You are on page 1of 10

Chemical Engineering Science 62 (2007) 38393848

www.elsevier.com/locate/ces
Hollowself-inducing impellers: Flowvisualization and CFDsimulation
B.N. Murthy, N.A. Deshmukh, A.W. Patwardhan, J.B. Joshi

Department of Chemical Engineering, Institute of Chemical Technology, University of Mumbai, Matunga, Mumbai 400 019, India
Received 20 August 2006; received in revised form 20 February 2007; accepted 23 March 2007
Available online 13 April 2007
Abstract
The experimental uid dynamics (EFD) as well as computational uid dynamics (CFD) studies were performed for hollow self-inducing
agitator system. The system is described in detail elsewhere [Deshmukh, N.A., Patil, S.S., Joshi, J.B., 2006. Gas induction characteristics
of hollow self-inducing impeller. Transactions of the Institution of Chemical Engineers 84(A2), 124132]. The low pressure regions on the
impeller blade were determined by single phase CFD simulations. These results were validated with the experimentally measured power number
(N
P
). Then with the help of single phase CFD results, the orices were drilled on the impeller blade in low pressure regions. Gas induction
rate (Q
G
), power consumption (P) and overall fractional gas hold-up (
G
) were measured experimentally. The comparison of gas induction
rates for both the hole locations has been presented. The two phase CFD simulations yielded satisfactory predictions and proved to be most
promising for the design of efcient gas-inducing system.
2007 Elsevier Ltd. All rights reserved.
Keywords: Hollow self-inducing impeller; Gas induction rate; Pitched blade turbines; CFD
1. Introduction
Gas-inducing impellers are advantageous in situations where
internal recycle of unreacted gas is desirable. This condition
arises in a number of industrially important reactions such as
hydrogenation, alkylation, ethoxylation, ammonolysis, oxida-
tion with pure oxygen, hydrochlorination, etc. In these cases
the use of a gas-inducing impeller may be more benecial than
a recycle gas compressor for the reasons of safety, economy
and reliability. There are many types of gas-inducing impellers
reported in the literature, such as statorrotor type and hollow
impeller type. The present work is concerned with the hol-
low impellers and is in continuation with the earlier work of
Deshmukh et al. (2006).
When the impeller speed is zero, the level of liquid in hol-
low pipe and in the vessel is the same (in Deshmukh et al.,
2006, Fig. 1). As the impeller speed increases, at any point on
the impeller, pressure reduction occurs due to an increase in
the kinetic head according to Bernoullis equation. At critical

Corresponding author. Tel.: +91 22 2414 5616; fax: +91 22 2414 5614.
E-mail address: jbj@udct.org (J.B. Joshi).
0009-2509/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2007.03.043
impeller speed (N
CG
), the reduction in pressure at the outlet
of orice on each impeller blade is sufcient to overcome the
static head of liquid and the gas is just induced. This mecha-
nism is explained in detail elsewhere (Deshmukh et al., 2006).
The determination of gas induction rate (Q
G
) is an important
step in designing the hollow impeller system. The gas induction
rate depends on the pressure driving force (local pressure at
the oricehead space pressure) generated due to impeller ro-
tation. From the above discussion it is clear that the local pres-
sure at the orice is an important parameter that determines the
rate of gas induction. Once the gas starts to induce into the liq-
uid the local pressure eld itself is altered. This occurs mainly
due to the gasliquid interactions in the impeller zone. Hence
it was thought that it was desirable to estimate the pressure
eld in the gasliquid dispersion in the impeller region using
computational uid dynamics (CFD) simulation. This could
reduce the burden of experimentation considerably. Earlier the
attempts were made by various authors (Martin, 1972; Joshi
and Sharma, 1977; Joshi, 1980; Baczkiewicz and Michalski,
1988; Rielly et al., 1992; Forrester and Rielly, 1994; Rigby
et al., 1994; Mundale and Joshi, 1995) to increase gas induction
rate by (i) putting holes at low pressure region, (ii) minimizing
the pressure drop associated with gas ow and (iii) maximizing
3840 B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848
Fig. 1. Experimental set-up with impeller having holes on lower edge of each impeller blade.
the pressure differential by changing shape and/or design of the
impeller. In the present work, for the rst time, CFD support
has been employed to identify the location of low pressure
region in order to achieve maximum gas induction rate under
identical operating conditions. For gas-induction, Raidoo et al.
(1987) and Mundale and Joshi (1995) have shown that the
pitched blade downow turbine (PBTD) design is superior to
disc turbine as well as other axial ow impellers. Hence in this
case PBTD impeller was selected for further investigations.
In CFD, the ow eld is estimated by solution of the trans-
port equations along with suitable turbulence models. Joshi
et al. (1999) initiated the efforts for hollow impeller CFD sim-
ulations. However, full potential of CFD has still remained
unutilized for developing a systematic procedure for optimum
design of self-inducing impellers. The objective of this paper
is to (i) locate the low pressure region on the impeller blade
in single phase ow through CFD simulations, (ii) validate the
outcome with experimental results and (iii) study the effect of
presence of gas on location and magnitude of low pressure
zone. Also an attempt has been made to calculate the gas in-
duction rate (Q
G
) from local pressure at the orice, obtained
from two phase CFD simulations.
2. Experimental setup
The tank geometry employed in this work was a at-
bottomed cylindrical tank, (T =H =0.50 m) with four equally
spaced bafes (b = T/10) and it is the same as that used for
experimental work by Deshmukh et al. (2006). Six bladed 30

,
45

and 60

hollow PBTDs with D=T/2 diameter were used.


The impeller is located at a distance of T/3 from the base of
the tank. The shaft and the impeller were hollow. The inter-
nal diameter of shaft was 20 mm. In self-induction mode the
rate of gas induction (Q
G
) was measured by a precalibrated
turbine type anemometer. The impeller speed was measured
by magnetic proximity probe. Power drawn by the impeller
was estimated by measuring the torque exerted by the uid on
the frictionless torque table. The torque table was restrained
from rotating, and the force was measured by connecting it
to a cantilever type load cell. Multiplication of this force and
radius of the torque table (R
T
) gives torque acting on the table
(M=wgR
T
). Power drawn by the impeller was calculated by
using the measured value of torque (P =2NM) (Fig. 1).
3. CFD model
For the estimation of ow pattern and the pressure eld,
CFD simulations were performed. In the present case, three-
dimensional simulations have been carried out for the both sin-
gle phase and gasliquid multiphase system. Multiple reference
frame (MRF) model has been used to simulate the interaction
between the rotating impeller and the stationary bafes. In or-
der to reduce computational efforts and to keep accuracy, the
B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848 3841
MRF seems to be highly justiable (Kerdouss et al., 2006).
In single phase simulations turbulence was modeled using the
standard kc model. This model is essentially a high Reynolds
number model and assumes the existence of isotropic turbu-
lence and the spectral equilibrium. Details of the ow govern-
ing equations are discussed elsewhere (Sahu and Joshi, 1995;
Sahu et al., 1998).
3.1. Two phase simulations
Gasliquid ow was modeled using the two-uid Eulerian
Eulerian approach. This model solves continuity and momen-
tum equations for each phase separately. The coupling of the
equations is achieved through the pressure and interphase ex-
change coefcients.
In FLUENT, the derivation of the conservation equations for
mass and momentum for each phase is done by phase weighted
Favre-averaging (Viollet and Simonin, 1994) the local instan-
taneous balances for each of the phases, and then no addi-
tional turbulent dispersion term is introduced into the continuity
equation.
The mass conservation for L phase is written as follows:
j
jt
(
L
j
L
) +.(
L
j
L

t
L
) =0. (1)
The momentum conservation equation for the phase L after
averaging is written as follows:
j
jt
(
L
j
L

t
L
) +.(
L
j
L

t
L

t
L
)
=
L
p +.t
L
+(j +j
t,L
).

t
L
+
L
j
L

g
+

R
GL
+

F
D
+

F
L
, (2)
where t
L
is the Lth phase stressstrain tensor
t
L
=
L
j
L
(

t
L
+

t
T
L
) +
L
(z
L

2
3
j
L
).

t
L
I. (3)
Here j
L
and z
L
are the shear and bulk viscosity of phase
L,

F
L
represents the Coriolis and centrifugal forces ap-
plied in the rotating reference frame,

R
GL
is an interaction
force between phases, and p is the pressure shared by all
phases.
In the present work the standard kc model for single phase
ows has been extended for two phase owwith extra terms that
include interphase turbulent momentumtransfer (Elgobashi and
Abou-arab, 1983) to take into account the effects of turbulence.
The present model is the most general multiphase turbulence
model which solves a set of k and c transport equations for each
phase. This turbulence model is the appropriate choice when
the turbulence transfer among the phases plays a dominant role.
The turbulent viscosity (j
t,L
) was computed at each point in
the ow via the solution of the following governing equations
for the k and c:
j
jt
(
L
j
L
k
L
) +.(
L
j
L
k
L

t
L
)
=.

j +
j
t,L
o
k

k
L

+(
L
G
k,L

L
j
L
c
L
)
+K
GL
(C
GL
k
G
C
LG
k
L
) K
GL
(

t
G

t
L
)
.
j
t,G

G
o
GL

G
+K
GL
(

t
G

t
L
).
j
t,L

L
o
GL

L
, (4)
j
jt
(
L
j
L
c
L
) +.(
L
j
L
c
L

t
L
)
=.

j +
j
t,L
o
k

c
L

+
c
L
k
L

(C
c1

L
G
k,L
C
c2

L
j
L
c
L
) +C
c3

K
GL
(C
GL
k
G
C
LG
k
L
) K
GL
(

t
G

t
L
)
.
j
t,G

G
o
GL

G
+K
GL
(

t
G

t
L
)
.
j
t,L

L
o
GL

.
(5)
The turbulent viscosity is then related to k and c by the expres-
sion
j
t,L
=j
L
C
j
k
2
L
c
L
. (6)
The terms C
GL
and C
LG
can be approximated as
C
GL
=2, C
LG
=2

p
GL
1 +p
GL

, (7)
where p
GL
is dened as the ratio between the Lagrangian in-
tegral time scale and the characteristic particle relaxation time
(FLUENT 6.2, 2005). G
k,L
is the production of turbulent ki-
netic energy and c
L
is the turbulent kinetic energy dissipation
rate. Standard and default values of all model parameters were
used. The kc model constants C
j
, C
c1
, C
c2
, o
c
and o
k
were
set as 0.09, 1.44, 1.92, 1.3 and 1.0, respectively (Launder and
Spalding, 1972).
3.2. Interfacial momentum exchange
Eq. (2) needs to be closed with appropriate expressions for
the interphase forces. The most important interphase force is
the turbulent drag force acting on the bubble resulting from the
mean relative velocity between the two phases and an additional
contribution resulting from turbulent uctuation in the volume
fractions due to averaging of the momentum equation. In the
present simulations, the drag force was modeled through Morsi
and Alexander model (FLUENT 6.2, 2005). Following are the
details:
f =
C
D
Re
24
,
3842 B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848
where
C
D
=a
1
+
a
2
Re
+
a
3
Re
3
, (8)
Re =
j
L
|

t
L

t
G
|d
b
j
L
,
a
1
, a
2
, a
3
=

0, 24, 0, 0 <Re <0.1,


3.690, 22.73, 0.0903, 0.1 <Re <1,
1.222, 29.26, 3.89, 1 <Re <10,
0.617, 46.50, 116.67, 10 <Re <100,
0.364, 98.33, 2778, 100 <Re <1000,
0.357, 148.62, 47, 500, 1000 <Re <5000,
0.46, 490.546, 578, 700, 5000 <Re <10, 000,
0.52, 1662.5, 5, 416, 700, Re10, 000.
Turbulent uctuations in the volume fraction have been
modeled using the dispersion force term:

F
D
=K
GL

j
t,G
o
GL

j
t,L
o
GL

, (9)
where j
t,G
and j
t,L
are diffusivities, and o
GL
is a dispersion
Prandtl number. Here the diffusivities j
t,G
and j
t,L
are com-
puted from the transport equations and the default value for
o
GL
is 0.75.
Other forces such as lift and added mass force have not been
included in the present simulation because of the low velocity
gradients in the bulk region (Aubin et al., 2004; Montante et al.,
2006). They have carried out detailed PIVmeasurements for the
gasliquid ows generated by axial ow impeller (PBTD45)
and radial ow impeller (Rushton turbine) in bafed stirred
vessel, respectively, and it has been observed that the liquid
velocity proles tend to be more or less at in the bulk region.
However, in the impeller region, velocity and pressure gradients
are appreciable. Khopkar et al. (2005) have shown that, in the
impeller region, the drag force dominates the bubbles motion.
In their work, an order of magnitude analysis indicated that
the magnitude of the lift force and added mass force are much
smaller than the interphase drag force. Further work on these
aspects is in progress.
3.3. Computational geometry, grid and method of solution
In the present study, entire geometry has been consid-
ered for the simulations. For all the simulations, the bound-
ary of the rotating domain was positioned at r = 0.16 and
0.10 mz0.24 m. Tetrahedral elements were used for mesh-
ing the geometry and a good quality of mesh was ensured
throughout the computational domain using the GAMBIT
mesh generation tool.
As regards to the particularly mesh quality, we have been
restricted to use tetra mesh element due to complex geom-
etry. However, in this study a very high quality of mesh
(skewness <0.7) has been ensured throughout the computa-
tional domain. The number of grid elements in all the three
directions in both the impeller and outer zone were systemati-
cally increased. When rening the mesh, care was taken to put
most additional mesh element in the regions of high gradient
around the blades and discharge region. The effect of grid
Fig. 2. (The grid independence study) Pressure contours for both the single
and two phase simulations at an impeller speed of 7 rps.
independence has been shown in Figs. 2A and B. It can be
seen that the predicted pressure distribution is practically the
same for grid elements of 550,000667,800. Further, for these
two numbers of grids, the values of power number and ow
number were also found to be constant.
Simulations were carried out for the PBTD impeller of 60

blade angle for various combinations of speeds of agitation


(4.17, 5, 6 and 7 rps). Each impeller blade had six holes of 3 mm
diameter through which gas is induced into the system. The
tank walls, the impeller surfaces and bafes have been treated
as no-slip boundary condition with standard wall functions. The
gas ow rate at the hole is initially specied as pressure outlet
boundary condition and gas ow rate has been calculated and
which was subsequently used for the two phase simulations
and a detailed discussion has been given in Section 4.2.2. The
bubble size distribution in the stirred tank reactor depends on
many design parameters. Unfortunately, available experimental
data of bubble size distribution in the present case is not avail-
able. Hence, the mean bubble size of 3 mm has been used for
all the simulations. At a liquid surface, a small gas zone was
added at the free surface of water, a method that has been re-
ported to dampen instabilities (FLUENT 6.2, 2005) and only
gas is allowed to escape using pressure outlet boundary con-
dition which means top surface being exposed to atmospheric
pressure. All terms of the governing equations are discretized
using the second-order upwind differencing and the rst-order
implicit scheme for the time integration. The SIMPLE algo-
rithm has been employed for the pressurevelocity coupling.
The convergence criterion (sum of normalized residuals) was
set at 510
4
for all the equations. Unsteady simulations have
been performed for the ow time of 8 s with a time step size of
B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848 3843
0.001 s. Every time step took around 25 internal iterations to get
converged. All the simulations have been carried on AMD64
dual processors with clock speed of 2.4 GH and 2 GB memory.
Total simulation time for each was around 120 h. All the simu-
lations were conducted using the commercially available CFD
software FLUENT 6.2.
4. Results and discussions
4.1. Single phase ows
In this section validation of power number and mean quali-
tative ow pattern has been presented. Further, validated sim-
ulations have been used for the identication of low pressure
region.
4.1.1. Validation for N
P
The power consumption P is calculated as the product of
torque on the impeller blades and the angular velocity. This is
then used for the estimation of power number and it can be
expressed as follows:
N
P
=
2NM
jN
3
D
5
, (10)
where torque (M) exerted on all blades was computed from
the total momentum vector which is computed by summing
the cross products of the pressure and viscous force vectors for
each facet on the impeller with the moment vector.
The power number (N
P
) calculated by Eq. (10) for single
phase simulations was found to be N
P
= 5.2. This compares
well with the experimental value of N
P
=5.1.
4.1.2. Flow eld
The stirred tank of given geometry was simulated for both
single and gasliquid multiphase ows for different impeller
rotational speeds (4.17, 5, 6 and 7 rps). Fig. 3 shows the vector
ow eld at the impeller rotational speed of 7 rps. It can be
seen that this impeller pumps the liquid in the outward direction
through the vertical periphery of the impeller, which is contrary
to the ow patterns generated by the axial ow impeller with
blade angels in the range 30

60

. Patwardhan and Joshi (1999)


and Aubin et al. (2004) have studied ow pattern generated by
PBTD45 with D/T ratio 0.5. For this purpose, they have found
two circulation cells, one is primary. Asmall secondary cell was
also observed below the impeller. Therefore, for PBTD60

also
one can expect two clear circulation loops which are shown in
Fig. 3. A primary circulation loop exists just below the impeller
plane and extends up to almost
2
3
of the vessel. In the upper
part of the tank, the liquid velocities are low and circulation is
poor (Fig. 2).
4.1.3. Pressure eld
Figs. 4AD show the contours of static pressure on impeller
blades for various speeds of operation for single phase ow. As
shown in Fig. 4 it was observed that for all the impeller speeds,
the low pressure region is formed at the upper edge of the rear
face of the blade and not at the lower edge of the blade. It has
Fig. 3. Axial velocity (ms
1
) ow eld for liquid phase in the mid-plane
between two bafes.
been found that the static pressure decreases with an increase
in the impeller speed which should lead to an increase in the
gas induction rate. It also becomes clear that the pressure at the
lower face of the impeller is higher than that at rear face.
4.2. Two phase ows
It is envisaged from the single phase simulations that the
gas induction rate can be enhanced if holes are placed in the
low pressure zone (rear face of the blade near the upper edge).
Therefore, the impeller design was altered by drilling holes in
the low pressure region which is shown in Fig. 5. In order to
assess the gas-induction performance, power number and gas
induction rate were measured experimentally and compared
against previous impeller design with bottom holes. Further,
CFD simulations have been performed for the same system
(i) to see the change in the magnitude of low pressure in the
presence of gas, (ii) verify that whether the low pressure region
remains at the same location even in the presence of gas and
(iii) to develop a predictive procedure for the gas-induction rate
using the values of local pressure.
4.2.1. Validation for N
P
In case of two phase simulations the experimentally mea-
sured and predicted values of gassed power numbers are given
in Table 1. It is evident that, in the case of two phase ows, CFD
simulations tend to underpredict the measured power number,
and it may be due to complex hydrodynamics at higher gas
hold-up values in the given system.
4.2.2. Two phase ow eld
In the case of gas induction, the liquid velocity vectors show
that the same qualitative ow pattern as in case of single phase
ow remains but, the magnitude is relatively less which is
shown in Fig. 6. This is attributed to the formation of a gas
3844 B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848
Fig. 4. Contour plots of static pressure (Pascal) on impeller blade for PBTD 60

(A: N =4 rps, B: N =5 rps, C: N =6 rps, D: N =7 rps).


Fig. 5. Modied impeller geometry.
Table 1
Experimental and simulated values of power number and gas hold-up for two
phase simulations
Impeller speed (s
1
) Power number (N
PG
)
Experimental Simulated
4.17 4.1 3.3
5 4.0 3.2
6 3.67 2.8
7 3.4 2.5
cavity behind the impeller which effectively reduces the power
given to the liquid. The at, inclined blade of the PBT leads
to the formation of a low pressure trailing vortex at the rear of
the blade.
4.2.3. Gas induction rate
The single phase CFD simulations have revealed that
the low pressure zone is at the upper edge of rear face of the
impeller blade (new location) and not at the lower face of the
blade. Hence the hole of similar diameter and spacing as those
at lower edge were drilled at new locations and experiments
were performed. The rate of gas-induction was measured for
three impeller blade angles (viz. 30

, 45

and 60

) and various
speed of operation and it has been compared against impeller
design where holes are on the bottom edge. The experimental
results have been shown in Fig. 7. It can be seen that the gas
induction rate for the new location is 100200% higher than
that when the holes were drilled on the bottom edge. Thus, the
identication of low pressure regions by CFD has enabled a
substantial improvement in the performance of gas-inducing
impeller.
Further, an attempt has been made to predict the gas induc-
tion rate using CFD simulations to minimize cumbersome and
expensive experimentation. The present work adopted an itera-
tive method for calculating the gas induction rate. This method
involves carrying out the CFD simulations for the two indepen-
dent geometries i.e., single phase simulations for hollow shaft
and impeller and two phase simulations for the stirred tank.
The following procedure was adopted:
(i) The pressure drop in the gas pathway for a given gas ow
rate was estimated as follows:
(a) Pressure loss (P
G
), due to induced gas ow in the
gas phase involves frictional pressure loss along the gas
B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848 3845
Fig. 6. Axial velocity (ms
1
) vectors for liquid in presence of gas in the
mid-plane between two bafes.
8
6
4
2
0 R
A
T
E

O
F

G
A
S

I
N
D
U
C
T
I
O
N
>

Q
0

x

1
0
4

(
m
3
/
s
)
4 5 6 7 8 9 10
IMPELLER SPEED, N (rps)
Fig. 7. Comparison of rate of gas induction (m
3
s
1
), (: PBTD 30

(lower
edge), : PBTD 45

(lower edge), : PBTD 60

(lower edge), : PBTD 30

(new location), : PBTD 45

(new location), : PBTD 60

(new location).
pathway through the shaft, blade and pressure drop across
the gas orice on the blade surface. For this purpose, the
following equation was used:
P
G
=C
G
1
2
j

Q
G
A
0

2
. (11)
For the estimation of C
G
a separate set of experiments
was performed in which the measured quantity of gas
(by turbine anemometry, Q
G
) was introduced at the inlet
(Fig. 1) and the pressure drop was measured. In the ab-
sence of liquid in the vessel the pressure drop was de-
noted by P
G
. A data set of about 10 measurements of
10
8
6
4
2
0
R
A
T
E

G
A
S

I
N
D
U
C
T
I
O
N
.

Q
0

x

1
0
-
3

(
m
3
/
s
)
3 4 5 6 7 8
IMPELLER SPEED, N (rps)
Fig. 8. Comparison of experimental and predicted values of rate of gas
induction (m
3
s
1
), experimental, CFD predictions.
P
G
, Q
G
enables the estimation of C
G
in Eq. (11) and
was found to be 3 10
05
.
(ii) In the presence of liquid, the pressure drop (P) consists
of the following components:
P =P
G
+P
L
+ P
H
, (12)
where P
H
is the hydrostatic head of liquid above the
orice and P
L
is the pressure drop required for creating
gasliquid interface (at the orice outlet) and for imparting
energy to the liquid phase. P
L
was correlated by the
following equation:
P
L
=C
L
Q
A
G

H
S
A
0

B
(2Nr)
C
. (13)
The values of C
L
, A, B, and C were estimated from the
data set of P versus Q
G
and were found to be 0.093,
0.12, 0.29, and 1.89, respectively.
The pressure drop in the gas path and at the orice outlet
(P
T
) is given by
P
T
=P
G
+P
L
. (14)
(iii) The pressure driving force is the pressure difference be-
tween the gas inlet (P
i
, point A in Fig. 1) and the pressure
at the outlet of the orice (P

). The value of P

was es-
timated by CFD using the following stepwise procedure:
(a) For a given hole location, P

was obtained from CFD


solution assuming that the gas phase is absent and the
impeller moves only in liquid. The (P
i
P

) becomes
the pressure driving force, P
T
.
(b) For P
T
obtained in step (a) the gas ow rate was esti-
mated using Eqs. (11)(14).
(c) For the gas ow rate and the hole location in step (b), the
two phase CFD simulation gives improved value of P

and hence P
T
.
(d) The steps (b) and (c) were repeated till the driving and
the resistive forces become equal.
3846 B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848
Fig. 9. Contour plots of static pressure (Pascal) of mixture on impeller blade for PBTD 60

(A: N=4 rps, Q


G
=0.001 m
3
s
1
, B: N=5 rps, Q
G
=0.003 m
3
s
1
,
C: N =6 rps, Q
G
=0.0055 m
3
s
1
, D: N =7 rps, Q
G
=0.008 m
3
s
1
).
Fig. 10. Contours of gas volume fraction (N =250 rpm, Q
G
=0.001 m
3
s
1
)
in the mid-plane between two bafes.
The predicted gas ow rate has been compared with the ex-
perimental value of gas induction rate for PBTD60

impeller
for all operating speeds in Fig. 8. For further details of the above
expressions, reader may refer Deshmukh et al. (2006). In the
present work, the pressure driving force has been estimated us-
ing CFD. The comparison is excellent at low speeds (250 and
300 rpm). At higher speeds (above 5 rps) the present iterative
technique based on CFD slightly underpredicts the gas induc-
tion rate. Overall the comparison is fairly good considering the
complexity of two phase ow eld in the vicinity of impeller.
Fig. 11. Contours of gas volume fraction (N =250 rpm, Q
G
=0.001 m
3
s
1
)
in the horizontal impeller center plane.
4.2.4. Pressure eld
After the commencement of gas induction, pressure eld gets
modied which is depicted in the form of pressure contours in
Fig. 9 AD. It can be seen that the low pressure region was
still located near the top edge of the each impeller blade. This
result is very important because the location of holes can be
estimated by only single phase simulations and the complex and
expensive two phase simulations need not be performed in a
majority of the cases. The present modication of placing holes
near the top edge of the impeller rear face enabled enhancement
in the induction rate.
B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848 3847
Table 2
Experimental and predicted values of overall gas hold-up
Impeller speed (s
1
) % Gas hold-up (
G
)
Experimental Predicted
4.17 7.5 8.5
5 13.21 15.14
6 15.1 17.28
7 24.2 27.3
4.3. Gas hold-up
In addition to gas induction, its dispersion in the liquid
also plays an important role in the performance of the reactor.
Figs. 10 and 11 show (for N =250 rpm, induced gas ow rate
of 0.001 m
3
s
1
) the computed gas volume fraction in a verti-
cal plane containing the impeller and in the horizontal impeller
center plane, respectively. It can be observed from the gures
that the gas induced through the holes follows impeller dis-
charge stream. The ow pattern shows the gas accumulation in
the recirculating ow regions both above and below impeller
center plane. Further, in the low pressure region behind the im-
peller blades gas tends to accumulate by forming the so-called
gas cavities and the gas leaving the orices was entering the
cavity directly without discrete bubbling or jetting. It is clear
from Fig. 10 that the high gas fractions are encountered in the
near shaft zone, where liquid ows downwards and drags the
gas. Fig. 11 shows the accumulation of gas behind the bafes
due to recirculation of liquid. The total gas hold-up predicted
by the present CFD simulations is shown in Table 2. A satis-
factory agreement can be seen between the simulated overall
gas hold-ups and the experimentally observed values.
5. Conclusions
(1) In the present work, an optimization procedure has been
suggested for the rate of gas induction using CFD.
(2) For the maximization of the gas induction rate, the outlet
holes on the impeller need to be located in the low pres-
sure region. The present investigation gives a systematic
procedure for the identication of low pressure region.
(3) The location of low pressure regions (as estimated from
CFD) was found to be the same using single and two phase
simulations. This important result is useful because the
complex and expensive two phase simulations need not
be performed as the single phase simulations sufce the
objective of optimization.
(4) The CFD predictions were compared with the experimen-
tal values of gassed power number and overall gas hold
up. It has been observed that the gassed power number
was under predicted at all the above mentioned operating
conditions. Whereas, the overall hold-up values have been
over predicted.
(5) CFD simulations for single phase and multiphase ows
have been extended to calculate the gas inductions rate it-
eratively. The predicted gas induction rates were found to
be in excellent agreement with the experimentally mea-
sured values.
Notation
a
1
, a
2
, a
3
draglaw constants
A empirical constant
A
0
cross sectional area of orice, m
2
b bafe width, m
B empirical constant
C
j
, C
c1
, C
c2
, C
c3
turbulence model constants
C empirical constant
C
D
drag coefcient
C
G
constant
C
L
constant
C
P
pressure coefcient
d
b
bubble diameter, m
D impeller diameter, m
f drag function

F volumetric force, Nm
3
g acceleration due to gravity, ms
2
G
k
production of turbulent kinetic energy,
m
2
s
3
H liquid height, m
I unit tensor
k turbulent kinetic energy, m
2
s
2
K exchange coefcient, kg m
3
s
1
M torque, Nm
N impeller rotation speed, s
1
N
CG
critical impeller speed for onset of gas in-
duction, s
1
N
P
impeller power number
N
PG
gassed impeller power number
p pressure, Nm
2
P power consumption,W
P
i
pressure at shaft inlet, Nm
2
P

pressure at the outlet of the orice, Nm


2
P
G
pressure loss in gas phase, Nm
2
P
L
pressure loss in liquid phase, Nm
2
P
D
total pressure driving force generated by
liquid ow, Nm
2
P
T
total pressure loss associated with gas
ow, Nm
2
Q
G
rate of gas induction, m
3
s
1
r orice distance from the impeller axis, m

R interphase force, Nm
3
Re Reynolds number
R
T
radius of the torque table, m
t time, s
T tank diameter, m
t average velocity, ms
1
V volume of liquid, m
3
w weight of the load cell, kg
z axial coordinate, m
Greek letters
volume fraction
c turbulent energy dissipated per unit mass,
m
2
s
3
3848 B.N. Murthy et al. / Chemical Engineering Science 62 (2007) 38393848
z bulk viscosity, kg m
1
s
1
j viscosity, kg m
1
s
1
j density of uid, kg m
3
o
k
turbulent Prandtl number for the turbulent ki-
netic energy
o
c
turbulent Prandtl number for the dissipation rate
o
GL
dispersion Prandtl number
t stress tensor, kg m
1
s
2
Subscripts
G gas phase
L liquid phase
t turbulent
Acknowledgment
B.N. Murthy and N.A. Deshmukh gratefully acknowledge
the nancial support during this work by Department of Atomic
Energy (DAE), and Board of Research in Nuclear Sciences
(BRNS), Government of India, respectively.
References
Aubin, J., Le Sauza, N., Bertrand, J., Fletcher, D.F., Xuereb, C., 2004. PIV
measurements of ow in an aerated tank stirred by a down-and an up-
pumping axial ow impeller. Experimental Thermal and Fluid Science 28
(5), 447456.
Baczkiewicz, J., Michalski, M., 1988. Oxygen transfer during mixing of acetic
acid fermentation medium of self aspirating tube agitator. In: Proceedings
of the Seventh European Conference on Mixing, Pavia, Italy, pp. 2426.
Deshmukh, N.A., Patil, S.S., Joshi, J.B., 2006. Gas induction characteristics of
hollow self-inducing impeller. Transactions of the Institution of Chemical
Engineers 84 (A2), 124132.
Elgobashi, S.E., Abou-arab, T.W., 1983. A two-equation turbulence model
for two-phase ows. Physics of Fluids 26, 931938.
FLUENT 6.2, 2005. Users Manual to FLUENT 6.2. Fluent Inc., Lebanon,
USA.
Forrester, S.F., Rielly, C.D., 1994. Modeling the increased gas capacity of
self-inducing impellers. Chemical Engineering Science 49, 57095718.
Joshi, J.B., 1980. Modications in the design of gas inducing impellers.
Chemical Engineering Communications 5, 109114.
Joshi, J.B., Sharma, M.M., 1977. Mass transfer and hydrodynamic
characteristics of gas inducing type of agitated contactors. Canadian Journal
of Chemical Engineering 65, 683695.
Joshi, J.B., Patwardhan A.W., Patil S.S., 1999. CFD modeling for the design
of hollow pipe gas-inducing impellers. In: Proceedings of the Third
International Symposium on Mixing in Industrial Processes, ISMIP3,
Osaka, Japan, pp. 297304.
Kerdouss, F., Bannari, A., Proulx, P., 2006. CFD modeling of gas dispersion
and bubble size in a double turbine stirred tank. Chemical Engineering
Science 61, 33133322.
Khopkar, A.R., Rammohan, A.R., Ranade, V.V., Dudukovic, M.P., 2005.
Gasliquid ow generated by a Rushton turbine in stirred vessel:
CARPT/CT measurement and CFD simulations. Chemical Engineering
Science 60, 22152229.
Launder, B.E., Spalding, D.B., 1972. Lectures in Mathematical Model of
Turbulence. Academic Press, London, England.
Martin, G.Q., 1972. Gas inducing agitator. Industrial and Engineering
Chemistry Process Design and Development 11, 397404.
Montante, G., Ghadge, R.S., Magelli, F., Paglianti, A., 2006. PIV
measurements of gas and liquid ow in aerated stirred vessels. In:
Proceedings of the 10th International Conference on Multiphase Flow in
Industrial Plant, Tropea (VV), Italy, pp. 421430.
Mundale, V.D., Joshi, J.B., 1995. Optimization of impeller design for gas-
inducing type mechanically agitated contactors. Canadian Journal of
Chemical Engineering 73, 161172.
Patwardhan, A.W., Joshi, J.B., 1999. Relation between ow pattern and
blending in stirred tanks. Industrial and Engineering Chemistry Research
38, 31313143.
Raidoo, A.D., Raghavrao, K.S.M.S., Sawant, S.B., Joshi, J.B., 1987.
Improvements in the gas inducing impeller design. Chemical Engineering
Communication 39, 4366.
Rielly, C.D., Evans, G.M., Davidson, J.F., Carpenter, K.J., 1992. Effect of
vessel scale-up on the hydrodynamics of a self aerating concave blade
impeller. Chemical Engineering Science 47, 33953402.
Rigby, G.D., Evans, G.M., Jameson, G.J., 1994. Inuence of uid pressure
eld on gas ow rate for a gas-inducing impeller. In: Proceedings of the
Eighth European Conference on Mixing, BHRA, Cambridge, England, pp.
180194.
Sahu, A.K., Joshi, J.B., 1995. Simulations of ow in stirred vessels with axial
ow impellers: effects of various numerical schemes and turbulence model
parameters. Industrial and Engineering Chemistry Research 34, 626639.
Sahu, A., Kumar, P., Joshi, J.B., 1998. Simulation of ow in stirred vessel
with axial ow impeller: zonal modeling and optimization of parameters.
Industrial and Engineering Chemistry Research 37, 21162130.
Viollet, P.L., Simonin, O., 1994. Modelling dispersed two-phase ows:
closure, validation and software development. Applied Mechanical Reviews
47 (6), S80S84.

You might also like