You are on page 1of 9

Particuology 7 (2009) 269–277

Contents lists available at ScienceDirect

Particuology
journal homepage: www.elsevier.com/locate/partic

Drag models for simulating gas–solid flow in the turbulent fluidization


of FCC particles
Peng Li, Xingying Lan, Chunming Xu, Gang Wang, Chunxi Lu, Jinsen Gao ∗
State Key Laboratory of Heavy Oil Processing, China University of Petroleum, Beijing 102249, China

a r t i c l e i n f o a b s t r a c t

Article history: This paper examines the suitability of various drag models for predicting the hydrodynamics of the
Received 27 November 2008 turbulent fluidization of FCC particles on the Fluent V6.2 platform. The drag models included those of
Accepted 22 March 2009 Syamlal–O’Brien, Gidaspow, modified Syamlal–O’Brien, and McKeen. Comparison between experimental
data and simulated results showed that the Syamlal–O’Brien, Gidaspow, and modified Syamlal–O’Brien
Keywords: drag models highly overestimated gas–solid momentum exchange and could not predict the formation
Turbulent fluidized bed
of dense phase in the fluidized bed, while the McKeen drag model could not capture the dilute charac-
FCC particle
teristics due to underestimation of drag force. The standard Gidaspow drag model was then modified
Drag model
CFD
by adopting the effective particle cluster diameter to account for particle clusters, which was, however,
proved inapplicable for FCC particle turbulent fluidization. A four-zone drag model (dense phase, sub-
dense phase, sub-dilute phase and dilute phase) was finally proposed to calculate the gas–solid exchange
coefficient in the turbulent fluidization of FCC particles, and was validated by satisfactory agreement
between prediction and experiment.
© 2009 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of
Sciences. Published by Elsevier B.V. All rights reserved.

1. Introduction (1966), Syamlal and O’Brien (1989), and Gidaspow drag models
(Gidaspow, 1994). Many researchers have successfully simulated
The turbulent fluidized bed (TFB) is widely used in various com- the circulating fluidized bed of FCC particles using the classical
mercial processes including fluid catalytic cracking, regeneration drag models (Benyahia, Arastoopour, Knowlton, & Massah, 2000;
and particle drying, because of its excellent gas–solid contacting, Chan, Guo, & Lau, 2005; Neri & Gidaspow, 2000; Zheng, Wan,
favorable heat transfer, and relatively low axial dispersion of gas. In Qian, Wei, & Jin, 2001). However, few successful simulations were
spite of its many applications, turbulent fluidization has received reported on dense fluidization of Geldart A particles. The CFD mod-
much less attention than bubbling and fast fluidization due to the eling of a bubbling FCC fluidized-bed reactor by Zimmermann and
current deficiencies in experimental and theoretical work. Optimal Taghipour (2005) showed that the drag models of Syamlal–O’Brien
design and scale-up of TFB require fundamental understanding of and Gidaspow overestimated the momentum exchange between
its hydrodynamic behavior. the gas and the solid phase and overpredicted bed expansion in
Recently, computational fluid dynamics (CFD) has become a comparison to experimental data. McKeen and Pugsley (2003) sim-
useful tool for understanding the hydrodynamics and transfer ulated a freely bubbling bed of FCC particles at a superficial gas
mechanisms in multiphase flow systems. Previous simulations velocity of 0.05–0.20 m/s using a two-fluid CFD model. They found
indicated that the drag force between particle and fluid plays that the generally poor simulation results for Geldart A particles
an essential role in the prediction of the flow structure of a flu- could be attributed to the existence of significant cohesive inter-
idization bed (Beetstra, van der Hoef, & Kuipers, 2007; Helland, particle forces. Even earlier, Massimilla and Donsi (1976) reported
Bournot, Occelli, & Tadrist, 2007; McKeen & Pugsley, 2003; Yang, that the cohesive force between particles of 40–100 ␮m played
Wang, Ge, & Li, 2003; Zimmermann & Taghipour, 2005). Several an important role in stabilizing the flow behavior. Such cohesive
drag models have been developed to calculate the inter-phase force led to cluster formation of FCC particles, resulting in larger
momentum exchange in fluidized bed, such as the Wen and Yu effective particle sizes and hence reduced fluid-particle drag forces.
Hence, the standard drag correlations must be corrected to properly
account for the clusters of Geldart A particles.
Lettieri, Newton, and Yates (2002) analyzed the homogeneous
∗ Corresponding author. Tel.: +86 10 89733993. expansion data for FCC particles in terms of the Richardson–Zaki
E-mail address: jsgao@cup.edu.cn (J. Gao). equation. They observed that the experimentally obtained termi-

1674-2001/$ – see front matter © 2009 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.partic.2009.03.010
270 P. Li et al. / Particuology 7 (2009) 269–277

by Jiradilok, Gidaspow, Damronglerd, Koves, and Mostofi (2006)


Nomenclature to predict the hydrodynamics of turbulent fluidization of FCC
particles.
CD drag coefficient
The above evidences suggest that various drag models have been
d diameter of particle
successfully applied to some extent to respective systems. However,
dp∗ effective mean diameter of particle
there are currently no universal properly validated drag models
e restitution coefficient
for gas–solid fluidized beds. In this study, we sought for an appro-
gi acceleration due to gravity
priate drag model to describe the hydrodynamics of the turbulent
g0 radial distribution function
fluidization of FCC particles. The performance of Syamlal–O’Brien,
h bed height
Gidaspow, modified Syamlal–O’Brien and McKeen drag models
p pressure
was first examined. Then, various modified Gidaspow drag mod-
P parameter of Syamlal–O’Brien drag model
els based on effective particle cluster diameters were compared to
Q parameter of Syamlal–O’Brien drag model
investigate their suitability. Finally, we proposed a four-zone drag
r radial direction
model to predict the gas–solid inter-phase exchange coefficient in
R diameter of bed
the turbulent fluidization of FCC particles.
Rep particle Reynolds number
Rep∗ cluster Reynolds number
t time 2. Hydrodynamic model
u velocity
u instantaneous turbulent component of velocity The Eulerian–Eulerian (two-fluid) model based on the kinetic
ug superficial gas velocity theory of granular flow is applied to the simulation of gas–solid tur-
ut terminal velocity bulent flow. Both the gas phase and the solid phase are modeled as
u∗t experimental terminal velocity interpenetrating continua with similar conservation equations. The
x direction coordinate interactions between the two phases are expressed as additional
source terms added to the conservation equations. The kinetic the-
Greek symbols ory of granular flow is used to define the fluid properties of the solid
˛ volume fraction phase through constitutive equations.
ˇ inter-phase momentum exchange coefficient
 collisional dissipation of energy fluctuation
2.1. Conservation equations
p solid bulk viscosity
 granular temperature
The conservation equations for mass and momentum for the gas
 solid phase shear viscosity
and solid phases as well as the particulate phase fluctuating energy
 density
are as follows.
ϕ switch function
Mass conservation equation of gas (g) and particulate phases
 stress tensor
(p):
diffusion coefficient for the energy fluctuation
∂(˛g g ) ∂
Subscripts + (˛g g ug j ) = 0, (1)
∂t ∂xj
i, j, k direction coordinate
g gas phase ∂(˛p p ) ∂
p particulate phase + (˛p p upj ) = 0. (2)
∂t ∂xj

Each computational cell is shared by the interpenetrating


nal velocity was much higher than the calculated. If a homogeneous phases, so that the sum of their volume fractions is unity.
bed was assumed to be characterized by the presence of clus-
˛g + ˛p = 1. (3)
ters, the diameter of clusters, dp∗ , can be back-calculated from the
experimental values. Lettieri et al. obtained the effective particle
Momentum conservation equation of gas and particulate
diameters in the range of 200–474 ␮m for FCC catalysts with Sauter
phases:
mean diameters from 49 to 71 ␮m. Gao, Chang, Xu, Lan, and Yang
(2008) modified the classical Gidaspow drag model using an effec- ∂ ∂
tive diameter of 265 ␮m instead of the actual particle diameter of (˛g g ugi ) + (˛g g ugi ugj )
∂t ∂xj
58 ␮m, and successfully simulated the flow behavior in FCC strip-
pers. ∂p ∂g,ij
= −˛g + − ˇ(ugi − upi ) + g ˛g gi , (4)
McKeen and Pugsley (2003) also proposed an empirical method ∂xi ∂xj
to reduce the Gibilaro drag (Gibilaro, Di Felice, Waldram, &
Foscolo, 1985) using a constant scale factor, adjustable to take
into account the effect of inter-particle cohesive forces on parti- ∂ ∂
(˛p p upi ) + (˛p p upi upj )
cle agglomeration. They concluded that the Gibilaro drag model ∂t ∂xj
with a correction factor of 0.25, corresponding to an effective
∂p ∂p,ij
particle agglomerate diameter of 135–170 ␮m for FCC parti- = −˛p + − ˇ(upi − ugi ) + p ˛p gi . (5)
∂xi ∂xj
cles with actual mean diameter of 75 ␮m, could adequately
predict the experimentally observed bubbling fluidization behav-
ior. Yang et al. (Yang et al., 2003; Yang, Wang, Ge, Wang, The fluctuations that occur in the particulate phase are modeled
& Li, 2004) presented a drag model based on the energy- from the kinetic theory of gases modified to account for inelastic
minimization multi-scale (EMMS) approach, which was applied collisions between particles. The equation for the turbulent fluctu-
P. Li et al. / Particuology 7 (2009) 269–277 271

Table 1
Constitutive equations.
 
∂ugj ∂ugi
Stress tensor of gas phase g,ij = g + (7)
∂xi ∂xj
   
∂upj ∂upi 2 ∂upk
Stress tensor of particulate phase p,ij = p + + p − p ıij − pp ıij (8)
∂xi ∂xj 3 ∂xk
Particulate pressure pp = ˛p p [1 + 2(1 + e)˛p g0 ] (9)
  1/3 −1
˛p
Radial distribution function g0 = 1− (10)
˛p,max

1  
Granular temperature = u u  (11)
3 p p
√  2 
10p dp  4 4 2 
Solids phase shear viscosity p = 1+ (1 + e)g0 ˛p + ˛ p dp g0 (1 + e) (12)
96(1 + e)g0 5 5 p

4 2 
Solids bulk viscosity p = ˛ p dp g0 (1 + e) (13)
3 p
√  2 
150p dp  6 
Granular energy diffusion coefficient  = 1+ (1 + e)g0 ˛p + 2˛2p p dp g0 (1 + e) (14)
384(1 + e)g0 5
  
4  ∂upk
Collision dissipation energy  = 3(1 − e2 )˛2p p g0  − (15)
dp ∂xk

ating energy of the particulate phase is expressed as: determined minimum fluidization velocity.

3 ∂ ˛p p  ∂ ˛p p upk  ˛g g
+ vg = Ret (24)
2 ∂t ∂xk ds g

 
∂ ∂ ∂upk ∂upi ∂upk ∂upk Ret = vr,s Rets (25)
=  + p + − pp
∂xk ∂xk ∂xi ∂xk ∂xi ∂xk
A + 0.06BRets
  ∂u 2 vr,s = (26)
2 pk 1 + 0.06Rets
+ p − p − . (6)
3 ∂xk

Table 2
2.2. Constitutive equations Drag models.

Syamlal–O’Brien drag model (Syamlal & O’Brien, 1989)


 
CD u ៝ g
Several closure models have been proposed to define appro- 3˛p ˛g g
ˇ= ៝p − u (16)
priate constitutive equations for multi-phase flows based on the 4v2r,p dp
kinetic theory of granular flow. The constitutive equations for
 2
4.8
closing the above conservation relations are described in Table 1 CD = 0.63 +  (17)
(Gidaspow, 1994), in which e in Eqs. ((9), (12)–(15)) is the restitution Rep /ur,p
coefficient, which quantifies the elasticity of inter-particle collision ៝p − u
g dp |u ៝ g|
resulting in kinetic energy dissipation. The restitution coefficient Rep = (18)
g
ranges from 0 for inelastic collisions with complete energy dissipa-   
2
tion to 1 for fully elastic collisions with no energy dissipation. ur,p = 0.5 A − 0.06Rep + (0.06Rep ) + 0.12Rep (2B − A) + A2 (19)


P˛1.28
g (˛g ≤ 0.85)
A = ˛4.14 , B= (20)
2.3. Drag models g
˛Qg (˛g > 0.85)
P = 0.8, Q = 2.65
A proper drag mode is required to close ˇ in Eqs. (4) and (5).
To investigate the suitability of drag laws for modeling the tur- ⎧
Gidaspow drag model (Gidaspow, 1994)  
⎪  ˛ u ៝ −u៝ 
bulent fluidization of FCC particles, the classical drag models of ⎨ ˇ = 150 ˛p (1 − ˛2g ) g + 1.75 g p p g for ˛g ≤ 0.8
Syamlal–O’Brien and Gidaspow were first investigated. Then, the ˛g dp  dp
(21)
⎪  ៝ g
៝p − u
modified drag models including McKeen drag model, modified ⎩ ˇ = 3 CD ˛p ˛g g u
˛−2.65 for ˛g > 0.8
Syamlal–O’Brien drag model were examined. The correlations of 4 dp g

Syamlal–O’Brien, Gidaspow and McKeen drag models are summa-  24


rized in Table 2. The modification of the Syamlal–O’Brien drag law (1 + 0.15Rep0.687 ) (Rep ≤ 1000)
CD = Rep (22)
is based on the minimum fluidization conditions (Zimmermann 0.44 (Rep > 1000)
& Taghipour, 2005). The parameter P in Eq. (20) is related to the McKeen drag model (McKeen & Pugsley, 2003)
   
g u ៝ g
minimum fluidization velocity through the velocity-voidage cor- ៝p − u
relation and the terminal Reynolds number, Ret . P is varied until 17.3
ˇ=C + 0.336 ˛p ˛g −1.8 (23)
Rep dp
vg , expressed by the following equation, equals the experimentally
272 P. Li et al. / Particuology 7 (2009) 269–277

Table 3
Modified Gidaspow drag models based on the effective cluster diameter.

Void fraction ≤0.80 0.8−0.99 >0.99


     
˛p (1 − ˛g ) g g ˛p u ៝ g
៝p − u
3 ˛p ˛g g u ៝ g
៝p − u
3 ˛p g u ៝ g
៝p − u
Drag A ˇ = 150 + 1.75 ∗ ˇ = CD ˛−2.65
g ˇ = CD

2 dp 4 dp 4 dp
˛g dp
     
˛p (1 − ˛g ) g g ˛p u ៝ g
៝p − u
3 ˛p ˛g g u ៝ g
៝p − u
3 ˛p g u ៝ g
៝p − u
Drag B ˇ = 150 + 1.75 ˇ = CD ∗
˛−2.65
g ˇ = CD
˛g dp2 dp 4 dp 4 dp
     
˛p (1 − ˛g ) g g ˛p u ៝ g
៝p − u
3 ˛p ˛g g u ៝ g
៝p − u
3 ˛p g u ៝ g
៝p − u
Drag C ˇ = 150 + 1.75 ˇ = CD ˛−2.65
g ˇ = CD
2 dp∗ 4 dp∗ 4 dp
˛g dp∗

⎛  ⎞2 those of the actual experiment. In the following investigation on


4.82 + 2.52 4Ar/3 − 4.8 drag models, all the simulations were carried out under the con-
Rets = ⎝ ⎠ (27) dition of 1.0 m for initial bed height and 0.5 m/s for superficial gas
1.26
velocity.
(s − g )ds3 g g
Ar = (28) 3.1. Syamlal–O’Brien and Gidaspow drag models
g

The parameter Q in Eq. (20) has to be modified according to Eq. Fig. 1 depicts the flow structure of the bed by the classical
(29): Syamlal–O’Brien and Gidaspow drag models, showing that there
1.28 + log(P) is no formation of the dense phase, the flow being more charac-
Q = . (29) teristic of fast fluidization. Fig. 2 compares the bed density profiles
log(0.85)
predicted by classical drag models with experimental data, indicat-
Accounting for the effect of clustering of FCC particles on the drag ing underestimated bed density in the dense phase of the bed and
force between gas and solid phases, we modified the standard overestimated in the dilute phase. As reported by other researchers
Gidaspow drag model using an effective particle cluster diame- (McKeen & Pugsley, 2003; Zimmermann & Taghipour, 2005), the
ter. For example, the actual 60 ␮m FCC particle was replaced by an Syamlal–O’Brien and Gidaspow drag models highly over-predict
effective cluster diameter of 300 ␮m according to the experimental the momentum exchange between the gas and the solid phases
terminal velocity of FCC catalyst. Table 3 shows the various modi- for small Geldart A particles due to neglect of inter-particle cohe-
fied Gidaspow drag models based on the effective particle cluster sive forces and agglomeration. Hence, the Syamlal–O’Brien and
diameter for different void fractions. Gidaspow drag models fail to predict the hydrodynamics of FCC
particle (Geldart A) in a turbulent fluidized bed.
2.4. Simulation setup
3.2. Modified Syamlal–O’Brien and McKeen drag models
The governing equations were solved using the finite volume
method by Patankar (1980). The differential equations were dis- While the flow structure by the modified Syamlal–O’Brien and
cretized by a first-order upwind differencing scheme over the finite McKeen drag models is displayed in Fig. 3, Fig. 4 compares the
volume used, and solved by the commercial CFD package Fluent bed density profiles predicted by the McKeen and the modified
V6.2. The Phase Coupled SIMPLE (PC-SIMPLE) algorithm, which is Syamlal–O’Brien drag models with experimental data. The flow
an extension of the SIMPLE algorithm for multiphase flow, was used pattern by the modified Syamlal–O’Brien drag model is similar to
for the pressure–velocity coupling and correction. that by the classical drag models with the same overestimation.
All simulations were performed in 2D Cartesian space. The Although the modified Syamlal–O’Brien drag model based on the
dimensions of the computational domain in radial and axial direc- minimum fluidization conditions of FCC particles decreased the
tions were the same as those of the actual experimental fluidized drag between the gas and the solid phase, it still does not cap-
bed. The 2D computational domain was discretized by 19,400 ture the basic characteristics of the turbulent fluidization of FCC
(100 × 194) rectangular cells, which confirmed the mesh inde- particles.
pendency (Gao et al., 2009). Because of the usual instability and Although the McKeen drag model captures the significant
convergence for multiphase simulation, a very small time step dense-phase characteristics of bubble splitting and coalescence, it
(0.0001s) with about 20 iterations per time step was used. A misses the dilute-phase characteristics. The McKeen drag model
convergence criterion of 10−3 for each scaled residual compo-
nent was specified for the relative error between two successive
Table 4
iterations.
Properties of FCC particles and air.

3. Results and discussion FCC particles


Size range 20−130 ␮m
Mean diameter 60 ␮m
Experimental data needed for evaluating the various drag mod- Particle density 1500 kg/m3
els and hydrodynamic CFD predictions were acquired in a plexiglass Minimum fluidization velocity 0.0047 m/s
gas–solid fluidized bed of 0.5 m diameter and 4.0 m height. Detailed Terminal velocity 0.58 m/s
discussions of the experimental apparatus and measurement tech- Air
niques were provided elsewhere (Cao, 2006). The properties of Density 1.225 kg/m3
FCC particles and air (used for the CFD modeling setup) are sum- Viscosity 1.79 × 10−5 kg/(ms)
Superficial gas velocity 0.27–0.62 m/s
marized in Table 4. The simulation conditions are the same as
P. Li et al. / Particuology 7 (2009) 269–277 273

Fig. 1. Flow structure with classical drag models: (a) Syamlal–O’Brien and (b)
Gidaspow. Fig. 3. Flow structure with modified drag models: (a) modified Syamlal–O’Brien and
(b) McKeen.

decreases the standard drag law by a scaling factor of 0.2–0.3 to


eter of 300 ␮m instead of 60 ␮m for actual FCC particles. Fig. 6
account for the cohesive inter-particle forces (McKeen & Pugsley,
compares the resulting bed density profiles with experimental data,
2003). For turbulent fluidization of FCC particles, the scaling factor
showing that, similar to the classical drag models, the flow behav-
of 0.2–0.3 is yet too small to show the formation of the dilute phase.
ior predicted by drag model A involves the same overestimation
of fluidization. Drag model B overestimates the drag force for FCC
3.3. Modified Gidaspow drag models based on effective cluster particles, thus predicting larger bed expansion than experiment.
diameter Similar to the McKeen drag model, model C underestimates the
drag force and fails to reveal the dilute-phase characteristics. The
Fig. 5 illustrates the flow structure by the various modified discrepancy between modeling and experimental results indicates
Gidaspow drag models based on an effective particle cluster diam- that the rough modification of the Gidaspow drag model based

Fig. 2. Bed density profiles predicted with classical drag models. Fig. 4. Bed density profiles predicted with modified drag models.
274 P. Li et al. / Particuology 7 (2009) 269–277

Fig. 6. Bed density profiles predicted with variously modified Gidaspow drag mod-
els.

Helland et al. (2007) supposed that there co-exist two antag-


onistic drag effects on individual particles depending on the
inter-particle distance within a cluster. Dilute clusters lead to a
decrease of the drag coefficient on the total cluster area, while
dense clusters result in an increase of the drag coefficient on the
total cluster area. These authors applied this combined-drag sup-
position to observe a dense bottom zone, a dilute upper zone and
a sharp in-between transition zone in the bed.
To account for the fact that the drag force is affected by the
degree of clustering, and the degree of clustering is affected by the
Fig. 5. Flow structure with variously modified Gidaspow drag models: (a) Drag A,
void fraction, we identified four zones for the turbulent fluidization
(b) Drag B and (c) Drag C.
of FCC particles: dense phase, sub-dense phase, sub-dilute phase
and dilute phase zones, in accordance to bed density. The transition
on effective particle cluster diameter is inapplicable for turbulent points between zones are respectively at 300, 100 and 15 kg/m3 , the
fluidization of FCC particles. corresponding void fractions are 0.8, 0.933 and 0.99. The transition
point of 0.8 was referred to in the Gidaspow drag model. The zone
3.4. Four-zone drag model with bed density of less than 100 kg/m3 is generally believed to
be the dilute phase in experiments. In the turbulent fluidization of
The turbulent fluidized bed is characterized by two differ- FCC particles studied in this paper, the void fraction was 0.933 at the
ent coexisting regions: a bottom dense, bubbling region and a bed density of 100 kg/m3 , which was used as the transition point
top dilute, dispersed region (Berruti, Chaouki, Godfroy, Pugsley, & between the sub-dense and sub-dilute phases. When the void frac-
Patience, 1995). Various classical and modified drag models men- tion is larger than 0.99, the drag force on an individual particle is not
tioned above fail to predict the co-existence of the dense and dilute affected by other neighboring particles, thus not calling for modi-
phases in the turbulent fluidization of FCC particles. fication. Hence, the void fraction of 0.99 is regarded as a transition

Table 5
Four-zone drag model.

Bed density, kg/m3 Void fraction Drag correlations


 
˛p (1 − ˛g ) g g ˛p u ៝ g
៝p − u
≥300 ≤0.80 (dense) ˇ1 = 150 + 1.75 (30)
2 dp∗
˛g dp∗
 
៝ ៝ 
5 ∗ ˛p ˛g g up − ug
100–300 0.8–0.933 (sub-dense) ˇ2 = CD (31)
72 d∗ (1 − ˛g )
0.293
p
 
3 ˛p ˛g g u ៝ g
៝p − u
15–100 0.933–0.99 (sub-dilute) ˇ3 = CD ˛−2.65
g (32)
4 dp
 
3 ˛p ˛g g u ៝ g
៝p − u
<15 >0.99 (dilute) ˇ4 = CD (33)
4 dp
 24
(1 + 0.15Rep∗0.687 ) (Rep∗ ≤ 1000) ˛g g dp∗ |−

ug −−

u p|
where CD∗ = Rep∗ Rep∗ = ,
g
0.44 (Rep∗ > 1000)

 24
(1 + 0.15Rep0.687 ) (Rep ≤ 1000) ˛g g dp |−

ug −−

u p|
CD = Rep Rep = .
g
0.44 (Rep > 1000)
P. Li et al. / Particuology 7 (2009) 269–277 275

point between the sub-dilute and dilute phases. In the dense phase
zone (˛g ≤ 0.80) and sub-dense phase zone (0.8 < ˛g ≤ 0.933), there
coexist dense clusters and dilute clusters; so Ergun drag model
and ZP drag model (Cao, Gao, Zhang, Zheng, & Xu, 2004) modi-
fied with an effective diameter of 300 ␮m were used, respectively.
In the sub-dilute phase zone (0.933 < ˛g ≤ 0.990) and dilute phase
zone (˛g > 0.990), the Wen and Yu drag model and single particle
drag model were employed, respectively. These drag correlations
are presented in Table 5.
The step changes in the drag coefficient are observed at the
“crossover” void fractions of 0.8, 0.933, and 0.99, which can possibly
lead to difficulties in numerical convergence. To avoid the discon-
tinuous behavior, four drag correlations were stitched together by
the following equation:

ˇ = (1 − ϕ1 )ˇ1 + ϕ1 {(1 − ϕ2 )ˇ2 + ϕ2 [(1 − ϕ3 )ˇ3 + ϕ3 ˇ4 ]}, (34)


Fig. 7. Computed drag coefficients as a function of void fraction.
where ϕ1 , ϕ2 , ϕ3 were the stitching functions, which were calcu-
lated by the expression proposed by Lu and Gidaspow (2003). Fig. 7
shows the drag coefficients as a function of void fraction before and the bed is dense and the upper part is dilute, with significant bub-
after stitching, to illustrate the smooth transition between the drag ble splitting and coalescence in the former and dilute clusters of
correlations. particles in the latter. Fig. 9 compares the computed bed density
profile, averaged from 22 to 26 s, with experiment, showing satis-
arctan[150 × 1.75(˛g − ˛i )]
ϕi = + 0.5 (35) factory agreement between prediction and experiment, testifying
to the fact that the four-zone drag model is capable of correctly
where ˛1 , ˛2 , ˛3 were the transition points, ˛1 = 0.8, ˛2 = 0.933, describing the coexistence of the dense and dilute regimes for the
˛3 = 0.99. turbulent fluidization of FCC particles.
Fig. 8 shows the instantaneous flow structures predicted by the To further validate the four-zone drag model, Fig. 10 presents
four-zone drag model, clearly indicating that the bottom part of the instantaneous flow structure in the fluidized bed at the same

Fig. 8. Instantaneous flow structure with the four-zone drag model.


276 P. Li et al. / Particuology 7 (2009) 269–277

moment of 30 s for various gas velocities, and Fig. 11 compares


the bed density between simulated and experimental results, both
illustrating the two different coexisting regions: a bottom dense,
bubbling region and a dilute, dispersed flow region, highly consis-
tent with experimental phenomena. Moreover, the predicted bed
density profiles in the axial direction are in reasonable agreement
with experimental results. It can also be seen that with increasing
superficial gas velocity, turbulencies aggravated and the entrain-
ing capability of the gas phase improves, that is, more particles
are entrained into the dilute phase, thus decreasing the height and
density of the dense phase, and increasing the density of the dilute
phase.

4. Conclusion
Fig. 9. Bed density profile predicted with the four-zone drag model.
An Eulerian–Eulerian CFD model based on the kinetic theory
of granular flow, and using commercial CFD package Fluent V6.2,
was applied to simulate the hydrodynamics of the turbulent flu-
idization of FCC particles. Various drag models including those
of Syamlal–O’Brien, Gidaspow, modified Syamlal–O’Brien, McKeen
and the modified Gidaspow based on effective cluster diameter,
were examined and compared to study their respective suitabil-
ity for predicting the hydrodynamics of the turbulent fluidization
of FCC particles. The simulated results showed that they all failed
due to either underestimation or overestimation of the drag force
between the gas and the solid phase. To account for the fact that the
drag force is affected by the degree of clustering and the degree of
clustering is affected by the void fraction, a four-zone drag model
was proposed to calculate the gas–solid exchange coefficient in the
turbulent fluidization of FCC particles, and was validated with sat-
isfactory agreement between prediction and experiment. Results
of the present simulation demonstrate that the core of a proper
drag model lies in the simulation of the coexistence of both the
dense and the dilute regimes in a turbulent fluidized bed. How-
ever, the proposed four-zone drag model calls for further validation
by more experimental and modeling efforts, and the underlying
mechanisms need further exploration.

Acknowledgments

The authors acknowledge the supports by the National Natural


Science Foundation of China through the programs for Distin-
guished Young Scholars of China (Grant No. 20725620 and Grant No.
20525621) and the programs “Multiple Scale Analysis and Scaling-
Fig. 10. Flow structure at various superficial gas velocities: (a) 0.36 m/s, (b) 0.53 m/s
up of Direct Coupled Dual Gas-Solid Fluidized Reaction Systems”
and (c) 0.62 m/s. (Grant No. 20490202).

Fig. 11. Comparison of the predicted bed densities with experimental data at ug : (a) 0.36 m/s, (b) 0.53 m/s and (c) 0.62 m/s.
P. Li et al. / Particuology 7 (2009) 269–277 277

References Jiradilok, V., Gidaspow, D., Damronglerd, S., Koves, W. J., & Mostofi, R. (2006). Kinetic
theory based CFD simulation of turbulent fluidization of FCC particles in a riser.
Beetstra, R., van der Hoef, M. A., & Kuipers, J. A. M. (2007). Numerical study Chemical Engineering Science, 61, 5544–5559.
of segregation using a new drag force correlation for polydisperse systems Lettieri, P., Newton, D., & Yates, J. G. (2002). Homogeneous bed expansion of FCC
derived from lattice-Boltzmann simulations. Chemical Engineering Science, 62, catalysts, influence of temperature on the parameters of the Richardson–Zaki
246–255. equation. Powder Technology, 123, 221–231.
Benyahia, S., Arastoopour, H., Knowlton, T. M., & Massah, H. (2000). Simulation of Lu, H., & Gidaspow, D. (2003). Hydrodynamics of binary fluidization in a riser: CFD
particles and gas flow behavior in the riser section of a circulating fluidized bed simulation using two granular temperatures. Chemical Engineering Science, 58,
using the kinetic theory approach for the particulate phase. Powder Technology, 3777–3792.
112, 24–33. Massimilla, L., & Donsi, G. (1976). Cohesive forces between particles of fluid-bed
Berruti, F., Chaouki, J., Godfroy, L., Pugsley, T. S., & Patience, G. S. (1995). Hydrody- catalysts. Powder Technology, 15, 253–260.
namics of circulating fluidized bed risers: A review. Canadian Journal of Chemical McKeen, T., & Pugsley, T. (2003). Simulation and experimental validation of a freely
Engineering, 73, 579–602. bubbling bed of FCC catalyst. Powder Technology, 129, 139–152.
Cao, B. (2006). Study on the flow behavior of gas-solid fluidized bed with large dif- Neri, A., & Gidaspow, D. (2000). Riser hydrodynamics: Simulation using kinetic the-
ference mixing particles. Unpublished doctoral dissertation, China University of ory. AIChE Journal, 46, 52–67.
Petroleum, Beijing, China (in Chinese). Patankar, S. V. (1980). Numerical heat transfer and fluid flow. Washington, DC: Hemi-
Cao, B., Gao, J., Zhang, P., Zheng, X., & Xu, C. (2004). Numerical simulation sphere Publishing Corporation.
on the gas-particle flows in FCC regenerator. In In the second international Syamlal, M., & O’Brien, T. J. (1989). Computer simulation of bubbles in a fluidized
symposium on multiphase, non-Newtonian and reacting flows’04 Hangzhou, bed. AIChE Symposium Series, 85, 22–31.
China. Wen, C. Y., & Yu, Y. H. (1966). Mechanics of fluidization. Chemical Engineering Progress
Chan, C. K., Guo, Y. C., & Lau, K. S. (2005). Numerical modeling of gas-particle flow Symposium Series, 62, 100–111.
using a comprehensive kinetic theory with turbulence modulation. Powder Tech- Yang, N., Wang, W., Ge, W., & Li, J. (2003). CFD simulation of concurrent-up gas–solid
nology, 150, 42–55. flow in circulating fluidized beds with structure-dependent drag coefficient.
Gao, J., Chang, J., Xu, C., Lan, X., & Yang, Y. (2008). CFD simulation of gas solid flow in Chemical Engineering Journal, 96, 71–80.
FCC strippers. Chemical Engineering Science, 63, 1827–1841. Yang, N., Wang, W., Ge, W., Wang, L., & Li, J. (2004). Simulation of heteroge-
Gao, J., Lan, X., Fan, Y., Wang, G., Lu, X., & Xu, C. (2009). CFD modeling and validation neous structure in a circulating fluidized-bed riser by combining the two-fluid
of the turbulent fluidized bed. AIChE Journal, 55, 1680–1694. model with the EMMS approach. Industrial & Engineering Chemistry Research, 43,
Gibilaro, L. G., Di Felice, R., Waldram, S. P., & Foscolo, P. U. (1985). Generalized friction 5548–5561.
factor and drag coefficient correlations for fluid-particle interactions. Chemical Zheng, Y., Wan, X., Qian, Z., Wei, F., & Jin, Y. (2001). Numerical simulation of the gas-
Engineering Science, 40, 1817–1823. particle turbulent flow in riser reactor based on k–ε–kp –␧p – two-fluid model.
Gidaspow, D. (1994). Multiphase flow and fluidization: Continuum and kinetic theory Chemical Engineering Science, 56, 6813–6822.
description. New York: Academic Press. Zimmermann, S., & Taghipour, F. (2005). CFD modeling of the hydrodynamics and
Helland, E., Bournot, H., Occelli, R., & Tadrist, L. (2007). Drag reduction and clus- reaction kinetics of FCC fluidized-bed reactors. Industrial & Engineering Chem-
ter formation in a circulating fluidised bed. Chemical Engineering Science, 62, istry Research, 44, 9818–9827.
148–158.

You might also like