You are on page 1of 11

Advanced Powder Technology 25 (2014) 1737–1747

Contents lists available at ScienceDirect

Advanced Powder Technology


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

Original Research Paper

Numerical simulation of turbulent fluidized bed with Geldart B particles


Salma Benzarti a,⇑, Hatem Mhiri a, Hervé Bournot b, René Occelli b
a
National Engineering School of Monastir, Unit of Thermic and Thermodynamics of the Industrial Processes, Avenue Ibn El Jazzar, 5019 Monastir, Tunisia
b
Aix Marseille Université, CNRS, IUSTI UMR 7343, 13453 Marseille, France

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

Article history: A numerical study was carried out to explore the hydrodynamic behavior of gas–solid flow in a 2D tur-
Received 2 May 2013 bulent fluidized bed. The gas–solid flow was simulated by mean of an Eulerian–Eulerian two fluid model
Received in revised form 18 June 2014 incorporating the Kinetic Theory of Granular Flow (KTGF). The main purpose of this work was to examine
Accepted 26 June 2014
the ability of various models to predict the hydrodynamic of a turbulent fluidized bed filled with Geldart
Available online 9 July 2014
B particles on the Fluent V.6.2 platform. The effect of different drag models including those of Gidaspow,
Syamlal & O’Brien and Mckeen, were tested and the corresponding results were compared. A good level of
Keywords:
agreement was achieved with the Gidaspow model with reference to experimental data. Two viscous
Turbulent fluidization
CFD
models were tested and the standard k—e model performed the better prediction of the core annular
Drag model structure of the bed with reference to the experimental data. Different wall boundary conditions were
Turbulence also tested and their impact evaluated on the hydrodynamics of the bed.
Wall boundary conditions Ó 2014 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder
Technology Japan. All rights reserved.

1. Introduction regarding flow dynamics and computational models are required


to make it a standard tool in designing large scale industrial reactors.
The turbulent fluidized bed was widely used in industrial appli- There are two different classifications of CFD models in the lit-
cations during the last decades. It is characterized by its ability to erature for modeling gas–solid flows: The Lagrangian and the Eule-
handle the continuous powder, its vigorous gas–solids contacting, rian approach. In the Lagrangian approach, the Newtonian
its ability of favorable heat and mass transfer and by its relatively equations of motion are solved for each individual particle. This
low axial gas dispersion. approach also takes into account a collision model for commending
The turbulent fluidization regime has been commonly acknowl- the energy dissipation caused by non ideal particle–particle inter-
edged as a distinct flow regime occurring between two adjacent actions [6,7]. The second approach, namely the Eulerian approach,
flow regimes: the bubbling and the fast fluidization regimes. It is treats the two different phases mathematically as continuous and
characterized by two different coexisting regions: a dense bottom, fully interpenetrating [8–10]. Several comparisons between
bubbling region and a dilute, dispersed flow region. The turbulent numerical models based on the multifluid model and the discrete
fluidization is further characterized by its high value of dispersion particle model are available in the literature. Ibsen et al. [11], for
coefficient for the solids. According to the literature review, less example, made a comparison based on 2D simulations of a circu-
attention was dedicated to the turbulent fluidization due to cur- lating fluidized bed. They found that the discrete particle model
rent deficiencies in experimental and theoretical works [1–4]. was in better agreement with experimental findings, while the
Therefore, fundamental understanding of its hydrodynamic behav- Eulerian approach was by far the most efficient. Conservations
ior is required. equations (momentum, mass, energy balance) were derived to
Having achieved success in simulation of single phase flow, Com- obtain a set of equations that have similar structure in all phases.
putational Fluid Dynamics (CFD) is considered greatly promising for To close the equations’ system and to describe the rheology of
modeling multiphase flows [5]. Nevertheless, CFD is still at the ver- the solid phase, constitutive equations are necessary. The Kinetic
ification and validation stages for modeling multiphase flows, for Theory of Granular Flow (KTGF) has become a very promising
fluidized bed applications particularly, and more improvements model for the gas–particle fluidized beds. It is an extension of
the classical kinetic theory of gases described by Chapman and
Cowling [12] to dense particulate flows. Numerous studies incor-
⇑ Corresponding author. Tel.: +216 29344333. porating the KTGF model have shown its ability to simulate the
E-mail address: salma_benzarti@yahoo.fr (S. Benzarti). two-phase flow in circulating fluidized beds (CFBs) [9,13–16].

http://dx.doi.org/10.1016/j.apt.2014.06.024
0921-8831/Ó 2014 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
1738 S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747

Nomenclature

C 2e , C 1e turbulence model coefficients Greek letters


CD drag coefficient eg turbulence dissipation of gas phase, m2 s3
D riser width, m a volume fraction
ds particle diameter, m q density, kg m3
es particle–particle restitution coefficient ns solid bulk viscosity, Pa s
g gravitational acceleration, 9.81 m s2 u specularity coefficient
g0 radial distribution function rk ; re turbulent Prandtl numbers
Gk;g production of turbulence kinetic energy, W m3 b gas–particle interaction coefficient, kg m3 s1
Gs solids external mass flux, kg m2 s1 H granular temperature, m2 s2
H riser height, m s stress tensor, N m2
I unit tensor / angle of internal friction, °
I2D the second invariant of the deviatoric stress tensor Pkg ; Pg influence of the dispersed phases on the continuous
J granular energy transfer, kg m1 s3 phase
kg turbulence quantities of gas phase, m2 s2 c collisional dissipation of energy, kg m3 s1
kH the diffusion coefficient for granular energy
P pressure, N m2 Subscripts
Ps solid pressure, N m2 KTGF Kinetic Theory of Granular Flow
q diffusion of fluctuating energy, kg s3 Col collisional
Re Reynols number Kin kinetic
t time, s max maximum
u velocity, m s1 fr frictional
udr drift velocity, m s1 X lateral coordinate, m
Ug superficial gas velocity, m s1 s solid phase
g gas phase

In the Eulerian model, the momentum equations for both model with a scaling factor C varying between 0.2 and 0.3. The cor-
phases are connected by the drag coefficient, which represents rection factor of 0.25 proved to predict the experimental results
the momentum exchange between the gas and solid phases. The observed in a freely bubbling bed for FCC with a mean diameter
drag coefficient plays an important role in characterizing the over- of (75 lm).
all behavior of fluidization; justifying its frequent investigation in Yang et al. [26] adopted the energy minimization multiscale
the literature. Several drag models were mentioned in the litera- (EMMS) approach, proposed by Li and Kwauk [27] to calculate the
ture, including the Gidaspow [14], Syamlal–O’Brien [17] and drag coefficient from structure parameters, and to incorporate it
Wen-Yu [18] models. into the two fluid models. They claimed that this approach reason-
Many researchers have successfully simulated the hydrody- ably resolves the heterogeneous structures and suggested the feasi-
namics of Fluid Catalytic Cracking (FCC) particles, and of the coars- bility of using it as a sub-grid closure law for the drag coefficient.
est Geldart B [19] particles, respectively in circulating fluidized More recently, the EMMS drag model was applied by Jiradilok
beds and in bubbling fluidized beds by conventional drag models. et al. [2] to predict the hydrodynamic of the turbulent fluidization
Nevertheless, less successful simulations have been performed on of FCC particles. They found that the turbulent regime was cor-
gas fluidization of Geldart A particles or turbulent fluidization of rectly computed using the modified drag of Yang et al. [26].
Geldart B particles. Although several correlations exist in the literature for the drag
Qi et al. [20] reported that classical drag models perform satis- force, considerable uncertainties still remain for its prediction.
fying results for only low gas velocities and coarse particles, in In addition to the drag coefficient, the restitution coefficient
which case the terminal velocity was equal or close to the superfi- which accounts for the inelasticity of the particle–particle collision
cial gas velocity. may affect the hydrodynamic of a fluidized bed. Indeed, being an
Zimmermann and Taghipour [21] reported also that the drag input in the kinetic theory based CFD models, it has a direct impact
models of Gidaspow and Syamlal–O’Brien are not suitable, in their on the quantities they contain, such as particle pressure, particle
original form, for simulations of Geldart A particles. In fact, these viscosity and dissipation of particle temperature. Goldschmidt
models overestimate the momentum exchange between the two et al. [10], studied the effect of the restitution coefficient on the
phases and overpredict the bed expansion with reference to exper- hydrodynamic of a dense gas fluidized bed. They reported that, to
imental data. obtain realistic simulations using fundamental hydrodynamic mod-
Zhang and Vanderheyden [22] reported that the inter-particle els, it is important to correctly take the effect of non-ideal particle–
cohesive forces leading to the formation of mesoscale bubbling particle encounters into account. Taghipour et al. [28] found higher
or clustering structures, cause a significant change in drag forces. sensitivity to the restitution coefficient at U < U mf . Gao et al. [29]
They proposed then to correct the standard drag correlation by a indicated that there is sensitivity to this coefficient in the behavior
scaling factor of ðð1  cr Þ2 Þ, to account for the mesoscale bubbling of a turbulent bed filled with Geladart B particles when its values
or clustering structures. This method has been since then used range from 0.6 to 0.99. Although, the effect of the restitution coeffi-
by several authors (Ye et al. [23]; Mckeen and Pugsley [24]) to sim- cient has been extensively explored, the role of this parameter is not
ulate bubbling fluidized beds of Geldart A particles. yet fully understood and more explorations are still required.
Recently, Mckeen and Pugsley [24] found that the poor simula- In the riser of a turbulent fluidized bed, the multiphase flow is
tions results for Geldart A particles could be attributed to the exis- very turbulent and displays high Reynolds numbers. In order to
tence of significant cohesive forces between particles. To cope with make the simulation predictions more realistic, the time average
this weakness they brought modification on the Gibilaro [25] drag turbulent behavior and the turbulent interaction between phases
S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747 1739

should be taken into account. Different models, laminar and turbu- In the current work, a multifluid Eulerian model incorporating
lent models with kinetic theory, have been used by various authors the Kinetic Theory of Granular Flows has been carried out to sim-
to simulate the hydrodynamic of gas–particle multiphase flow. ulate the hydrodynamic of a 2D circulating fluidized bed filled with
Almuttahar and Taghipour [30] compared two different viscous Geldart B particles. These simulations were conducted using the
models, the k—e turbulence and the laminar model with Kinetic commercial software package Fluent V.6.2. In order to get the best
Theory of Granular Flow. They claimed that the laminar model pre- simulation of the hydrodynamic of the fluidized bed, the effect of
dictions give more consistent results than the turbulent models. various drag models including those of Gidaspow, Syamlal–O’Brien
Nevertheless, this comparison was carried out in 2D modeling and Mckeen have been investigated. The numerical results
and since turbulence fluctuations always have three-dimensional obtained are compared to experimental data available in the liter-
spatial character then, to draw a firm conclusion, 3D comparative ature. This work is further extended to explore the effect of differ-
studies are required. However, this requires excessive computa- ent wall boundary conditions (slip, no slip and partially slip) on the
tional time, hence the 2D modeling still takes preference over 3D. hydrodynamic of the gas–solid flow. Finally, the performance of
Further parameters may affect the simulation results of a turbu- two different viscous models (the k–e turbulence and the laminar
lent fluidized bed, such as the wall boundary conditions. Li et al. model) has been evaluated by comparing their predictions with
[31] investigated the impact of this parameter in a 2D simulation reference to experimental data.
of a bubbling fluidized bed, for gas and solid phases, over the gen-
erated flow hydrodynamics. According to their investigation, the 2. Experimental setup
wall boundary conditions need to be specified with great care
due to their high relevance over the hydrodynamic. Johnson and Our simulation is based on the experimental work conducted by
Jackson [32] have developed a set of boundary conditions to define the CFB team of the I.U.S.T.I., a Laboratory in Marseille (France).
the interactions between the particles and the wall. These bound- These experiments were carried out by Van den Moortel et al.
ary conditions have been widely applied in numerical simulations [37] and were recently extended by Zaabout et al. [38,39]. They
of gas–solid flows due to their relatively simple form [29,33]. In all used a rectangular Plexiglas riser with 2 m in height and
Johnson and Jackson boundary conditions, the specularity coeffi- 0:2 m in width. The solid particles used were glass spheres with
cient, which characterizes the tangential momentum transfer due a density of 2400 kg=m3 and a mean diameter of 120 lm. Air at
to collision between the particles and the wall, must be specified room temperature was used as the fluidizing gas. The static bed
with great care as well (Li et al. [31]). The effect of this coefficient height was 0:1 m and the superficial gas velocity was 1 m=s. The
on the numerical results for different solid particle distributions geometry of the fluidization loop is depicted in Fig. 1.
has been studied by several researchers [31,34]. The core-annulus
structure of the solid phase was reported to be very sensitive to the 3. Computational fluid dynamics simulation set up
choice of the specularity coefficient [31,35,36]. However, choosing
the right number is critical for most validation studies due to the Through the development of high performance computer and
difficulty in measurements. Further theoretical investigations are the advancement of computational methodologies, computational
then required in order to contribute to a better understanding of fluid dynamics analysis of multiphase systems has been largely
the gas–solid flows. enhanced and appears as an efficient and promising tool in the

Cyclone

Flow meter

Riser
PDA Valve

H
Buffer vessel Pump
Air outlet

D
Lifting screw
Y

Air distributor

Air inlet

Fig. 1. Schematic of the experimental apparatus of the turbulent fluidized bed.


1740 S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747

Table 1
Governing equations.

Equations Mathematical expressions


Mass conservation equations of gas and solids phases @ðag qg Þ (T1-1)
@t aq
þ r  ð g g~ ug Þ ¼ 0
@ðas qs Þ (T1-2)
@t þ r  ð s saq
~
usÞ ¼ 0
ag + as = 1 (T1-3)
Momentum conservation equations of gas and solids phases @ðag qg ~
ug Þ !! (T1-4)
@t þ r  ð g g~aq ug Þ ¼ r  ðg Þ  g rP  bð ug us Þ þ g g g
ug ~ s a aq
@ðas qs ~
us Þ ! ! (T1-5)
@t þ r  ð s s~ us Þ ¼ r  ðg Þ  s rP  bð ug us Þ þ s s g
aq
us~ s a aq
Granular Temperature H ¼ 3u 1 02 (T1-6)
 
Equation of conservation of solids fluctuating energy 3 @ðas qs HÞ
us HÞ ¼ ðP sI þ s Þ : r~
þ r  ð s s~aq s c
us  r  q   J (T1-7)
2 @t
Equation of conservation of solids fluctuating energy in algebraic form 0 ¼ ðP sI þ s
s Þ : r~
us  c (T1-8)

development and design of circulating fluidized beds. We propose included in Fluent 6.2 by default. However, the Mckeen drag model
in the current work to simulate the hydrodynamics of a gas–solid has been implemented into Fluent by using a user-defined function
CFB riser by means of a multifluid Eulerian model incorporating (UDF).
the Kinetic Theory of Granular Flows using the commercial CFD
software package, Fluent. With this approach, the two different 3.2. Kinetic Theory of Granular Flow (KTGF)
phases are mathematically treated as continuous and fully inter-
penetrating. However, to obtain a set of equations having similar The Two-Fluid-Model (TFM) needs closure equations to illus-
structures for both phases, the conservation equations are derived. trate the rheology of the solid phase (the solid pressure and the
A brief summary of the governing equations and its constitutive solid phase viscosity). There are two different approaches in the lit-
equations is provided respectively in Tables 1 and 2. erature brought to describe these stresses: the Constant Viscosity
Model (CVM) and the Kinetic Theory of Granular Flow (KTGF).
3.1. Gas–solid exchange coefficient b (drag models) The first approach treats the viscosity of the solid phase as a con-
stant and the solid pressure as a function of the local solid porosity
The momentum exchange between the gas and solid phases is only. The second approach was founded on the application of the
accounted for by the drag coefficient, which has a significant effect kinetic theory of dense gases to the particulate assemblies. It gives
on the prediction of the Eulerian–Eulerian model. Based on the the- more perspicacity in terms of interactions between particles by
ory and application, several expressions for this coefficient are assuming that binary collisions between hard spheres take place
available in the open literature. In the current work, the effect of instantaneously. Analogous to the thermodynamic temperature
various drag models including those of Gidaspow [14], Syamlal– for gases, the model introduces a granular temperature as a mean
O’Brien [17] and Mckeen [243] has been analyzed. to measure particle velocity fluctuations. Fluent uses the granular
The Gidaspow drag model is a standard model which combines temperature as in Eq. (T1-6). The kinetic fluctuation energy of solid
the Wen-Yu [18] and the Ergun [42] equations to cover the whole particles that describes the distribution of granular temperature, is
range of void fractions. The Ergun equation is used where the sus- expressed by: Eq. (T1-7). Where ðPs I þ ss Þ : r~
us is the generation
pension is dense, whereas the formulation by Wen-Yu is used of fluctuating energy due to the local acceleration of the particles, q
where the suspension is dilute. The drag model developed by Syam- is the diffusion of the fluctuating energy defined as q ¼ kH rH
lal and O’Brien [17] is based on the measurement of the terminal (where kH is the diffusion coefficient for granular energy [42]), c
velocities of particles in fixed or fluidized beds. This correlation is is the dissipation of fluctuating energy due to inelastic particle–
a function of the relative Reynolds number and of the volume frac- particle collisions and J is the exchange of the fluctuating energy
tion. When applying this correlation, the solid stresses should be between the gas and the solid phase.
defined according to O’Brien and Syamlal [43]. Finally, the Mckeen Instead of solving the full granular temperature equation, van
drag model [24] is based on a single compact function overall value Wachem et al. [44] proposed to simplify this equation by using
of voidage. The detailed mathematical depiction for these models is an algebraic form. They assumed that the granular energy is in a
shown in Table 3. The Gidaspow and Syamlal drag models, are steady state and dissipates locally, thus convection and diffusion

Table 2
Constitutive equations.

Equations Mathematical expressions


h i
Gas phase stress tensor (Lun et al. [41]) sg ¼ ag ðng  23 lg Þðr  ~
ug ÞI þ lg ððr~ ug Þ þ ðr~
T
ug Þ Þ (T2-1)
h i
Solid phase stress tensor (Lun et al. [41]) us ÞI þ ls ððr~
ss ¼ as ðns  23 ls Þðr  ~ us Þ þ ðr~
T
us Þ Þ (T2-2)

Solids pressure (Lun et al. [41]) P s ¼ as qs H þ 2g 0 2s


a qs Hð1 þ es Þ (T2-3)
Solids shear viscosity ls = ls,col + ls,kin + ls,fr (T2-4)
qffiffiffi
Collisional viscosity (Gidaspow et al. [14]) ls;col ¼ 45 as qs ds g 0 ð1 þ es Þ H (T2-5)
p
Kinetic viscosity (Gidaspow et al. [14]) pffiffiffiffiffiffiffiffi
qs ds  2 (T2-6)
ls;kin ¼ 10 4
96 Hp ð1þes Þas g 0 1 þ 5 g 0 as ð1 þ es Þ
pffiffiffiffiffiffi  
Kinetic viscosity (Syamlal et al. [18]) as qs ds Hp
ls;kin ¼ 6ð3es Þ 1þ 2
 as  ð1 þ es Þð3es  1Þ (T2-7)
5 g0
Frictional viscosity (Schaeffer et al. [40] ls;fr ¼ P2spsinffiffiffiffiffi
/ (T2-8)
I
2D
qffiffiffi
Solids bulk viscosity (Lun et al. [41]) ns ¼ 43 as qs ds g 0 ð1 þ es Þ H (T2-9)
p
Radial distribution function (Lun et al. [41])   1=3 1 (T2-10)
g 0 ¼ 1  as;max a s

 qffiffiffi
Collisional energy dissipation (Lun et al. [41]) cs ¼ 3ð1  e2s Þa2s qs g 0 H d4p Hp (T2-11)
S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747 1741

Table 3
Drag model equations.

Equations Mathematical expressions


Gidaspow drag model [14] ag > 0.8 (T3-1)
ag ð1ag Þ
b ¼ 34 C D0 ds
qg j~ us ja2:65
ug  ~
ag 6 0.8 (T3-2)
2
ð1a Þ lg q
b ¼ 150 ag g  1:75ð1  ag Þ dgs j~ug  ~
us j
ðds Þ2

24
½1 þ 0:15ðRe s Þ0:687
; Re s < 1000 (T3-3)
C D0 ¼ Re s
0:44; Res > 1000
Res ¼
ag qg j~
ug ~
us jds (T3-4)
lg
Syamlal and O’Brien drag model [17] 3 as ag qg !
! (T3-5)
b ¼ 4 #2 d C D jus  ug j
r;s s
2 (T3-6)
4:8
C D ¼ 0:63 þ pffiffiffiffiffiffiffiffiffiffiffiffi
Res =#r;s
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(T3-7)
#r;s ¼ 0:5ðA  0:06Res þ ð0:06Res Þ2 þ 0:12Res ð2B  AÞ þ A2 Þ
8 1:28
< P ag ðag 6 0:85Þ (T3-8)
A ¼ a4:14 ; B ¼ Q
g
: ag ðag > 0:85Þ

P = 0.8 et Q = 2.65
Mckeen drag model [24]   q j! ! (T3-9)
g u s  ug j
b ¼ C 17:3
Res þ 0:336 ds
as ag1:8

terms can be neglected. According to the work of van Wachem The Reynolds stress tensor for the gas phase, given in Eq. (T1-7),
et al. [44], this assumption is only valid for dense particulate flows. takes the following form:
However, several researchers have successfully simulated gas–
solid flows in circulating fluidized beds using the algebraic granu- 2 ! ! !
s00g ¼  ðag kg þ qg lt;g r  U g ÞI þ qg lt;g ðr U g þ r U Tg Þ ð1Þ
lar temperature approach such as Gomez et al. [45], Milinkumar 3
et al. [34] and Chen et al. [46]. They found that the simplification !
where U g denotes the gas-weighted velocity and lt;g is the gas tur-
of the granular temperature does not lead to significantly different
bulent viscosity given by the following expression:
results, however, it decreases greatly the computational effort, and
saves computational time (22% in total computational time). Gry-
kg
czka et al. [47] investigated similarly the effect of the algebraic lt;g ¼ qg C l ð2Þ
eg
granular temperature approach on spouted bed hydrodynamics.
It was observed that 2D calculations are less time-consuming
Here kg is the turbulent kinetic energy, eg is the dissipation rate of
and give satisfying granular temperatures. They further reported
the turbulent energy and C l is 0:09 .
that differences between the two approaches, full granular temper-
Turbulent predictions of the gas phase are obtained from the
ature equation and simplified algebraic approach were even
following equations of the modified k—e model:
expected to be efficient in three dimensional simulations. There-
fore, the algebraic form of the granular temperature was then used @ ! lt;g
in this study and described as in Eq. (T1-8). ðag qg kg Þ þ r  ðag qg U g kg Þ ¼ r  ðag rk Þ þ ag Gk;g
@t rk g
 ag qg eg þ ag qg Pkg ð3Þ
3.3. Turbulence modeling
@ ! lt;g e
As aforementioned, two different coexisting regions character- ðag qg eg Þ þ r  ðag qg U g eg Þ ¼ r  ðag re Þ þ ag g
@t re g kg
ize turbulent fluidized bed: the bottom dense region and the upper
dilute region. The bottom dense region displays low Reynolds  ðC 1e Gk;g  C 2e qg eg Þ
numbers and, the effect of turbulent behavior and the interaction þ ag qg Peg ð4Þ
between phases are not very consistent due to the dense solid
phase concentration. On the other hand, the upper dilute region where Pkg and Peg represent the influence of the dispersed phases
displays high Reynolds numbers and very important gas-turbulent on the continuous phase and Gk;g the production of turbulent kinetic
stresses. Therefore, it is difficult to clearly determine the impor- energy. These terms are given by the following equations:
tance of gas or solid phase turbulent stresses in the whole riser.
In the literature, satisfactory agreement with experimental data b
Pkg ¼ ½ksg  2kg þ ~
ugs  ~
udr  ð5Þ
has been reported in both approaches; laminar [29,48] and turbu- ag qg
lent models [49].
In the present work, comparison between these two approaches eg
has been established in a try to find out which leads to the best Peg ¼ C 3e Pkg ð6Þ
kg
prediction of the hydrodynamic behavior of a fluidized bed. The
dispersed turbulent model was used to describe the effects of tur-
T
bulent fluctuations of velocities and scalar quantities in the gas- Gk;g ¼ lt;g ½r~
ug þ rð~
ug Þ  : r~
ug ð7Þ
eous phase. It is based on the standard k—e model with
supplementary terms for the momentum exchange between the ksg and ~
ud r represent respectively the covariance of the velocities of
two phases. The calculation of the turbulence for the dispersed the gas and the dispersed phase, and the drift velocity. These terms
phase is achieved by application of the Tchen theory correlation and all the other parameters involved in the dispersed k—e model
[50] into Fluent. can be found in the Fluent User’s Guide [51].
1742 S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747

4. Initial and boundary conditions in the riser is set using the available experimental data. At the
walls, the no slip boundary condition is used for the gas phase,
The governing equations mentioned in the previous section are while the partial slip boundary condition of Johnson and Jackson
solved using the Finite Volume Method (FVM) employing the [32] is specified for the solid phase. Physical properties and simu-
phase coupled algorithm for pressure–velocity coupling (PCSIM- lation parameters are listed in Table 4.
PLE). The discretization scheme used in the current work is carried
out with the first order upwind scheme. A convergence criterion of 5. Results and discussion
1003 is specified as the relative error between successive itera-
tions. The commercial CFD package Fluent V.6.2 is used to provide 5.1. Study of modeling parameters
a numerical solution for these equations. The calculations are con-
ducted using a 2D Cartesian space. The 2D computational domain Since there is no certitude on the appropriate values of the
(shown in Fig. 2) is discretized using 18848 cells with a uniform model parameters such as the restitution coefficient, the specular-
quadratic mesh. ity coefficient and the drag coefficient, the main objective of the
At the inlet, the velocity-inlet boundary condition for both present section is to significantly analyze the effects of these
phases is used, while on the outlet the pressure outlet boundary parameters on the gas–solids flow.
condition is applied. For the solid phase, the initial solids inventory
5.1.1. Comparison of different drag models
In order to get an optimum drag model which lead to the best
simulation of the fluidized bed hydrodynamic, the drag models of
Gidaspow, Syamlal-O’Brien and Mckeen are explored and
compared.
The simulation results for the transient behavior of the axial
particle velocity under the same operating conditions and at differ-
ent heights are illustrated on Figs. 3 and 4.
The existence of the traditional core-annulus flow structure is
clearly predicted by all models. A mean axial particle velocity that
is negative near the wall and positive away from it characterizes
this kind of flow structure. In comparison, the Syamlal drag model
overpredicts the particle velocity in the central region and under-
predicts it near the wall over the two elevations. This should be
attributed to the lack of consideration of the effects of cohesive
forces and agglomeration, which results in higher drag forces,
and then higher solid velocities in the core region, and lower solid
velocities near the wall region.
The axial particle velocity profile shows reasonable qualitative
agreement with the Mckeen drag model when compared to exper-
imental data at a height of 1 m. At a height of 1:2 m this model
under predict the axial particle velocity near the central region
and reveals the typical annulus structure of the experimental data.
The discrepancy especially at the higher elevation may be attrib-
uted to the incapability of this model to capture the characteristics
of the dilute region. Indeed, Gao et al. [48] and Li et al. [52] report

Fig. 2. Schematic diagram of the simulated riser.


1
Axial particle velocity (m/s)

Table 4
Modeling parameters. 0
Description Value 0.5 0.6 0.7 0.8 0.9 1

Riser height H 2m (X/D)


Riser width 0.2 m -1
Static bed height H0 0.1 m
Gas density qg 1.2 kg/m3 Gidaspow
Particle density qs 2400 kg/m3 Syamlal
Particle diameter ds 120 lm
-2 Mckeen
Initial solid volume fraction a0 0.6
Inlet gas velocity Ug 0.9 m/s, 1 m/s, 1.1 m/s Exp (Van Den Moortel, 1998)
Solid flux Gs 0.1 kg/m2s, 0.22 kg/m2s, 0.413 kg/m2s
Angle of internal friction 30°
Restitution coefficient es 0.8, 0.9, 0.95 and 0.99
-3
Specularity coefficient u 0, 0.001, 0.25 and 1
Maximum particle packing limit 0.64 Fig. 3. The time-averaged axial particle velocity predicted along the radial direction
Time step 103 s by various drag models at Y ¼ 1 m (superficial gas velocity U g ¼ 1 m=s, k—e
turbulent model, es ¼ 0:9, u ¼ 1).
S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747 1743

2 experimental results. Some differences still exist however, and


may be justified by the unrealistic 2D simulation of a real 3D sys-
tem. According to these findings, the standard k—e model will then
be adopted in the rest of this work.
Axial particle velocity (m/s)

5.1.3. Restitution coefficient


The restitution coefficient, which represents the degree of elas-
ticity of particle–particle collisions, ranges from zero to 1, from
0 fully inelastic to fully elastic collisions respectively. With the aim
0.5 0.6 0.7 0.8 0.9 1 to confirm the value of this parameter, four different restitution
(X/D) coefficients (es ¼ 0:8, 0:9, 0:95 and 0:99) were selected and further
compared. The computed solid volume fraction contour and axial
-1 Gidaspow solid velocity profile at a height of 1:2 m are shown in Figs. 6 and
Syamlal 7, respectively. On Fig. 6, it can be observed that decreasing the res-
Mckeen titution coefficient lead to more and bigger bubbles in the bed,
Exp (Van Den Moortel, 1998)
causing more pronounced heterogeneities (thus instabilities).
Referring to Eq. (T2-11), the strong dependency of the dissipation
-2
of the granular temperature on the restitution coefficient appears
Fig. 4. The time-averaged axial particle velocity predicted along the radial direction in the term ð1  e2s Þ.
by various drag models at Y ¼ 1:2 m (superficial gas velocity U g ¼ 1 m=s, k—e Then, decreasing the restitution coefficient would lead to a
turbulent model, es ¼ 0:9, u ¼ 1). greater dissipation of mechanical energy dissipation, resulting in
more particles attaching to each other and more bubbles created
in the bed. Therefore, as the collisions become more inelastic the
that the Mckeen drag model cannot capture the dilute characteris-
particles bed becomes more compact in the dense region and the
tics of turbulent fluidized beds in spite of the better prediction for
bubbles become more present. We can further observe that by
the dense region.
increasing the restitution coefficient more particles enter into the
Gidaspow drag model provided good description of the hydro-
dilute dispersed region. In fact, increasing es leads to a lower dissi-
dynamic of fluidized beds filled with Geldart B particles (Lu et al.
pation of the kinetic energy fluctuation due to elastic particle–par-
[53]). In the present work, the same model gave reasonable quali-
ticle collisions and then particles are conveyed straightly forward
tative and quantitative agreement with reference to experimental
in the upward direction. Gao et al. [32] have also investigated
data in term of particle velocity profiles. We propose then to adopt
the effect of the restitution coefficient on the hydrodynamic behav-
the Gidaspow drag model in the rest of this work.
ior of a turbulent fluidized beds filled with Geldart B particles and
they found that with the increase of es more particles enter into the
5.1.2. Comparison of the laminar and the turbulence model dilute region, similarly to our observations (Fig. 6).
To study the effect of turbulence modeling on flow predictions, According to Fig. 7, the restitution coefficient of 0:99, represent-
simulations with laminar model and k—e turbulent model were ing the case of nearly ideal particle–particle interactions, shows an
conducted and further compared to experimental measurements.
Fig. 5 compares the time averaged axial solid velocity obtained
with both models to experimental data evaluated at Y ¼ 1:2 m.
In both models, the core-annulus structure of the gas–solid flow
is well predicted. However, the solid velocity profile predicted by
the standard k—e model provides a better agreement with the

2
Axial particle velocity (m/s)

0
0.5 0.6 0.7 0.8 0.9 1
(X/D)

Turbulent
-1
Laminar
Exp (Van Den Moortel, 1998)

-2

Fig. 5. The time-averaged axial particle velocity predicted along the radial direction Fig. 6. The computed instantaneous solid volume fraction for three different
by laminar and turbulent model at Y ¼ 1:2 m (superficial gas velocity U g ¼ 1 m=s, restitution coefficient at t ¼ 30 s (superficial gas velocity U g ¼ 1 m=s, Gidaspow
Gidaspow drag model, es ¼ 0:9, u ¼ 1). drag model, k—e turbulent model, u ¼ 1).
1744 S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747

2 2
Axial particle velocity (m/s)

Axial particle velocity (m/s)


1 1

0 0
0.5 0.6 0.7 0.8 0.9 1 0.5 0.6 0.7 0.8 0.9 1
(X/D) (X/D)

Rest. 0.8 Spec.coeff=0


-1 Rest. 0.9 -1
Spec.coeff=0.001
Rest. 0.95
Spec.coeff =0.25
Rest. 0.99
Spec.coeff=1
Exp (Van Den Moortel, 1998)
Exp (Van Den Moortel, 1998)
-2 -2

Fig. 7. The time-averaged axial particle velocity predicted along the radial direction Fig. 8. The time-averaged axial particle velocity predicted along the radial direction
by several restitution coefficients at Y ¼ 1:2 m (superficial gas velocity U g ¼ 1 m=s, by several specularity coefficients at Y ¼ 1:2 m (superficial gas velocity U g ¼ 1 m=s,
Gidaspow drag model, k—e turbulent model, u ¼ 1). Gidaspow drag model, k—e turbulent model, es ¼ 0:9).

overproduction of particles velocity within the riser’s core and near The axial particle velocities predicted by u ¼ 1 and u ¼ 0:001
the wall region. In fact, increasing the restitution coefficient are far away from the experimental data near the wall and within
reduces the presence of bubbles in the riser and then more parti- the core of the riser. However, the specularity coefficient close to
cles enter into the dilute dispersed region with high velocity. How- one shows a reasonable agreement for all models carried out,
ever, the best agreement between the model predictions and and hence is chosen to be the specularity coefficient for subsequent
experimental data is obtained with es ¼ 0:9: this value is to be simulations.
adopted in the rest of our CFD simulations.

5.2. Comparison of modeling and experimental data


5.1.4. Specularity coefficient
The specularity coefficient, u, represents the measurement of In order to produce an adequate description of the behavior of
the fraction of collisions that transfer momentum to the wall. the two phase flow in the riser section of a circulating fluidized
The value of u varies from zero (for smooth walls) to one (for bed unit operating in a turbulent fluidization regime, several sim-
rough walls). In the vicinity of zero, a free slip boundary condition ulations have been performed with different modeling parameters
for the solids tangential velocity is achieved at the walls, while in and criteria. The best agreement between model predictions and
the vicinity of one, a substantial amount of momentum transfers experimental data was obtained using: the k—e turbulent viscous
is obtained between the particles and the wall. Four different spec- model, the Gidaspow drag model and a no slip boundary condition
ularity coefficient values (u ¼ 0, 0:001, 0:25 and 1) are tested in as discussed earlier in Section 5.1. So, to investigate the limitations
the present work in a try to find out which fits best the handled and capabilities of this model, the CFD model was evaluated under
case. The time-averaged axial solid velocity at Y ¼ 1:2 m, are given different operating conditions. The same modeling parameters
in Fig. 8. In the case of the free slip boundary condition (u ¼ 0), were used for all cases with varying only the superficial gas veloc-
there is no friction between the particles and the wall. The free slip ity, U g and the solid mass flux Gs , to match the experimental oper-
of particles on the wall leads to more particles in the downward, ating conditions.
thus resulting in higher particle velocities and consequently higher Fig. 9 compares the numerical results to experimental data at
particles concentrations in the bottom of the riser. However, in the different superficial gas velocities (U g ¼ 0:9 m=s and U g ¼ 1:1 m=s).
case of the no slip boundary condition ðu ¼ 1Þ, friction is impor- A satisfying agreement is obtained for all cases. This is likely to
tant between the particles and the wall. This strong friction pre- provide further support for the effectiveness and accuracy of the
vents the downward of particles and then higher downward proposed model to simulate well the core-annulus flow in a turbu-
particles velocity near the wall. As observed in Fig. 8, the downflow lent fluidized bed.
of the particles at the wall increases when decreasing the specular- Fig. 10 shows the distribution of the computed time-averaged
ity coefficient as observed by Armstrong et al. [33] and Li et al. [26]. solid volume fractions at different heights and for U g ¼ 1 m=s. As
Both authors reported that a small specularity coefficient leads to a we can see, solid volume fractions appear high near the wall region
considerably higher velocity of the particle downflow near the and low in the center of the bed. These concentration profiles
wall. As to the axial particle velocity in the center (core region), clearly illustrate the inherent core-annular flow structure of the
it is found to increase with the specularity coefficient u. The spec- solid phase. This kind of flow structure is characterized by the
ularity coefficient affects consistently the particles’ movement accumulation of the solid phase moving downward at the wall
near the wall and in the core region. This study indicates then that and by the upward movement of a dilute gas–particles stream in
for a 2D simulation of a turbulent fluidized bed filled with Geldart the core of the riser. We can also observe that the solid volume
B particles, an arbitrary setting of the specularity coefficient may fraction decreases at higher locations.
lead to unreasonable predictions of the related hydrodynamic At heights of 0:07 m, 0:1 m and 0:3 m, the solid volume fraction
proprieties. profile attains a local minimum value near the wall region due to
The time-averaged axial solid velocity at Y ¼ 1:2 m is also com- solid entrance. This profile does no longer attains a minimum for
pared to the experimental data of Van den Moortel [37] in Fig. 8. heights beyond 0:3 m (Y ¼ 0:5, 0:7, 1 and 1:2 m). A similar
S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747 1745

2
Axial particle velocity (m/s)

0
0.5 0.6 0.7 0.8 0.9 1
(X/D)
H=1.2 m
-1
Exp (Ug=1.1 m/s)
Num (Ug=1.1 m/s)
Exp (Ug=0.9 m/s)
Num (Ug=0.9 m/s)
-2

Fig. 9. Comparison of the predicted axial particle velocity with the experimental
data at Y ¼ 1:2 m and for different superficial gas velocities (a) U g ¼ 0:9 m=s and (b)
U g ¼ 1:1 m=s. (Gidaspow drag model, k—e turbulent model, es ¼ 0:9, u ¼ 1).

Ug =1 m/s
0.3
YH=0.07 m
YH=0.1 m
YH=0.3 m Fig. 11. The computed instantaneous solid volume fraction for different superficial
YH=0.5 m gas velocities and at t ¼ 30 s (Gidaspow drag model, k—e turbulent model, es ¼ 0:9,
Solid Volume Fraction

YH=0.7 m u ¼ 1).
YH=1 m
0.2 YH=1.2 m

0.7
Standard deviation of axial

0.1
0.6
particle velocity (m/s)

0.5
0
0 0.2 0.4 0.6 0.8 1
(X/D)
0.4 Exp (Van den Moortel,1998) Ug=1 m/s
Fig. 10. Radial distribution of the axial solid volume fraction at different axial bed Ug=0.9 m/s
Ug=1 m/s
distances for U g ¼ 1 m=s. (Gidaspow drag model, k—e turbulent model, es ¼ 0:9,
Ug=1.1 m/s
u ¼ 1).
0.3

observation in the concentration profiles has been made by


Ernst-Ulrich Hartge et al. [30] in a circulating fluidized bed filled 0.5 0.6 0.7 0.8 0.9 1
with Geldart B particles. (X/D)
Fig. 11 displays the contours of the instantaneous solid concen-
tration in the turbulent fluidized bed at t ¼ 30 s for three different Fig. 12. The time-averaged standard deviation of axial particle velocity for different
gas velocities. We observe for all cases the formation of clusters in superficial gas velocities along the radial direction at Y ¼ 1 m (Gidaspow drag
model, k—e turbulent model, es ¼ 0:9, u ¼ 1).
the core region and near the wall. Clusters are denser at the wall
than in the core zone. Such flow patterns are attributed to the for-
mation, motion and split of clusters. When particles forming clus- region with the superficial gas velocity. Indeed, in the case of high
ters are grouped near the wall region, they fall downward along velocity (U g ¼ 1:1 m=s) there are more particles entrained into
the wall, collecting particles on their way. As a result the wall clus- the upper dilute region, as previously seen in Fig. 11. An increase
ters appear denser than the core zone clusters. In Fig. 11, we also of the superficial gas velocity leads then to further bubble coales-
observe that more particles are entrained into the dilute zone cence and to an increase of the drag forces acting on the particles.
under increasing gas velocities. The flow structure in the upper zone becomes then more heteroge-
Profiles of the standard deviation of the axial particle velocity at neous, leading to increasing particles velocity fluctuations.
a given height are plotted in Fig. 12 for three different gas velocities. We can also observe from Fig. 12 that the shape of the velocity
These fluctuations, which reflect the turbulence of the local suspen- fluctuations plots changes close to the wall by accusing a sudden
sion flow, are further analyzed in the following. As we can see from decrease due to the existence of a bulk annulus dense solids down-
the plots, the velocity fluctuations level increases in the center ward flow under all velocity cases.
1746 S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747

Fig. 12 also contains the experimental data of Van den Moortel [13] J.L. Sinclair, R. Jackson, Gas–particle flow in a vertical pipe with particle-
particle interactions, AIChE J. 35 (1989) 1473–1496.
[37] for the particular case of U g ¼ 1 m=s. A good agreement is
[14] D. Gidaspow, Multiphase Flow Fluidization: Continuum Kinetic Theory
obtained qualitatively as well as quantitatively. Some discrepan- Description, Academic Press, Boston, 1994.
cies are however observed near the wall; probably due to the [15] C. Pain, S. Mansoorzadeh, C.R.E.D. Oliveira, A.J.H. Goddard, Numerical
choice of the turbulence model. Elsewhere, the level of fluctuations modeling of gas–solid fluidized beds using the two-fluid approach, Int. J.
Numer. Meth. Fluids 36 (2001) 91–124.
in the core is around 35% of the mean velocity in both experimen- [16] F. Taghipour, N. Ellis, C. Wong, Experimental and computational study of gas–
tal and computed cases, and the decreasing shape of the plots are solid fluidized bed hydrodynamics, Chem. Eng. Sci. J. 60 (2005) 6857–6867.
very similar when exiting from the core to the annulus region. [17] M. Syamlal, T.J. O’Brien, Computer simulation of bubbles in a fluidized bed,
AIChE Symp. Ser. 85 (1989) 22–31.
[18] C.Y. Wen, Y.H. Yu, Mechanics of fluidization, Chem. Eng. Progr. Symp. Ser. 62
(1966) 100–111.
6. Conclusion
[19] D. Geldart, The effect of particle size and size distribution on the behavior of
gas-fluidized bed, Powder Technol. 6 (1972) 201–215.
The behavior of the solid phase in the upper dilute zone of a tur- [20] H. Qi, C. You, A. Boemer, U. Renz, 2000. Eulerian simulation of gas–solid two-
phase flow in a CFB-riser under consideration of cluster effects. in: D. Xu, S.
bulent fluidized bed riser was investigated by means of a multi-
Mori (Eds.), Fluidization 2000: Science and Technology, Xi’an Publishing
fluid Eulerian–Eulerian CFD model based on the Kinetic Theory of House, Xi’, pp. 231–237.
Granular Flows and using the commercial CFD package Fluent [21] S. Zimmermann, F. Taghipour, CFD modeling of the hydrodynamics and
V.6.2. Various drag models including those of Gidaspow, Syam- reaction kinetics of FCC fluidized-bed reactors, Ind. Eng. Chem. Res. 44 (2005)
9818–9827.
lal–O’Brien, and Mckeen were tested in order to get an optimum [22] D.Z. Zhang, W.B. Vanderheyden, The effects of mesoscopic structures on the
drag model for a better simulation of the hydrodynamic of the flu- macroscopic momentum equations for two-phase flows, Int. J. Multiph. Flow
idized bed. Compared to experimental data available in the litera- 28 (2002) 805–822.
[23] M. Ye, J. Wang, M.A. Van der Hoef, J.A.M. Kuipers, Two fluid modeling of
ture, the Gidaspow drag model gave the most reasonable Geldart A particles in gas-fluidized beds, Particuology 6 (2008) 540–548.
qualitative and quantitative agreement. Two viscous models were [24] T. Mckeen, T. Pugsley, Simulation and experimental validation of a freely
tested and best agreement with experimental data was obtained bubbling bed of FCC catalyst, Powder Technol. 129 (2003) 139–152.
[25] L.G. Gibilaro, R. Di Felice, S.P. Waldram, P.U. Foscolo, Generalized friction factor
with the k—e turbulent model. The computational results showed and drag coefficient correlations for fluid-particle interactions, Chem. Eng. Sci.
certain sensitivity to the restitution coefficient in the turbulent flu- 40 (1985) 1817–1823.
idized bed filled with Geldart B particles, and the simulated axial [26] N. Yang, W. Wang, W. Ge, J. Li, CFD simulation of concurrent-up gas–solid flow
in circulating fluidized beds with structure-dependent drag coefficient, Chem.
solid velocity fitted better the experimental data when using the
Eng. J. 96 (2003) 71–80.
restitution coefficient of 0.9. While analyzing the model prediction [27] J. Li, M. Kwauk, Particle–fluid Two-phase Flow—The Energy minimization
sensitivity to the specularity coefficient, it was found that a reason- Multi-scale Model, Metallurgy Industry Press, Beijing, 1994.
[28] F. Taghipour, N. Ellis, C. Wong, Experimental and computational study of gas–
able reproduction of the experimental results was reached for a
solid fluidized bed hydrodynamics, Chem. Eng. Sci. 60 (2005) 6857–6867.
specularity coefficient close to one. Finally, particle velocities, [29] X. Gao, C. Wu, Y.W. Cheng, L.J. Wang, X. Li, Experimental and numerical
radial solid volume fraction and standard deviation of axial parti- investigation of solids behavior in a gas–solid turbulent fluidized bed, Powder
cles velocity distributions were investigated for different gas veloc- Technol. 228 (2012) 1–13.
[30] A. Almuttahar, F. Taghipour, Computational fluid dynamics of high density
ities: the results provided further support for the effectiveness and circulating fluidized bed riser: study of modeling parameters, Powder Technol.
accuracy of the proposed models to simulate the core-annulus 185 (2008) 11–23.
two-phase flow in a turbulent fluidized bed. [31] T. Li, J. Grace, X. Bi, Study of wall boundary condition in numerical simulations
of bubbling fluidized beds, Powder Technol. J. 203 (2010) 447–457.
[32] P.C. Johnson, R. Jackson, Frictional–collisional constitutive relations for
References granular materials with application to plane shearing, J. Fluid Mech. 176
(1987) 67–93.
[33] C. Loha, H. Chattopadhyay, P.K. Chatterjee, 2013. Effect of coefficient of
[1] R. Andreux, T. Gauthier, J. Chaouki, O. Simonin, New description of fluidization
restitution in Euler–Euler CFD simulation of fluidized-bed hydrodynamics,
regimes, AIChE J. 51 (2005) 1125–1130.
Particuology, in press.
[2] V. Jiradilok, D. Gidaspow, S. Damronglerd, W.J. Koves, R. Mostofi, S.
[34] M.T. Shah, R.P. Utikar, M.O. Tade, V.K. Pareek, Hydrodynamics of an FCC riser
Nitivattananon, Kinetic theory based CFD simulation of turbulent fluidization
using energy minimization multiscale drag model, Chem. Eng. J. 168 (2011)
of FCC particles in a riser, Chem. Eng. Sci. 61 (2006) 5544–5559.
812–821.
[3] N. Ellis, H.T. Bi, C.J. Lim, J.R. Grace, Hydrodynamics of turbulent fluidized beds
[35] S. Benyahia, M. Syamlal, T.J. O’Brien, Evaluation of boundary conditions used to
of different diameters, Powder Technol. 141 (2004) 124–136.
model dilute, turbulent gas/solids flows in a pipe, Powder Technol. 156 (2005)
[4] J.W. Wang, Flow structures inside a large-scale turbulent fluidized bed of FCC
62–72.
particles: Eulerian simulation with an EMMS-based sub-grid scale model,
[36] X. Lan, C. Xu, J. Gao, M. Al-Dahhan, Influence of solid phase wall boundary
Particuology 8 (2010) 176–185.
condition on CFD simulation of spouted bed, Chem. Eng. Sci. 69 (2012) 419–
[5] G. Liu, P. Wang, S. Wang, L. Sun, Y. Yang, P. Xu, Numerical simulation of flow
430.
behavior of liquid and particles in liquid–solid risers with multi scale
[37] T. Van den Moortel, E. Azario, R. Santini, L. Tadrist, Experimental analysis of the
interfacial drag method, Adv. Powder Technol. 24 (2013) 537–548.
gas–particle flow in a circulating fluidized bed using a phase Doppler particle
[6] M.A. van der Hoef, M. van Sint Annaland, N.G. Deen, J.A.M. Kuipers, Numerical
analyzer, Chem. Eng. Sci. 53 (1998) 1883–1899.
simulation of dense gas–solid fluidized beds: a multiscale modeling strategy,
[38] A. Zaabout, H. Bournot, R. Occelli, A. Draoui, Solids behavior in dilute zone of a
Annu. Rev. Fluid Mech. 40 (2008) 47–70.
CFB riser under turbulent conditions, Particuology 9 (2011) 598–605.
[7] J. Xie, W. Zhong, B. Jin, Y. Shao, Y. Huang, Eulerian–Lagrangian method for
[39] A. Zaabout, H. Bournot, R. Occelli, B. Kharbouch, Local solid particle behavior
three-dimensional simulation of fluidized bed coal gasification, Adv. Powder
inside the upper zone of a circulating fluidized bed riser, Adv. Powder Technol.
Technol. 24 (2013) 382–392.
22 (2011) 375–382.
[8] A. Samuelsberg, B.H. Hjertager, An experimental and numerical study of flow
[40] D.G. Schaeffer, Instability in the evolution equations describing incompressible
patterns in a circulating fluidized bed reactor, Int. J. Multiphase Flow 22 (1995)
granular flow, J. Different. Eq. 66 (1987) 19–50.
575–591.
[41] C. Lun, S. Savage, D. Jeffrey, N. Chepumiy, Kinetic theories of granular flow:
[9] S. Benyahia, H. Arastoopour, T.M. Knowlton, H. Massah, Simulation of particles
inelastic particles in coquette flow and slightly inelastic particles in a general
and gas flow behaviour in the riser section of a circulating fluidized bed using
flow field, J. Fluid Mech. 140 (1984) 223–256.
the kinetic theory approach for the particulate phase, Powder Technol. 112
[42] S. Ergun, Fluid flow through packed columns, Chem. Eng. Prog. 48 (1952) 89–
(2000) 24–33.
94.
[10] M.J.V. Goldschmidt, J.A.M. Kuipers, W.P.M. van Swaaij, Hydrodynamic
[43] T.J. O’Brien, M. Syamlal, 1993. Particle cluster effects in the numerical
modelling of dense gas-fluidized beds using the kinetic theory of granular
simulation of a circulating fluidized bed. CFB-IV. Avidan AA, Ed. AIChE, New
flow: effect of restitution coefficient on bed dynamics, Chem. Eng. Sci. 56
York, pp. 430–435.
(2001) 571–578.
[44] B.G.M. van Wachem, J.C. Schouten, R. Krishna, C.M. van den Bleek, Eulerian
[11] C.H. Ibsen, E. Helland, B.H. Hjertager, T. Solberg, L. Tadrist, R. Occelli,
simulations of bubbling behavior in gas–solid fluidized beds, Comput. Chem.
Comparison of multifluid and discrete particle modeling in numerical
Eng. 22 (1998) 299–366.
predictions of gas particle flow in circulating fluidized beds, Powder Technol.
[45] C.L. Gomez, R.C. Silva, F.E. Milioli, Some modeling and numerical aspects of the
149 (2004) 29–41.
two-fluid simulation of the gas–solids flow in a CFB riser, Braz. J. Chem. Eng. 23
[12] S. Champan, T.J. Cowling, The Mathematical Theory of Non-Uniform Gases,
(4) (2006) 487–496.
Cambridge University Press, London, 1961.
S. Benzarti et al. / Advanced Powder Technology 25 (2014) 1737–1747 1747

[46] X.-Z. Chen, D.-P. Shi, X. Gao, Z.-H. Luo, A fundamental CFD study of the gas– [50] J.O. Hinze, Turbulence, McGraw-Hill, New York, 1975.
solid flow field in fluidized bed polymerization reactors, Powder Technol. 205 [51] Fluent User’s Guide, Fluent Inc., Lebanon, USA, 2005.
(2011) 276–288. [52] P. Li, X. Lan, C. Xu, G. Wang, C. Lu, J. Gao, Drag models for simulating gas–solid
[47] O. Gryczka, S. Heinrich, N.G. Deen, M. Van Sint Annaland, J.A.M. Kuipers, M. flow in the turbulent fluidization of FCC particles, Particuology 7 (2009) 269–
Jacob, L. Mörl, Characterization and CFD-modeling of the hydrodynamics of a 277.
prismatic spouted bed apparatus, Chem. Eng. Sci. 64 (2009) 3352–3375. [53] B. Lu, W. Wang, J. Li, Eulerian simulation of gas–solid flows with particles of
[48] J. Gao, X. Lan, Y. Fan, G. Wang, X. Lu, C. Xu, CFD modeling and validation of the Geldart groups A, B and D using EMMS-based meso-scale model, Chem. Eng.
turbulent fluidized bed, AIChE J. 55 (2009) 1680–1694. Sci. 66 (2011) 4624–4635.
[49] J. Wang, Flow structures inside a large-scale turbulent fluidized bed of FCC
particles: Eulerian simulation with an EMMS-based sub-grid scale model,
Particology 8 (2010) 176–185.

You might also like