You are on page 1of 10

Research Article 1827

Mehran Ehsani1
Effects of Restitution and Specularity
Salman Movahedirad2
Shahrokh Shahhosseini1 Coefficients on Solid-Liquid Fluidized
Mahshid Ashtiani2
Bed Hydrodynamics
1
Process Simulation and Control
Research Laboratory, School An Eulerian-Eulerian two-fluid model in combination with the kinetic theory of
of Chemical Engineering, Iran granular flow was applied to simulate a solid-liquid fluidized bed filled with stain-
University of Science and less-steel particles. The drag model of Gidaspow used in computational fluid
Technology, Tehran, Iran. dynamics simulations proved to be the appropriate model for predicting the bed
2
School of Chemical hydrodynamics. The effect of the restitution coefficient of the particles on the bed
Engineering, Iran University hydrodynamics was investigated numerically. For the solid phase wall boundary
of Science and Technology, condition according to Johnson and Jackson a specularity coefficient was intro-
Tehran, Iran. duced and evaluated. The change in specularity coefficient did not affect consider-
ably the bed behavior at the middle part, but at the walls it significantly influenced
the granular temperature. For specularity coefficients more than 0.5 the results do
not change noticeably.

Keywords: Computational fluid dynamics simulation, Fluidized bed, Hydrodynamics,


Restitution coefficient, Specularity coefficient
Received: December 08, 2014; accepted: July 06, 2015
DOI: 10.1002/ceat.201400744

1 Introduction fluidized bed behavior, there are only a few published works on
the effect of these parameters on the bed hydrodynamic
Solid-liquid fluidized beds are widely applied in many indus- [14, 15]. In one of the few studies in this area, Li et al. [15] ana-
tries [1]. As the bed hydrodynamics is of great importance in lyzed the impact of the specularity coefficient in a bubbling
design, scale-up, and operation of the solid-liquid fluidized fluidized bed and found a strong effect on the bed voidage
bed, a large number of attempts were made to comprehend it modeling.
theoretically and experimentally [2]. Among many models and In the present study, the Eu-Eu two-fluid model in combina-
software, which are used to simulate and predict fluidized bed tion with KTGF is employed to simulate a solid-liquid fluidized
behavior, computational fluid dynamics (CFD) tools are the bed. Since the effect of the drag coefficient on the fluidized bed
most promising ones to solve conservation and other required behavior is undeniable, two common drag models are applied
equations [3]. In general, there are two approaches to simulate in the simulations. In addition, CFD simulation data of the to-
hydrodynamics of solid-liquid fluidized beds by means of CFD tal bed voidage is compared with the experimental results. The
tools, i.e., Eulerian-Eulerian (Eu-Eu) and Eulerian-Lagrangian impact of restitution and specularity coefficients on the bed
(Eu-La). Most of the fluidized bed simulations use the Eu-Eu hydrodynamics is studied as well as the influence of several
two-fluid model due to its simplicity and short computational model input and solution parameters such as mesh size and
time. For prediction of particulate phase properties, kinetic time step. The effect of liquid superficial velocity on the bed
theory of granular flow (KTGF) is applied in most CFD simu- hydrodynamics is investigated by comparing the results with
lations [4–10]. the experimental values of the present study.
The results of the Eu-Eu two-fluid model depend on many
modeling parameters such as solid-liquid drag model, particle
restitution coefficient, specularity coefficient, boundary condi- 2 Experimental Setup
tions, and others. The effects of most of these parameters were
investigated by several researchers [11–13]. Although restitu- Experiments were performed to assure the accuracy of CFD
tion and specularity coefficients have great influences on the simulations. A schematic diagram of the experimental setup is
presented in Fig. 1. The fluidized bed dimensions are 0.6 and
0.08 m in length and diameter, respectively. Tap water is fed to
– the column bottom with a uniform flow rate of 1.01 ·10–3 m3s–1,
Correspondence: Dr. Salman Movahedirad (movahedirad@iust.ac.ir), measured by a precalibrated flow meter. Spherical stainless-
School of Chemical Engineering, Iran University of Science and Tech- steel particles with narrow size distribution and small diameter
nology, P.O. Box 16765-163, Tehran, Iran. in comparison to the bed diameter are filled into the bed.

Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836 ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1828 Research Article

bulk viscosity, particulate phase pressure, and energy dissipa-


tion coefficient were derived from Lun et al. [19]. For predic-
tion of frictional viscosity, collision viscosity, and radial distri-
bution function, the equations of Schaeffer [20], Gidaspow
et al. [21], and Ding and Gidaspow [22] were taken, respective-
ly. Mass transfer between the phases, virtual mass, lift and ex-
ternal forces other than gravity were assumed to be negligible.
Generally speaking, adding the turbulence model to the sim-
ulation can make prediction more realistic, especially in the
case of high Reynolds numbers. Since the turbulence model
does not have a significant effect on the bed hydrodynamics
simulations [14], the standard k-e model was employed in the
fluidized bed simulation [23].

3.1 Boundary Conditions

The initial particle velocity was set to zero and the initial parti-
cle volume fraction (ep)1) was considered to be 0.6 according to
Figure 1. Schematic representation of the experimental setup. the experimental data. At the bottom inlet, water flows into the
bed with a specific velocity and exits the bed under atmospher-
Pressure transducers are placed at the top and bottom of the ic pressure. For the liquid and particulate phases, no slip wall
bed. The particles are initially packed in the bed up to 40 % of and partially slip wall conditions were considered, respectively,
the bed height with the initial bed viodage of 0.4. All experi- as proposed by Johnson and Jackson [24]:
ments were done at a constant temperature of 25 C. Details of
the experimental fluidized bed are listed in Tab. 1. ¶up;w
up;w ¼ A (18)
¶n
Table 1. Experimental conditions.
where A is the slip coefficient which depends on the specularity
Bed diameter [m] 0.08 coefficient and is defined as:
Bed height [m] 0.6 6mp ep; max
A ¼ pffiffiffiffiffi (19)
Particle material Stainless steel 3qpfp ep rp go
Particle diameter [m] 0.003
In the above equations, ep and ep,max denote the particulate
Particle density [kg m–3] 8030 volume fraction and its maximum value, respectively. q is the
granular temperature, which represents the particle collision
intensity. fp and tp,w are the specularity coefficient and partic-
3 Numerical Simulations ulate velocity, respectively. rp and mp signify the particulate
density and viscosity, respectively. In order to examine the
The Eu-Eu approach is applied for simulations and the kinetic effect of slip condition, different values of the specularity coeffi-
theory of granular flow is used to model the particle flow. It is cient are used in this study.
presumed that the particles are spherical in shape and have a
uniform size. The liquid phase is assumed to be incompressible
and its other physical properties are considered to be constant. 3.2 Solution Procedure and Method: Investigation
The detailed equations of momentum and mass conservation of Modeling Parameters
for each phase and other effective equations are summarized in
Tab. 2. Conservations of mass and momentum for both phases The effective equations, mentioned in the previous sections,
were solved during the simulations, see Eqs. (1)–(4) in Tab. 2. were solved under appropriate boundary conditions by means
Kinetic theory of granular flow provides a model for simulation of CFD techniques. The simulations were run for 30 s in un-
of the particles in Eqs. (11)–(15) in Tab. 2. steady state mode and time-averaged quantities were reported
To obtain the kinetic part of granular viscosity and conduc- for the last 25 s for reducing start-up effects. A first-order im-
tivity, the equation proposed by Syamlal et al. [16] was em- plicit upwind scheme was used for discretization of momen-
ployed. Eqs. (16) and (17) from Tab. 2 were used for momen- tum, granular temperature, and energy equations. The phase-
tum exchange between the liquid and solid phases. Therefore, coupled SIMPLE algorithm, which is a multiphase flow model,
two different drag models, i.e., from Wen and Yu [17] and is taken for pressure-velocity coupling. The focus of this section
Gidaspow [18], were applied. The particulate phase pressure
according to Eq. (5) from Tab. 2 was embedded into the mo- –
mentum conservation equation for the solid phase. Granular 1) List of symbols at the end of the paper.

www.cet-journal.com ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836
Research Article 1829

Table 2. Mathematical model of a solid-liquid fluidized bed.


 
Particle continuity equation ¶ ep rp  
þ  ep rp vp ¼ 0 (1)
¶t

Liquid continuity equation ¶ðel rl Þ


þ ðel rl vl Þ ¼ 0 (2)
¶t
 
Particle momentum equation ¶ ep rp vp    
þ  ep rp vp vp ¼ ep p þ pp þ tp þ ep rp g þ blp vl  vp (3)
¶t

Liquid momentum equation ¶ðel rl ul Þ  


þ ðel rl ul ul Þ ¼ el p þ tl þ el rl g þ bpl vp  vl (4)
¶t
  
Particulate phase pressure pp ¼ ep rp q 1 þ 2g0;pp ep 1 þ epp (5)

2
Particulate phase shear stress tp ¼ ep mp ðvp þ vp T Þ þ ep ðlp  mp Þvp I – (6)
3

4
mp ¼ ep rp dp g0;pp ð1 þ epp Þ
5 pffiffiffiffiffiffi
rffiffiffi
q ep rp dp qp 2
þ ½1 þ ð1 þ epp Þð3epp  1Þep g0;pp  (7)
p 6ð3 þ epp Þ 5
pp sin w
þ pffiffiffiffiffiffiffi
2 I2D

1
Liquid shear stress tl ¼ el ml ðvl þ vl T Þ  el ml ðvl ÞI (8)
3
rffiffiffi
4 q
Particle bulk viscosity lp ¼ ep rp dp go;pp ð1 þ epp Þ (9)
5 p
2 !1=3 31
Radial distribution function
ep
g0;pp ¼ 41  5 (10)
ep; max



Granular temperature 3 ¶     
e r q þ  ep rp vp q ¼ pp I þ tp : vp þ  kp q  gp þ yp (11)
2 ¶t p p
pffiffiffiffiffiffi

15dp rp ep qp 12 2 16
kp ¼ 1þ h ð4h  3Þep g0;pp þ ð41  33hÞhep g0;pp (12)
4ð41  33hÞ 5 15p

1
h ¼ ð1 þ epp Þ (13)
2

12ð1  e2pp Þg0;pp


gp ¼ pffiffiffi rp e2p q3=2 (14)
ds p

yp ¼ 3kp q (15)

Particle liquid drag equations 3 ð1  el Þel


bpl ¼ CD rl vl  vp e2:65
l , ðWen and Yu ½17Þ (16)
4 dp
8
3 ð1el Þel
>
< CD rl vl  vp e2:65
l el > 0:8
4 dp
bpl ¼ 2
ð1el Þ ml ð1el Þrl jvl vp j (Gidaspow ½18Þ (17)
>
: 150 þ 1:75 el £ 0:8
el dp2 dp

Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836 ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1830 Research Article

lies on the investigation of modeling parameters, i.e., mesh size


and time step, which have a significant effect on the computa-
tional time and simulation precision [25]. The convergence cri-
terion for residuals was considered to be 10–3 for all conserva-
tion equations except for continuity equations, which was set
to 10–5 to increase accuracy.

3.2.1 Mesh Size

Prior to CFD calculations, the fluidized bed geometry was de-


fined and the effect of the mesh size on the modeling results
was investigated. As 2D structured mesh provides proper re-
sults with an acceptable computational cost, it was chosen to
simulate the fluidized bed. As can be seen in Fig. 2, all four
mesh sizes show a similar qualitative trend and reach steady
state in 3 s after the fluid starts to flow. A refining mesh size less
than 1 ·10–3 m changed the results negligibly and the relative Figure 3. Bed voidage versus time by employing different time
error of simulation from experimental results decreased from steps at vf = 2.19 ·10–1 m s–1.
0.7 to 0.6 %, which is not noticeable. Moreover, the bed voidage
prediction relative error in the case of 2 ·10–3 and 5 ·10–3 m The dimensionless Courant number, which relates the mesh
mesh size is 1.2 and 1.9 %, respectively. Therefore, the size of size to the time step, is defined as follows:
1 ·10–3 m, which gives a mesh-independent solution was em-
ployed for the rest of simulations. Dt
NC ¼ vl (20)
Dy

where nl is the fluid velocity, Dt is the time step, and Dy is


the mesh size. The results indicate that employing 10–2 and
5 ·10–3 time steps leads to some fluctuations and gives a Cou-
rant number more than 1 which is unacceptable. Consequently,
10–3 and 2 ·10–3 time steps can be considered for further inves-
tigations. Moreover, some of the simulation works in the field
of fluidized beds have reported a fixed time step of 10–3 [26–29]
or even less [8, 11, 14], but our simulations showed that choos-
ing a time step of 2 ·10–3 provides reliable results. Since
decreasing the time step to less than 2 ·10–3 does not have a
noticeable effect on the result accuracy and causes higher com-
putational cost, it was applied in the simulations.

4 Results and Discussion

Figure 2. Bed voidage as a function of time in different mesh 4.1 Effect of Drag Coefficient
size at vf = 2.19 ·10–1 m s–1.
In fluidized beds, the drag force between the fluid and particles
has an important impact on the bed hydrodynamic, thus em-
3.2.2 Time Step ploying a proper drag model plays a significant role in simulat-
ing fluidized beds [27]. Two common drag models, which are
Several fixed time steps were used for the prediction of bed voi- widely used in solid-liquid interaction problems, are presented
dage. Fig. 3 shows the overall bed voidage over 30 s for different by Wen and Yu [17] and Gidaspow [18] and either of them is
time steps. As can be seen in Fig. 3 for the selected mesh size, proper for a limited particle fraction [30]. The Wen and Yu
time steps range from 10–3 to 5 ·10–3 and give very similar re- model [17] uses the correlation of Richardson and Zaki [31],
sults under the steady-state condition that are close to the which is valid when the internal forces are negligible [32]. The
experimental value. Obviously, a large time step of 10–2 and Gidaspow drag model [18] is based on the Ergun equation for
even 5 ·10–3 leads to some fluctuations. On the other hand, de- void fractions less than 0.8 and on an empirical correlation for
creasing the time step to 10–3 increases the computational time void fractions more than 0.8. Huilin and Gidaspow [33] pro-
considerably without significant enhancement in the accuracy posed a switch function to consider the continuity of these drag
or precision. correlations from dilute to dense regime.

www.cet-journal.com ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836
Research Article 1831

In the present work, two different drag models have been


incorporated to investigate their effect on the fluidized bed hy- a)
drodynamics and the results are compared with experimental
data in Fig. 4. In this figure, the Gidaspow drag model [18] is
fitted into the experimental results adequately well. While the
Wen and Yu model [17] could not correctly predict the results,
the Gidaspow drag model [18] was employed for all further
studies.

b)

Figure 4. Bed voidage as a function of time at vf = 2.19 ·10–1 m s–1.

4.2 Effect of Restitution Coefficient

The restitution coefficient, which represents the particle colli-


sion elasticity, ranges from 0 (inelastic) to 1 (elastic) and is of
great importance in the bed hydrodynamics prediction, espe-
cially in the case of dense fluidized beds. Determining the resti-
tution coefficient of the particles, which appears in the momen- c)
tum and the granular temperature conservation equations, is
difficult [26].
The influence of the restitution coefficient on the fluidized
bed hydrodynamics was studied by many researchers [34–38]
who stated that even a small deviation from elastic collision
leads to a huge difference in the bed hydrodynamic behavior.
Du et al. [38] demonstrated by increasing the restitution coeffi-
cient from 0.85 to 0.95 that the difference of simulation results
from experimental data becomes higher. In this study, the effect
of employing three different restitution coefficients, i.e., 0.85,
0.9, and 0.99, and specularity coefficients, i.e., 0.1, 0.5, and 0.9,
on the bed hydrodynamics is investigated. The effect of the
specularity coefficient on the bed hydrodynamics will be dis-
cussed in detail in the next section.
Fig. 5 depicts the effect of different specularity coefficients at
various restitution coefficients on the bed hydrodynamics. As Figure 5. Overall bed voidage prediction as a function of time
can be seen in all curves of Fig. 5, at a very high restitution co- at different restitution coefficients for a specularity coefficient of
efficient of e = 0.99, the overall bed voidage was overpredicted. (a) 0.1, (b) 0.5, and (c) 0.9.
Moreover, by decreasing the restitution coefficient to 0.85, the
overall bed voidage prediction was underestimated in this case bed voidage prediction. It is also obvious that for all specularity
study. This behavior can be explained considering the defini- coefficients (curves of Fig. 5a–c), the restitution coefficient of
tion of the restitution coefficient, which reveals that higher 0.9 predicted more reliable results compared to the
e-values lead to higher particle collision intensities and overall experimental values of the overall bed voidage.

Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836 ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1832 Research Article

Fig. 6 illustrates the mean squared error (MSE) of the bed


voidage prediction from the experimental data for different res- a)
titution coefficients. The MSE is a measurement of the average
of the squared errors. This parameter can be a good criterion
for choosing a restitution coefficient which shows the mini-
mum deviation of bed voidage prediction from the experimen-
tal data. According to Fig. 6, e = 0.9 led to the minimum MSE
bed voidage prediction from the experimental value in this case
study.

b)

Figure 6. Mean squared error of bed voidage prediction from


the experimental value for different restitution coefficients and
f = 0.1.

4.3 Effect of Specularity Coefficient

The specularity coefficient, which represents the partial slip


wall boundary condition and shows the nature of collision be- c)
tween the boundaries and particles, was rarely mentioned in
the available experimental data. A value of zero for the specu-
larity coefficient implies smooth frictionless walls and unity
means a very rough wall. This parameter has an important ef-
fect on the bed hydrodynamics behavior especially in the case
of laboratory-scale beds [14, 15]. Therefore, selecting the cor-
rect specularity coefficient is of great importance in the accu-
rate simulation of fluidized bed hydrodynamics. Benyahia et al.
[39] declared that a low specularity coefficient is necessary for
the accurate prediction of the bed hydrodynamics. To investi-
gate the influence of the particle-wall boundary condition,
three different specularity coefficients were employed in this
study.
Fig. 7 presents the results of employing different specularity
coefficients on the overall bed voidage as a function of time.
Obviously, the predicted overall bed voidage in the case of a Figure 7. Overall bed voidage prediction as a function of time
restitution coefficient of 0.99 is overestimated. On the other at different specularity coefficients for a restitution coefficient of
hand, by increasing the specularity coefficient in all curves of (a) 0.85, (b) 0.9, and (c) 0.99.
Fig. 7, the predicted overall bed voidage is reduced. It is also
evident that reducing this coefficient to 0.1 improves the agree- be observed, at e = 0.9, which is the best restitution coefficient
ment between simulation and experimental results. in the current case, the specularity coefficient of 0.1 shows the
Fig. 8 illustrates the MSE of bed voidage prediction from ex- minimum MSE from the experimental value of the bed voidage
perimental values for different specularity coefficients. As can prediction.

www.cet-journal.com ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836
Research Article 1833

Figure 9. The overall bed voidage at different superficial liquid


Figure 8. Mean squared error of bed voidage prediction from velocities predicted by simulation and Richardson and Zaki cor-
the experimental value for different specularity coefficients and relation [31] in comparison to experimental values.
e = 0.9.

As in the solid-liquid fluidized bed, for the liquid that is sur- Moreover, it is obvious that the overall bed voidage is higher by
rounding the particles and prevents perfectly elastic particle increasing the fluid velocity, which is consistent with the results
collisions, the restitution coefficients differ from unity. On the of our previous work [40].
other hand, Figs. 5–8 indicate that a restitution coefficient of
0.9 and a specularity coefficient of 0.1 fitted the experimental
results well. 4.5 Distribution of Granular Temperature

The granular temperature is a measure of particle collision in-


tensity and greatly depends on particle velocity fluctuations.
4.4 Effect of Superficial Liquid Velocity
This parameter expressed the macroscopic dissipation energy
To compare the overall bed voidage for different liquid superfi- of inelastic collision and drag between the fluid and particles
cial velocities, the CFD and experimental results were also [27]. Generally speaking, the granular temperature is due to
compared with the predictions of Richardson and Zaki [31]: the oscillation of particles in a small region for a short time
period [35]. It is low at dilute particle volume fraction due to
v 1 fewer particle collisions. On the other hand, it is also low at
z ¼ ð Þa high volume fraction due to the difficulty in particle movement
vt
dp and short mean free path. The mean free path is the average
where a ¼ 4:65 þ 20 Ret < 0:2 distance traveled by a particle between two consecutive colli-
db
sions [7].
dp
a ¼ ð4:4 þ 18 ÞRe0:03 0:2 < Ret < 1 Fig. 10 depicts the time-averaged granular temperature at
db t (21) different distances from the column axis for a few values of res-
dp titution and specularity coefficients. As can be seen in Fig. 10 a,
a ¼ ð4:4 þ 18 ÞRe0:1 1 < Ret < 200
db t the granular temperature is high near the wall and tends to
a ¼ 4:4Re0:1
t 200 < Ret < 500 decrease in the region near the center. It is also obvious that
when the restitution coefficient is increased, the granular tem-
a ¼ 2:4 Ret > 500 perature rises. Although the specularity coefficient is not of
great importance in relation to the granular temperature pre-
In Eq. (21), z and a are the bed voidage and the Richardson diction at the column axis, it has an undeniable effect in the
and Zaki exponent, n and nt denote the particle velocity and its prediction of the granular temperature near the wall. Fig. 10 b
terminal value, respectively. Fig. 9 depicts the results of the illustrates the time-averaged granular temperature distribution
overall bed voidage prediction of the mentioned correlation versus dimensionless distance for four values of specularity
and CFD simulation in comparison to the experimental results. coefficients near the wall at the middle of the maximum bed
As can be seen in the figure, the Richardson and Zaki equation expansion height. The granular temperature becomes higher by
[31] predicts the experimental results with 6.85 ·10–4 MSE val- increasing the specularity coefficient near the wall, which is
ue, which is noticeable in comparison with the CFD simulation consistent with the result of Loha et al. [14]. Moreover, as men-
results with 3.99 ·10–4 MSE. In addition, it is obvious that both tioned in Sect. 4.2 and 4.3, if an increment in the specularity
the correlation and CFD model can predict the experimental coefficient is more than 0.5, it exerts no significant effect on the
results with an acceptable error at low superficial velocities. simulation results.

Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836 ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1834 Research Article

a) tal data demonstrated that a time step of 2 ·10–3 s


and a mesh size of 1 ·10–3 can predict the results
well with an acceptable mesh independency.
Two different drag models were employed in the
simulation to evaluate the effect of drag force on
the overall bed voidage prediction. As the particle
volume fraction of the present fluidized bed is less
than 0.8, the Gidaspow drag model provides a
more realistic prediction in comparison to the
model of Wen and Yu.
There are few researches conducted in the field
of specularity and restitution coefficient effects on
the bed hydrodynamics due to the difficulty in their
value measurements. The effect of applying differ-
ent values of specularity coefficients and also resti-
tution coefficients on the bed hydrodynamic be-
havior was investigated. The simulation results
showed that a specularity coefficient of 0.1 and a
restitution coefficient of 0.9 led to accurate predic-
tion with a relative error of less than unity.
The results demonstrated that the CFD simula-
tion method could predict the overall bed voidage
b) reasonably well at different superficial liquid veloc-
ities. The MSE value of the overall bed voidage pre-
diction in the case of the Richardson and Zaki
correlation is 6.85 ·10–4 and for simulations
3.99 ·10–4 compared to our experimental results.
Further investigation confirmed that increasing
either the specularity or restitution coefficients
causes the time-averaged granular temperature to
rise. Exploring the influence of convergence criteria
and also the inter-related effect of specularity and
restitution coefficient will be the subject of future
work.

The authors have declared no conflict of interest.

Symbols used
CD [–] drag coefficient
d [m] diameter
e [–] restitution coefficient
Figure 10. Time-averaged granular temperature as a function of dimensionless
distance for different (a) restitution coefficients, (b) specularity coefficients at g [m s–2] gravitational acceleration
h/hmax = 0.5 and vf = 2.19 ·10–1 m s–1. g0 [–] radial distribution
coefficient
h [m] expanded bed height
hmax [m] maximum expanded bed height
5 Conclusions I [–] unit tensor
I2D [–] second invariant of the deviatoric stress
An Eu-Eu two-phase model was employed to simulate a solid- tensor
liquid fluidized bed by means of the CFD method. To investi- k [–] diffusion coefficient for granular energy
gate the effect of different modeling parameters on the bed n [–] normal vector
hydrodynamics, different mesh sizes and time steps were eval- p [Pa] pressure
uated. The results showed that the mesh size has a stronger R [m] column radius
influence on the bed hydrodynamics prediction. Moreover, a t [s] time
time step, which could significantly increase the computational v [m s–1] velocity
cost, has no noticeable effect on the overall bed voidage predic- x [m] radial distance
tion. Comparison of the simulation results and the experimen- y [m] mesh size

www.cet-journal.com ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836
Research Article 1835

Greek letters [9] Y. Cheng, J. Zhu, Chem. Eng. Sci. 2008, 63, 3201–3211. DOI:
10.1016/j.ces.2008.03.036
a [–] Richardson-Zaki exponent [10] Y. Behjat, S. Shahhosseini, M. A. Marvast, Int. Commun.
b [–] interphase exchange coefficient Heat Mass Transfer 2011, 38, 100–109. DOI: 10.1016/
g [kg m–1s–3] collision dissipation of energy j.icheatmasstransfer.2010.09.014
e [–] volume fraction [11] E. Doroodchi, K. P. Galvin, D. F. Fletcher, Powder Technol.
h [Pa s] dynamic viscosity 2005, 156, 1–7. DOI: 10.1016/j.powtec.2005.05.057
q [m2s–2] granular temperature [12] F. Gevrin, O. Masbernat, O. Simonin, AIChE J. 2010, 56,
z [–] bed voidage 2781–2794. DOI: 10.1002/aic.12209
l [Pa s] bulk viscosity [13] A. Neri, D. Gidaspow, AIChE J. 2000, 46, 52–67. DOI:
m [Pa s] shear viscosity 10.1002/aic.690460108
r [kg m–3] density [14] C. Loha, H. Chattopadhyay, P. K. Chatterjee, Particuology
t [Pa] stress tensor 2013, 11, 673–680. DOI: 10.1016/j.partic.2012.08.007
f [–] specularity coefficient [15] T. Li, J. Grace, X. Bi, Powder Technol. 2010, 203, 447–457.
y [kg m–1s–3] transfer rate of kinetic energy DOI: 0.1016/j.powtec.2010.06.005
w [] angle of internal friction [16] M. Syamlal, W. Rogers, T. J. O’Brien, Technical Note, DOE/
t [m s–1] particulate velocity METC-94/1004, NTIS/DE94000087, National Technical In-
formation Service, Springfield, VA 1993.
[17] C. Y. Wen, Y. H. Yu, AIChE J. 1966, 12, 610–612. DOI:
Subscripts 10.1002/aic.690120343
[18] D. Gidaspow, Multiphase Flow and Fluidization: Continuum
b bed
and Kinetic Theory Descriptions, Academic Press, New York
l liquid phase
1994.
p particle
[19] C. K. K. Lun, S. B. Savage, D. J. Jeffrey, N. Chepurniy, J. Flu-
t terminal
id. Mech. 1984, 140, 223–256. DOI: 10.1017/S002211208
w wall
4000586.
[20] D. Schaeffer, J. Differential Equations 1987, 66, 19–50. DOI:
10.1016/0022-0396(87)90038-6
Abbreviations
[21] D. Gidaspow, R. Bezburuah, J. Ding, in Fluidization VII.
CFD computational fluid dynamics (Eds: O. E. Potter, D. J. Nicklin), Engineering Foundation,
Eu-Eu Eulerian-Eulerian New York 1992, 75–82.
KTGF kinetic theory of granular flow [22] J. Ding, D. Gidaspow, AIChE J. 1990, 36, 523–538. DOI:
Eu-La Eulerian-Lagrangian 10.1002/aic.690360404
MSE mean squared error [23] R. K. Reddy, J. B. Joshi, Chem. Eng. Sci. 2009, 64, 3641–3658.
DOI: 10.1016/j.ces.2009.05.004
[24] P. C. Johnson, R. Jackson, J. Fluid. Mech. 1987, 176, 67–93.
References DOI: 10.1017/S0022112087000570
[25] K. Zhang, Y. Guan, X. Yao, Y. Li, X. Fan, S. Brandani, Powder
[1] R. K. Reddy, M. J. Sathe, J. B. Joshi, K. Nandakumar, G. M. Technol. 2013, 235, 180–191. DOI: 10.1016/j.powtec.2012.
Evans, Chem. Eng. Sci. 2013, 92, 1–12. DOI: 10.1016/j.ces. 09.046.
2012.11.017 [26] S. Zimmermann, F. Taghipour, Ind. Eng. Chem. Res. 2005,
[2] B. G. M. van Wachem, A. E. Almstedt, Chem. Eng. J. 2003, 44, 9818–9827. DOI: 10.1021/ie050490
96, 81–98. DOI: 10.1016/j.cej.2003.08.025 [27] J. Jung, D. Gidaspow, I. K. Gamwo, Chem. Eng. Commun.
[3] J. R. Grace, F. Taghipour, Powder Technol. 2004, 139, 99– 2006, 193, 946–975. DOI: 10.1080/00986440500351982
110. DOI: 10.1016/j.powtec.2003.10.006 [28] S. Movahedirad, M. Ghafari, A. M. Dehkordi, Chem. Eng.
[4] J. T. Cornelissen, F. Taghipour, R. Escudié, N. Ellis, J. R. Technol. 2012, 35, 929–936. DOI: 10.1002/ceat.201100383
Grace, Chem. Eng. Sci. 2007, 62, 6334–6348. DOI: 10.1016/ [29] S. Movahedirad, M. Ghafari, A. Molaei Dehkordi, Chem.
j.ces.2007.07.014 Eng. Technol. 2014, 37, 103–112. 10.1002/ceat.201300432.
[5] K. Zhang, G. Wu, S. Brandani, H. Chen, Y. Yang, Powder [30] X. Huang, Particuology 2011, 9, 441–445. DOI: 10.1016/
Technol. 2012, 227, 104–110. DOI: 10.1016/j.powtec. j.partic.2011.03.005
[31] J. F. Richarson, W. N. Zaki, Trans. Inst. Chem. Eng. 1954, 32,
2012.01.030
[6] S. Wang, Y. Zhao, X. Li, L. Liu, L. Wei, Y. Liu, J. Gao, Adv. 35–53. DOI: 10.1016/0009-2509(54)85015-9
[32] J. Lundberg, B. M. Halvorsen, in Proc. of the 49th Scandina-
Powder Technol. 2014, 25, 1103–1110. DOI: 10.1016/j.apt.
vian Conf. on Simulation and Modeling, Oslo University
2014.02.009
[7] L. Guodong, W. Peng, W. Shuai, S. Liyan, Y. Yunchao, College, Oslo 2008.
[33] L. Huilin, D. Gidaspow, Chem. Eng. Sci. 2003, 58, 3777–
X. Pengfei, Adv. Powder Technol. 2013, 24, 537–548. DOI:
3792. DOI: 10.1016/S0009-2509(03)00238-0
10.1016/j.apt.2012.10.007
[34] M. J. V. Goldschmidt, J. A. M. Kuipers, W. P. M. van Swaaij,
[8] P. Lettieri, R. Di Felice, R. Pacciani, O. Owoyemi, Powder
Chem. Eng. Sci. 2001, 56, 571–578. DOI: 10.1016/S0009-
Technol. 2006, 167, 94–103. DOI: 10.1016/j.powtec.2006.
2509(00)00262-1
06.012

Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836 ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.cet-journal.com
1836 Research Article

[35] J. Jung, D. Gidaspow, I. K. Gamwo, Ind. Eng. Chem. Res. [39] S. Benyahia, M. Syamlal, T. J. O’Brien, Powder Technol.
2005, 44, 1329–1341. DOI: 10.1021/ie0496838 2005, 156, 62–72. DOI: 10.1016/j.powtec.2005.04.002
[36] C. Loha, H. Chattopadhyay, P. K. Chatterjee, Chem. Eng. Sci. [40] Y. Behjat, S. Shahhosseini, S. H. Hashemabadi, Int. Commun.
2012, 75, 400–407. DOI: 10.1016/j.ces.2012.03.044 Heat Mass Transfer 2008, 35, 357–368. DOI: 10.1016/
[37] X. Lan, C. Xu, J. Gao, M. Al-Dahhan, Chem. Eng. Sci. 2012, j.icheatmasstransfer.2007.09.011
69, 419–430. DOI: 10.1016/j.ces.2011.10.064
[38] W. Du, X. Bao, J. Xu, W. Wei, Chem. Eng. Sci. 2006, 61,
4558–4570. DOI: 10.1016/j.ces.2006.02.028

www.cet-journal.com ª 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eng. Technol. 2015, 38, No. 10, 1827–1836

You might also like