Chemical Engineering Science 66 (2011) 41274137

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Three-dimensional CFD study of liquidsolid ow behaviors in tubular loop

polymerization reactors: The effect of guide vane
Wei-Cheng Yan, De-Pan Shi, Zheng-Hong Luo n, Ying-Hua Lu
Department of Chemical and Biochemical Engineering, College of Chemistry and Chemical Engineering, Xiamen University, Xiamen 361005, China

a r t i c l e i n f o abstract

Article history: A three-dimensional (3D) computational uid dynamics (CFD) model, using an EulerianEulerian two-
Received 18 October 2010 uid model incorporating the kinetic theory of granular ow, is adopted to describe the steady-state
Received in revised form liquidsolid two-phase ow under conditions employed in a tubular loop propylene polymerization
10 May 2011
reactor composed of loop and axial ow pump. The model is validated by comparing its simulation
Accepted 24 May 2011
Available online 1 June 2011
result with the classical calculated data as well as a set of data collected from certain pilot plant in
China. The entire ow behaviors and the effects of guide vane on them in the reactor are also
Keywords: investigated numerically. Especially, the whole eld in the loop reactor with the guide vane is obtained
Chemical reactors via the above model. The results show that a guide vane weakens the turbulent intensity, reduces the
Complex uids
component of the rotating velocity, and contributes to the uniform distribution of the particles in the
Computation
reactor. The second ow phenomenon is successfully predicted in the loop reactor with the guide vane.
Multiphase ow
Polyolen loop reactor & 2011 Elsevier Ltd. All rights reserved.
Guide vane

1. Introduction reactors (Traube, 1967; Tang and Zhu, 2004; Naqvi et al., 2007;
Kim and Suh, 2008; Kim et al., 2009; Chi et al., 2010).
A computational uid dynamics (CFD) model is adopted to Refer to Mahecha-Botero et al. (2006, 2008, 2009a, b, c),
describe the steady-state liquidsolid two-phase ow in tubular Ranade et al. (Khopkar and Ranade, 2006; Gunjal and Ranade,
loop propylene polymerization reactors recently (Shi et al., 2007; Utikar and Ranade, 2007; Kulkarni et al., 2008; Kasat et al.,
2010a,b). The entire ow behaviors described by the solid holdup 2008; Rampure et al., 2009) and us (Shi et al., 2010a, b) for a
distributions in the tubular loop reactors are obtained. It is noted survey of the papers on ow elds in reactors. There are still few
that the simulated object is a pilot-plant scale polypropylene loop studies on the ow elds in loop polymerization reactors, espe-
reactor consisting of loop and axial ow pump with post-guide cially for the inuence of adding a guide vane on the ow elds in
vane, and some important two-phase ow behavior parameters, loop polymerization reactors. Recently, Liu et al. (2007) used a
such as the turbulent energies, the velocity vectors and the uid three-dimensional (3D) EulerianEulerian model to study the
trajectories, are not mentioned, although many ow features, liquidsolid ow in a loop polymerization reactor. They only
such as the excellent solid mixing, heat and mass transfer proper- simulated part of the loop reactor, ignoring the axial ow pump.
ties, can be related to these behavior parameters (Boothroyd, Mohanarangam et al. (2008) reported a numerical investigation
1971; Akilli et al., 2001; Sheintuch and Nekhamkina, 2004). into the physical characteristics of dilute gas-particle ow over a
Furthermore, the inuence of adding a guide vane installed on square-sectioned 901 bend based on a modied Eulerian two-uid
the back of the axial ow pump is also not considered. Adding a model. The modied EulerianEulerian model was found to yield
guide vane may decrease the turbulent intensity (Traube, 1967; good agreement with the measured values. Tian et al. (2008) also
Tang and Zhu, 2004), however, a guide vane is an important investigated the dilute gassolid two-phase ows in a curved 901
energy recovery apparatus, which increases the efciency of duct bed by the CFD simulation. Their objective was to investigate
conversion from rotating kinetic energy to axial kinetic energy the effect of wall roughness on the particle ow properties.
(Traube, 1967; Tang and Zhu, 2004). The inuence of adding a Therefore, besides the gas velocity distributions, the uid trajec-
guide vane installed on the back of the axial ow pump especially tories in the bed were also obtained. More recently, our group
needs to be further investigated for proper scale-up and design of (Gao et al., 2010) used an EulerianEulerian model incorporating
the kinetic theory of granular ow, the energy balance and the
heat transfer equations, to describe the steady-state temperature
n
Corresponding author. Tel.: 86 592 2187190; fax: 86 592 2187231. eld in a loop propylene polymerization reactor. Unfortunately,
E-mail address: luozh@xmu.edu.cn (Z.-H. Luo). the ow behaviors in the loop reactors are not systemically

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.05.039
4128 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137

studied in all the above reports. Furthermore, as far as we know, Pro-ENGINEER 4.0 is adopted to built the 3D physical models and
the effect of a guide vane on the ow in the loop polymerization their meshes are constructed in Gambit 2.3.16 (Ansys Inc., US).
reactors is still not studied.
This paper focuses on the systemic 3D CFD study of the liquid
solid ow in loop polymerization reactors. The model is validated 3. CFD Model
by comparing the simulation result with the classical calculated
as well as actual data. The entire ow characteristics in the An EulerianEulerian two-uid model that incorporates the
reactors are thus investigated for giving insight into the scale- kinetic theory of granular ow is used to describe the liquidsolid
up and improvement of the loop polymerization reactors. Parti- two-phase ow in loop polymerization reactors.
cular attention is paid to the inuence of the guide vane on the
liquidsolid ow to optimize the axial ow pump in the loop
3.1. EulerianEulerian two uid equations
polymerization reactor.
The continuity equations for phase n (n l for the liquid phase,
s for the solid phase) can be written as
2. 3D model of the tubular loop reactor
rUal rl !
v l 0, 1
The Spheripol technology is one of the most widespread
commercial methods used to produce polypropylene. Its key part rUas rs !
v s 0: 2
consists of two liquid-phase loop reactors and a gas-phase
uidized bed reactor (FBR). In this work, a pilot-plant polypropy- The momentum balance equations for the liquid and solid
lene loop reactor of the Spheripol technology in certain Chinese phases can be written as
--
chemical plant (shown in Fig. 1) is selected as our object. The rUal rl v l v l al rp rUtl Ksl ! !
v s  v l al rl g, 3
reactor has an inner-diameter of 0.15 m, a height of 3.3 m and a
length of 1.4 m, which is the same as that reported in our T
previous work (Shi et al., 2010a, b). In order to investigate the t l al ml r! !
v l r v l , 4
effect of the guide vane, a loop reactor of the same size and shape
without a post-guide vane, comparing with the loop reactor with rUas rs ! ! ! !
v s v s as rprps rUts Kls v q  v s as rs g, 5
a post-guide vane, is also selected in this work. More details on
T
the loop reactor congurations are shown in Fig. 1. Furthermore, t s as ms r! ! !
v s r v s as ls 23ms rU v s I : 6
the ow systems in the loop reactors are both supposed to be a
mixture of liquid-phase and solid-phase (Gao et al., 2010; Shi
et al., 2010a, b).
During the polymerization in the loop reactor, the growth rate 3.2. Kinetic theory of granular ow (KTGF)
of the polymer particles is very slow, and their growth in
diameter is mainly determined by their residence time in the The two-uid model requires constitutive equations to
reactor. Accordingly, under the steady-state conditions, the poly- describe the particulate phase viscosity and the particulate phase
merization kinetics is not considered in this work (Shi et al., pressure gradient. Due to the domination of collision interactions,
2010a, b). The polymerization temperature in the reactor remains the concept from the uid kinetic theory is used to describe the
constant and the liquidsolid system in the reactor is simulated. effective stresses in the solid phase resulting from particle
In this study, a 3D physical model of the reactor system is streaming collision contribution (Boemer et al., 1997). Constitu-
built rst to simulate the 3D reactors. The commercial CAD code tive relations for the solid-phase stress based on the kinetic
theory concepts have been derived by Lun et al. (1984). In
addition, corresponding equations derived by Lun et al. (1984)
are used in this present work.
ps as rs Ys 1 2g0 as 1 es , 7
r
4 2 Ys
ls as rs ds g0 1 es ; 8
3 p
where
1
g0 , 9
1as =as,max 1=3

1 0 0
Ys mm: 10
3 s s
Furthermore, a transport equation for the granular tempera-
ture is also needed. The following equation has been proposed by
Ding and Gidspow (1990):

3 , ,
2rUrs as v s Ys ps I ts : rv s rUkYs rYs gYs fls : 11

where kYs is given according to the model of Syamlal et al. (1993)

p  
p Ys 12 2 16
kYs 15rs ds as 1 Z 4Z3as g0 4133ZZas g0 ;
44133Z 5 15p
Fig. 1. Loop reactor and axial ow pump. 12
 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137 4129

where The general boundary conditions for granular temperature at

the wall follow the equation suggested by Johnson and Jackson
Z 121 es : 13
(1987). In this work, the equation is also applied as the corre-
The collisional dissipation of energy, gYs , is modeled using the sponding boundary conditions.
correlation by Lun et al. The corresponding equations are as p p  
3prs g0 as Ys 3Y
follows (Lun et al., 1984): qs fv2s,w  1e2w , 25
6as,m 2
121e2s g0
gYs p rs a2s Y1:5
s , 14 where qs represents the exchange of pseudo-thermal energy
ds p
between the particles and the wall.
fls 3Kls Ys : 15
There are a number of similar models for the solid phase 3.5. CFD modeling strategy
dynamic viscosity. The model selected in the present work is as
follows (Lun et al., 1984; Schaeffer, 1987; Gidaspow et al., 1992): As described earlier, the CFD with the EulerianEulerian
approach is adopted to study the liquidsolid interactions in this
ms ms,col ms,kin ms,fr ; 16
work. The standard ke model is adopted to describe the turbu-
where lence and the kinetic theory of granular ow is employed to close
r the momentum balance equation for the solid phase. The above
4 Ys equations are solved by the commercial CFD code FLUENT 6.3.26
ms,col as rs ds go 1 es , 17
5 p (Ansys Inc., US) in double precision mode. The phase coupled
p  2 SIMPLE algorithm is used to couple pressure and velocity, and the
Ys p 4 Multiple Reference Frame model (MRF) is used to simulate the
ms,kin 10ds rs 1 1 es as g0 , 18
96as 1 es g0 5 axial ow pump and they are associated by interface. In addition,
a commercial grid-generation tool, GAMBIT 2.3.16 (Ansys Inc., US)
sp sin y
ms,fr p 19 is used to generate the 3D grids of the loop reactor model created
2 I2D : in Pro-ENGINEER 4.0. Initial grid sensitivity studies are carried out
(results not shown for the sake of brevity), ve kinds of grids with
total cells of 55,000, 610,000, 750,000, 870,000, and 970,000 have
been studied, and a total of 610,000 cells are sufcient to resolve
3.3. Drag force model
the 3D ow eld. The applied grid for the loop reactor in this work
is shown in Fig. 2. Furthermore, a Pentium 4 CPU of 2.83 GHz
Similar to our previous works (Shi et al., 2010a, b), the transfer
cooperating with 4 GB RAM performs the simulations.
of force between the liquidsolid two-phase is described by an
empirical drag law based on the Gidaspow et al.s (1992) model.
The applied equations are as follows:

3 aS al rl 9! !
v s  v l 9 2:65
at al 4 0:8, Ksl CD al , 20
4 ds
where
"  0:687 #
24 3
CD 1 al Res , 21
al Res 20

rl ds 9! !
v s v l9
Res , 22
ml
! !
as 1al ml 7 as rl 9 v s  v l 9
at al r 0:8, Ksl 150 : 23
al d2s 4 ds

3.4. Boundary conditions

The effects of the inlet and outlet of the loop reactor on the
hydrodynamics are ignored due to a comparatively higher circula-
tion ux in the loop than in the inlet/outlet ux. Both the inlet and
outlet of the reactor are considered as wall. In addition, the
continuous phase is assumed to obey the no slip boundary condition
at the wall. For the solid phase, a partial slip model suggested by
Johnson and Jackson (1987) is used and shown as follows:
p
@vs,w pfrs as g0 Ys
ms p vs,w , 24
@n 2 3as,m
where f is the reection coefcient of wall, which describes the
interaction between the solid phase and the wall. Its value ranges
from 01. Among them, 0 describes the condition of free slip and Fig. 2. The applied grid for the whole loop reactor in this work: (a) the radial
1 describes the condition of no slip. section, (b) the elbow containing axial ow pump, (c) the vertical pipe.
4130 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137

4. Simulation conditions pressure gradient data from certain pilot plant in China are
collected to further validate the CFD model. Fig. 3 gives the
The simulation results dependence on the range of parameter comparisons of pressure gradient among the data at different
values are presented in Eqs. (1)(25). Most of the parameters are circulation ow velocities. In addition, the data shown in Fig. 3
directly linked to the properties of the liquid and solid phases. All includes the guide vane. From Fig. 3, one knows that the simula-
above parameter values adopted in this work are listed in Table 1. tion results are in good agreement with not only the data
In addition, we also point out that three important parameters obtained by the classical Newitt model well, but also the data
including the restitution coefcient (es), particlewall restitution from the pilot plant (Newitt and Richardson, 1961; Shi et al.,
coefcient (ew) and specularity coefcient (f) were investigated 1985). As a whole, the average deviations of the CFD model and
in our previous works (Shi et al., 2010a, b). Therefore, the same the classical model are 5.38% and 4.16%. In addition, Fig. 3 also
values of the three parameters are applied and listed in Table 2. illustrates that the actual gradient data are higher than the
Unless otherwise noted, the parameters used for the following simulated data of the two groups.
simulation are those in Tables 1 and 2.

5.2. The effect of the guide vane on the ow eld

5. Results and discussion As described earlier, the investigation of the effect of the guide
vane installed on the back of the axial ow pump on the ow eld
5.1. Model verication in the loop reactor is one of our objects in this work. In this
section, its effects on the liquid-phase ow velocity, the turbulent
As described in Section 3, the suggested model in this work is energy, the velocity vector, the uid trajectory and the solid
similar to those reported in our previous works (Shi et al., 2010a, holdup distribution are discussed. When the rotating speed of the
b). The model is also veried in our previous works via the axial ow pump is 400 rad/s, the average velocity reaches to
comparison of pressure gradient between the predictions of CFD 7.34 m/s. In addition, only the ow elds in the vertical ascending
model and the classical Newitt model (Newitt and Richardson, pipe and/or the elbow containing the axial ow pump are
1961; Shi et al., 1985, 2010a, b). However, no practical data are simulated here because the effects of the guide vane represents
adopted, which may not be convincible enough to conclude that more obviously in them than other positions of the loop reactor.
the CFD model agrees well with the observation. In this study, the

Table 1 5.2.1. The effect of the guide vane on the liquid-phase ow velocity
Main physical property and operation parameters. Fig. 4 shows the ow velocity contours of the liquid phase in
the vertical ascending pipe and the elbow containing the axial
ds (m) R (rad s  1) as,i rl (kg m  3) rs (kg m  3) ml (pa s)
ow pump. From Fig. 4, the velocity distribution of the liquid
5  10 4
400 0.35 417 900 5.54  10  5 phase in the loop with a guide vane is similar to that in the loop
without a guide vane. As a whole, in the two loop reactors, the
Where R is the rotating speed of the axial ow pump and as,i is the initial solid liquid-phase ow velocity near the axial ow pump is higher than
phase volume fraction.
that at the other positions of the reactor and a uniform velocity

Table 2
Main parameters used in KTGF.

ew es as,max f

0.9 0.9 0.63 0.0001

Fig. 3. Comparisons of the CFD simulated data with the pilot plant data and the Fig. 4. Liquid phase velocity contour of the vertical ascending pipe and elbow
classical calculated data according to the empirical equation. contains axial ow pump.
 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137 4131

distribution can be obtained when the uid ows into the vertical motions shown in Fig. 7b, it brings out strong centrifugal force
ascending pipe. acting on the solid particles in the reactor without a guide vane.
Accordingly, the particles move toward the wall due to the action
5.2.2. The effect of the guide vane on the turbulent energy of the centrifugal force and congregate near the wall.
Fig. 5 illustrates the effect of the guide vane on the turbulent
energies. As described in Fig. 5, in both reactors, which are with
and without the guide vane, the turbulent energy near the reactor
wall is higher than that in the reactor center as a result of the
higher uctuation of the ow velocity near the wall. In addition, it
is also shown that the turbulent energy near the wall of the loop
reactor without a guide vane is larger than that with a guide vane,
namely, more energy loss occurs in the reactor when without a
guide vane. Accordingly, the simulated results prove that adding a
guide vane can weaken the turbulent intensity and reduce the
energy loss in the loop reactor.

5.2.3. The effect of the guide vane on the velocity vector

The owing motions of the rotating torus in the loop reactors
with and without the guide vane are also simulated via the above
model. The simulated velocity vectors in three typical positions in
the two loop reactors are shown in Fig. 6. From Fig. 6, it is found
that a more straight velocity vector prole can be obtained in the
loop reactor with a guide vane, than that without a guide vane. In
addition, an obvious curving prole can be found in Fig. 6b, which
means that adding the guide vane reduces the component of the
rotating velocity in the reactor. Furthermore, according to Fig. 6b, Fig. 6. Velocity vector of vertical ascending pipe. ((a) With post-guide vane. (b)
Without post-guide vane; top, middle, bottom: top, middle, bottom of the vertical
with the owing of the uid along the reactor, the component of pipe after the pump).
the rotating velocity decreases.

5.2.4. The effect of the guide vane on the uid trajectory

The uid trajectories in the two reactors with/without the
guide vane are obtained and shown in Fig. 7. In practice, Fig. 7
describes the uid path lines behind the axial ow pump in the
steady state simulation. According to Fig. 7, in the loop reactor
without a guide vane the uid system comprising of the liquid
and solid phases ows with obvious rotating motions. However,
the uid system ows along obvious line paths in the loop reactor
with a guide vane, which is shown in Fig. 7a. As to the rotating Fig. 7. Fluid path lines. ((a) With post-guide vane. (b) Without post-guide vane).

Fig. 5. Turbulent energy contour of vertical ascending pipe. (a: With post-guide vane; b: Without post-guide vane; top, middle, bottom: top, middle, bottom of the vertical
pipe after the pump).
4132 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137

5.2.5. The effect of the guide vane on the solid holdup distribution 5.3. The whole ow eld in the loop reactor with the guide vane
The above agglomeration phenomenon can be proved via
describing the solid holdup distribution in the loop reactor. According to the investigation in Section 5.2, adding a guide
Fig. 8 illustrates the solid holdup distributions in the vertical vane weakens the turbulent intensity and reduces the component
ascending pipes of the two reactors with and without a guide of the rotating velocity. In addition, as described earlier, the guide
vane. From Fig. 8, one knows that there is an obvious particle vane is an important energy recovery apparatus and can increase
congregating phenomenon near the wall of the reactor without a the efciency of conversion from rotating kinetic energy to axial
guide vane. Such a phenomenon leads to a low solid holdup in the kinetic energy (Traube, 1967; Tang and Zhu, 2004). In practice,
center and a high solid holdup near the wall of the loop reactor the loop reactor with a guide vane is widely applied in the
yet. In addition, as shown in Fig. 8a, the solid holdup distribution polymerization industry. In order to further understand the
in the loop reactor with a guide vane is more uniform when inuence of a guide vane, the whole ow eld in the loop reactor
compared with that in Fig. 8b. In addition, Fig. 8a also shows that with a guide vane is described.
due to the action of the centrifugal force there is still a small
particle congregating phenomenon in the top of the vertical
5.3.1. The velocity distribution of the liquid phase in the whole loop
ascending pipe of the loop reactor with a guide vane.
The velocity of the liquid phase is one of the most important
parameters in the loop reactors. In the above simulation, only the
velocity distributions of the liquid phase in certain positions of the
loop reactor are simulated. Here, the velocity distribution in the
whole loop reactor is obtained. From Fig. 9, the velocity distribution
within the loop with a guide vane is uniform as a whole. In addition,
an obvious high velocity area is found at the inner side of bends,
which is shown in Fig. 9a. With the uid owing along the loop, the
velocity distribution develops toward uniform. Specically, the velo-
city distribution of the liquid phase in the down bend/elbow situated
the bottom of the loop is described and shown in Fig. 9b. According to
Fig. 9b, one observes that there is an obvious velocity gradient within
the bend. It means that the second ow appears in the loop
(Mohanarangam et al., 2008; Tian et al., 2008). In Section 5.3.2, the
second ow phenomenon is described in detail. In addition, some
simulated positions in this loop/bend are also described in Fig. 9.

5.3.2. The second ow in the loop

The velocity distribution of the liquid phase in the left bottom
bend shown in Fig. 9b is rstly simulated. Fig. 10 describes the
velocity distribution of the liquid phase in the down bend. From
Fig. 10, the velocity proles of the liquid phase are uniform before it
ows into the elbow (s/D  0.5) after the development in the
vertical descending pipe. But when the uid ows into the elbow
(y 01), the velocity at the inner side of elbow accelerates; however,
Fig. 8. Particle volume fraction contour of vertical ascending pipe. ((a) With post- it will slow down at the outer side. Therefore, a high velocity region
guide vane. (b) Without post-guide vane). is formed next to the inner wall. Along the elbow (y 01-301), this

Fig. 9. Liquid phase velocity contour of the whole tubular loop reactor. ((a) Whole tubular loop reactor. (b) Left bottom elbow).
 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137 4133

Fig. 10. Liquid phase velocity distribution of the elbow.

phenomenon becomes more and more evident. Between y 301 and is nally greater than that of inner side. The formation of this
601, the velocity proles almost keep the same. In the rest part of the secondary ow behavior is attributed to changes in local pressure
elbow (y 601-901), the velocity at the outer side of elbow gradients (Mohanarangam et al., 2008; Tian et al., 2008). In order to
accelerates, but slows down at the inner side. When the slurry explain this phenomenon, the pressure distribution in the elbow is
ows out of the elbow, the high-speed area gradually moves toward simulated and the simulated results are shown in Fig. 11. As shown
the outside of the elbow, and the velocity of the outside of the elbow in Fig. 11, the pressure (here, the simulation pressure is a relative
4134 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137

pressure in order to reect a pressure gradient) at the outside phase are almost the same, and there is no velocity slip between the
of elbow is much greater than that at the inner, which leads to two phases. At the initial interface of the vertical ascending pipe,
an acceleration of the velocity at the inner side of elbow when shown at Fig. 12a, the velocity in the inner side of the tubular loop is
y 451901, and a deceleration when y 451901. higher than that in the outer side. In the middle of the vertical pipe, as
The velocity distributions at the initial interface, middle and the shown at Fig. 12b, the velocities are fully developed. When it comes
end of straight pipes which are marked in the Fig. 9a will be discussed to the end of the vertical ascending pipe, the velocity at the inner side
as following. Fig. 12 describes the velocity distributions of the liquid of the tubular loop is higher than that at the outer side again.
and solid phases in the vertical ascending pipe. From Fig. 12, in the Fig. 13 shows the velocity distributions of the liquid and solid
vertical ascending pipe, the velocities of the liquid phase and solid phases in the top horizontal pipe. From Fig. 13d, at the beginning
of the horizontal pipe, there is some velocity slip between the
liquid phase and solid phase in the outer side of the tubular loop,
and the solid phase velocity is higher than the liquid-phase
velocity, which is because the solid phase suffers a larger
centrifugal force compared with the liquid phase. From Fig. 13e,
the velocities in the middle of the top horizontal pipe are not
completely developed. And to be the same as the vertical
ascending pipe, from Fig. 13f, the velocities at the inner side of
the tubular loop are higher than that at in the outer side.
Fig. 14 shows the velocity distributions of the liquid and solid
phases in vertical descending pipe. Similar to the top horizontal
pipe, from Fig. 14g, it is found that in the initial interface of the
vertical descending pipe, there is some velocity slips between the
liquid phase and solid phase in the outer side of the tubular loop,
and the solid phase velocity is higher than the liquid-phase velocity.
Along the pipe, the slurry velocity begins to develop gradually into
fully developed, but there is a low velocity region in the center of the
pipe. The same results can be observed in the above two pipes, the
velocities in the inner side of the tubular loop are higher than those
Fig. 11. Static pressure contour of the elbow. at the outer side in the end of the vertical descending pipe.

Fig. 12. Liquid and solid phase velocity distribution of the vertical ascending pipe. (r/R 1: the inside of tubular loop reactor, r/R  1: the outside of tubular loop reactor).
 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137 4135

Fig. 13. Liquid and solid phase velocity distribution of the top horizontal pipe. (r/R 1: the inside of tubular loop reactor, r/R  1: the outside of tubular loop reactor).

Fig. 14. Liquid and solid phase velocity distribution of the vertical descending pipe. (r/R1: the inside of tubular loop reactor, rr/R  1: the outside of tubular loop reactor).
4136 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137

6. Conclusions in this work. The authors also thank the anonymous referees for
comments on this manuscript.
In this study, a 3D CFD model is adopted to describe the steady- The simulation work are implemented by advanced software
state liquidsolid two-phase ow under conditions employed in a tools (FLUENT 6.3.26 and GAMBIT 2.3.16) provided by China
tubular loop propylene polymerization reactor. The model is validated National Petroleum Corporation and its subsidiary company.
by comparing the simulation result with the classical calculated data
as well as the data collected from certain pilot plant in China. The
predicted pressure drop data are found to be in good agreement with References
both of them. The entire ow behaviors and the effects of a guide
vane on them in the reactor are also investigated numerically. Akilli, H., Levy, E.K., Sahin, B., 2001. Gassolid ow behaviour in a horizontal pipe
The simulation results show that adding a guide vane weakens after vertical-to-horizontal elbow. Powder Technol. 116, 43.
the turbulent intensity, reduces the component of the rotating Boemer, A., Qi, H., Renz, U., 1997. Eulerian simulation of bubble formation at a jet
in a two-dimensional uidized bed. Int. J. Multiphase Flow 23, 927.
velocity in the reactors, and contributes to the uniform distribu-
Boothroyd, R.G., 1971. Flowing GasSolids Suspensions. Chapman & Hall, London,
tion of the particles in the reactor. In addition, the second ow UK, pp. 120230.
phenomenon can be successfully simulated in the reactor with a Chi, J.L., Wang, B., Zhang, S.L., Xiao, Y.H., 2010. Off-design performance of a
guide vane. Furthermore, the simulated results show that there is chemical looping combustion (CLC) combined cycle: effects of ambient
temperature. J. Therm. Sci. 19, 87.
a high velocity area within the bend of the loop reactor with the Ding, J., Gidspow, D., 1990. A bubbling uidization model using kinetic-theory of
guide vane due to the effect of centrifugal force. granular ow. AIChE. J. 36, 523.
Gao, X., Shi, D.P., Chen, X.Z., Luo, Z.H., 2010. Three-dimensional CFD model of the
temperature eld for a pilot-plant tubular loop polymerization reactor.
Powder Technol. 203, 574590.
Gidaspow, D., Bezburuah, R., Ding, J., 1992. Hydrodynamics of circulating uidized
Notation
beds: kinetic theory approach. In: Fluidization VII, Proceedings of the 7th
Engineering Foundation Conference on Fluidization. pp 7582.
Cd drag coefcient Gunjal, P.R., Ranade, V.V., 2007. Modeling of laboratory and commercial
scale hydro-processing reactors using CFD. Chem. Eng. Sci. 62, 55125526.
dp particle diameter (m) Johnson, P.C., Jackson, R., 1987. Frictionalcollisional constitutive relations for
D pipe diameters (m) granular materials, with application to plane shearing. Fluid Mech. J. 176, 67.
es particleparticle restitution coefcient Kasat, G.R., Khopkar, A.R., Ranade, V.V., Pandita, A.B., 2008. CFD simulation of
liquid-phase mixing in solidliquid stirred reactor. Chem. Eng. Sci. 63,
ew particlewall restitution coefcient 38773885.
g gravitational acceleration (m s  2) Khopkar, A.R., Ranade, V.V., 2006. CFD simulation of gas-liquid stirred vessel: VC,
I identity matrix S33, and L33 ow regimes. AIChE J. 52, 16541672.
Kim, K.T., Suh, J.M., 2008. Development of an advanced PWR fuel for OPR1000s in
I2D second invariant of the deviatoric stress tensor
Korea. Nucl. Eng. Des. 238, 2606.
Kls interphase exchange coefcient (kg m2 s  1) Kim, K.T., Jang, Y.K., Kim, J.I., 2009. The study on the impact of fsi on the fuel
p pressure (Pa) assembly design optimization. Pres. Ves. Pip. Div. Am. Soc. Mech. Eng., 1519.
ps particulate phase pressure (Pa) Kulkarni, A.A., Ranade, V.V., Rajeev, R., Koganti, S.B., 2008. CFD simulation of ow
in vortex diodes. AIChE J. 54, 11391152.
R rotating velocity of loop reactor (rad s  1) Liu, Y.B., Chen, J.Z., Yang, Y.R., 2007. Number simulation of liquid-solid two-phase
Res particles Reynolds number ow in tubular loop reactor. J. Chem. Eng. Chin. Univ. 21, 790.
vl liquid velocity (m s  1) Lun, C.K.K., Savage, S.B., Jeffrey, D.J., Chepurniy, N., 1984. Kinetic theories for
granular ow-inelastic particles in Couette-ow and slightly inelastic particles
vs solid velocity, (m s  1) in a general ow eld. J. Fluid Mech. 140, 223.
vs,w solid velocity at wall, (m s  1) Mahecha-Botero, A., 2009a. Comprehensive Modelling and its Application to
al volume fraction of liquid phase Simulation of Fluidized-Bed Reactors for Efcient Production of Hydrogen
and other Hydrocarbon Processes. Ph.D. Dissertation, University of British
as volume fraction of solid phase Columbia, London, UK.
as,m maximum volume fraction of solid phase Mahecha-Botero, A., Boyd, T., Gulamhusein, A., Comyn, N., Lim, C.J., Grace, J.R.,
f specularity factor Shirasaki, Y., Yasuda, I., 2008. Pure hydrogen generation in a uidized bed
membrane reactor: experimental ndings. Chem. Eng. Sci. 63, 2752.
ml viscosity of liquid phase (Pa s) Mahecha-Botero, A., Grace, J.R., Elnashaie, S.S.E.H., Lim, C.J., 2009b. Advances in
ms solid shear viscosity (Pa s) modeling of uidized-bed catalytic reactors: a comprehensive review. Chem.
ms,col solid collisional viscosity (Pa s) Eng. Commun. 196, 1375.
Mahecha-Botero, A., Grace, J.R., Lim, C.J., Elnashaie, S.S.E.H., Boyd, T., Gulamhus-
ms,kin solid kinetic viscosity (Pa s)
sein, A., 2009c. Pure hydrogen generation in a uidized bed membrane
ms,fr solid frictional viscosity (Pa s) reactor: application of the generalized comprehensive reactor model. Chem.
y angle of internal friction (deg.) Eng. Sci. 64, 3826.
Ys granular temperature (m2 s  2) Mahecha-Botero, A., Grace, J.R., Elnashaie, S.S.E.H., Lim, C.J., 2006. Comprehensive
modelling of gas uidized-bed reactors allowing for transients, multiple
gYs collisional dissipation of energy (m2 s  2) ow regimes and selective removal of species. Int. J. Chem. React. Eng. 4,
tl shear stress of liquid phase (N m  2) A11.
ts shear stress of solid phase (N m  2) Mohanarangam, K., Tian, Z.F., Tu, J.Y., 2008. Numerical simulation of turbulent gas-
particle ow in a 901 blend: EulerianEulerian approach. Comput. Chem. Eng.
ls solid bulk viscosity (Pa s) 32, 561571.
rl liquid density (kg m  3) Naqvi, R., Wolf, J., Bolland, O., 2007. Part-load analysis of a chemical looping
rs solid density (kg m  3) combustion (CLC) combined cycle with CO2 capture. Energy 32, 360.
Newitt, D.M., Richardson, J.F., 1961. Hydraulic conveying of solids in vertical pipes.
Trans. Inst. Chem. Eng. 39, 93.
Rampure, M.R., Mahajani, S.M., Ranade, V.V., 2009. CFD Simulation of bubble
columns: modeling of nonuniform gas distribution at sparger. Ind. Eng. Chem.
Res. 48, 81868192.
Acknowledgments Schaeffer, D.G., 1987. Instability in the evolution equations describing incompres-
sible granular ow. J. Differ. Equations 66, 19.
Sheintuch, M., Nekhamkina, O., 2004. Comparison of ow-reversal, internal-
The authors thank National Natural Science Foundation of recirculation and loop reactors. Chem. Eng. Sci. 59, 40654072.
China (No. 21076171) and China National Petroleum Corporation Shi, D.P., Luo, Z.H., Guo, A.Y., 2010a. Numerical simulation of the gas-solid ow in
for supporting this work. We would also like to thank Dr. Fox uidized-bed polymerization reactors. Ind. Eng. Chem. Res. 49, 40704076.
Shi, D.P., Luo, Z.H., Zheng, Z.W., 2010b. Numerical simulation of liquidsolid two-
R.O. and Dr. Ram G.R. (Department of Chemical & Biological phase ow in a tubular loop polymerization reactor. Powder Technol. 198,
Engineering, Iowa State University) for their valuable discussion 135.
 W.-C. Yan et al. / Chemical Engineering Science 66 (2011) 41274137 4137

Shi, J., Wang, J.D., Feng, B.H., 1985. Chemical engineering handbook uid ows. Tian, Z.F., Inthavong, K., Tu, J.Y., Yeoh, G.H., 2008. Numerical investigation into the
Chemical Engineering Press(Chinese), Beijing, China, pp. 23124. effects of wall roughness on a gas-particle ow in a 901 bend. Int. J. Heat Mass
Syamlal, M., Rogers, W., OBrien, T.J., 1993. MFIX Documentation: volume 1, Trans. 51, 1238.
Theory Guide. National Technical Information Service. Springeld.
Traube, K, 1967. Anordnung von kuhlwasserumw alzpumpen innerhalb des
Tang, S.P., Zhu, S.M., 2004. Equipment & Machine Foundation for Chemical
druckgefasses von siedewasserreaktoren. Nucl. Eng. Des. 5, 161.
INDUSTRY. East China. University of Science and Technology Press, Shanghai, Utikar, R.P., Ranade, V.V., 2007. Singlejet uidized beds: experiments and CFD
China, pp. 12118. simulations with glass and polypropylene particles. Chem. Eng. Sci. 62, 167183.