Powder Technology 286 (2015) 516–526

Contents lists available at ScienceDirect

Powder Technology

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

A new structural parameters model based on drag coefficient for

simulation of circulating fluidized beds
Wenming Liu a,b, Hongzhong Li a,⁎, Qingshan Zhu a,⁎, Quanhong Zhu a
a
State Key Laboratory of Multi-phase Complex Systems, Institute of Process Engineering, Chinese Academy of Sciences, P.O. Box 353, Beijing 100190, PR China
b
University of Chinese Academy of Sciences, Beijing 100049, PR China

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

Article history: This work presented a new scheme to establish structural parameters model, and the model was used to solve
Received 13 January 2015 structural parameters based on the available structure-based drag model. By combining with the Eulerian two-
Received in revised form 27 August 2015 fluid model, the hydrodynamics of circulating fluidized beds (CFBs) was simulated. Different combinations of
Accepted 31 August 2015
clusters properties, including the cluster voidage and diameter, were adopted to fit for Geldart A and B particles
Available online 2 September 2015
and to close the insufficient solving equations, respectively. The simulated solid mass flux, radial and axial
Keywords:
voidage profiles were in agreement with the experimental data. The dilute-top/dense-bottom and the core-
Fluidization annular flow structure were also captured. Moreover the spatiotemporal fluctuation of clusters can be observed
Simulation from those simulations. Simulation results showed the combination of the structural parameters model with the
Hydrodynamics available structure-based drag model can predict well the hydrodynamics for Geldart A and B particles in CFBs.
CFBs © 2015 Elsevier B.V. All rights reserved.
Structure-based drag model

1. Introduction drag models have already been successfully developed to understand

Circulating fluidized beds (CFBs) are widely used in various modern
industrial processes, such as fluid catalytic cracking and coal combus-
tion. Because of high flow rate of gas–solid and intensive contacting,
CFBs have favorable hydrodynamics, reactions and heat transfer charac-
teristics. Understanding of the hydrodynamic characteristics is the key
to design and the scale-up of such reactors. With the increase of the
computational ability, computational fluid dynamics (CFD) has been a
valuable tool to predict the fluid dynamics, which is of vital importance
to the scale-up of CFBs.
For heterogeneous gas–solid flows in CFBs, drag models have a sig-
nificant effect on the simulation of bed hydrodynamics [1–3]. The com-
petition and compromise in different spatiotemporal scales lead to the
formation of multiscale structures [4,5]. Conventional drag models,
such as Gidaspow [6] and Syamlal-O’Brien [7], have been proved to
over-predict the drag coefficient because they neglect the multiscale
structures [5,8,9] in CFBs. Multiscale structures of CFBs are characterized
by clusters, and heterogeneous flows can be divided into the cluster
phase, the dispersed phase and the inter-phase which are usually de-
fined by 8–10 structure parameters [10–12]. Because conventional
two-fluid drag models assume homogeneous conditions inside the con-
trol volume, they cannot predict structure parameters. Structure-based
⁎ Corresponding authors.
E-mail addresses: hzli@ipe.ac.cn (H. Li), qszhu@ipe.ac.cn (Q. Zhu).
the multiscale nature of heterogeneous two-phase flows in CFBs [13].
The key to solve structure-based drag models is to solve the structural
parameters. Due to the lack of enough number of equations to solve
these parameters, many researchers adopted minimization stability
condition to close the insufficient solving equations, such as the energy
minimization multi-scale model (EMMS) [12] and the cluster structure-
dependent (CSD) drag coefficient model [14]. Adopting other measures
to close the equations of structural parameters models has received
much less attention and remains a challenge for the complex heteroge-
neous flows of CFBs.
With respect to the system of multiscale structures, clusters play a
dominant role in gas–solid interaction [15–17]. Clusters have two
major properties, including diameter and voidage; equations of clusters
properties have been used by many researchers to improve drag
models. Gao et al. [15] adopted the cluster diameter, derived from the
experimental particles terminal velocity, to replace the particle diame-
ter of Gidaspow drag model, and the simulation results were in agree-
ment with experimental data. Lu et al. [18] used equations of clusters
properties to modify Gidaspow drag model and conservation equations
of the kinetic theory of granular flow. As for EMMS model, Nikolopoulos
et al. [19] claimed that the predicted cluster diameter was smaller than
the diameter of a single particle or negative for averaged solid volume
fraction εs less than 0.01 or greater than 0.44, respectively, and intro-
duced Gu and Chen’s [20] correlation of the cluster diameter to confront
these problems. Wang et al. [10] extended EMMS model from Geldart A
to Geldart B particles by integrating the equation of the cluster voidage.
Therefore, clusters properties are of prime importance to the simulation

http://dx.doi.org/10.1016/j.powtec.2015.08.049
0032-5910/© 2015 Elsevier B.V. All rights reserved.
 W. Liu et al. / Powder Technology 286 (2015) 516–526 517

of CFBs, and an improved structural parameters model should be able to
describe the heterogeneous multiscale structures caused by the clusters.
In the present work, a new structural parameters model was
established incorporating with equations of clusters properties, which
can not only close the insufficient solving equations but also take into
consideration the effect of clusters. By incorporating it into an Eulerian
two-fluid model, the gas–solid flow hydrodynamics of CFBs was simulat-
ed by using CFD software (FLUENT6.2.16). Simulation results were com-
pared with the experimental data available in literature to validate its
feasibility.

2. Mathematical model

2.1. Structural parameters model

Fig. 1 shows multiscale resolution of structure and gas–solid interaction proposed by Wang et al. [10]. The complex heterogeneous flow structures
of CFBs can be divided into three simple homogeneous phases, including the cluster phase, the dispersed phase and the inter-phase. Based on the
homogeneous assumption and Matsen’s [21] investigation, the voidage in dispersed phase εd is defined as 0.9997. The heterogeneous structures
have 9 unknown parameters, namely, Ufd, Upd and αd for the dispersed phase, εc, dc, Ufc, Upc, f and αc for the cluster phase. The solving equations
can be detailed as follows, and the related variables involved in these equations are summarized in Appendix A.
(1) Force balance in cluster phase
The forces on the cluster include cluster-fluid drag force FDcn inside cluster phase, cluster-fluid drag force FDcf outside cluster phase, the collision
force Fpdc from particles in the dispersed phase outside cluster phase and apparent gravity.
The drag force FDcn equals to the drag force of single particle multiplying the effective number of particles inside cluster phase.

F Dcn ¼ nF Dc ð1Þ

The drag force of single particle FDc can be expressed by

π 2
F Dc ¼ dp C Dc ρ f jU sc jU sc ð2Þ
8

The force on the particles of outside surface of the cluster is different from those inside the cluster. The former force is caused by high gas velocity
from the dispersed phase, and the latter force derives from low gas velocity of the cluster phase which means only half of particles of outside surface
are relatively effective. Therefore, the effective number of particles inside cluster phase is calculated by

π 3
3

dp ð1−ε c Þ 1 πd 2 ð1−ε Þ dc dp
n¼6 π
c c
− π 2 ¼ ð1−ε c Þ 1−2 ð3Þ
dp3 2 dp dp dc
6 4

Fig. 1. Resolution of structure and gas–solid interaction proposed by Wang [10].

518 W. Liu et al. / Powder Technology 286 (2015) 516–526

Then, Eq. (1) is written as:

3

π 2 dc dp
F Dcn ¼ dp C Dc ρ f jU sc jU sc ð1−ε c Þ 1−2 ð4Þ
8 dp dc

The drag force outside cluster phase FDcf is expressed by

π 2
F Dc f ¼ dc C Di ρ f jU si jU si ð5Þ
8

When the cluster moves upward at the velocity Upc/(1 − εc), particles in the dispersed phase move upward at the velocity Upd/(1 − εd). Because
the cluster diameter is much bigger than the particle diameter, the particle velocity is faster than the cluster velocity. The particles below the cluster
chase and bump the bottom of the cluster, then the particle velocity would reduced to be Upc/(1 − εc). The transverse collision would be offset be-
cause of symmetry, and the particles above the cluster would not be considered. Therefore, the mass flow rate of particles below the cluster is
expressed as follows:


π 2 U pc U pd
Gpdc ¼ dc − ρ ð1−ε d Þ ð6Þ
4 ð1−ε c Þ ð1−εd Þ p

According to the momentum conservation, the collision force Fpdc is written as:

2
  π U pc U pd
F pdc ¼ Gpdc upd −upc ¼ d2c ρp ð1−ε d Þ − ð7Þ
4 ð1−εc Þ ð1−ε d Þ

Based on the force balance in the cluster phase, the equation is calculated by

π 3  
F Dcn þ F Dc f þ F pdc ¼ dc ð1−ε c Þ ρp −ρ f ðg þ α c Þ ð8Þ
6

Then substituting Eqs. (4), (5) and (7) into Eq. (8), the following equation can be obtained:

3

π d dp π
ð1−ε c Þ c 1−2 C Dc ρ f jU sc jU sc þ dc2 C Di ρ f jU si jU si
8 dp dc 8 ð9Þ

2  
π 2 U pc U pd π 3
þ dc ρp ð1−ε d Þ − ¼ dc ð1−ε c Þ ρp −ρ f ðg þ α c Þ
4 ð1−ε c Þ ð1−ε d Þ 6

(2) Force balance in dispersed phase

For unit volume of the bed, the forces on the particles in dispersed phase include the particles–fluid drag force FDdn, the collision force from the
cluster Fpdcn and apparent gravity.
The particles–fluid drag force FDdn can be expressed as a similar way with Eq. (1)

F Ddn ¼ n F Dd ð10Þ

The number of particles in the dispersed phase is written as:

ð1−f Þð1−ε d Þ
n¼ π 3 ð11Þ
d
6 p

The single particle drag force FDd is calculated by

π 2
F Dd ¼ d C ρ jU jU ð12Þ
8 p Dd f sd sd

Therefore, Eq. (10) is changed as:

ð1−f Þð1−εd Þ 3 ð1−f Þð1−ε d Þ

F Ddn ¼ F Dd π 3 ¼ C Dd ρ f jU sd jU sd ð13Þ
dp 4 dp
6

Based on Eq. (6), the collision force Fpdcn is expressed as:

2
f 3 f U pd U pc
F pdcn ¼ F pdc π ¼ ρp ð1−ε d Þ − ð14Þ
dc
3 2 dc 1−ε d 1−ε c
6
 W. Liu et al. / Powder Technology 286 (2015) 516–526 519

According to the force balance in the dispersed phase, the following equation is obtained:

2
3 ð1−f Þð1−ε d Þ 3 f U pd U pc
C Dd ρ f jU sd jU sd − ρp ð1−ε d Þ −
4  dp  2 dc 1−ε d 1−εc ð15Þ
¼ ð1− f Þð1−ε d Þ ρp −ρ f ðg þ α d Þ

(3) Mass balance of gas and solid

U f ¼ f U fc þ ð1−f ÞU f d ð16Þ

U p ¼ f U pc þ ð1− f ÞU pd ð17Þ

ε f ¼ f ε c þ ð1−f Þεd ð18Þ

(4) The superficial slip velocity

In this work, particles and clusters were assumed to be homogeneously distributed in each phase, which was in accordance with the characteristic
of particulate fluidization, so the well-known Richardson and Zaki equation [22] was selected.

U sd ¼ U t εm
d ð19Þ

U sc ¼ U t εm
c ð20Þ

Ut is the terminal particle velocity, and the exponent m is calculated by the correlation suggested by Garside [23].

5:1 þ 0:28Re0:9
t
m¼ ð21Þ
1 þ 0:10Re0:9
t

(5) Equations for clusters properties

Clusters manifest fluctuations in velocity, voidage, diameter and density over a wide range of space and time scales. The voidage and diameter of
clusters are two major properties which affect the flow behavior of CFBs, and many researchers has performed lots of experiments and theoretical
study [24–27] to propose equations of the cluster voidage and diameter. Different combinations of equations for the cluster voidage and diameter
have a significant effect on the drag coefficient, so different combinations were adopted to fit for Geldart A and B particles, respectively.
For Geldart A particles, adopting the similar way with Gao’s [1] investigation which proposed different drag force correlations by classifying the
bed into four zones by the voidage differences, in this work only two zones (dispersed phase and dense phase) was divided, and the voidage of 0.933
was also used as the dividing criteria.
The equations of the cluster voidage and diameter are as follows:
For voidage εf ≤ 0.933, the equations of the cluster diameter dc of Harris et al. [28] and the cluster voidage εc of Gu [29] are

εs
dc jHarris ¼ ð22  aÞ
40:8−94:5ε s
 
ε c jGu ¼ 1−0:64 1−ð1−εs =ε sm Þ3:4 ð23  aÞ

For voidage εf N 0.933, the equations of the cluster diameter dc of Gu [20] and the cluster voidage εc of Harris et al. [28] are
 
dc jGu ¼ dp þ 0:027−10dp εs þ 32ε6s ð22  bÞ

0:58ε 1:48
ε c jHarris ¼ 1− s
ð23  bÞ
0:013 þ ε1:48
s

To avoid discontinuity of these equations, Lu et al. [30] used a switch equation as follows:

arctan½150  1:75ðε i −ε s Þ
φ¼ þ 0:5
π

εi equals to the voidage for the dividing criteria 0.933. Therefore, equations of the cluster voidage and diameter are written as:

dc ¼ φdc jHarr is þ ð1−φÞdc jGu ð22  cÞ

ε c ¼ φεc jGu þ ð1−φÞε c jHarris ð23  cÞ

For Geldart B particles, the equations of the cluster diameter dc of Subbaro [31] and the cluster voidage εc of Harris et al. [28] are

 pffiffiffiffiffiffi 2 
   1=3 2
dc ¼ 1−ε f = ε f −ε c 2ut g 1 þ u2t 0:35 gDt þ dp ð22  dÞ
520 W. Liu et al. / Powder Technology 286 (2015) 516–526

0:58ε s 1:48
εc ¼ 1− ð23  dÞ
0:013 þ εs 1:48

2.2 Computation scheme

For a given system with specified Up, Uf, εf and physical parameters ρg, ρf, μf, dp, to solve 9 independent variables (Ufd, Ufc, Upd, Upc, f, dc, εc, αd, αc), we
adopt the scheme as follows:
1, Calculate εc, dc from Eqs. (22), (23).
2, Combined with εd and εc, calculate f from Eq. (18).
3, Calculate Usc, Usd, from Eqs. (19), (20).
4, Combined with the definition of Usc, Usd, and Eqs. (16), (17), calculate Ufd, Ufc, Upd, Upc,then Usi can be calculated by its definition.
5, Calculate αd, αc, from Eqs. (9), (15).
The values of parameters of Usc, Usd, Usi, αd and αc as a function of voidage are summarized in Appendix B.
According to the Hou et al. [2] analysis, heterogeneous drag coefficient can be expressed as follows:
2 3


6ð1−f Þð1−ε d Þ 1 π dp f ð1−ε c Þ 1 π f 1 π 27
4 π 3 C Dd ρ f jU sd jU sd dp2 þ 1−2 2
π 3 C Dc 2 ρ f jU sc jU sc 4 dp þ π 3 C Di 2 ρ f jU si jU si 4 dc 5  ε f
2

dp 2 4 dc dp dc
β¼ 6 6  6 ð24Þ
ug −us

Theoretically, the values of Up, Uf, εf for the every local unit volume are substituted into the present structural parameters model, and every struc-
tural parameters can be calculated. Then the heterogeneous drag coefficient can be obtained and is incorporated into Fluent using the User-Defined
Functions(UDF). But this way, it will lead to increasing of calculation for every local unit volume. Therefore, the heterogeneous index Hd is adopted
which is correlated with local voidage in order to reduce the computational cost and is coupled with UDF to solve the drag coefficient.
For comparison, the Wen and Yu drag coefficient is considered to be a contrast factor, that is
 
3 1−ε f ε f

β0 ¼ ρ f u f −up
C D0 ε−2:7
f ð25Þ
4 dp

The heterogeneous index is defined as the ratio between the heterogeneous drag coefficient β and the Wen and Yu drag coefficient β0, and that is

β
Hd ¼ ð26Þ
β0

3. CFD model and simulation method
All the simulations were performed by CFD software (Fluent 6.2.16)
in 2-D Cartesian space. By employing the kinetic theory of granular
flows, the solid viscosity, pressure and thermal conductivity can be
expressed by the granular temperature. Appendix C shows the
governing equations for two-fluid model and constitutive equations.
Fig. 2 shows the geometry of the riser section, and Fig. 2(a) and
(b) are used to simulate Geldart A particles and Geldart B particles, re-
spectively. Inlet at the bottom was designated as velocity inlet for
both gas and solid phases, and pressure outlet was fixed at a reference
value (atmospheric) for the boundary condition outlet. The solids circu-
lated from outlet to inlet with the same mass flux which was dynamical-
ly calculated. The no-slip wall condition was set for the gas phase, and
the partially slip wall condition was set for the solid phase [33]. Accord-
ing to the literature, initially a certain amount of particles were filled
within the bed with a given voidage 0.5, and the initial bed height was
1.225 m for riser (a) [12] and 2.64 m for riser (b) [10]. Table 1 summa-
rizes the detailed simulation parameters. The simulation for each case is

Table 1
Parameters setting for the simulation.

Geldart A particles Geldart B particles

Particle diameter(μm) 54 180

Particle density( kg/m3) 930 1420
Grid size Δx (m) 0.00225 0.00508
Grid size Δy (m) 0.035 0.025
Riser height (m) 10.5 16.5
Superficial gas velocity Uf (m/s) 1.52, 2.1 4.28, 4.78
Initial bed height H0 (m) 1.225 2.64
Time step (s) 5.0e-4 5.0e-4
Convergence criteria 10e-3 10e-3
Maximum solid packing volume fraction 0.63 0.63
Fig. 2. Geometry of 2D riser: (a) case 1 [10]; case 2 [12].
 W. Liu et al. / Powder Technology 286 (2015) 516–526 521

Fig. 3. Comparison of solid mass flux of Geldart A particles: (a) Uf = 1.52m/s; (b) Uf = 2.1m/s.

performed up to 40 s. In order to maintain sufficient statistical stabiliza-
tion, the time-averaged samples are computed covering a period of 20–
40 s.
4. Results and discussion
Figs. 3 and 4 present the fluctuation of solid mass flux with time,
which can manifest the value of the drag coefficient, and the experi-
mental data and simulated solid mass fluxes are summarized in
Table 2. It can be seen that for drag model G (Gidaspow drag model),
the average values of solid mass flux are greater than the experimental
data [34,35]. This is mainly because drag model G is based on average
approach and neglects the effects of multiscale structures, leading to
the over-prediction of the drag coefficient [5,8,9]. Whereas, for drag
model L (the present drag model), the average values of solid mass
flux close to the experimental data. Besides, errors for drag model G
are quite large, but errors for drag model L range from 7.9% to 40%
which are very small, demonstrating that the improved structural pa-
rameters model can predict well the drag coefficient. Whether solving
correctly drag coefficient or not greatly depends on considering the ef-
fects of multiscale structures. Therefore, the structural parameters
model, considering the effect of heterogeneous flow structures, can
correctly compute multiscale parameters, and provide a suitable drag
coefficient for CFBs.
Fig. 5 shows the simulated distribution of instantaneous solids con-
centration for drag model G and L of Geldart A particles at Uf =
1.52m/s. Fig. 6 displays the simulated distribution of instantaneous
solids concentration for drag model L of Geldart B particles at Uf =
4.78m/s. As shown in Fig. 5(a), the whole bed seems to be homogeneous
whatever in the radial and axial direction because of the over-predicted
drag coefficient calculated by drag model G. However, for drag model L
from Fig. 5(b), the core-annular flow structure which means more par-
ticles gather near the wall than in the center can be clearly observed in
the radial direction, and a dilute-top/dense-bottom flow structure can
also be seen in the axial direction. Therefore, the conventional drag
model G cannot simulate heterogeneous structures of CFBs, but the
present drag model L can correctly simulate the dilute-top/dense-bot-
tom and core-annular flow structure. Moreover, the spatiotemporal
fluctuation of clusters can be observed from those simulations, indicat-
ing the formation and disaggregation of clusters with time. Clusters in
the dense phase resist the up-flowing of gas and solid and result in
the wriggling movement, which has been observed by Rhodes et al.
[36] from the experiment. When clusters move into the upper dilute
phase, they are blown into particles and leave out the bed or fall against
the wall. During the fluctuating movement of clusters, the cluster

Fig. 4. Comparison of solid mass flux of Geldart B particles: (a) Uf = 4.28m/s; (b) Uf = 4.78m/s.
522 W. Liu et al. / Powder Technology 286 (2015) 516–526

Table 2
Summary of experimental data and simulated solid mass fluxes.

Geldart A particles Geldart B particles

Uf (m/s) 1.52 2.1 4.28 4.78

Exp. (Gs, kg/m2/s) 14.3 24.1 13.0 22.0
Drag model L (c, kg/m2/s) 11.6 26.0 16.2 30.8
Drag model G (Gs, kg/m2/s) 58.5 73.3 232.6 208.5
Drag model L error (%) 18.9 7.9 24.6 40
Drag model G error (%) 309.1 204.1 957.3 1503.8

diameter and voidage change with the variation of voidage in the dense
or dilute phase, and it is agreement with the trend of the Eqs. (22) and
(23) in the structural parameters model. Fig. 6 also shows the similar
phenomenon which accords with the typical hydrodynamic character-
istics of CFBs. It can be found that drag model L can correctly simulate
the heterogeneous flow structures of CFBs.
Figs. 7 and 8 show the axial profile of time-averaged voidage at dif-
ferent riser heights using drag model L, and their comparison with the
experimental data [34,35], respectively. As can be seen from Figs. 7
and 8, drag model L can predict the flow structure of dilute-top and
dense-bottom which can also be seen in Figs. 5 and 6, and is reasonable
agreement with experimental results. The trends of curves are similar to
other simulation results [3,10,12,37], although the values have quantita-
tive discrepancies at the dense bottom bed which may be attributed to
the inaccurate evaluation of solid inventory or incomplete 2D descrip-
tion [12,37]. The interaction between gas and solid is determined by
drag force, and suitable drag coefficient could predict well the
Fig. 6. Simulated voidage distribution of Geldart B particles at U f = 4.78m/s for drag
model L.
distribution of voidage for the whole bed. It can be seen that by consid-
ering multiscale structures, the computed drag coefficient can predict
well the voidage in the axial direction for Geldart A and B particles at dif-
ferent riser heights and superficial gas velocity.
Figs. 9 and 10 show the radial profile of time-averaged voidage at
various riser heights between the simulation results and the experi-
mental correlations of Tung et al. [38]. It can be found that the computed
results of drag model L are in agreement with the experimental correla-
tions, and core-annular flow structure, which can also be found in Figs. 5
and 6, can be reflected by the trends that the voidage in the core zone is
greater than that in the annulus zone. Therefore, suitable structural pa-
rameters model and drag coefficient play a vital role in the prediction of
voidage in the CFBs.

5. Conclusions

The structure-based drag model has recently been an effective

approach to understand the multiscale nature of heterogeneous two-
phase flows in CFBs, and the key to solve structure-based drag

Fig. 5. Simulated voidage distribution of Geldart A particles at Uf = 1.52m/s: (a) drag

model G; (b) drag model L. Fig. 7. Axial voidage profile of Geldart A particles: (a) Uf = 1.52m/s; (b) Uf = 2.1m/s.
 W. Liu et al. / Powder Technology 286 (2015) 516–526 523

in CFBs were simulated by the CFD software with the structure-based

drag model.
The solid mass flux, radial and axial voidage profile were obtained. It
was shown that the present drag model predicted the experimental
data considerably better than the Gidaspow drag model. Moreover,
the heterogeneous hydrodynamics of CFBs, including the dilute-top/
dense-bottom and core-annular flow structure, could be simulated by
the present drag model. Besides, the formation and disaggregation of
clusters with time can be observed. Understanding the properties of
clusters is of prime importance to study the gas–solid interaction and
establish drag models of CFBs. Therefore, the present structural param-
eters model incorporating the equations of clusters properties can com-
pute correctly the drag coefficient, and is a valuable tool to simulate the
hydrodynamics of gas–solid flows in CFBs.

Nomenclature
CD drag coefficient
CDc drag coefficient in cluster phase
CDd drag coefficient in dispersed phase
CDi drag coefficient between cluster and gas in dispersed phase
CD0 drag coefficient for single particle in gas flow
CD averaged drag coefficient
dc diameter of cluster, m
dp diameter of particle, m
FD total drag force of flow gas on the particles in unit volume of
flow, N/m3
Fig. 8. Axial voidage profile of Geldart B particles: (a) Uf = 4.28m/s; (b) Uf = 4.78m/s. FDc drag force of flow gas on single particle in cluster phase, N
FDcn drag force of flow gas in cluster phase on the particles in sin-
coefficient is to establish structural parameters model. An improved gle cluster, N
structural parameters model should be able to describe the heteroge- FDcu drag force of gas flow on particles in unit volume of cluster
neous multiscale structures caused by the clusters. In this work, a new phase, N/m3
structural parameters model based on the available structure-based FDd drag force of flow gas on the single particle in dispersed
drag model was established incorporating with equations of clusters phase, N
properties. The hydrodynamic characteristics of gas–solid flow behaviors FDi drag force of flow gas in dispersed phase on the single cluster, N

Fig. 9. Radial voidage profile of Geldart A particles: (a) Uf = 1.52m/s, H = 8 m; (b) Uf = 1.52m/s, H = 5 m; (c) Uf = 2.1m/s, H = 8 m; (d) Uf = 2.1m/s, H = 5 m.
524 W. Liu et al. / Powder Technology 286 (2015) 516–526

Fig. 10. Radial voidage profile of Geldart B particles: (a) Uf = 4.28m/s, H = 10 m; (b) Uf = 4.28m/s, H = 5 m; (c) Uf = 4.78m/s, H = 10 m; (d) Uf = 4.78m/s, H = 5 m.

Fpdc collision force, N εc voidage in cluster phase

Fpdcn collision force exerting the dispersed phase in unit volume, εd voidage in dispersed phase
N/m3 εf averaged voidage
f volume friction of cluster phase εi voidage of inter-phase
Gs solids circulation flux, kg/m2/s εmf voidage at minimum fluidization
HD heterogeneity index εmin minimum particle voidage at packing
Rep Reynolds Number for particle εs averaged solid volume fraction
n the number of particles μf viscosity of gas kg/(m s)
Uf superficial gas velocity, m/s ρf density of gas, kg/m3
uf gas velocity, m/s ρp density of particle, kg/m3
Ufc superficial gas velocity in cluster phase, m/s
Ufd superficial gas velocity in dispersed phase, m/s Acknowledgements
Up superficial particle velocity, m/s
up particle velocity, m/s The authors are grateful to the financial support from National Nat-
Upc superficial particle velocity in cluster phase, m/s ural Science Foundation of China under Grant No. 21325628 and from
Upd superficial particle velocity in dispersed phase, m/s the State Key Development Program for Basic Research of China (973
Us superficial slip velocity between gas and particles, m/s Program) under Grant 2015CB251402.
Usc superficial slip velocity between gas and particle in cluster
phase, m/s Appendix A. Variables involved in the structural parameters model
Usd superficial slip velocity between gas and particle in dispersed
phase, m/s Variables involved in the structural parameters model are given
Usi superficial slip velocity between gas in dispersed phase and below.
cluster, m/s Superficial slip velocity in different phases:
Ut the terminal particle velocity, m/s
εc
U sc ¼ U fc −U pc ðA1Þ
Greek symbols 1−ε c
αd the accelerated velocity of particles in the dispersed phase

m/s2 U fd U pc
U si ¼ − ε ð1−f Þ ðA2Þ
αc the accelerated velocity of particles in the cluster phase m/s2 εd 1−ε c d
β drag coefficient including the heterogeneous flow structure
β0 Wen and Yu drag coefficient εd
U sd ¼ U f d −U pd ðA3Þ
ε voidage 1−ε d
 W. Liu et al. / Powder Technology 286 (2015) 516–526 525

Reynolds number in different phases: Appendix C. Governing equations and constitutive equations

ρ f dp U sc Governing equations and constitutive equations are given below.

Rep ¼ ðA4Þ
μf Continuity equation (k = g, s):

ρ f dc U si ∂ðεk ρk Þ
Rep ¼ ðA5Þ þ ∇ðεk ρk uk Þ ¼ 0 ðC1Þ
μf ∂t

εs þ εg ¼ 1 ðC2Þ
ρ f dp U sd
Rep ¼ ðA6Þ
μf
Momentum equation (k = g, s; l = s, g):

Standard drag coefficient: ∂ðεk ρk uk Þ

þ ∇ðε k ρk uk uk Þ ¼ −εk ∇pg þ εk ρk g þ ∇τ k þ βðul −uk Þ ðC3Þ
8 ∂t
< 0:44 Rep N 1000
C D0 ¼ 24  
ðA7Þ
: 1 þ 0:15Re0:687
p Rep ≤ 1000 Gas phase stress:
Rep
τg ¼ 2μ g Sg ðC4Þ

Effective drag coefficient in different phases:

Solid phase stress:
8
> −4:7
< C D0 εc εc ≥ 0:8
τs ¼ ½−ps þ λs ∇μ s δ þ 2μ s Ss ðC5Þ
C Dc ¼ ð1−ε c Þμ f 7 ðA8Þ
>
: 200 ε3 ρ d U þ 3ε 3 εc b 0:8
c f p sc c
Deformation rate:
8
> −4:7
< C D0 εi εi ≥ 0:8 1h i 1
ð1−ε i Þμ f Sk ¼ ∇uk þ ð∇uk ÞT − ∇uk δ ðC6Þ
C Di ¼ 7 ðA9Þ 2 3
: 200 ε3 ρ d U þ 3ε 3
> εi b 0:8
i f p si i
Solid phase pressure:
8
> −4:7
< C D0 εd εd ≥ 0:8 ps ¼ εs ρs Θ½1 þ 2ð1 þ eÞεs g 0  ðC7Þ
C Dd ¼ ð1−ε d Þμ f 7 ðA10Þ
: 200 ε 3 ρ d U þ 3ε 3
> εd b 0:8
d f p sd d Solid phase shear viscosity:
rffiffiffiffi  2
4 2 Θ 2μ s:dilute 4
Appendix B. The values of Usc, Usd, Usi, αd and αc as a function μs ¼ εs ρs dp g 0 ð1 þ eÞ þ 1 þ ð1 þ eÞεs g 0 ðC8Þ
3 π ð1 þ eÞg 0 5
of voidage
5 pffiffiffiffiffiffiffi
μ s:dilute ¼ ρ dp πΘ ðC9Þ
The values of Usc, Usd, Usi, αd and αc can be solved by the nonlinear 96 s
equations under a given system with specified Up, Uf, εf and physical pa-
rameters ρg, ρf, μf, dp. Traverse εf within [εmf, 1], those values as a func- Solid phase bulk viscosity:
tion of voidage εf can be obtained as follows: rffiffiffiffi
4 2 Θ
λs ¼ εs ρs dp g 0 ð1 þ eÞ ðC10Þ
3 π
Geldart A particles:

U sc ¼ ε
20:97ε f 2 −30:93ε f 2 þ11:55
3 −179:8ε 2 −150:2ε þ346:7
Uf = 1.52 m/s, Uf = (B1) Radial distribution functions:
f f f
2.1 m/s
Usd = 0.0825 Uf = 1.52 m/s, Uf = (B2) "
1=3 #−1
εg
2.1 m/s g0 ¼ 1− ðC11Þ
U si ¼ −23:9ε f 3 þ 42:92ε f 2 −22:21ε f þ 4:146 Uf = 1.52 m/s (B3) εsm
U si ¼ −23:85ε f 3 þ 38:16ε f 2 −14:53ε f þ 1:654 Uf = 2.1 m/s (B4)
ε −0:5603 2 Uf = 1.52 m/s (B5) Granular temperature equation:
α c ¼ 9:79  105 ; exp½−ð f 0:2123 Þ 
6 ε f −0:5623 2
α c ¼ 1:863  10 ; exp½−ð 0:2109 Þ 
Uf = 2.1 m/s (B6)  
3 ∂ðεs ρs ΘÞ
ε −0:4323
α d ¼ −7:691  109 ; exp½−ð f 0:2047 Þ 
2 Uf = 1.52 m/s (B7) þ ∇ðεs ρs us ΘÞ ¼ τs : ∇us −∇q−γ þ βC g C−3βΘ ðC12Þ
2 ∂t
ε −0:4341 2 Uf = 2.1 m/s (B8)
α d ¼ −1:454  1010 ; exp½−ð f 0:2041 Þ 
Collisional energy dissipation:
Geldart B particles:
20:75ε f 2 −43:22ε f 2 þ23:05 Uf = 4.28 m/s, Uf = (B9)
" rffiffiffiffiffiffiffiffiffi #
U sc ¼ ε 3 þ723:7ε 2 −1455ε þ731:4  
2 2 4 Θ
γ ¼ 3 1−e ε s ρs g 0 Θ −∇us ðC13Þ
f f f
4.78 m/s
Usd = 0.9707 Uf = 4.28 m/s, Uf = (B10) dp π
4.78 m/s Flux of fluctuating energy:
U si ¼ −161:3ε f 3 þ 333:8ε f 2 −220:3ε f þ 49:12 Uf = 4.28 m/s (B11)
U si ¼ −198:2ε f 3 þ 414:3ε f 2 −277ε f þ 62:48 Uf = 4.78 m/s (B12) q ¼ −k∇Θ ðC14Þ
ε f −0:3778 2 Uf = 4.28 m/s (B13)
α c ¼ 3:378  107 ; exp½−ð 0:2311 Þ 

α c ¼ 4:076  107 ; exp½−ð

ε −0:3838 2
f
Þ 
Uf = 4.78 m/s (B14) Conductivity if the fluctuating energy:
0:2299

k 2
12 ε −0:1139 2 Uf = 4.28 m/s (B15)
α d ¼ −3:375  10 ; exp½−ð f 0:2449 Þ 
2k 1 þ 65 ð1 þ eÞεs g0 c
α d ¼ −3:894 
ε −0:1212 2
1012 ; exp½−ð f 0:244 Þ 
Uf = 4.78 m/s (B16) k¼ þk ðC15Þ
ð1 þ eÞg 0
526 W. Liu et al. / Powder Technology 286 (2015) 516–526

k 75 pffiffiffiffiffiffiffi [15] J. Gao, J. Chang, X. Lan, Y. Yang, C. Lu, C. Xu, CFD modeling of mass transfer and strip-
k ¼ d ρ Θπ ðC16Þ ping efficiency in FCCU strippers, AICHE J. 54 (2008) 1164–1177.
384 p s
[16] J. Gao, J. Chang, C. Xu, X. Lan, Y. Yang, CFD simulation of gas solid flow in FCC strip-
pers, Chem. Eng. Sci. 63 (2008) 1827–1841.
rffiffiffiffi
Θ [17] S. Wang, H.L. Lu, Z.H. Hao, F.X. Zhao, L.Y. Sun, Q.H. Zhang, Modeling of reactive gas–
c
k ¼ 2ε2s ρs dp g 0 ð1 þ eÞ ðC17Þ solid flows in riser reactors using a multi-scale chemical reaction model, Chem. Eng.
π Sci. 116 (2014) 773–780.
[18] H.L. Lu, Q.Q. Sun, Y.R. He, Y.L. Sun, H.M. Ding, L. Xiang, Numerical study of particle
Drag coefficient for Gidaspow et al. [32]: cluster flow in risers with cluster-based approach, Chem. Eng. Sci. 60 (2005)
6757–6767.
8

s [19] A. Nikolopoulos, D. Papafotiou, N. Nikolopoulos, P. Grammelis, E. Kakaras, An ad-
>
>
> εg εs μ g ρg εs
ug −us
vanced EMMS scheme for the prediction of drag coefficient under a 1.2 MWth
>
< 150 2
þ 1:75 εg b0:8 CFBC isothermal flow-Part I: Numerical formulation, Chem. Eng. Sci. 65 (2010)
εg dp dp
β¼

ðC18Þ 4080–4088.
> 3 εs εg ρg
ug −us
−2:65
>
> [20] W. Gu, J. Chen, A model for solid concentration in circulating fluidized beds, Fluidi-
>
: CD εg εg ≥0:8 zation IX, 1998 501–508.
4 dp [21] J.M. Matsen, Mechanisms of choking and entrainment, Powder Technol. 32 (1982)
21–33.
Drag coefficient for this work is Eq. (24). [22] J.F. Richardson, W.N. Zaki, Sedimentation and fluidization, Trans. Inst. Chem. Eng. 32
(1954) 35–53.
[23] J. Garside, M.R. Al-Dibouni, Velocity–voidage relationships for fluidization and sedi-
References mentation in solid–liquid systems, Ind. Eng. Chem. Process Des. Dev. 16 (1977)
206–214.
[1] J. Gao, X. Lan, Y. Fan, J. Chang, G. Wang, C. Lu, C. Xu, CFD modeling and validation of [24] M. Rhodes, H. Mineo, T. Hirama, Particle motion at the wall of a circulating fluidized-
the turbulent fluidized bed of FCC particles, AICHE J. 55 (2009) 1680–1694. bed, Powder Technol. 70 (1992) 207–214.
[2] B. Hou, H. Li, Relationship between flow structure and transfer coefficients in fast [25] A.K. Sharma, K. Tuzla, J. Matsen, J.C. Chen, Parametric effects of particle size and gas
fluidized beds, Chem. Eng. J. 157 (2010) 509–519. velocity on cluster characteristics in fast fluidized beds, Powder Technol. 111 (2000)
[3] W. Wang, J. Li, Simulation of gas–solid two-phase flow by a multi-scale CFD ap- 114–122.
proach – of the EMMS model to the sub-grid level, Chem. Eng. Sci. 62 (2007) [26] R. Cocco, F. Shaffer, R. Hays, S.B.R. Karri, T. Knowlton, Particle clusters in and above
208–231. fluidized beds, Powder Technol. 203 (2010) 3–11.
[4] J.D. Gibbon, E.S. Titi, Cluster formation in complex multi-scale systems, Proc. R. Soc. [27] A.T. Andrews, P.N. Loezos, S. Sundaresan, Coarse-grid simulation of gas–particle
London, Ser. A 461 (2005) 3089–3097. flows in vertical risers, Ind. Eng. Chem. Res. 44 (2005) 6022–6037.
[5] J. Li, M. Kwauk, Exploring complex systems in chemical engineering – the multi- [28] A. Harris, J. Davidson, R. Thorpe, The prediction of particle cluster properties in the
scale methodology, Chem. Eng. Sci. 58 (2003) 521–535. near wall region of a vertical riser (200157), Powder Technol. 127 (2002) 128–143.
[6] D. Gidaspow, Multiphase flow and fluidization: continuum and kinetic theory [29] W. Gu, Diameters of catalyst clusters in FCC, AIChE symposium series, American
descriptions, Academic Press, 1994. Institute of Chemical Engineers, 1971–c2002, New York, NY 1999, pp. 42–47.
[7] M. Syamlal, O’Brien T.J., The derivation of a drag coefficient formula from velocity- [30] H.L. Lu, D. Gidaspow, Hydrodynamics of binary fluidization in a riser: CFD simula-
voidage correlations, U. S. Department of Energy, Office of Fossil Energy, Technical tion using two granular temperatures, Chem. Eng. Sci. 58 (2003) 3777–3792.
Report 1994. [31] D. Subbarao, A model for cluster size in risers, Powder Technol. 199 (2010) 48–54.
[8] J. Li, M. Kwauk, Multiscale nature of complex fluid-particle systems, Ind. Eng. Chem. [32] D. Gidaspow, R. Bezburuah, J. Ding, Hydrodynamics of circulating fluidized beds, ki-
Res. 40 (2001) 4227–4237. netic theory approach, Fluidization VII, Proceedings of the 7th Engineering Founda-
[9] J. Ma, W. Ge, X. Wang, J. Wang, J. Li, High-resolution simulation of gas–solid suspen- tion conference on fluidization 1992, pp. 75–82.
sion using macro-scale particle methods, Chem. Eng. Sci. 61 (2006) 7096–7106. [33] J. Ding, D. Gidaspow, A bubbling fluidization model using kinetic theory of granular
[10] J. Wang, W. Ge, J. Li, Eulerian simulation of heterogeneous gas–solid flows in CFB flow, AICHE J. 36 (1990) 523–538.
risers: EMMS-based sub-grid scale model with a revised cluster description, [34] J. Li, Particle-fluid two-phase flow: the energy-minimization multi-scale method,
Chem. Eng. Sci. 63 (2008) 1553–1571. Metallurgical Industry Press, 1994.
[11] S. Wang, H. Lu, G. Liu, Z. Sheng, P. Xu, D. Gidaspow, Modeling of cluster structure- [35] E.R. Monazam, L.J. Shadle, L.O. Lawson, A transient method for determination of
dependent drag with Eulerian approach for circulating fluidized beds, Powder saturation carrying capacity, Powder Technol. 121 (2001) 205–212.
Technol. 208 (2011) 98–110. [36] M.J. Rhodes, M. Sollaart, X.S. Wang, Flow structure in a fast fluid bed, Powder
[12] N. Yang, W. Wang, W. Ge, J. Li, CFD simulation of concurrent-up gas–solid flow in Technol. 99 (1998) 194–200.
circulating fluidized beds with structure-dependent drag coefficient, Chem. Eng. J. [37] N. Yang, W. Wang, W. Ge, L. Wang, J. Li, Simulation of heterogeneous structure in a
96 (2003) 71–80. circulating fluidized-bed riser by combining the two-fluid model with the EMMS
[13] J. Wang, A review of Eulerian simulation of Geldart A particles in gas-fluidized beds, approach, Ind. Eng. Chem. Res. 43 (2004) 5548–5561.
Ind. Eng. Chem. Res. 48 (2009) 5567–5577. [38] Y. Tung, J. Li, M. Kwauk, Radial voidage profiles in a fast fluidized bed, in: M. Kwauk,
[14] W. Shuai, L. Huilin, L. Guodong, S. Zhiheng, X. Pengfei, D. Gidaspow, Modeling of D. Kunii (Eds.), Fluidization’88: science and technology, Science Press, Beijing 1988,
cluster structure-dependent drag with Eulerian approach for circulating fluidized pp. 139–145.
beds, Powder Technol. 208 (2011) 98–110.