Sie sind auf Seite 1von 16

Chemical Engineering Science 61 (2006) 7535 7550

www.elsevier.com/locate/ces

Measurement of phase holdups in liquidliquidsolid three-phase stirred


tanks and CFD simulation
Feng Wang1 , Zai-Sha Mao , Yuefa Wang, Chao Yang
Institute of Process Engineering, Chinese Academy of Sciences, Beijing 100080, China
Received 9 November 2005; received in revised form 3 August 2006; accepted 13 August 2006
Available online 5 September 2006

Abstract
Phase holdup is an important hydrodynamic characteristic of multiphase systems relevant to optimization and scale-up of related process
equipment. In the present article, measurements of phase distribution of solid particles and oil droplets are conducted in a lab-scale stirred
tank by sample withdrawal under various operating conditions. A EulerianEulerian three-uid model is established for the prediction of phase
distribution of two dispersed phases in the agitated liquidliquidsolid dispersion system. The turbulence structure in the system is described
by an extension of the standard k. turbulence model to three-phase ow including the inuence of presence of two dispersed phases as an
additional source of turbulent kinetic energy. Momentum exchange between continuous and dispersed phase as well as between the two dispersed
phases are incorporated into the model formulation. Comparison of model predictions with experimental data suggests reasonable agreement
for the dispersed oil phase. The predicted distribution of solid particles shows some discrepancies in comparison with the measurements, but
the agreement is signicantly improved for higher impeller speeds.
2006 Elsevier Ltd. All rights reserved.
Keywords: Stirred tank; Three-phase ow; Turbulence; Numerical simulation

1. Introduction
Liquidliquidsolid three-phase stirred tanks are common in
process industry. Typical applications include reactive occulation and solid catalyzed liquidliquid reaction, etc. The knowledge of the hydrodynamic characteristics, such as suspension
of solid particles, dispersion of dispersed liquid phase and their
spatial distribution in the stirred tank, is essential in determination of rates of heat/mass transfer and desired chemical reactions, and consequently of vital importance for the reliable
design and scale-up of such chemical reactors.
Design and scale-up of multiphase stirred tanks are mainly
based on empirical and semi-empirical correlations gained from
experimental data so far. Extrapolating use of those empirical
correlations beyond the original operating conditions is highly
Corresponding author. Tel.: +86 10 6255 4558; fax: +86 10 6255 1822.

E-mail addresses: wangfeng@hqcec.com (F. Wang),


zsmao@home.ipe.ac.cn (Z.-S. Mao).
1 Present address: Process & System Department, China HuanQiu Contracting & Engineering Corp., Beijing 100029, China.
0009-2509/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2006.08.046

risky. The strategy of stagewise scale-up is costly and timeconsuming, and the satisfactory scale-up to large-scale reactor
is not guaranteed.
Scientic and reliable design of a liquidliquidsolid reactor requires full understanding of the hydrodynamics and transport properties in such multiphase systems, including the phase
holdups and spatial distributions. This demands sufcient experimental and theoretical investigations, but up to date little
has been reported in the open literature.
Among three-phase systems in stirred tanks, the gasliquid
solid system is the most investigated one. Chapman et al.
(1983ac) investigated the effects of the presence of solid particle on gasliquid hydrodynamics as well as the aeration on
the suspension of particles. More recently, Dohi et al. (2004)
performed experimental measurements on the power consumption and solid suspension in gasliquidsolid stirred tanks.
However, other hydrodynamic characteristics, in particular
the distribution of two dispersed phases in three-phase stirred
tanks, are scarcely reported in the literature.
The computational uid dynamics (CFD) approach has
attracted intensive attention in recent years for its powerful

7536

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

capacity for understanding the physical reality of multiphase


ows. Nevertheless, CFD simulation of the hydrodynamics in
stirred tanks is very difcult because the ow eld in the tank
is complex and highly unsteady due to the interaction of the
rotating impeller with multiphase dispersion. Most of the literature on numerical simulation of ow eld in stirred tanks is
limited to single-phase and two-phase ows. When it comes to
three-phase ow, the interactions between two dispersed phases
and the contribution of dispersed phases on the turbulence of
the continuous phase make the numerical solution of the governing equations more challenging.
A few papers have been published to date on the numerical prediction of three-phase ow, mostly focused on
gasliquidsolid three-phase ow in the bubble column reactors, e.g. Krishna et al. (2000), Michele and Hempel (2002),
Mitra-Majumdar et al. (1997, 1998), Padial et al. (2000). In
general, the EulerianEulerian multiuid approach was employed unanimously to formulate the turbulent gasliquidsolid
three-phase ows. The k. two-equation turbulence model was
adopted to represent the turbulence of the continuous phase.
For modeling the momentum exchange between continuous
and dispersed phases as well as between two dispersed phases,
very limited information is available in the literature and no
convergent opinion on the expression of the inter-phase interaction has been reached, yet. To our best knowledge, no
three-dimensional numerical simulation of liquidliquidsolid
three-phase ow eld in stirred tanks has been reported to date.
In addition, experimental data is also scarce for the purpose
of validation of numerical simulation of liquidliquidsolid
three-phase ow.
The objective of the present study is to provide experimental
measurements of axial and radial variations of phase holdups of
two dispersed phases in a stirred tank under different operating
conditions. Also a computational approach is established based
on the EulerianEulerian multi-uid approach to describe the
motion of each phase, and then to predict the phase distribution
of two dispersed phases in the three-phase stirred tank by numerical solution of Reynolds time-averaged mass and momentum conservation equations (RANS). The experimental data
obtained in this work are compared with the model predictions
to validate the computational approach. Preliminary effort of
this work has been reported previously in a CFD conference
presentation (Wang et al., 2005).
2. Experiments
Experiments are carried out in a cylindrical, at-bottomed
stirred tank made of Perspex with the inner diameter T =
0.154 m, equipped with four equally spaced bafes with the
width B = 0.1T . The impeller is a six at blade disc (Rushton) turbine with the diameter D = 0.062 m. The conguration of the impeller and the stirred tank are shown in Fig. 1. It
is noted that a relatively large impeller (D = 0.4T ) is chosen
in this study in comparison to the standard conguration with
D = T /3, in order to provide good suspension and dispersion
of the two dispersed phases. The clearance between the center plane of the impeller disc and the tank bottom is set to be

b
B
w

D
d
D
T

Fig. 1. Congurations of stirred tank and Rushton impeller: T = 0.154 m,


H = T , B = 0.1T , D = 0.062 m, d = 0.75D, b = 0.25D, w = 0.2D.

7
6
5

2
1

Fig. 2. Experimental setup. 1. Stirred tank, 2. impeller, 3. motor, 4. sampling


tubes, 5. graduated cylinder, 6. peristaltic pump, 7. valve.

C = T /3. Nine copper sampling tubes with 4.5 mm inner diameter ( 6.0 mm outside diameter) are mounted midway between two bafes vertically along the tank wall from the bottom
to the free liquid surface (z = 0.1H 0.9H ). Sample tubes can
reach the tank interior from r/R = 0.1 to 0.9, if not prohibited
by the sweeping impeller blades.
The experimental setup is illustrated in Fig. 2. The stirred
tank is charged with tap water as continuous phase and n-hexane
as oil phase. Glass beads with a density of s =2550 kg m3 and
a mean diameter of ds = 110 m are the solid phase. The total
height of three-phase mixture is always equal to the diameter of
the tank (T =H ). The impeller speed is measured using a digital
laser tachometer and is accurate to 5 rpm. In all experimental
runs, 1 h is needed to achieve a steady state of dispersion in the
present investigation, although Armenante and Huang (1992)
claimed that the equilibration period was only 1015 min for agitated liquidliquid dispersion. To avoid the entrainment of gas
into the liquidliquidsolid system, the impeller speed tested
is between N = 300 and 500 rpm, which is still below that the
critical speed for full suspension of solid particles.
A peristaltic pump is activated after the steady dispersion
state is achieved and the dispersion is partly circulated through

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

the sample tube back to the tank (to ensure the representative
samples of tank holdup being collected). Then, the ow is directed to the graduated cylinder and about 15 ml of three-phase
dispersion is collected. Since there is signicant difference in
volatility between oil and water phases, the sample is evaporized at 80 C after centrifuged and weighted to determine the
amount of oil phase, given the lighter phase (n-hexane) which
boils at 78 C under atmospheric pressure and always acts as
dispersed phase in all experimental runs. The remained sample (water and glass beads) is dried in the oven and weighted
to obtain the local holdup of solid phase. The total quantity of
the samples in each run is about 3% of the total system and the
bulk ow is deemed to not being changed signicantly by the
sample withdrawal. Adequate amounts of water, oil and glass
beads are added into the tank to keep the individual phase inventory of the system at the original values. The reproducibility
of the measurements is about 3.3%. The same procedure is repeated for various axial and radial positions, impeller speeds,
and phase fractions of two dispersed phases.

3. Mathematical model
In the present study, the mathematical model is formulated
based on the EulerianEulerian multi-uid model. Water, oil
and solid phases are all treated as different continua, interpenetrating and interacting with each other everywhere in the
domain under consideration. The oil and solid phases are in
the form of spherical-dispersed droplets and particles, respectively. The effect of breakup and coalescence of droplets is ignored. The pressure eld is presumed to be shared by three
phases, which are exerted, respectively, by the pressure gradient in proportion to their volume fraction. Motion of each phase
is governed by respective mass and momentum conservation
equations.
The RANS version of the governing equations for threephase ow is derived from averaging the instantaneous conservation equations of physical variables over the space or time.
In the present work, the time averaging method of Ishii (1975)
is followed. Reynolds averaging procedure leads to the mean
ow equations by decomposing the instantaneous variables into
a mean value and a uctuating component since the multiphase
ow in stirred tanks is usually turbulent. The resulted mass
conservation equation for phase k is written as
j
j
j
( k ukj ) =
(  u )
( k ) +
jt k
jxj k
jxj k k kj

(1)

7537







ju
ju
juki jukj
j
kj
ki

k  k
+ k k
+
+
+
jxj
jxj
jxj
jxi
jxi
2 j

3 jxj

jukj
jukj
k k ij
+ k ij k
jxj
jxj

+ k k gi + Fki .

(2)

The uctuating pressure P  and the uctuating inter-phase


momentum exchange term Fki as well as other third-order terms
are neglected in time-averaging process. The viscous shear
stress is also omitted as compared with large magnitude of turbulent shear terms.
Several turbulent uctuation correlation terms with second
order appear in Eqs. (1) and (2), namely the Reynolds stresses,
are modeled by invoking the Boussinesq gradient transport hypothesis:


jukj
2
juki
uki ukj = kt
+ kij ,
+
(3)
jxj
3
jxi
where kt is the turbulent kinematic viscosity.
The correlation of the uctuating velocity and uctuating
holdup, k uki appearing in both mass and momentum conservation equations, represents the transport of mass and momentum by dispersion. The simplest method to evaluate the correlation is to assume gradient transport as (Elaghobashi and
Abou-Arab, 1983):
k uki =

kt jk
,
t jxi

(4)

where t is the turbulent Schmidt number for phase dispersion, which is set to 1.0 in the present work after preliminary
numerical trials.
The inter-phase momentum exchange term Fki represents
the interaction between phases, which is composed of a linear combination of different momentum exchange mechanisms:
the drag force, the added mass force and the lift force, etc.
Only the contribution of drag force is taken into account in this
study since other forces are testied to be less signicant for
both the liquidliquid and liquidsolid ows in stirred tanks
(Ljungqvist and Rasmuson, 2001; Wang and Mao, 2005). The
drag force between continuous phase and two dispersed phases
is expressed following the practice of two-phase ows:

Fci,drag =
Cdrag (udi uci ),
(5)
k

where Cdrag is the momentum exchange coefcient:


and the momentum conservation for phase k reads
j
j
(k k uki + k k uki ) +
( k uki ukj )
jt
jxj k
= k

j
jP

( k uki ukj + k uki k ukj + k ukj k uki


jxi
jxj k

+ k k uki ukj )

Cdrag =

3c c d CD |ud uc |
.
4dd

(6)

For determination of the drag coefcient CD , the following


correlation is used as Clift et al. (1978) recommended:
24

(1 + 0.15Re0.687
), Red 1000,
d
CD = Red

0.44,
Red > 1000,

7538

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

where
Red =

Here, G is the turbulent kinetic energy production term given


by

c dd |ud uc |
.
c,lam

Different from two-phase ows, there are still interactions between dispersed droplets and solid particles in
liquidliquidsolid three-phase ows, which have to be taken
into account as well. However, this factor has not been modeled
so far in the literature. In the computational model developed
by Padial et al. (2000) for gasliquidsolid three-phase bubble
column, the drag between solid particles and gas bubbles was
modeled identically to drag between liquid and gas bubbles
based on the notion that particles in the vicinity of gas bubbles
tend to follow the liquid, while in the model used by Michele
and Hempel (2002), the momentum exchange terms between
the dispersed gas and solid phases are expressed as
Fgs,i = Fsg,i =

3g s cgs |us ug |(usi ugi )


4dp

(7)

The combination of cgs |us ug | was dened as a tting parameter, whose value was determined by tting model predictions to measured local solid holdups.
Since two dispersed phases are assumed to be continua as
mentioned above, it is reasonable to expect the drag between
solid particles and droplets behaving in a similar way as the
drag force between the continuous and the dispersed phase:
Fos,drag,i = Fso,drag,i
=

3o o s CD,os |us uo |(usi uoi )


.
4ds

(8)

The modeling of a multiphase turbulent ow is basically an


unresolved problem as predicting the hydrodynamic characteristics of multiphase ow is concerned, mainly because the inuence of the presence of dispersed phases on turbulence of the
continuous phase is difcult to model. In the context of CFD
modeling, it is assumed that the turbulence in multiphase stirred
tanks is dominated by the continuous phase and in multiphase
ows it is more often described using an extension of the widely
accepted standard k. two-equation turbulence model in analogy to the single-phase one. The effects of dispersed phases on
turbulence structure are thought to generate additional turbulent kinetic energy. In this work, the conservation equations for
the turbulent kinetic energy of the continuous phase is written
as follows:
j
j
(c c uci k)
(c c k) +
jt
jxi


j
ct jk
=
c
+ c [(G + Ge ) c ]
jxi
k jxi

juci
jxj

(11)

and Ge is an extra production source term representing the inuence of the dispersed phase, which is taken to be proportional to the production of drag force and slip velocity between
two phases as Kataoka and Serizawa (1989) suggested:

Ge = Cb |Fdrag |

(udi uci )2

1/2

(12)

where Cb is an empirical constant with the value of 0.02 in this


study.
The model parameter C1 for the impeller region is modied
as follows to describe the effect of strongly swirling ow (Zhou,
1993):
C1 = 1.44 +

0.8Rf c 
G + Ge

(13)

with


1
j  u 



ur u r
,
jr r
Rf = 

0,

1.5w < z < 1.5w,

(14)

z < 1.5w, z > 1.5w.

The values of other model parameters are taken as common


practice in single-phase ow (Launder and Spalding, 1974):
C = 0.09,

C2 = 1.92,

k = 1.0,

 = 1.3.

The turbulent viscosity of the continuous phase, ct is expressed as


ct =

C  c k 2
.


(15)

Also the turbulent viscosity of dispersed phases is estimated


based on its counterpart of the continuous phase via a factor K
(Grienberger and Hofmann, 1992):
dt = Kct

(16)

with

(9)

K=

d udi udi
c uci uci

Gosman et al. (1992) introduced a correlation coefcient to


relate the uctuating velocities of the dispersed phase to those
of the continuous phase by the expression:

and the turbulent energy dissipation rate is expressed as




j
 j
j
j
(c c ) +
c ct
(c c uci ) =
jt
jxi
jxi
k jxi

+ c [C1 (G + Ge ) C2 c ].
k

G = c uci ucj

(10)




t1
udi = uci 1 exp
,
td

(17)

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

7539

where t1 = 0.41k/ is the mean eddy lifetime and td is the


particle response time given by
td =

4d dd
.
3c CD d |ud uc |

It is well known that ow in the stirred tank is always unsteady due to interaction of the rotating impeller blades with the
stationary wall bafes. However, the ow pattern will become
axisymmetrically repeating once it is fully developed. Ranade
and van den Akker (1994) suggested to ignore the time derivatives in the governing equations without introducing much error in most part in the tank except in the impeller swept volume. A snapshot of the ow can describe the ow within the
impeller region at a particular instant. In the present work, the
ow eld in the impeller swept volume is simulated in a noninertial reference frame rotating with the impeller, where ow
may be steady and the time-dependent terms disappear. Thus,
the resulting formulation of the mass and momentum conservation equations for phase k in general form in a cylindrical
coordinate system reads
 1 j 


1 j 
j 
rk k ukr +
 k uk  +
 k ukz
r jr
r j k
jz k




j
1 j
1 j k
eff j
=
rk
eff
+
r jr
jr
r j
r
j


j
j
+
k
eff
(18)
+ S ,
jz
jz
where

eff =
lam +
t .
For the complete set of conservation equations and corresponding source terms it may be referred to the counterparts
for the two-phase turbulent ow as tabulated in the authors
previous papers (Wang and Mao, 2002; Wang et al., 2004).
It is still necessary to specify the laminar viscosity of the
solid phase in the momentum conservation equations. In the
present study, the solid laminar viscosity is set equal to that of
the continuous phase as s,lam =103 Pa s. This is considered as
a reasonable approximation since test calculations showed that
variation of solid laminar viscosity between 101 and 104 Pa s
do not yield signicant difference in the simulation results. This
was also validated by the examination of Michele and Hempel
(2002).
It is worthy to note that the computational model developed
in this study does not involve any adjustable parameters to
match the model predictions with the experimental data, and
all relevant constants are cited from the literature as is.
4. Numerical procedure
The phase distributions of two dispersed phases is numerically predicted as a part of the solution of three-dimensional
turbulent ow in a liquidliquidsolid three-phase stirred tank

Fig. 3. The computational grid of inner and outer domains.

under the ow conditions in coincidence with the experiment


in the present study. The computational domain is chosen as a
half of the tank since the time-averaged ow eld is periodical
repeating in the azimuthal direction. The computational grid
adopted for the simulations is depicted in Fig. 3, which consists of 36 36 90 elements in radial, azimuthal and axial
directions, respectively. The action of impeller is modeled using a modied innerouter iterative procedure that has been
detailed in Wang and Mao (2002) and is not described here. Its
main advantage is that the calculated ow parameters on the
surface of the inner and outer regions are not averaged in
the present procedure, differing from the original innerouter
approach developed by Brucato et al. (1998), to preserve the
pseudo-periodical turbulent properties.
The governing equations of mass and momentum of each
phase together with the turbulent kinetic energy and energy
dissipation rate transport equations of the continuous phase
listed above are solved by a nite volume technique (Patankar,
1980). The discretization of the equations is implemented using
a power-law differencing scheme in a staggered grid system.
The most important concern in the numerical procedure is
the manner in which the pressure eld is computed and its effect on the continuity. The extensively used SIMPLE algorithm
(Patankar, 1980) is adopted here to solve the pressure eld,
although other modied algorithms like SIMPLER (Patankar,
1980), SIMPLEST (Van Doormal and Raithby, 1984) and SIMPLEC (Spalding, 1980) give faster convergence than the SIMPLE algorithm at the cost of increased computational load as
suggested by Ekambara and Joshi (2003).
The pressure eld shared by three phases is obtained using
a pressure-correction formula derived by combining the three
continuity equations together after being normalized with the

7540

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

respective phase density as Carver (1984) suggested:


 
 
 
ap
ap
ap
+
P
+
 c
 o
 s p




 anb 
anb
anb
=
+
+
P
 c
 o
 s nb
nb
 
 
 
f
f
f
+
+
+
 c
 o
 s
 
 
 
D
D
D
+
+
+
,
 c
 o
 s

while for the continuity equation,



|b|
,
R(m) =
0.5c DHT

where 0.5c DHT is a reference liquid mass ux in proportion to the pumping capacity of the impeller. The solution is
deemed convergent when the relative residuals of the continuity equation is less than 104 and for others less than 106 .
The boundary conditions are:
(19)

where fk is the turbulent diffusive term due to the asymmetry


of the mass ow, and Dk is the mass imbalance term indicating
the extent that the continuity equation is not satised since the
pressure eld is updated gradually from the initial guess in the
SIMPLE algorithm.
For the numerical simulation of liquidliquidsolid threephase ow, at least two continuity equations have to be solved to
obtain the spatial distribution of the phase holdup in the whole
ow domain. In this article, the continuity equations of the
two dispersed phases are modied by subtracting the continuity
equation of the continuous phase and using the relation c =
1 o s . This gives


 Foi
Fci
o
+
o
c
i
out








Foi
Fci
Fci
o
=
+
(1 s )
+
o
c
c
i
i
in
out


 
 

Fci
f
f

(1 s )
+

(20)
c
 o
 c
i

in

and similarly,


 Fsi
Fci
s
+
s
c
i
out








Fsi
Fci
Fci
s
=
+
+
(1 o )
s
c
c
i
i
in
out







Fci
f
f

(1 o )
+

.
(21)
c
 s
 c
i

in

(23)

(1) Symmetry axis (r=0): ucr =uor =usr =0, uc =uo =us  =0,
j /jr = 0 (  = ucr , uor , usr , uc , uo , us  ).
(2) Free surface: The free surface of the stirred tank is assumed
to be at, then ucz = uoz = usz = 0, j /jz = 0 (  =
ucz , uoz , usz ).
(3) Solid surface: The wall function is applied to all solid
surface of the stirred tank, including the tank wall, bottom,
and the impeller.
5. Results of measurement and modeling
5.1. Experimental results of phase distribution
For liquidliquidsolid three-phase ow in stirred tanks, it
is desired to gain information on the state of dispersion of the
oil phase and solid particles for diagnosis and optimization of
operation of such reactors. Effects of impeller speed on the axial and radial variation of local holdup of solid and oil phases
are investigated. The present measurement results normalized
with the respective phase volume fraction are presented in
Fig. 4 for the system with phase volume fractions of oil and
solid phases being o,av = 0.10 and s,av = 0.10. It can be seen
that, at relative lower impeller speed (see Fig. 4(a)), the local holdup of the solid phase shows a maximum near the tank
bottom, while for the dispersed oil phase, the maximum local holdup appears at the top of the tank. It implies that both
solid and oil phases are not sufciently dispersed at such impeller speed. Increasing the impeller speed will signicantly
promote the dispersion of the oil phase, as seen from Fig. 4(b).
It is also observed that the local holdup of the oil phase below the impeller plane is larger than those at the upper section,
which can be attributed to the fact that some droplets adhere
to the surface of solid particles as a result of better wettability
to oil, and thus move downwards under gravity. This indicates
that the introduction of a solid phase will affect the dispersion characteristics of the oil phase. However, the variation of
impeller speed has only marginal effect on the suspension of
solid particles within the speed range employed in the present
study.
The axial distribution of solid and oil phases is shown in
Fig. 5 for o,av = 0.30 and s,av = 0.10. Experiments show that

The trial calculation shows faster convergence in this manner as compared with directly solving the original continuity
equation of the two dispersed phases.
The sets of discretized equations for physical variables are
solved iteratively through an ADI algorithm. The non-linearity
in the phase momentum and turbulence equations is tackled
with suitable under-relaxation. The convergence of the solution is assessed by relative residuals R( ). For momentum and
turbulence transport equations

|ap p (aE E + aW W + aN N + aS S + aT T + aB B + b)|
R( ) = 
[|ap p | + |aE E | + |aW W | + |aN N | + |aS S | + |aT T | + |aB B | + |b|]

(22)

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

1.0

7541

1.0
0.2(r/R)
0.3

0.8

0.8

0.5
0.7
0.9

0.6
z/H

z/H

0.6

0.4

0.2(r/R)

0.4

0.3
0.2

0.5
0.7

0.2

0.9
0.0

0.0
1

o/o,av

s/s,av

(a)
1.0

1.0
0.2(r/R)
0.3

0.8

0.8

0.5
0.7
0.9

0.6
z/H

z/H

0.6

0.4

0.4

0.2(r/R)
0.3
0.5

0.2

0.2

0.7
0.9

0.0

0.0
0

(b)

s/s,av

0.0

0.5

1.0

1.5

2.0

2.5

3.0

o/o,av

Fig. 4. Axial proles of normalized holdup of solid and oil phases at different impeller speed (s,av = 0.10 and o,av = 0.10). Left: solid; right: oil.
(a) N = 300 rpm, (b) N = 500 rpm.

the increase of volume fraction of the oil phase yields more


homogeneous distribution of solid particles, especially at
high impeller speeds, presumably because the adherence
of droplets to solid particles gives more buoyancy to solid
particles.
An increase in the amount of solid particles tends to enhance the homogeneity of the oil phase dispersion, as shown
in Fig. 6. The axial distribution of the oil phase becomes more
uniform at s,av = 0.10, compared to the experimental results
at s,av = 0.05.
The variation of phase holdups of two dispersed phases with
inuencing factors is more easily shown by the phase holdup
contour maps drawn from the experimental data across a vertical section mid-way between two wall bafes with the reso-

lution of 10 mm 20 mm (Figs. 79, impeller shaft at the left


side is not marked).

5.2. Model prediction of mean ow eld and phase distribution


Numerical simulations are conducted to examine the hydrodynamic characteristics in the liquidliquidsolid three-phase
stirred tank under the ow conditions identical to the present
experiment. The model prediction of the mean velocity elds
for the continuous and the dispersed phases are presented in
Fig. 10 as velocity vector plots corresponding to different impeller speeds. The phase volume fractions of oil and
solid phases are all 0.10. The well-documented ow pattern

7542

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

1.0

1.0
0.2(r/R)
0.3

0.8

0.8

0.5
0.7
0.9

0.6
z/H

z/H

0.6

0.4

0.4

0.2 (r/R)
0.3
0.5

0.2

0.2

0.7
0.9

0.0
0.0

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.5

1.0

1.5

2.0

2.5

3.0

o/o,av

s/s,av

(a)
1.0

1.0
0.2(r/R)
0.3

0.8

0.8

0.5
0.7
0.6

0.9
z/H

z/H

0.6

0.4

0.4

0.2(r/R)
0.3
0.5

0.2

0.2

0.7
0.9

0.0

0.0
0.0
(b)

0.5

1.0

1.5

2.0

2.5

3.0

s/s,av

0.0

0.5

1.0

1.5

2.0

2.5

3.0

o/o,av

Fig. 5. Effects of the volume fraction of oil phase on the axial proles of normalized local holdup of solid and oil phases at different impeller speeds
(s,av = 0.10 and o,av = 0.30). Left: solid; right: oil. (a) N = 300 rpm, (b) N = 500 rpm.

generated by a disc turbine in stirred tanks is clearly illustrated,


that is, two large ring vortices exist, respectively, above and
below the impeller plane, and a high-velocity radial impeller
stream is also predicted. Overall, the ow elds of three phases
are very similar to each other in most parts of the domain.
The velocity eld of the oil phase shows a trend of drifting
upwards to the top of the tank at lower impeller speed. This
might be attributed to the fact that the oil phase with lower
density accumulates easily in the top section of the tank. The
time-averaged ow eld of the solid phase reveals a small
recirculation zone above the center of the tank bottom, indicating that the solid particles tend to settle down in this zone.

With the increase of impeller speed, the mean velocity ow


elds of three phases become more consistent.
The calculated normalized local holdups of two dispersed
phases are illustrated using the contour plots as shown in
Fig. 11. It is easy to observe that the distributions of oil and
solid phases are all less homogeneous at low impeller speeds
(N = 300 rpm). The maximum oil phase holdup is located
in the center of the free surface due to the ring eddy in the
upper bulk zone, in qualitative agreement with the experimental observations. The distribution of all phases becomes
more uniform at higher impeller speed. The maximum solid
concentration occurs at the center of the tank bottom due to

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

1.0

7543

1.0
0.2(r/R)
0.3
0.5
0.7

0.8

0.8

0.9

0.6
0.2(r/R)
0.3
0.5
0.7
0.9

z/H

z/H

0.6

0.4

0.4

0.2

0.2

0.0

0.0
0.0

0.4

0.8

1.2

1.6

2.0

0.0

0.5

s/s,av

(a)

1.0

1.5

2.0

2.5

3.0

o/o,av
1.0

1.0
0.2(r/R)
0.3

0.8

0.8

0.5
0.7
0.6

0.6

z/H

z/H

0.9

0.2(r/R)
0.3

0.4

0.4

0.5
0.7

0.0

0.0
0
(b)

0.9

0.2

0.2

s/s,av

0.0

0.4

0.8

1.2

1.6

2.0

o/o,av

Fig. 6. Effects of the volume fraction of solid phase on the axial proles of solid and oil phases normalized local holdups at different impeller speeds
(s,av = 0.05 and o,av = 0.10). Left: solid; right: oil. (a) N = 300 rpm, (b) N = 500 rpm.

the density difference and the ring eddy at the bottom, which
is conrmed by the circulation ow depicted in the velocity
vector plot.
The numerically computed axial proles of dispersed oil
and solid phases are compared in Fig. 12 with the present
measurements. The simulation results are generally in reasonable agreement with the experimental data, especially for
higher impeller speeds. The comparison indicates that the
computational approach adopted here is suitable for predicting the dispersed phase distribution in the liquidliquidsolid
three-phase stirred tank. The model prediction for the solid
phase, however, is notably above the experimental data especially at low impeller speeds. With the increase of N , the
agreement improves slightly. The discrepancy between the
prediction and the experimental results is probably because

the model of inter-phase momentum exchange employed in


the present work is too simple to describe the real complex inter-phase interaction coupling in liquidliquidsolid
three-phase ows. Additionally, the isotropic k. twoequation turbulence model is decient in describing the
well-recognized anisotropic nature of turbulent ow in stirred
tanks.
Fig. 13 shows the comparison between the simulated axial
proles of the holdup of two dispersed phases and the experimental results obtained in the present work with s,av = 0.05
and o,av = 0.10. Also, the numerical prediction agrees reasonably well with the experimental data. Decreasing the amount
of solid particles leads to more accurate quantitative agreement
with the experimental results, as can be seen from Figs. 3 and
4. A possible reason may be that the lower volume fraction of

7544

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

0.15

0.15
4.50

4.003.50

4.00

0.12

0.12

3.50

3.00
1.50

3.50
2.50

2.00

2.50
2.00

z (m)

z (m)

3.00
0.09

1.50

0.06

0.09

0.06

1.00
0.03

1.00
0.50

0.50

2.00 1.50 1.00 0.50


3.50 2.50

0.03

0.00
0.00 0.02 0.04 0.06

0.00
0.00 0.02 0.04 0.06

r (m)

r (m)

(a)

0.15

0.15

0.09

0.30
0.60

0.06

1.80
3.30

1.20 1.50

0.60

2.70 3.00
2.10
0.90

3.50
3.30
3.00
2.70
2.40
2.10
1.80
1.50
1.20
0.90
0.60
0.30
0

1.05
1.20

0.12

0.09

1.00
1.00

0.06
1.05

0.03
1.10
1.15

1.30
1.25
1.20
1.15
1.10
1.05
1.00
0.95
0.90
0.85
0.80

0.00
0.00 0.02 0.04 0.06

0.00
0.00 0.02 0.04 0.06

r (m)

r (m)

(b)

1.00
0.95
0.95

z (m)

z (m)

0.12

0.03

5.00
4.50
4.00
3.50
3.00
2.50
2.00
1.50
1.00
0.50
0

Fig. 7. Experimental results of solid and oil phase normalized holdup at different impeller speeds (s,av = 0.10 and o,av = 0.10). Left: solid; right: oil.
(a) N = 300 rpm, (b) N = 500 rpm.

0.15

0.15

0.09

0.06

0.03

0.50
2.00 1.00
1.50
4.50 3.50
2.50
4.00

0.00
0.00 0.02 0.04 0.06
r (m)

1.15
0.12

1.10

1.05
1.00
1.05

z [m]

z (m)

0.12

6.00
5.50
5.00
4.50
4.00
3.50
3.00
2.50
2.00
1.50
1.00
0.50
0

0.09

1.00

1.05
1.00
0.95

0.06

0.90
0.95
1.00

0.03
0.90

0.85
0.80

0.00
0.00 0.02 0.04 0.06
r [m]

Fig. 8. Experimental results of solid and oil phase normalized holdup at N = 500 rpm (s,av = 0.10 and o,av = 0.30). Left: solid; right: oil.

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550


0.15

0.15

0.30

z (m)

0.60

0.09
0.60

0.06

0.03

1.80

0.30

0.90

3.002.10
1.50 1.20

0.00
0.00 0.02 0.04 0.06
r (m)

3.50
3.30
3.00
2.70
2.40
2.10
1.80
1.50
1.20
0.90
0.60
0.30
0

1.05

1.20
1.19

0.12

1.12
z (m)

0.12

7545

0.09

1.05

1.12
1.05
0.98

1.12

0.98
1.05

0.06

0.84
0.91
0.98

0.03

0.91

0.77

0.77

0.84
1.12

0.70

0.00
0.00 0.02 0.04 0.06
r (m)

Fig. 9. Experimental results of solid and oil phase normalized holdup at N = 500 rpm (s,av = 0.05 and o,av = 0.10). Left: solid; right: oil.

Fig. 10. The velocity vector plots of the continuous, solid and oil phases. Impeller shaft at the left side is not marked. Left: continuous; middle: solid; right:
oil. (a) N = 300 rpm, (b) N = 500 rpm.

7546

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

0.15

0.15

1.50

4.50

4.50
0.5

0.12

4.00

0.12

3.50

4.00
1.00

3.50
3.00

2.50
2.00

z (m)

z (m)

3.00
0.09

1.50

0.06
1.00

1.00
0.50

0.03

0
0.50

0.00
0.00 0.02 0.04 0.06
r (m)

0.00
0.00 0.02 0.04 0.06
r (m)

0.15

0.15

0.60

0.09
1.00
0.80
1.00
1.20
1.40

0.03
1.20

1.60

2.40
2.20
2.00
1.80
1.60
1.40
1.20
1.00
0.80
0.60
0.40
0.20

1.60
1.55
0.12

1.40

1.40
0.80

z (m)

0.60
0.80

z (m)

1.50

1.50

1.00

2.00
2.50
3.00

0.06

2.50
2.00

0.06

0.50

0.03

0.12

1.00

1.00

1.50

(a)

0.09

1.25
1.10

0.09

1.100.9

0.95
0.80

1.25

0.06
0.65

0.03

0.95
0.80

0.65
0.50
0.35

0.95

0.20
0.80

1.80

0.00
0.00 0.02 0.04 0.06
(b)

r (m)

0.00 0.65
0.00 0.02 0.04 0.06
r (m)

Fig. 11. Model predicted contour plots of normalized holdup of solid and oil phases (s,av = 0.10 and o,av = 0.10). Impeller shaft at the left side not shown.
Left: solid; right: oil. (a) N = 300 rpm, (b) N = 500 rpm.

solid phase makes the complete suspension of solid particles


easier and the interaction between solid particles and droplets
becomes less important.
Fig. 14 shows the model predicted holdup axial proles corresponding to the oil and solid phase with o,av = 0.30 and
s,av = 0.10. The numerical simulation reproduces quite well
the general trend of phase distribution of measurements.

6. Conclusions
Experimental measurement and numerical simulation are
conducted to investigate the phase distribution of dispersed
phases in the liquidliquidsolid three-phase stirred tank. The
sample withdrawal method is employed in the experimental
study. It is found that the presence of solid particles has significant inuence on the liquidliquid dispersion, and vice versa,
the introduction of the lighter oil phase eases the suspension

of solid particles. In addition, the measured local holdup of


the oil phase at high impeller speed below the impeller plane
is larger than that in the upper section of the tank. This is
due to adherence of the oil droplets to the surface of solid
particles as a result of better wetting. Wettability is a unique
factor of dispersion of the liquidliquidsolid three-phase
system.
A three-uid model is developed based on the Eulerian
Eulerian approach for numerical simulation of a liquidliquid
solid three-phase ow. The inter-phase momentum exchange
between continuous and dispersed phases as well as between
two dispersed phases is modeled using a unied drag force
expression. Furthermore, an extension of the standard k. turbulence model including the inuence of the presence of two
dispersed phases is suggested to describe the turbulent effects.
The computational model is implemented to predict the threedimensional turbulent ow eld in the three-phase stirred tank
as well as the phase holdup distribution of two dispersed phases.

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550


1.0

7547

1.0
Simulation
Experiment.
r/R=0.5

0.8

0.8

0.6
z/H

z/H

0.6

0.4

0.4

0.2

0.2

Simulation.
Experiment.
r/R=0.5

0.0

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

o/o,av

s/s,av
1.0

0.8

0.8

0.6

0.6
z/H

z/H

1.0

0.4

0.4
Simulation
Experiment.
r/R=0.9

0.2

Simulation.
Experiment.
r/R=0.9

0.2
0.0

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0

0.5

1.0

s/s,av

(a)

1.5

2.0

2.5

3.0

o/o,av
1.0

1.0
Simulation.
0.8

0.8

Experiment.
r/R=0.5

0.6

z/H

z/H

0.6
0.4

0.4

0.2

0.2

0.0

Simulation.
Experiment.
r/R=0.5

0.0
0

0.0

0.5

1.0

s/s,av
1.0

2.0

2.5

3.0

1.0
Simulation.
Experiment.
r/R=0.9

0.8

0.8

0.6
z/H

z/H

0.6

0.4

0.4

0.2

0.2

0.0

Simulation.
Experiment.
r/R=0.9

0.0
0.0

(b)

1.5
o/o,av

0.5

1.0

1.5

2.0

2.5

3.0

s/s,av

0.0

0.5

1.0

1.5

2.0

2.5

3.0

o/o,av

Fig. 12. Comparison of simulated normalized holdup proles of oil and solid phases with the experimental data for different impeller speeds. (s,av = 0.10
and o,av = 0.10). Left: solid; right: oil. (a) N = 300 rpm, (b) N = 500 rpm.

In general, the numerical simulation is capable of reproducing


the trend of the axial holdup prole of the oil and solid phases
and of giving an approximation of the phase holdup values
comparable to the experimental results.
Further improvement of the present model is necessary
so as to obtain quantitative agreement with measurements,

including rening the interaction model for inter-phase momentum exchange and incorporating the breakup and coalescence of droplets. A well-grounded anisotropic multiphase
turbulence model for three-phase ow is also desired for
a more accurate description of the turbulent ow in stirred
tanks.

7548

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

1.0

1.0
Simulation.
Experiment.

0.8

0.8

r/R=0.5

0.6

z/H

z/H

0.6
0.4

0.4

0.2

0.2

Simulation.
Experiment.
r/R=0.5

0.0

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0

0.5

1.0

s/s,av

1.5

2.0

2.5

3.0

o/o,av

1.0

1.0
Simulation.
Experiment.
r/R=0.9

0.8

0.8

0.6
z/H

z/H

0.6

0.4

0.4

0.2

0.2

0.0

Simulation.
Experiment.
r/R=0.9

0.0
0.0

0.5

1.0

2.0

1.5

2.5

3.0

0.0

0.5

1.0

s/s,av

1.5

2.0

2.5

3.0

o/o,av

Fig. 13. Comparison of simulated normalized holdup proles of solid and oil phases with the experimental data at s,av = 0.05 and o,av = 0.10 (N = 500 rpm).
Left: solid; right: oil.

1.0

1.0
Simulation.
Experiment.

0.8

0.8

r/R=0.5

0.6
z/H

z/H

0.6
0.4

0.4

0.2

0.2

0.0

Simulation.
Experiment.
r/R=0.5

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0

0.5

1.0

s/s,av
1.0

2.0

2.5

3.0

1.0
Simulation.
Experiment.

0.8

0.8

r/R=0.9

0.6

0.6
z/H

z/H

1.5
o/o,av

0.4

0.4

0.2

0.2

Simulation.
Experiment.
r/R=0.9

0.0

0.0
0.0

0.5

1.0

1.5
s/s,av

2.0

2.5

3.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

o/o,av

Fig. 14. Comparison of simulated normalized holdup proles of solid and oil phases with the experimental data at s,av = 0.10 and o,av = 0.30 (N = 500 rpm).
Left: solid; right: oil.

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

Notation
a
b
B
C
C1 , C2 , C 
Cb
CD
d
D
Dk
fk
F
Fdrag
Fcen
Fcor
g
G
Ge
H
k
K
N
p
P
P
r
r
R
Rf
R( )
Red
S
t
t1
td
T
u
u
u
w
z

coefcient in algebraic equation


source term in algebraic equation
width of bafe, m
clearance of impeller plane to bottom of
stirred tank, m
constants in k. model
constant in extra turbulent kinetic energy generation term
drag coefcient
diameter of droplet, m
diameter of disc of impeller, m
diameter of impeller, m
mass imbalance term in continuity equation,
kg s1
turbulent diffusive term in continuity equation, kg s1
interphase force, N
drag force, N
centrifugal force, N
Coriolis force, N
gravity acceleration, m s2
turbulent generation term, kg m1 s3
extra turbulent generation term, kg m1 s3
height of liquid in stirred tank, m
turbulent kinetic energy, m2 s2
proportional factor
impeller rotation speed, rpm
pressure correction, Pa
pressure, Pa
uctuating correction, Pa
radial coordinate
radial vector
radius of stirred tank, m
Richardson number dened by Eq. (14)
relative residual
particle Reynolds number (Red = c dd |ud
uc |/c )
source term
time, s
mean eddy lifetime, s
particle response time, s
diameter of stirred tank, m
mean velocity component, m s1
uctuation of velocity component, m s1
velocity vector, m s1
height of impeller blade, m
axial coordinate starting from the tank bottom

Greek letters


ij


holdup
diffusion coefcient, Pa s
Kronecker delta
turbulent energy dissipation rate, m2 s1





k ,  
t


7549

azimuthal coordinate, rad


viscosity, Pa s
kinematic viscosity, m2 s1
density, kg m3
constants in k. model
Schmidt number
shear stress, Pa
general variable
angular speed, rad s1

Subscripts
av
c
d
drag
eff
g
i, j
k
lam
nb
o
p, E, W, S,
N,T,B
r, , z
s
t

averaged
continuous phase
dispersed
drag force
effective
gas
coordinate axes
kth phase
laminar
neighboring node
oil phase
center, east, west, south, north, top and
bottom neighboring nodes of a cell
radial, azimuthal and axial direction
solid phase
turbulence

Acknowledgments
Financial support from the National Natural Science Foundation of China (Nos. 20006015, 20236050) and the National Basic Research Priorities Program (No. 2004CB217604) is gratefully acknowledged.
References
Armenante, P.M., Huang, Y.T., 1992. Experimental determination of the
minimum agitation speed for complete liquidliquid dispersion in
mechanically agitated vessels. Industrial and Engineering Chemistry
Research 31, 13981406.
Brucato, A., Ciofalo, M., Grisa, F., Micale, G., 1998. Numerical prediction of
ow elds in bafed stirred vessels: a comparison of alternative modeling
approaches. Chemical Engineering Science 53, 36533684.
Carver, M.B., 1984. Numerical computation of phase separation in two uid
ow. Journal of Fluids Engineering 106, 147153.
Chapman, C.M., Nienow, A.W., Cooke, M., Middleton, J.C., 1983a.
Particlegasliquid mixing in stirred vessels. Part I. Particleliquid-mixing.
Chemical Engineering Research and Design 61, 7181.
Chapman, C.M., Nienow, A.W., Cooke, M., Middleton, J.C., 1983b.
Particlegasliquid mixing in stirred vessels. Part II. Gasliquid-mixing.
Chemical Engineering Research and Design 61, 8295.
Chapman, C.M., Nienow, A.W., Cooke, M., Middleton, J.C., 1983c.
Particlegasliquid mixing in stirred vessels. Part III. Three phase mixing.
Chemical Engineering Research and Design 61, 167181.
Clift, R., Grace, J.R., Weber, M.E., 1978. Bubbles, Drops, and Particles.
Academic Press, New York.

7550

F. Wang et al. / Chemical Engineering Science 61 (2006) 7535 7550

Dohi, N., Takahashi, T., Minekawa, K., Kawase, Y., 2004. Power
consumption and solid suspension performance of large-scale impellers in
gasliquidsolid three-phase stirred tanks reactors. Chemical Engineering
Journal 97, 103114.
Ekambara, K., Joshi, J.B., 2003. CFD simulation of mixing and dispersion
in bubble columns. Chemical Engineering Research and Design 81A,
9871002.
Elaghobashi, S.E., Abou-Arab, T.W., 1983. A two-equation turbulence model
for two-phase ows. Physics of Fluids 26, 931938.
Gosman, A.D., Issa, R.I., Lekakou, C., Looney, M.K., Politis, S., 1992.
Multidimensional modeling of turbulent two-phase ows in stirred tanks.
The American Institute of Chemical Engineers Journal 38, 19461956.
Grienberger, J., Hofmann, H., 1992. Investigations and modeling of bubble
columns. Chemical Engineering Science 47, 22152220.
Ishii, M., 1975. Thermo-Fluid Dynamics Theory of Two-Phase Flow. Eyrolles,
Paris.
Kataoka, I., Serizawa, A., 1989. Basic equations of turbulence in gasliquid
two-phase ow. International Journal of Multiphase Flow 15, 843855.
Krishna, R., van Baten, J.M., Urseanu, M.I., 2000. Three-phase Eulerian
simulations of bubble column reactors operating in the churn-turbulent
regime: a scale up strategy. Chemical Engineering Science 55, 32753286.
Launder, B.E., Spalding, D.B., 1974. The numerical computation of turbulent
ows. Computer Methods in Applied Mechanics and Engineering 3,
269289.
Ljungqvist, M., Rasmuson, A., 2001. Numerical simulation of the two-phase
ow in an axially stirred vessel. Chemical Engineering Research and
Design 79A, 533546.
Michele, V., Hempel, D.C., 2002. Liquid ow and phase holdup
measurement and CFD modeling for two- and three-phase bubble columns.
Chemical Engineering Science 57, 18991908.
Mitra-Majumdar, D., Farouk, B., Shah, Y.T., 1997. Hydrodynamics modeling
of three-phase ows through a vertical column. Chemical Engineering
Science 52, 44854497.
Mitra-Majumdar, D., Farouk, B., Shah, Y.T., Macken, N., Oh, Y.K., 1998.
Two- and three-phase ows in bubble columns: numerical predictions and

measurements. Industrial and Engineering Chemistry Research 37,


22842292.
Padial, N.T., Van der Heyden, W.B., Rauenzahn, R.M., Yarbro, S.L., 2000.
Three-dimensional simulation of a three-phase draft-tube bubble column.
Chemical Engineering Science 55, 32613273.
Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow. McGraw-Hill,
New York.
Ranade, V.V., van den Akker, H.E.A., 1994. A computational snapshot of
gasliquid ow in bafed stirred reactors. Chemical Engineering Science
49, 51755192.
Spalding, D.B., 1980. Mathematical modeling of uid mechanics, heat transfer
and chemical reaction processes. Imperial College Report HTS/80/1,
London.
Van Doormal, J.P., Raithby, G.D., 1984. Enhancements of the simple method
for prediction incompressible uid ows. Numerical Heat Transfer 7,
147168.
Wang, F., Mao, Z.-S., 2005. Numerical and experimental investigation of
liquidliquid two-phase ow in stirred tanks. Industrial & Engineering
Chemistry Research 44, 57765787.
Wang, F., Wang, W.J., Mao, Z.-S., 2004. Numerical study of solidliquid
two-phase ow in stirred tanks with Rushton impeller, Part I. Formulation
and simulation of ow eld. Chinese Journal of Chemical Engineering
12, 599609.
Wang, F., Mao, Z.-S., Wang, Y.F., Yang, C., 2005. Numerical simulation and
measurement of phase holdups in liquidliquidsolid three-phase stirred
tanks, Fourth International Conference on CFD in the Oil and Gas,
Metallurgical and Process Industries, Trondheim, Norway, 68 June 2005.
Wang, W.J., Mao, Z.-S., 2002. Numerical simulation of gasliquid ow in
a stirred tank with a Rushton impeller. Chinese Journal of Chemical
Engineering 10, 285295.
Zhou, L.X., 1993. Theory and Numerical Modeling of Turbulent GasParticle
Flows and Combustion. Science Press, CRC Press, Beijing, Florida.

Das könnte Ihnen auch gefallen