You are on page 1of 9

CFD simulation of hydrodynamics of gassolid multiphase ow in

downer reactors: revisited


Yong Nam Kim, Changning Wu, Yi Cheng
n
Department of Chemical Engineering, Beijing Key Laboratory of Green Chemical Reaction Engineering and Technology, Tsinghua University, Beijing 100084, PR China
a r t i c l e i n f o
Article history:
Received 23 September 2010
Received in revised form
2 May 2011
Accepted 25 July 2011
Available online 29 July 2011
Keywords:
Hydrodynamics
Multiphase ow
CFD
Simulation
Turbulence
Downer
a b s t r a c t
In the present work, a k
1
e
1
k
2
k
12
two-uid model based on the kinetic theory of granular ow (KTGF)
was employed to predict the ow behavior of gas and solids in downers, where the particles of small
size as 70 mm in diameter apparently interact with the gas turbulence. The turbulence energy
interaction between gas and solids was described by different k
12
transport equations, while the
particle dissipation by the large-scale gas turbulent motion was taken into account through a drift
velocity. JohnsonJackson boundary condition was adopted to describe the inuence of the wall on the
hydrodynamics. The simulation results by current CFD model were compared with the experimental
data and simulation results reported by Cheng et al. (1999. Chem. Eng. Sci. 54, 2019) and Zhang and Zhu
(1999. Chem. Eng. Sci. 54, 5461). Good agreement was obtained based on the PDE-type k
12
transport
equation. The results demonstrated that the proposed model could provide good physical under-
standing on the hydrodynamics of gassolid multiphase ow in downers. Using the current model, the
mechanism for formation and disappearance of the dense-ring ow structure and the scale-up
characteristics of downers were discussed.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Downer reactor, in which both gas and solids move down-
wards co-currently, has attracted great attention from both
academia and industry because of its unique features such as
relatively uniform ow structure in the radial direction and near
plug-ow reactor performance in comparison with the other gas
solids uidized bed reactors, e.g., bubbling bed, turbulent bed, and
riser (Zhu et al., 1995; Cheng et al., 2008). The ow characteristics
of downer reactor shows good potential applications in diverse
fast reaction processes with intermediates as desired products,
typically as fast catalytic conversions of feedstock (e.g., heavy oil
and other hydrocarbons) and pyrolysis process of solid materials
(e.g., biomass, coal, and solid waste).
In the last two decades, a number of studies have been
published on the experimental measurements and computational
uid dynamics (CFD) predictions of downer hydrodynamics. For
example, Wang et al. (1992) gave an early report on the radial
distributions of solid volume fraction and the axial distributions
of pressure and particle velocity in a downer. A higher solid
volume fraction was illustrated in the wall region than the one in
the central region and the existence of three axial sections
reecting on a variety of radial ow structures was suggested.
Their experiments showed that the radial distributions of solid
volume fraction and particle velocity in the fully developed region
in a downer are much more uniform than the ones in a riser with
comparable geometry and operating conditions for these two
reactors. Zhang and Zhu (1999) made a systematic study of radial
ow structure and ow development along the length of downer
from entrance to the exit under different operating conditions.
Besides the experimental investigations, several CFD models
have been proposed to predict the gassolids ows in downers
based on the kinetic theory of granular ow (KTGF). Cheng et al.
(1999, 2001) developed a gas turbulencesolid turbulence model
(keYk
p
), taking into account the fast and dense gassolids
ow in downers. The model comprised a ke turbulence model for
gas phase, a k
p
turbulence model and a kinetic theory description
of solid stresses characterized by the granular temperature (Y)
for solid phase. The proposed keYk
p
model successfully
predicted axial and radial distributions of local solids fraction,
local particle velocity and pressure measured in a downer (0.14 m
i.d. and 7 m in height), quantitatively validated by the corre-
sponding experimental data. A dense-ring ow structure had
been found to appear and sustain for wide operating ranges in
the fully developed region of downers with 0.070.50 m i.d. by
the proposed keYk
p
model, which had very good consistency
with the experimental ndings in the literature (see review article
by Cheng et al., 2008). On the basis of understanding important
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/ces
Chemical Engineering Science
0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.07.036
n
Corresponding author. Fax: 86 10 62772051.
E-mail address: yicheng@tsinghua.edu.cn (Y. Cheng).
Chemical Engineering Science 66 (2011) 53575365
inuence of the wall effect on the gassolids ows in vertical
pipes, Cheng et al. (1999, 2001) described the inuence of the
reactor wall on the hydrodynamics through an empirical radial
function of the coefcient of restitution.
Jian and Ocone (2003) proposed a reduced steady-state two-
uid model based on KTGF for predicting the fully developed ow
in downers, and attempted to validate their model predictions by
the experimental data published by Cheng et al. (1999). Within
their model, the inter-particle cohesive force for solid phase was
taken into account by modifying the solid stress tensor via the
introduction of a counter-diffusive term, and the gas phase
turbulence was considered using a mixing-length model. It is
noted that the assumption of perfect elastic collisions was used in
the simulations and the experimental data on the solids friction at
wall was adopted in the boundary conditions (BC). The simulation
results showed that the impact of gas phase turbulence is not
signicant but the counter-diffusive solid concentration term
plays the dominant role in the prediction capability for the
features of dense-ring ow structure.
The dense-ring ow structure is a time-averaged result of the
transient cluster phenomena inherently existing in downers as
well as the core-annulus structure in risers. When the ow eld is
simulated by steady-state models, unrealistic sensitivity to
the inelasticity of particleparticle collisions would be manifested
and required to introduce terms to account for the effects of
time-smoothing (Pita and Sundaresan, 1991; Benyahia et al.,
2007). So, it is preferable to use an unsteady model to investigate
hydrodynamics in downers. On the other hand, the formation
mechanisms of this unique ow structure are mainly due to
particleparticle and particlewall collisions, gasparticle inter-
action and effect of gravity. Therefore, the hydrodynamic model
must take into account simultaneously the gasparticle and
particleparticle interactions in the mean and uctuating motions
by entrainment and inter-particle collision mechanisms, while
the boundary conditions are required to describe the collisional
exchanges of momentum and kinetic energy between particles
and wall. Vaishalia et al. (2008) presented a EulerianEulerian
CFD model and carried out two-dimensional (2D) axi-symmetric
unsteady simulations to investigate the radial ow structure in
gassolids downer under a framework of standard ke model for
the gas phase turbulence with a granular temperature conserva-
tion equation for the solid phase turbulence. The idea of matching
slip velocities and the trends thereof with solids fraction was
emphasized as the key to developing a robust CFD model, which
has predictive capability over a wide variety of ow conditions.
The dense-ring ow structure was observed in the time-averaged
simulation results. However, evident differences could be found
between the model predictions and the experimental data, which
might be caused by assuming 2D axi-symmetric ow, neglecting
the interaction between the uctuating elds of gas phase and
solid phase, and some other key model assumptions.
The approach used in this study is a k
1
e
1
k
2
k
12
transient
two-uid model developed by Simonins group (Balzer et al.,
1996; Simonin, 1996), combined with the boundary condition
derived by Johnson and Jackson (1987). The uidparticle velocity
covariance k
12
describes turbulent momentum transfer between
the uid phase and solid phase uctuating motions, which is
given by an additional transport equation. The transport proper-
ties of the solid phase are obtained by applying the kinetic theory
of gases while taking into account the inuence of the interstitial
gas (Cao and Ahmadi, 1995; Balzer et al., 1996; Peirano and
Leckner, 1998). Additionally, a drift velocity is considered as the
relative velocity to reect the particle dissipation by the large-
scale uid turbulent motion (Balzer et al., 1996; Peirano and
Leckner, 1998; Balzer, 2000; Peirano et al., 2001; Ferschneider
and Mege, 2002; Zhang and Reese, 2003). In detail, the ow of
uid catalytic cracking (FCC) catalyst (r
s
1400 kg/m
3
,
d
p
70 mm) and air in fully developed region of downers is
investigated by the k
1
e
1
k
2
k
12
transient two-uid model,
where a partial differential equation (PDE) formulation of the
k
12
transport equation is focused. The model predictions are
compared with the experimental and simulated results reported
by Cheng et al. (1999). The simulations are implemented in open-
source software MFIX, a platform developed by National Energy
Technology Laboratory (NETL) (Syamlal et al., 1993).
2. Mathematical model and simulation conditions
In two-uid model framework, both gas and particle phases
are considered to be continuous and fully interpenetrating. The
equations describing gassolid turbulent two-phase ow can be
derived by conditional phase averaging method for the gas and
the kinetic theory formalism for the discrete particles (Balzer
et al., 1996; Peirano and Leckner, 1998). The k
1
e
1
k
2
k
12
transient two-uid model was found to be able to reproduce
both the statics (bed height and probability density function of
the spatial distribution of particles) and the dynamics (power
spectrum of pressure uctuations) of a cold circulating uidized
bed operated at low uidization velocities (Peirano et al., 2001).
Also, the model has been used to predict the core-annulus
structure and the existence of clusters in riser reasonably well
with appropriate boundary conditions (BC) (Ferschneider and
Mege, 2002; Benyahia et al., 2005).
The model equations used in this study are summarized in
Appendix A. In the governing equations, the uidparticle velo-
city covariance k
12
is dened as k
12
/u
0
1i
u
0
2i
S
2
, where
u
0
mi
u
mi
U
mi
is the uctuational velocity component of the
phase m (m1 for gas phase, 2 for particle phase), and
U
mi
/u
mi
S
m
represents the mean velocity component of the
phase m dened by the conditional volumetric phase average
/S
m
.
The derivation of k
12
transport equation is a direct application
of kinetic theory of granular ow (KTGF) and of the models of
uidparticle interaction based on the quasi-isotropy assumption
(Boussinesq approximation) (Peirano and Leckner, 1998). For the
ow is highly anisotropic, the second-order closure models
should be used to model the second-order velocity moment
S
12ij
as listed in Eq. (A13).
The closure models proposed by Simonin (1996) were used in
this study, as listed in Eqs. (A17)(A26). In Benyahia et al. (2005)s
simulations of risers using Johnson and Jackson wall boundary
conditions (as listed in Table A3), it could be found that experi-
mental data were close to the small or no friction limit, corre-
sponding to specularity coefcients at range of 0.0080.02. For
downers in this study, the specularity coefcients were empiri-
cally determined to be at range of 0.00050.003 due to the fact
that the collisional angle between solid particles and the wall is
relative small in gasparticle co-current downows.
Three mathematical models (see Table B1 in Appendix B) were
compared to simulate a two-dimensional (2D) channel and a
three-dimensional (3D) pipe. In the A-model, the algebraic for-
mulation listed in Table B1 was used to calculate k
12
. In the
B-model, the PDE formulation (i.e., Eq. (A7)) without drift velo-
city, was coupled into the governing equations to determine k
12
,
while in the C-model the PDE formulation was used with drift
velocity.
In the 2D channel and the 3D pipe, the gas and solid phases
travel along the direction of gravity co-currently. The 2D and 3D
simulations for the fully developed regime using periodic bound-
aries in the streamwise direction were performed. The 2D vertical
channel was 0.10 or 0.14 m in width and 0.42 m in height, while
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5358
the 3D vertical pipe (i.e., downer) was 0.10 or 0.14 m in diameter
and 0.28 m in height. The specularity coefcient was set 0.003 for
2D case but 0.0005 for 3D case, with a major consideration of the
different particlewall collision probability.
For all the simulation cases in this work, the initial conditions
were uniform ow elds. The initial granular temperature was
0.1 m
2
/s
2
and the gas turbulent energy and dissipation were set at
0.01 m
2
/s
2
and 0.1 m
2
/s
2
, respectively.
The summary of the parameters used in simulations is listed in
Table 1. The grid sensitivity was tested and the grid systems of
80160 (x y) for 2D cases and 301209 (r y y) for 3D
cases were nally used, respectively.
All the time-averaged variables were evaluated from the
transient simulation results over a period of 20 s, which did not
include the initial 5 s for the sufcient development of the gas
solids ows.
3. Results and discussion
Fig. 1 shows the time-averaged local solid volume fractions in
the 3D downer (0.14 m in diameter), predicted by the three
different models mentioned in Section 2, i.e., A-model, B-model,
and C-model. Corresponding to the case of d
p
70 mm as shown in
Fig. 1(a), it can be seen that the spatial distributions of solids
volume fraction are different among the three models. The high-
est peak is predicted by A-model, followed by the B-model
and then the C-model, which shows that the inter-phase
exchange of the kinetic turbulence energy and the particle energy
dissipation by the large-scale uid turbulent motion could
not be neglected. All the model predictions show good agreement
with the experimental data reported by Cheng et al. (1999)
but the predictions by C-model look a little better. However, for
the case of d
p
300 mm (see Fig. 1(b)), almost no difference could
be found among the predictions by the three models. It is
noted that the A-model is obtained from the B-model with the
assumptions of homogeneous/stationary two-phase turbulence
and Z
r
b1 (i.e., the particle motion is slightly affected by
the gas turbulence). The differences between Fig. 1(a) and
(b) indicate that the gas turbulence will take apparent effect
on the particle phase when the particle size is small as 70 mm
but much smaller effect when the particle size is large as
300 mm.
Fig. 2 shows the time-averaged local solids volume fractions in
the 2D channel (0.14 m in width), predicted by the three different
models. The similar trends can be observed with the simulation
results of the 3D downer. For the case of d
p
70 mm (see Fig. 2(a)),
the predicted peak by the A-model is apparently higher than the
ones by the B-model and C-model. For the case of d
p
300 mm
(see Fig. 2(b)), small differences could be found among the
predicted proles by the three models, which is similar to the
observations in the 3D downer simulations. The peaks in the 3D
downer appear at a radial location closer to the wall than the ones
in 2D channel with the same size, which is mainly caused by the
spatial difference between the at channel and the cylindrical
pipe.
Fig. 3 plots the proles of local solid volume fraction and
particle velocity from current 3D predictions by the C-model,
together with the simulation results by the keYk
p
model and
the experimental data (Cheng et al., 1999, 2000), at two kinds of
operating conditions. The model predictions have good agree-
ments with the simulation results by the keYk
p
model and the
experiment data. Also, the predicted proles by the C-model show
better agreement with the experimental data than the ones by the
keYk
p
model, e.g., uniform solid volume fraction and mono-
tonically increasing particle velocity along radial direction in the
central region (r/R at the range of 00.7).
For further validation, the predicted local solids volume frac-
tion and particle velocity in a downer (0.10 m in diameter) by the
C-model are compared with the experimental data reported by
Zhang and Zhu (1999) at two different operating conditions. In
the simulations, the particle diameter is 67 mm, the particle
density is 1400 kg/m
3
, and the solid circulation rate is 101 kg/
m
2
/s. For the case of U
g
3.7 m/s, the cross-sectional averaged
solid volume fractions are 0.0167 from the model prediction
and 0.0114 from the experiments, while the corresponding
Table 1
Parameters used in the simulations.
Parameters 2D channel 3D downer
Process temperature (K) 298 298
Process pressure (kPa) 101.325 101.325
Air density (kg/m
3
) 1.2 1.2
Particle diameter (mm) 70, 300 70, 300
Particle density (kg/m
3
) 1400 1400
Inter-particle coefcient of restitution 0.95 0.95
Particlewall coefcient of restitution 0.90 0.90
Specularity coefcient 0.003 0.0005
0.00
0.01
0.02
0.03

2
r/R
3D downer (D=0.14m)
Ug = 4.33 m/s, Gs = 70 kg/m
2
s
dp = 70 m
A-model
B-model
C-model
Expt. data (Cheng et al., 1999)
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
0.00
0.01
0.02
0.03

2
r/R
3D downer (D=0.14m)
Ug = 4.33 m/s, Gs = 70 kg/m
2
s
dp = 300 m
A-model
B-model
C-model
Fig. 1. Time-averaged local solid volume fraction in the 3D downer simulation.
(a) d
p
70 mm, (b) d
p
300 mm.
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5359
cross-sectional averaged particle velocities are 4.9 and 7.7 m/s,
respectively. For the case of U
g
7.2 m/s, the cross-sectional
averaged solid volume fraction is 0.0100 vs. 0.0071, while the
cross-sectional averaged particle velocity is 7.5 vs. 9.2 m/s. The
model predictions show about 40% larger solids volume fraction
but about 2040% smaller particle velocity than the reported
experimental data based on the cross-sectional average, which is
probably due to the model assumptions (e.g., uniform particle
diameter distribution), the key model parameters (e.g., particle
restitution and specularity coefcients), the key sub-models (e.g.,
gassolid drag force calculation), the experimental errors, and so
forth. Despite of the observed differences, the radial non-uni-
formity of particle phase could be clearly illustrated by the
proles of normalized local solid volume fraction and particle
velocity, either from the model predictions or the experimental
data as shown in Fig. 4. When the gas velocity is raised from
3.7 m/s to 7.2 m/s, the dense-ring ow structure (i.e., the peak in
the prole) is predicted to disappear as the experimental ndings
by Zhang and Zhu (1999).
Different from the force balance on particles in the fully
developed region in risers, the gas phase in downers acts resisting
force on particles, in balance with the driving force, i.e., the force
due to gravity. When the supercial gas velocity, U
g
, increases, the
resisting effect of gas phase on the particles decreases relatively,
which could be illustrated by the cross-sectional averaged
particlegas slip velocity of 1.2 vs. 0.3 m/s for the cases of model
predictions and 4.0 vs. 2.0 m/s for the cases of experiments at U
g
of 3.7 vs. 7.2 m/s. In the case of U
g
3.7 m/s, the dense-ring ow
structure has been observed near the wall, which could be
explained by the mechanism of cluster formation as follows.
Due to the relatively strong gas-to-particles resisting effect and
0.00
0.01
0.02
0.03

2
2D channel (W=0.14m)
Ug = 4.33 m/s, Gs = 70 kg/m
2
s
dp = 70 m
A-model
B-model
C-model
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
0.00
0.01
0.02
0.03

2
2D channel (W=0.14m)
Ug = 4.33 m/s, Gs = 70 kg/m
2
s
dp = 300 m
A-model
B-model
C-model
x/( W)
x/( W)
Fig. 2. Time-averaged local solid volume fraction in the 2D channel simulation.
(a) d
p
70 mm, (b) d
p
300 mm.
0.0 0.2 0.4 0.6 0.8 1.0
0.00
0.01
0.02
0.03
0.04
r/R

2
Downer (D=0.14m)
Ug = 4.33 m/s, Gs = 70 kg/m
2
s
Expt. data
Model Prediction (Cheng et al.,1999)
Model Prediction (C-model)
0.0 0.2 0.4 0.6 0.8 1.0
0.00
0.01
0.02
0.03
0.04
Downer (D=0.14m)
Ug = 6.10 m/s, Gs = 108 kg/m
2
s
Expt. data
Model Prediction (Cheng et al., 2000)
Model Prediction (C-model)
r/R

2
0.0 0.2 0.4 0.6 0.8 1.0
0
1
2
3
4
5
6
Downer (D=0.14m)
Ug = 4.33 m/s, Gs = 70 kg/m
2
s
Expt. data
Model Prediction (Cheng et al.,1999)
Model Prediction (C-model)
r/R
V
p

(
m
/
s
)
Fig. 3. Comparison of local solid volume fraction and particle velocity between the
3D model predictions and the experimental data reported by Cheng et al. (1999,
2000). (a) solid volume fraction at U
g
4.33 m/s and G
s
70 kg/m
2
s, (b) solid
volume fraction at U
g
6.10 m/s and G
s
108 kg/m
2
s, and (c) particle velocity at
U
g
4.33 m/s, and G
s
70 kg/m
2
s.
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5360
gassolids interaction in the central region, corresponding to
higher k
12
as shown in Fig. 5(c), the particles in the central region
tend to migrate toward the wall region. On the other hand, the
particle turbulent kinetic energy will be produced through
particlewall friction as considered in current C-model with a
given value of the specularity coefcient (f) in the Johnson
Jackson boundary condition. The introduced particlewall friction
results in an increasing particle turbulent kinetic energy near the
wall (see Fig. 5(b)) and pushing the particles away from the wall.
The above two factors coexist and lead to more frequent forma-
tion and residence of clusters in near wall region, therefore the
appearance of the dense-ring ow structure. Under free slip
boundary condition, the maximum of solid volume fraction
appears at the wall (Zhang and Zhu, 1999). However, in the case
of U
g
7.2 m/s, due to the greatly reduced gas-to-particles resist-
ing effect and gassolids interaction in central region, correspond-
ing to much lower k
12
in central region but close value in wall
region compared with the case of U
g
3.7 m/s as shown in
Fig. 5(c), the particles in the wall region tend to migrate toward
the central region, undergoing a reverse tendency of radial
migration. Although the turbulence kinetic energies of the gas
phase and particle phase (k
1
and k
2
, respectively) in the central
region increase evidently when U
g
is raised from 3.7 m/s to 7.2 m/s
(see Figs. 5(a) and (b)), the time-averaged solid volume fraction in
this region is raised from below the cross-sectional average to
beyond the average value, corresponding to a transition of uid
particle velocity covariance (k
12
). The systematic and quantitative
investigations on the inuence of parameters on the dense-ring
ow structure in downers are still an important and interesting
issue in the further work.
Fig. 6 shows the radial proles of solid volume fraction,
particle velocity, particle turbulence kinetic energy, and uid
particle velocity covariance at different supercial gas velocity in
the downer (0.14 m i.d.), which is operated at G
s
70 kg/m
2
s.
0.0 0.2 0.4 0.6 0.8 1.0
1.0
1.5
2.0
0.0
0.5
1.0
1.5
p
p
V
V
2
2

Downer (D=0.10m)
Expt. data (U
g
=3.7m/s)
Model Prediction
r/R
0.0 0.2 0.4 0.6 0.8 1.0
1.0
1.5
2.0
0.0
0.5
1.0
1.5
Downer (D=0.10m)
Expt. data (U
g
=7.2m/s)
Model Prediction
r/R
2
2

p
p
V
V
Fig. 4. Comparison of normalized local solid volume fraction and particle velocity
between the 3D model predictions and the experimental data reported by Zhang
and Zhu (1999). (a) U
g
3.7 m/s and G
s
101 kg/m
2
s, and (b) U
g
7.2 m/s and
G
s
101 kg/m
2
s.
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
k
1

(
m
2
/
s
2
)
r/R
Ug (m/s)
3.7
7.2
D=0.10m G
s
=101kg/(m
2
s)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.1
0.2
0.3
0.4
D=0.10m G
s
=101kg/(m
2
s)
Ug (m/s)
3.7
7.2
k
2

(
m
2
/
s
2
)
r/R
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.1
0.2
0.3
0.4
D=0.10m G
s
=101kg/(m
2
s)
Ug (m/s)
3.7
7.2
k
1
2
(
m
2
/
s
2
)

r/R
Fig. 5. Radial proles of turbulence kinetic energy at different supercial gas
velocity in the 3D model predictions. (a) gas turbulence kinetic energy, (b) particle
turbulence kinetic energy, and (c) uidparticle velocity covariance.
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5361
When the diameter of the downer is enlarged from 0.10 m to
0.14 m, the transitional gas velocity at which the dense-ring ow
structure disappears will be increased for a given solid ux. For
the case of G
s
70 kg/m
2
s, the dense-ring ow structure still
sustains when the supercial gas velocity is increased from 4.33
to 6.10 m/s except that the solid volume fraction is decreased, as
shown in Fig. 6(a). The peak of solid volume fraction is evident
and a bit far from the disappearance of dense-ring ow structure,
which could be supported by the similar radial distributions
of particle velocity, particle turbulence kinetic energy, and
uidparticle velocity covariance, as shown in Fig. 6(b)(d),
respectively.
The predicted radial proles of solid volume fraction in the
downers with different diameter by C-model are plotted in
Fig. 7(a). The simulation results for scale-up characteristics are
close to the ones predicted by the keYk
p
model (Cheng et al.,
1999), as shown in Fig. 7(b). When the downer diameter is small
as 70 mm, the radial distribution of time-averaged solid volume
fraction in the fully developed region in the downer is relatively
uniform.
4. Conclusions
The k
1
e
1
k
2
k
12
transient two-uid model developed by
Simonins group had been adopted in the present work to predict
the dense-ring characteristics in 2D and 3D gassolids downers
by considering the inuence of gas turbulence on the ow
behavior of small size particles. Within the proposed model, the
turbulence energy interaction between gas and particles was
taken into account by a PDE (partial differential equation)
formulation of the k
12
transport equation together with a drift
velocity. To evaluate the inuence of gas turbulence on uid
dynamics of gassolid ows in dower, the 3D model predictions
using different approaches of the k
1
e
1
k
2
k
12
two-uid model,
i.e., algebraic type model (A-model), PDE-type model either
without or with drift velocity (B-model or C-model, respectively),
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
0.00
0.01
0.02
0.03
0.04
D=0.14m
Ug=4.33m/s
Ug=6.10m/s

2
r/R
0
2
4
6
8
10
D=0.14m
Ug=4.33m/s
Ug=6.10m/s
V
p

(
m
/
s
)
r/R
0.00
0.02
0.04
0.06
0.08
0.10
D=0.14m
Ug=4.33m/s
Ug=6.10m/s
k
2

(
m
2
/
s
2
)
r/R
0.00
0.04
0.08
0.12
0.16
0.20
k
1
2

(
m
2
/
s
2
)
Ug=4.33m/s
Ug=6.10m/s
D=0.14m
r/R
Fig. 6. Radial proles of solid volume fraction, particle velocity, particle turbulence kinetic energy and uidparticle velocity covariance at different supercial gas velocity
in the 3D model predictions. D0.14 m, G
s
70 kg/m
2
s.
0.00
0.01
0.02
0.03
0.04
0.05
Ug = 4.33 m/s
Gs = 70 kg/m
2
s
D (m)
0.07
0.09
0.14

2
r/R
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
0.00
0.01
0.02
0.03
0.04
0.05
Ug = 4.33 m/s
Gs = 70 kg/m
2
s
D (m)
0.07
0.09
0.14
0.50

2
r/R
Fig. 7. Radial proles of solid volume fraction in the downers with different
diameter. (a) model predictions in this study, and (b) model predictions by Cheng
et al. (2000).
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5362
were compared with each other. The simulation results showed
that the particles of small size as 70 mm in diameter apparently
interact with the gas turbulence. Using the presented C-model,
the mechanism for formation and disappearance of the dense-
ring ow structure and the scale-up characteristics of downers
were discussed. Also, the model predictions were compared with
the reported experimental data (Cheng et al., 1999, 2000; Zhang
and Zhu, 1999) for the hydrodynamics and showed good agree-
ment with the experimental data and ndings, especially for
formation and disappearance of the dense-ring ow structure at
different operating conditions.
Nomenclature
C
m
, C
1e
, C
2e
, C
3e
constants in the gas turbulence model with
values: 0.09, 1.44, 1.92, and 1.2, respectively
d
p
particle diameter, m
D
12
t
binary dispersion coefcient, m
2
/s
e coefcient of restitution for particleparticle collision
e
w
coefcient of restitution for particlewall collision
g gravity constant (9.8 m/s
2
)
g
0
radial distribution function at contact
I
mi
momentum exchange, N/m
3
k
m
turbulent kinetic energy of phase m, m
2
/s
2
k
12
uidparticle velocity covariance, m
2
/s
2
K
t
2
conductivity of solids turbulent energy, m
2
/s
P
m
pressure of phase m, Pa
S
mij
mean strain-rate tensor, 1/s
S
12ij
uidparticle strain-rate tensor, 1/s
t time, s
u
mi
local instantaneous velocity, m/s
u
ri
local instantaneous relative velocity, m/s
u
0
ri
uctuating relative phase velocity, m/s
U
di
drift velocity, m/s
U
g
supercial gas velocity, m/s
U
mi
mean velocity of phase m, m/s
U
ri
mean relative velocity, m/s
x location, m
Greek letters
a
m
volume fraction of phase m
C
d
drag coefcient
Dx width of computational cell next to the wall, m
e
1
turbulent energy dissipation in the gas phase, m
2
/s
3
e
12
dissipation term in the k
12
equation, m
2
/s
3
e
2
dissipation of solids uctuating energy due to inter-
particle collisions, m
2
/s
3
Z
r
ratio between Lagrangian and particle relaxation time
scales
f specularity coefcient
k Von Karmen constant with value: 0.42
l
2
bulk viscosity in the solids phase, kg/m/s
m
t
1
turbulent eddy viscosity for phase m, kg/m/s
n
t
12
uidparticle turbulent viscosity, m
2
/s
n
t
2
turbulent kinematic viscosity for phase m, m
2
/s
P turbulence exchange terms
r
m
density of phase m, kg/m
3
y angle between mean particle velocity and mean relative
velocity
Y
s
granular temperature equal to 2k
2
/3
s
1ij
viscous stress tensor of phase m, Pa
s
k
, s
e
constants in the gas turbulence model with values: 1.0,
1.3, respectively
S
mij
effective stress tensor, Pa
t
x
12
particle relaxation time scale, s
t
t
12
Eddy-particle interaction time scale, s
t
x
1
energetic turbulent eddies time scale, s
Subscripts
col collisional
d drift
i, j, k indices used to represent spatial direction and in Ein-
stein summation convention
m phase m, taking values 1 and 2 for gas and solids phases
max maximum packing
kin kinetic
r relative
s, p solids or particulate phase.
w wall
Acknowledgment
This work is nancially supported by NSFC under the Grants of
no. 20976091 and no. 20806045. Dr. Yi Cheng would like to thank
the Program for New Century Excellent Talents in universities
(NCET).
Appendix A
The gassolids ow models used in this study are summarized
in this appendix. Benyahia et al. (2005) model was used in most of
this study. The governing equations, constitutive relations and
wall boundary conditions are briey summarized in Tables
A1A3, respectively. The Johnson and Jackson (1987) wall bound-
ary condition was used for the particle phase.
Appendix B
See Table B1 for more details.
Table A1
Governing equations for gassolids ows.
Continuity equation for phase m (m1 for gas phase, 2 for particle phase)
@
@t
amr
m

@
@xi
amr
m
U
mi
0 (A1)

m
am 1
(A2)
Momentum equation for phase m
amr
m
@Umi
@t
U
mi
@Umi
@xj
_ _
am
@P1
@xi

mij
@xj
I
mi
amr
m
g
i (A3)
Modied ke turbulence model for the gas phase
a
1
r
1
@k1
@t
U
1j
@k1
@xj
_ _

@
@xi
a
1
m
t
1
sk
@k1
@xi
_ _
a
1
S
1ij
@Ui
@xj
P
k1
a
1
r
1
e
1 (A4)
a
1
r
1
@e1
@t
U
1j
@e1
@xj
_ _

@
@xi
a
1
m
t
1
se
@e1
@xi
_ _
a
1
e1
k1
C
1e
S
1ij
@U1i
@xj
r
1
C
2e
e
1
_ _
P
e1 (A5)
k
2
k
12
turbulence model for the particle phase
a
2
r
2
@k2
@t
U
2j
@k2
@xj
_ _

@
@xi
a
2
r
2
K
t
2
@k2
@xi
_ _
a
2
r
2
S
2ij
@U2i
@xj
P
k2
a
2
r
2
e
2 (A6)
a
2
r
2
@k
12
@t
U
2j
@k
12
@x
j
_ _

@
@x
i
a
2
r
2
n
t
12
s
k
@k
12
@x
i
_ _
a
2
r
2
S
12ij
@U
2i
@x
j

@U
1j
@x
i
_ _
P
k12
a
2
r
2
e
12
(A7)
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5363
References
Balzer, G., Simonin, O., Boelle, A., Lavieville, J., 1996. A unifying modelling
approach for the numerical prediction of dilute and dense gassolid ow. In:
Kwauk, J., Li, J. (Eds.), Circulating Fluidized Bed V. Science Press, Beijing, China,
pp. 432439.
Balzer, G., 2000. Gassolid ow modelling based on the kinetic theory of granular
media: validation, applications and limitations. Powder Technology 113, 299.
Benyahia, S., Syamlal, M., OBrien, T.J., 2005. Evaluation of boundary conditions
used to model dilute, turbulent gas/solids ows in a pipe. Powder Technology
156, 62.
Benyahia, S., Syamlal, M., OBrien, T.J., 2007. Study of the ability of multiphase
continuum models to predict core-annulus ow. A.I.Ch.E. Journal 53, 2549.
Cao, J., Ahmadi, G., 1995. Gasparticle two-phase turbulent ow in a vertical duct.
International Journal of Multiphase Flow 21, 1203.
Cheng, Y., Guo, Y.C., Wei, F., Jin, Y., Lin, W.Y., 1999. Modeling the hydrodynamics of
downer reactors based on kinetic theory. Chemical Engineer Science 54, 2019.
Cheng, Y., Wei, F., Zheng, Y., Jin, Y., Guo, Y.C., Lin, W.Y., 2000. Computational uid
dynamic modeling of hydrodynamics in downer reactors. Journal of Chemical
Industry and Engineering (China) 51, 344 (in Chinese).
Cheng, Y., Wei, F., Guo, Y.C., Jin, Y., 2001. CFD simulation of hydrodynamics in the
entrance region of a downer. Chemical Engineering Science 56, 1687.
Cheng, Y., Wu, C.N., Zhu, J.X., Wei, F., Jin, Y., 2008. Downer reactor: from
fundamental study to industrial application. Powder Technology 183, 364.
Enwald, H., Almstedt, A.E., 1999. Fluid dynamics of a pressurized uidized bed:
comparison between numerical solutions from two-uid models and experi-
mental results. Chemical Engineering Science 54, 329.
Ferschneider, G., Mege, P., 2002. Dilute gassolid ow in a riser. Chemical
Engineering Journal 87, 41.
Jian, H., Ocone, R., 2003. Modelling the hydrodynamics of gassolid suspension in
downers. Powder Technology 138, 73.
Johnson, P.C., Jackson, R., 1987. Frictionalcollisional constitutive relations for
granular materials with application to plane shearing. Journal of Fluid
Mechanics 176, 67.
Peirano, E., Leckner, B., 1998. Fundamentals of turbulent gas-solid ows applied to
circulating uidized bed combustion. Progress in Energy Combustion Science
24, 259.
Peirano, E., Delloume, V., Leckner, B., 2001. Two- or three-dimensional simulations
of turbulent gassolid ows applied to uidization. Chemical Engineering
Science 56, 4787.
Pita, J.A., Sundaresan, S., 1991. Gassolid ow in vertical tubes. A.I.Ch.E. Journal 37,
1009.
Simonin, O., Deutsch, E., Minier, J.P., 1993. Eulerian prediction of the uid/particle
correlated motion in turbulent two-phase ows. Applied Scientic Research
51, 275.
Simonin, O., 1996. Continuum modeling of dispersed two-phase ows. In:
Combustion and Turbulence in Two-Phase Flows. Von Karman Institute of
Fluid Dynamics Lecture Series, pp. 147.
Table A2
Constitutive relations for gassolids ows.
Inter-phase momentum exchange terms
I
2i
I
1i
a
2
r
2
U
ri
=t
x
12
(A8)
U
ri
U
2i
U
1i
U
di
Formulation of drift velocity, U
di
U
di
/u
0
1i
S
2
D
t
12
1
a1
@a1
@xi

1
a2
@a2
@xi
_ _
(A9)
(Simonin et al., 1993)
D
t
12

1
3
k
12
t
t
12
, t
t
12

t
t
1

1Cbx
2
r
_ , C
b
1:81:35cos
2
y, cos
2
y
Ur U2
9Ur 99U29
x
2
r
3UrUr =2k
1
(Enwald and Almstedt, 1999)
Formulation of particle relaxation time scale, t
x
12
1
t
x
12

3
4
r
1
r
2
Cd
dp
a
1:7
1
/9ur9Sa
2
if a
2
o0:2
r
1
r
2
a
2
150
Re
1:75
_
1
dp
/9ur9S if a
2
Z0:2
_
_
_
(A10)
C
d

24
Re
10:15Re
0:687
_
, Re a
1
/9ur9Sdp=n
1
/9ur9S

U
ri
U
ri
/u
0
ri
u
0
ri
S
2
_
, /u
0
ri
u
0
ri
S
2
2k
1
k
2
k
12

(Sinclair and Jackson, 1989)


Reynolds stress terms
S
1ij
2m
t
1
S
1ij

2
3
d
ij
r
1
k
1
m
t
1
@U1k
@xk
_ _
(A11)
S
1ij

1
2
@U1i
@xj

@U1j
@xi
_ _
, m
t
1
r
1
Cm
k
2
1
e1
S
2ij
n
t
2
S
2ij

2
3
d
ij
@U
2k
@x
k
_ _
P
2
l
2
@U
2k
@x
k
_ _
d
ij
(A12)
S
2ij

1
2
@U2i
@xj

@U2j
@xi
_ _
S
12ij

1
3
k
12
d
ij

Z
r
1Z
r
2
^
S
1ij

1
1Z
r
n
t
12
^
S
12ij
(A13)
(Peirano et al., 2001)
^
S
12ij
S
12ij
S
12mm
d
ij
=3, S
12ij

@U1i
@xj

@U2j
@xi
_ _
, Z
r

t
t
12
t
x
12
, n
t
12
k
12
t
t
12
=3
Turbulence interaction terms
P
k1
a
2
r
2
1
t
x
12
k
12
2k
1
U
2i
U
1i
U
di
_
, P
e1
C
3e
e1
k1
P
k1 (A14)
P
k2
a
2
r
2
1
t
x
12
2k
2
k
12

(A15)
P
k12
a
2
r
2
1
t
x
12
1X
21
k
12
k
1
2X
21
k
2
_
, X
21

a2r
2
a1r
1

1
X12
(A16)
e
2

1
3
1e
2

t
c
2
k
2
, e
12

k12
t
t
12
Other closure models
Solids pressure
P
2

2
3
a
2
r
2
k
2
12a
2
g
0
1e (A17)
Solids shear viscosity
n
t
2
n
kin
2
n
col
2
(A18)
Solids kinetic viscosity
n
kin
2

2
3
k
12
Z
r
k
2
1z
c
a
2
g
0

_
2
t
x
12

B
t
c
2
_ _
1
(A19)
z
c

2
5
1e3e1, B
1
5
1e3e
Solids collisional viscosity
n
col
2

4
5
a
2
g
0
1e n
kin
2
dp

2k2
3p
_
_ _
(A20)
Solids bulk viscosity
l
2

5
3
a
2
r
2
n
col
2
(A21)
Solids granular conductivity
K
t
2
K
kin
2
K
col
2
(A22)
Solids kinetic turbulent conductivity
K
kin
2

2
3
9
10
k
12
Z
r
k
2
1$c a
2
g
0

_
9
5t
x
12

xc
t
c
2
_ _
1
(A23)
$c
1
100
1e
2
2e1
Solids collisional turbulent conductivity
K
col
2

6
5
a
2
g
0
1e K
kin
2

5
9
dp

2k2
3p
_

_
(A24)
The radial distribution function
g
0
1
a2
g
max
2
_ _
1=3
_ _
1
(A25)
Constants in ke model
s
k
, se, Cm, C
1e
, C
2e
, C
3e
1:0, 1:3, 0:09, 1:44, 1:92 and 1:2, respectively (A26)
Table A3
Wall boundary conditions.
Gas phase wall boundary condition (Benyahia et al., 2005)
@U1
@x

w

r
1
kU1C
1=4
1m
k
1=2
1
m
1
m
t
1
lnEx
n

, x
n

r
1
C
1=4
1m
k
1=2
1
Dx
2m
1
(A27)
Particle phase wall boundary conditions (Johnson and Jackson, 1987)
For particle velocity
n
t
2
@U2
@x

w
fpU
2
g
0

2=3k2
p
2

3
p
a
max
2
0
(A28)
For uctuation energy
K
t
2
@k2
@x

fpU
2
2
g0

2=3k2
p
2

3
p
a
max
2

3
p
pg01e
2
w
2=3k2
3=2
4a
max
2
0
(A29)
A Dirichlet wall boundary condition for k
1
, e
1
and k
12
k
1

w
0, e
1
j
w
0, k
12

w
0 (A30)
Table B1
Description of differences for three models used in this study.
Model name Calculation method of k
12
Major assumptions
A-model Solving the algebraic
formulation, k
12

2Z
r
k1 X12k2
1 1X12Z
r
instead of Eq. (A7)
Homogeneous and
stationary two-phase
turbulence in the case
of very massive particle
B-model Solving Eq. (A7) with U
di
0 Neglected effect of drift velocity
C-model Solving Eq. (A7) Full consideration of the
uctuations of uid velocity
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5364
Sinclair, J.L., Jackson, R., 1989. Gasparticle ow in a vertical pipe with particle
particle interactions. A.I.Ch.E. Journal 35, 1473.
Syamlal, M., Rogers, W.A., OBrien, T.J., 1993. MFIX Documentation and Theory
Guide. DOE/METC-94/1004. NTIS DE94000087. Electronically Available at:
/http://www.mx.org/documentation/Theory.pdfS.
Vaishalia, S., Roy, S., Mills, P.L., 2008. Hydrodynamic simulation of gassolids
downow reactors. Chemical Engineering Science 63, 5107.
Wang, Z.W., Bai, D.R., Jin, Y., 1992. Hydrodynamics of cocurrent downow
circulating uidized bed (CDCFB). Powder Technology 70, 271.
Zhang, Y., Reese, J., 2003. Gas turbulence modulation in a two-uid model for gas
solid ows. A.I.Ch.E. Journal 49, 3048.
Zhang, H., Zhu, J.X., 1999. Hydrodynamics in downow uidized beds. (1) Solids
concentration proles and pressure gradient distributions. Chemical Engineer-
ing Science 54, 5461.
Zhu, J.X., Yu, Z.Q., Jin, Y., Grace, J.R., Issangya, A., 1995. Cocurrent downow
circulating uidized bed (downer) reactorsa state of the art review.
Canadian Journal of Chemical Engineering 73, 662.
Y.N. Kim et al. / Chemical Engineering Science 66 (2011) 53575365 5365

You might also like