T ) 2 (12)
1 Xv
km
2
av = l / , where X v = s / l and Xvb
X vb Here, tm is the turbulent viscosity and can be
denotes the ratio of solid to liquid volume fraction written as
in the solid sediment layer, Xvb=1.083.
m C k 2
Among different turbulence models used to tm = (13)
estimate the turbulent flow in a stirred tank, the
standard k- turbulence model is the most The standard set of parameters usually adopted
common one. It is generally thought that the k- with the k- model was used in the computations,
model, based on isotropic turbulence assumption, namely: C1=1.44, C2=1.92, C=0.09, k=1.0 and
could not accurately simulate the physics of the =1.3. For all the simulations, the turbulent
turbulent flow in a stirred tank. However, it has Schmidt number for the liquid phase was taken as
been proven in many studies that, the predicted equal to 0.8, according to the literature suggestion
results obtained using this turbulent model have (Montante et al., 2001).
good agreement with the experimental results.
Montante and Maqelli (2005) reported that the 3.2 Simulation methods
standard k- turbulence model with mixture
properties gave a qualitatively good In this study, the commercial CFD software
representation of the solid distribution in stirred (Fluent 6.3) was used to simulate the solid-liquid
vessels. Ochieng and Onyango (2008) indicated flow in a side-entering agitated vessel. Because of
that for flows in low solid loading systems, the the symmetry of the tank configuration and
solid-liquid mixing was dominated by the bulk periodicity of the flow, only a quarter section of
fluid flow and for such flow the k- model was the tank was modeled as the solution domain. The
still an effective approach. Derkson (2003) and origin of the coordinate is on the centre of the
Guha et al. (2008) used the large eddy simulation vessel bottom and the Z-direction is set along the

285
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

axis of the vessel, as shown in Fig1. The multiple profiles along this line for the three grids were
reference frame (MRF) approach was employed compared with each other. The results showed
to model the rotation of the impeller in the stirred that the medium and fine grids made similar
tank. A rotating frame was defined for the region predictions with the velocity discrepancy up to
containing the agitator while a stationary frame 3.6%. The velocity magnitude of the coarse grid
was used for the outer zones containing the tank was markedly different from those of the medium
wall. The rotating zone was a cylinder of 1.5D and fine meshes, with the maximum difference up
diameter and 0.6D height, with the rotating axis to 14.3%. The calculating time difference
set along the shaft of the impeller. A steady between the medium and fine grids was more than
transfer of calculating data was made at the MRF 20%. Based on these results, the medium grid was
interface as the solution progressed. The tank wall employed for the simulation in this work, as
and stirring shaft were set to be no-slip wall shown in Fig. 2. Preliminary simulations were run
boundary condition and the standard wall with water only, and the glass particles were then
functions were used (Wang et al., 2010). The introduced into the fully developed liquid flow
surface of the liquid was defined as free surface field to calculate the solid suspension.
boundary condition where the shear stress was set
to zero. The two parting plane of the quarter
section of the tank was defined as periodic
boundary condition in the azimuthal direction.
Owing to the complex structure and inclination
angles of the propeller, a non-uniformed grid with
combined tetrahedral and prism unstructured
elements were performed to mesh the solution
domain. Mesh refinements near the impeller and
tank bottom were accomplished using the size
function. The second upwind discretization
scheme was chosen for the convection terms of
the momentum, turbulent kinetic energy and
energy dissipation rate equations, which was
superior to the first upwind scheme (Aubin et al.,
2004). The SIMPLE algorithm was used to couple
the pressure and momentum equations. Fig. 2 Mesh of solution domain.
Simulations were considered converged when the
residuals for all parameters was below 110-4. 4. RESULTS AND DISCUSSION
Further convergence checks were performed by
verifying that the velocity magnitude and solid 4.1 Flow field
concentrations of some monitoring points in the The single liquid phase flow in a vessel equipped
tank remained constant. with four side-entry impellers has been discussed
A preliminary grid convergence study was and the CFD simulation of flow field was
implemented so as to verify that the simulation validated using the PIV experiment data by Fang
was grid independent. The optimum grid was et al. (2011). Figures 3a and b show the velocity
obtained by decreasing the size to a final value vectors of liquid phase on the X-Z plane for the
below which the discrepancies in the velocity and liquid-only (Fang et al., 2011) and solid-liquid
kinetic energy profiles were less than 3% (Seyed (s=2%) system when the impeller speed N=400
et al., 2010). The numbers of grids in both the rpm. The velocity vectors are normalized to the
inner and outer zones were increased so that three impeller tip speed, Utip. The corresponding
grids were generated. They were defined as Reynolds numbers of a single impeller for the
coarse, medium and fine mesh containing single phase and solid-liquid system are 66547
294,336, 438,167 and 825,052 elements, and 57108, respectively. The introduction of solid
respectively. When refining the mesh, additional particles into the fluid increases the equivalent
meshing was included in the regions of high viscosity of fluid in the tank increased. According
velocity gradient around the impeller and to Eq. 8, the average density of the mixture m
discharge zone. A sample line which passes increases but the growth amplitude of the average
through two points (0.16, 0, 0) and (0.16, 0, viscosity of the mixture av is greater, leading to a
0.464) was chosen to check the independence of decrease of the Re value in the solid-liquid
the three grids. The velocity and solid holdup system. For the liquid-only system shown in Fig.

286
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

3a, a main circulation loop exists in the vessel,


which starts from the impeller, flows through the
axis of the tank up to the liquid surface, and
finally returns to the impeller suction region along
the tank wall. The center of this main circulation
loop is about 0.74H above the tank bottom. In
Fig. 3b, the introduction of the solids into the
vessel results in a decrease in the axial velocity
component compared with Fig. 3a. The impeller
generates one loop where the flow leaving the
impeller is directed downward toward the bottom
of the tank, which is directly responsible for the
suspension of the particles. The main circulation
loop reaches only about 0.7H high, with its center (a) vertical plane in Fang et al. (2011)
approximately 0.5H high. Secondary circulation
loops are revealed in the upper part of the vessel,
which are mainly circumferential flows around
the axis of the vessel. This zone has a quite low
velocity level and contains few solid particles.
There is little energy exchange between this
secondary loop zone and the main loop zone in
the lower part of the vessel. Therefore, this
secondary loop zone is not conducive to the
efficient solid-liquid mixing of the whole tank.
The liquid velocity distributions at the level of
0.215H high above the bottom for both the liquid-
only and solid-liquid systems are shown in Fig. 3c
and d, respectively. It can be seen from Fig. 3c (b) vertical plane for solid-liquid system
that, for the liquid-only system, the flow pattern at
the level of impeller mounting height is quite
complex. At the center part, an upward flow
toward the liquid surface exists. Around the
center there is an impingement region formed by
the discharge flows of the four impellers, and a
clockwise circulation loop exists in the periphery
of the plane because of the horizontally inclined
angle . When the solid was introduced, shown in
Fig. 3d, the coverage area of the four impeller
high flow zones was reduced. This is because the
drag effect of solid particles attenuates the
velocity of the impeller discharge more quickly. (c) z/H=0.215 in Fang et al. (2011)
And then the turbulent intensity around the
impeller zone is weakened. As the Re value
mainly indicates the turbulent intensity in the
tank, especially for the region around the
impeller, the introduction of solid decreases the
Re value compared with the single phase system.
The axial flow around the tank center is
weakened, and the velocity of the zones which lie
near the tank wall and between each of the two
impellers also becomes lower.
The comparison of axial velocity profiles along
two sample lines in the X-Z plane (z=0.3H and
0.7H) between the liquid-only and solid-liquid (d) z/H=0.215 for solid-liquid system
systems is shown in Fig. 4. It can be seen in Fig. Fig. 3 Flow fields in different planes for single and
4a that, at the high level of 0.3H which is within solid-liquid systems.

287
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

decreases. As for the high level of 0.7H shown in


Fig. 4b, the solid particles obviously affect the
liquid flow pattern. In the liquid-only system, the
pattern is a typical up-down axial flow. But in the
solid-liquid system, the axial velocity values have
been intensively reduced, most of which getting
close to zero. According to Fig. 3c, the height of
the main circulation loop is reduced to about 0.7H
high because of the solid particles. Therefore, the
0.7H high level is within the transition region of
the main circulation loop and secondary loop in
the upper part of vessel, and the flow pattern
around this level starts to change from axial loop
(a) z=0.3H to circumferential loop.

4.2 Solid particle distribution


The solid suspension is generally divided into
three types: incomplete suspension, just off-
bottom suspension and homogeneous suspension.
When the spherical solid particles are
accumulated into a single layer, the maximum
solid volume fraction value is found to be no less
than /6 (approximately 0.52). So when the local
solid hold-up value is greater than 0.52, solid
sediment zone exists in the vessel. Therefore, in
this CFD simulation study, when the predicted
maximum solid volume fraction in the solution
domain is less than 0.52, no solid sediment is
(b) z=0.7H
expected to exist in the vessel. Fig. 5 illustrates
Fig. 4 Axial velocity profiles of liquid phase when the contours of the solid concentrations in the X-Z
N=400 rpm for single and solid-liquid plane with the solid loading of 2 v/v% at N=400
systems. rpm. It can be seen from Fig. 5 that the solid
concentration shows ladder-like distributions
from the tank bottom to the liquid level. The solid
concentration decreases as the distance to the
bottom increases. As the maximum solid hold-up
is over 0.52, solid sediment zones exist in the
vessel, which lie mainly behind and below the
agitator on the tank bottom, as well as the center
part of the bottom. This can be attributed to the
weak interaction between the liquid flow and the
tank bottom around these areas according to Fig.
3b. The cloud height is about 0.67H, which is
relative to the height of the main loop 0.7H in Fig.
3b, and few solid particles exist in the region
above this height. The increase of the impeller
speed makes the relative distribution of solid
Fig. 5 Particle volume fraction in X-Z plane for
N=400 rpm. concentration more uniform in the X-Z plane, and
at the same time leads to an increase of the cloud
the impeller high velocity region, solid particles height. The suspension height changes from
have little effect on the liquid axial velocity 0.67H to 0.85H as the speed N increases from 400
values around the impeller pumping flow, while to 600 rpm.
the axial flow near the vessel axis has been The solid distribution on the tank bottom obtained
greatly reduced. The magnitude of reduction from both the experiment and CFD simulation
increases as the distance to the vessel axis results are shown in Fig. 6 (s=2%). It can be seen

288
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

from the experimental photo in Fig. 6 that low on the particle and the gravitational force also
solid concentration areas exist below each exits. On the tank bottom, as the local solid
impeller when N=400 rpm. This is caused by the holdup and solid holdup gradient are relatively
interaction between the impeller discharge flow high, the theory of turbulent dispersion energy
and the tank bottom. However, a certain amount may play a more important role. The regions
of particles remains on the bottom, as this speed below and between the impellers are located
was far below Njs. The sediments mainly exist in between the collision zones formed by the
the zones below each agitator which lie on the impeller discharge flow and the tank bottom, in
opposite side of the horizontal angle , and strip which the flow rate and turbulent kinetic energy
zones between each of the two impellers are relatively weak, so the particles in these
interaction regions on the bottom. These regions regions are most difficult to be suspended.
are all low velocity zones as shown in Fig. 3d.
The area of the sediment zones decreases with the
increase of impeller speed. When N=500 rpm, the
strip zones between impellers disappear and the
sediment zones on the side of each impeller
shrink but still exist. As N increases to 600 rpm,
the solid sediment basically disappears and the
solid particles show relatively uniform
distribution on the bottom, except for some
unstable high solid hold-up zones along the
bottom edge formed by the tank wall and tank
bottom. As for the CFD results shown also in Fig.
6, the rough position and shape of the solid
sediment zones are fairly similar to the
Fig. 6 Particle distribution on bottom of vessel by
experimental ones. At the same impeller speed,
experiment and CFD.
the area of the sediment zones obtained by the
CFD simulation is slightly lower than that of 4.3 Solid suspension studies
experiment. Besides, the solid sediments along
the bottom edge formed by the tank wall and tank The critical impeller speed for just off-bottom
bottom in the experimental picture do not exist in suspension Njs is an important design parameter
the CFD results. This may be attributed to the used currently by engineers for the design and
limitation of standard wall function used in this scale-up of stirred vessels. Different criteria are
simulation. Ochieng and Lewis (2006) indicated available to evaluate the Njs. In this work, it is
that for low solid loading mixing (<5%) in the difficult to use Zwieterings criterion in the CFD
stirred tank, the solid suspension theory based on simulation to determine the critical impeller speed
the assumption that the particles are suspended for off-bottom suspension with the Eulerian-
when there is a balance between the force of the Eulerian approach used. Instead, the standard
bulk fluid acting on the particles and the deviation method proposed by Bohnet and
gravitational force was more reliable. While for Niesmak (1980) has been used, which is based on
high solid loading systems (>10%), the a standard deviation of the solid concentration
assumption that solids suspension results from the and more applicable for steady state simulations.
balance between the suspension potential energy The same methodology has successfully been
and turbulent kinetic energy of the flow is more employed by Khopkar et al. (2008) and Murthy et
suitable. It is possible that both of the two theories al. (2007). The suspension quality is quantified
work in the solid-liquid mixing system, and using the standard deviation written as
whichever one plays the major role depends on 2
the local solid holdup in the tank. In this work, as 1 n s
the global solid loading is relatively low (=2%),
= 1
n 1 avg
(14)

the solids suspension is mainly governed by the
bulk fluid flow. At the cloud height level in the where n denotes the number of sampling locations
upper region of the tank with low solid holdup, applied for measuring the solid volume fraction.
the axial flow rate is rather low and the balance The increase in the degree of uniformity is
between the upward axial velocity and the settling indicated by the decrease of the value. Based on
velocity of the particle makes them remain at this the suspension quality, the range of the standard
height, where the balance between the drag force deviation is divided into three ranges. For

289
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

homogeneous suspension, the value of the


standard deviation is lower than 0.2 (<0.2).
While for the just suspension condition, the value
of the standard deviation is between 0.2 and 0.8
(0.2<<0.8), and for the incomplete suspension,
>0.8 (Oshinowo and Bakker, 2002).
Khopkar et al. (2006 and 2008) found that the
impeller speed corresponding to =0.8 agreed
well with the critical impeller speed obtained
using the correlation proposed by Zwietering
(1958). In the present work, the standard
deviation was calculated from the values of the
solid volume fraction stored at all computational
elements using Eq. 14. The simulations of the
solid suspension (with a size 100 m and solid Fig. 7 Predicted values of standard deviation with
hold-up of 2% v/v) were carried out for impeller respect to impeller speed.
speeds ranging from 300 to 1000 rpm and the
corresponding standard deviations are shown in
Fig. 7. It can be seen that there is a sharp
reduction in the value as the impeller speed
comes close to Njs. The experimentally measured
value of Njs is 546 rpm. When the predicted value
of equals to 0.8, the corresponding impeller
speed is 531 rpm, which shows good agreement
with a deviation of about -2.75%. Therefore, the
critical impeller speed for off-bottom suspension
Njs in this calculation is considered to be reached
when the value of is 0.8. It can also be seen
from Fig. 7 that, with a further increase in the
impeller rotating speed, the value of decreases
quite slowly. The mixing system did not achieve
the homogeneous suspension even when the Fig. 8 Influence of solid loading on critical impeller
impeller speed was increased to 1000 rpm speed.
(=0.28). Therefore, it was difficult to achieve the
homogeneous suspension in the side-entering for off-bottom suspension well. The CFD
stirred vessel by means of merely increasing the predicted values of Njs are generally lower than
impeller speed. the experimental data, especially for the high
In top-entering stirred tanks with a low solid solid loading system. The critical impeller speed
loading, the critical impeller speed increases with increases with an increase of the solid loading,
an increase in the solid hold-up, the correlation and shows a linear relation in double logarithm
according to the Zwietering correlation being coordinates. Based on the experimental data, the
fitted correlation of this linear relation can be
N js 0.13 (15) represented by
Here, denotes the mass fraction of the solid. N js 0.25 (16)
The influence of the solid loading on the critical
impeller speed in the side-entering vessel was Compared with Eq. 15, the correlation between
shown in Fig. 8, with solid loadings of 1~5 v/v%, the critical impeller speed and the solid loading in
i.e., 2.5~13.2 wt%. It can be seen from Fig. 8 that the side-entering vessel is similar to that in the
good agreement exists between the CFD predicted top-entering vessel, but with a higher value of the
values and experimental values of Njs, with a exponent of the solid mass fraction . This
maximum deviation of 3.7%. Therefore, the CFD indicates that the solid loading in the side-
simulation method used in this work with the entering tank has a greater influence on the
Brucato drag model and standard deviation critical impeller speed than that in the top-
method could predict the critical impeller speed entering one.

290
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

4.4 Effect of some installation parameters on


critical impeller speed
Concerning the structure of a side-entering stirred
tank with multi-agitators, there are some
installation parameters which have influence on
the solid suspension, including the inclined angles
and , the plunging length and mounting height
of the shaft. The downward inclined angle and
horizontally inclined angle could make impacts
on the flow pattern and thus the solid suspension
in the tank. To study the effect of the inclined
angles of the impeller on the critical impeller
speed, simulations were carried out with various
inclined angles and . Fig. 9a shows the critical
(a) different inclined angles
impeller speed variations with different
horizontally inclined angles ranging from 3 to
10 when the downward inclined angle is fixed
at 10 (s=2%). The Njs initially decreases and
then increases with an increase of the inclined
angle and the minimum speed occurs when
=5. Moreover, the difference between the
maximum value of Njs when =10 and the
minimum one is about 15.1%. The main effect of
the horizontally inclined angle is to form a
circumferential loop around the axis of tank so as
to make each of the low velocity zones between
two impellers join the main loop and keep the
particles in these zones suspended. When the
angle is too small, the solid suspension (b) different lengths of shaft l
performance in the zones between two impellers
as well as near the tank wall would be poor.
While when the angle is too large, the particles
on the center of tank bottom would be easy to
deposit. Therefore, the horizontally inclined angle
opt should have an optimum value, which is 5
according to Fig. 9a. The critical impeller speeds
with different downward inclined angles in the
range of 5 to 14 when =opt is also shown in
Fig. 9a. The trend is similar to that of but the
variation amplitude is greater, with the deviation
of the maximum from minimum value being
about 24.7%. Since the impeller is an axial flow
type one, the downward inclined angle makes
the impeller discharge stream also have a (c) different mounting heights h
downward direction, which could make more Fig. 9 Critical impeller speeds for different
solid particles suspended from the tank bottom. installation parameters.
When the angle is too small, the suspension
condition below the impellers would be poor, inclined angle , the downward inclined angle
while when the angle is too large, the bulk has a greater effect on the critical impeller speed.
circulation height would be affected and the According to engineering experience, when the
global suspension height would be reduced. plunging length of the shaft l is too short, the
Hence, the optimum angle could improve both solid suspensions on the center of the tank bottom
the suspension on the bottom of the tank and the would be weak. While when it is too long, the
suspension height, which is 10 as shown in Fig. suspension around the tank bottom would become
9a. Furthermore, compared with the horizontally poor. Fig. 9b illustrates the critical impeller

291
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

speeds for different shaft lengths, with l/D zones by the CFD simulation are similar to the
extending from 0.7 to 1.2 (s=2%), when the experimental ones, with the CFD model slightly
inclined angles and are fixed at 10 and 5, under-predicting the sediment area value.
respectively. The optimum shaft length l should 3. The CFD predicted critical impeller speeds
be 0.9D, and the difference between the agreed well with the experimentally measured
maximum and the minimum speed is 29.7%. Fig. ones for different solid loadings, with a maximum
9c shows the critical impeller speeds for the deviation of 3.7%. Meanwhile, the predicted
mounting height h/D in the range of 0.7 to 1.2 values of Njs were generally lower than the
(s=2%), when =10, =5 and l=0.9D. As experimental data, especially for the high solid
shown in Fig. 9c, the mounting height h also has a loading system. The solid loading has a greater
significant effect on the critical impeller speed. influence on the critical impeller speed in the
High h/D ratio is particularly bad for the solid side-entering stirred tank than that in the top-
suspension, because it could make the collision entering tank.
between the impeller discharge flows and the tank 4. Compared with the horizontally inclined angle
bottom weaker. The optimum mounting height h , the downward inclined angle has a greater
is 1.2D, and the maximum speeds difference is effect on the critical impeller speed and the
34.6%. optimum angles opt and opt of the mixer shaft are
10 and 5 in this work, respectively. The
5. CONCLUSIONS plunging length l and mounting height h also have
significant effects on the critical impeller speed
The solid-liquid flow and solid suspension in a and the optimum l/D and h/D ratios are 0.9 and
cylindrical stirred tank equipped with four side- 1.2, respectively.
entering impellers were investigated by Despite some deficiencies, the computational
experimental and CFD simulation methods. model developed in the present study successfully
Experiments were carried out to characterize the captured the key features of the solid suspension
regularity of the solid sediment distribution on the in this side-entering stirred tank, and thus could
tank bottom and critical impeller speed for off- be useful for the design and optimization of
bottom suspension. The Eulerian-Eulerian two- stirred tanks with side-entering agitators.
phase model, standard k- turbulence model and However, a highly accurate measurement
the Brucato drag model have been used to technique for solid volume fraction is required in
perform the solid-liquid simulation using the order to obtain detailed information about the
software Fluent 6.3. Based on the validation by solid distribution in the side-entering stirred tank.
the experimentally measured results, the model Further work needs to be carried out to investigate
was further extended to discuss the influence of the influence of D/T ratio and tank scale on the
some installation parameters including the solid suspension in the side-entering stirred tank.
inclined angles and , the plunging length l and
mounting height h of the shaft on the critical
ACKNOWLEDGEMENTS
impeller speed. The key conclusions are listed as
follows:
The authors thank the Innovation Project of
1. Compared with the single phase system, the
Graduate Students of Jiangsu Province of China
introduction of solid particles resulted in an
(Grant No.CX09B_130Z) and Science-
attenuation in the axial liquid flow in the side-
Technology Project of Jiangsu Province of China
entering stirred tank. The coverage range of the
(Grant No.be2007076) for their support.
main circulation loop was reduced from the whole
tank to the suspension zone. A secondary
circulation loop was revealed to exist in the upper NOMENCLATURE
part of the tank, which was mainly
circumferential flow around the axis of the vessel. CD drag coefficient
2. The solid concentration decreased as the CD0 drag coefficient in stagnant liquid
distance to the tank bottom increased, showing a CTD coefficient for turbulent dispersion
ladder like distribution in the vertical direction. D impeller diameter, m
When the impeller speed was lower than Njs, the Dm diffusion coefficient
solid sediment mainly existed in the zones below dp diameter of solid particle, m
each agitator and between each of the two FB centrifugal and Coriolis forces, N/m3
impeller interaction areas on the tank bottom. The FI interfacial force, N/m3
rough position and shape of the solid sediment FD drag force, N/m3

292
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

FND non-drag force, N/m3 Engineering Applications of Computational


FTD turbulent dispersion force, N/m3 Fluid Mechanics 3(1):135-146.
FL lift force, N/m3 2. Alajbegovic A, Drew DA, Lahey RTJ (1999).
FVM virtual mass force, N/m3 An analysis of phase distribution and
Gkm turbulence generation, (kgm)/s3 turbulence structure in dispersed particle
g acceleration of gravity, m/s2 liquid flows. Chemical engineering
H tank height, m Communications 174(1):85-133.
h mounting height of agitator, m 3. Aubin J, Fletcher DF, Xuereb C (2004).
k turbulent kinetic energy, m2/s2 Modeling turbulent flow in stirred tanks with
l shaft length stretched into the tank, m CFD: the influence of the modeling approach,
N impeller rotational speed, /s turbulence model and numerical scheme.
Njs critical impeller speed, /s Experimental Thermal and Fluid Science
p pressure, Pa 28:431-445.
R impeller radius, m 4. Bohnet M, Niesmak G (1980). Distribution of
Re Reynolds number solids in stirred suspension. German
Re p Chemical Engineering 3(1):57-65.
particle Reynolds number
5. Brucato A, Grisafi F, Montante G (1998).
T diameter of stirred tank, m Particle drag coefficient in turbulent fluids.
U velocity, m/s Chemical Engineering Science 45:3295-3314.
Um mean velocity of the mixture, m/s 6. Derksen JJ (2003). Numerical simulation of
Utip impeller blade tip velocity, m/s solids suspension in a stirred tank. AIChE
Xv ratio of solid and liquid volume Journal 49(11):2700-2714.
fraction 7. Dskhel AA, Rahimi M (2004). CFD
Xvb the ratio of solid to liquid volume simulation of homogenization in large-scale
fraction in the solid sediment layer crude oil storage tanks. Journal of Petroleum
X, Y, Z coordinates Science and Engineering 43(3-4):151-161.
x, y, z coordinate values, m 8. Fang J, Ling X, Sang ZF (2011).
Experimental and numerical studies of the
Greek letters flow field in a stirred tank equipped with
multiple side-entering agitators. Chemical
volume fraction Engineering & Technology 34(10):1619-
horizontal inclination angle, 1629.
vertical inclination angle, 9. Ford C, Ein-Mozaffari F, Bennington CPJ,
standard deviation Taghipour F (2006). Simulation of mixing
turbulent energy dissipation rate, m2/s3 dynamics in agitated pulp stock chests using
dynamic viscosity, Pas CFD. AIChE Journal 52(10):3562-3569.
average dynamic viscosity of mixture, 10. Gomez C, Bennington CPJ, Taghipour F
av (2010). Investigation of the flow field in a
Pas
Kolmogorov length scale, m rectangular vessel equipped with a side-
density, kg/m3 entering agitator. Journal of Fluids
mass fraction Engineering, Transaction of the ASME
tm turbulent viscosity, Pas 132(5):0511061-05110613.
q ,ij stress sensor, N/m3 11. Guha D, Ramachandran PA, Dudukovic MP,
Derksen JJ (2008). Evaluation of large eddy
simulation and Euler-Euler CFD models for
Subscripts
solids flow dynamics in a stirred tank reactor.
AIChE Journal 54(3):766-778.
l liquid 12. Hicks MT, Mayers KJ, Bakker A (1997).
s solid Cloud height in solid suspension agitation.
q phase number Chemical Engineering Communications
160:137-155.
REFERENCES 13. Jirout T, Rieger F (2011). Impeller design for
mixing of suspensions. Chemical Engineering
1. Al-Qaessi F, Abu-Farah L (2009). Prediction Research and Design 89(7):1144-1151.
of mixing time for miscible liquids by CFD 14. Khopkar AR, Kasat GR, Pandit AB, Ranade
simulation in semi-batch and batch reactors. VV (2006). Computational fluid dynamics

293
Engineering Applications of Computational Fluid Mechanics Vol. 7, No. 2 (2013)

simulation of solid suspension in a stirred 26. Seyed H, Dineshkumar P, Farhad EM,


slurry reactor. Industrial and Engineering Mehhrab M (2010). Study of solid-liquid
Chemical Research 45(12):4416-4428. mixing in agitated tanks through
15. Khopkar AR, Kasat GR, Pandit AB, Ranade computational fluid dynamics modeling.
VV (2008). CFD simulation of liquid-phase Industrial and Engineering Chemical
mixing in solid-liquid stirred reactor. Research 49(9):4426-4435.
Chemical Engineering Science 63(15):3877- 27. Sharma RN, Shaikh AA (2003). Solid
3885. suspension in stirred tanks with pitched blade
16. Kipke K (1984). Suspension by side-entering turbines. Chemical Engineering Science
agitators. Chemical Engineering and 58(10):2123-2140.
Processing 18(4):233-238. 28. Wang LC, Zhang YF, Ling XG, Zhang Y
17. Ljungqvist M, Rasmuson A (2001). (2010). Experimental investigation and CFD
Numerical simulation of the two-phase flow simulation of liquid-solid-solid dispersion in a
in an axially stirred vessel. Chemical stirred reactor. Chemical Engineering Science
Engineering Research and Design 79(5):533- 65(20):5559-5572.
546. 29. Wesselingh JA (1975). Mixing of liquids in
18. Montante G, Maqelli F (2005). Modelling of cylindrical storage tanks with side-entering
solids distribution in stirred tanks: analysis of propellers. Chemical Engineering Science
simulation strategies and comparison with 30(8):973-981.
experimental data. International Journal of 30. Wu BX (2012). Computational fluid
Computational Fluid Dynamics 19(3):253- dynamics study large-scale mixing systems
262. with side-entering impellers. Engineering
19. Montante G, Micale G, Maqelli F, Brucato A Applications of Computational Fluid
(2001). Experiments and CFD predictions of Mechanics 6(1):123-133.
solid particle distribution in a vessel agitated 31. Wu J, Zhu Y, Pullum L (2001). Impeller
with four pitched blade turbines. Chemical geometry effect on velocity and solids
Engineering Research and Design suspension. Chemical Engineering Research
79(8):1005-1010. and Design 79(8):989-997.
20. Murthy BN, Ghadge RS, Joshi JB (2007). 32. Zwietering TN (1958). Suspending of solid
CFD simulation of gas-liquid-solid stirred particles in liquid by agitators. Chemical
reactor: prediction of critical impeller speed Engineering Science 8:244-253.
for solid suspension. Chemical Engineering
Science 62(24):7184-7195.
21. Ochieng A, Lewis AE (2006). Nickel solids
concentration distribution in a stirred tank.
Minerals Engineering 19(2):180-189.
22. Ochieng A, Onyango MS (2008). Drag
models, solids concentration and velocity
distribution in a stirred tank. Powder
Technology 181(1):1-8.
23. Oshinowo IM, Bakker A (2002). CFD
modeling of solid suspension in stirred tanks.
Symposium on Computational Modeling of
Metals, Minerals and Materials, TMS Annual
Meeting, Seattle, WA, 205-215.
24. Saeed S, Ein-Mozaffari F, Upreti SR (2007).
Using computational fluid dynamics
modeling and ultrasonic doppler velocimetry
to study pulp suspension mixing. Industrial
and Engineering Chemistry Research
46(7):2172-2179.
25. Sardeshpande MV, Sagi AR, Juvekar VA,
Ranade VV (2009). Solid suspension and
liquid phase mixing in solid-liquid stirred
tanks. Industrial and Engineering Chemical
Research 48(21):9713-9722.

294

You might also like