You are on page 1of 9

Particuology 15 (2014) 151–159

Contents lists available at ScienceDirect

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

A new drag model for TFM simulation of gas–solid bubbling fluidized


beds with Geldart-B particles
Yingce Wang a,b , Zheng Zou a,b , Hongzhong Li a,∗ , Qingshan Zhu a
a
State Key Laboratory of Multi-phase Complex Systems, Institute of Process Engineering, Chinese Academy of Sciences, Beijing 100190, China
b
Graduate School of Chinese Academy of Science, Beijing 100049, China

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

Article history: In this work, a new drag model for TFM simulation in gas–solid bubbling fluidized beds was proposed,
Received 5 January 2013 and a set of equations was derived to determine the meso-scale structural parameters to calculate the
Received in revised form 24 April 2013 drag characteristics of Geldart-B particles under low gas velocities. In the new model, the meso-scale
Accepted 10 July 2013
structure was characterized while accounting for the bubble and meso-scale structure effects on the drag
coefficient. The Fluent software, incorporating the new drag model, was used to simulate the fluidization
Keywords:
behavior. Experiments were performed in a Plexiglas cylindrical fluidized bed consisting of quartz sand
Fluidization
as the solid phase and ambient air as the gas phase. Comparisons based on the solids hold-up inside the
Bubbling fluidized bed
CFD
fluidized bed at different superficial gas velocities, were made between the 2D Cartesian simulations, and
Geldart-B particles the experimental data, showing that the results of the new drag model reached much better agreement
Drag model with experimental data than those of the Gidaspow drag model did.
© 2013 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of
Sciences. Published by Elsevier B.V. All rights reserved.

1. Introduction1 In the DEM, the gas is considered to be a continuous medium,


while discrete particle motions are tracked individually using
Fluidized beds are widely used in chemical and physical pro- Lagrange coordinates and solved by Newtonian equations of
cesses in large-scale operations such as the manufacturing of motion. With this approach, the particle–particle collision mecha-
polyethylene and polypropylene, the synthesis of fuels and chemi- nism can be described by either soft-sphere or hard-sphere models.
cals, and in coating, drying, roasting, and heat exchangers (Hosseini, In the soft-sphere model, particles are permitted to suffer minute
Ahmadi, Rahimi, Zivdar, & Esfahany, 2010). Some of the distinct deformations, which can be used to calculate the elastic, plastic,
advantages of gas–solid bubbling fluidized bed reactors are con- and frictional forces between particles. This model has been exten-
trollable handling of solids, isothermal conditions due to good sively used to study various phenomena, such as particle packing,
solids mixing, large thermal inertia of solids, high heat flow, and transport properties, heaping/piling, hopper flow, mixing, and
advantageous reaction rates between gas and solids because of granulation (Cundall & Strack, 1979; Tsuji, Kawaguchi, & Tanaka,
the large gas–particle contact area (Rüdisüli, Schildhauer, Biollaz, 1993; Zhu, Zhou, Yang, & Yu, 2007). In the hard-sphere model,
& van Ommen, 2012). Continuous efforts have been made during the coefficients of restitution and friction are used to describe the
the past decades to gain a thorough understanding of the funda- particle–particle collisions, and this hard-particle method is more
mentals of gas–solid fluidized beds. The development of bubbling useful in the study of rapid granular flows (Lu et al., 2005; Wang,
fluidized bed simulations has become a popular research topic, Zhou, Wang, Xiong, & Ge, 2010). Because they contain few approx-
facilitated in part due to rapid increases in computational capa- imations and adjustable parameters, these discrete models are
bilities. Broadly speaking, direct numerical simulation approaches relatively rigorous. However, simulations of commercial-scale flu-
of two-phase flow in a bubbling fluidized bed can be classified idized beds using these models are prohibitively costly with current
into Euler–Lagrange discrete particle model (DEM) and Euler–Euler computing technology.
two-fluid model (TFM). In recent years, much research based on computational fluid
dynamics (CFD) has been reported, in which the TFM has been
shown to successfully predict the hydrodynamics of Geldart-B par-
ticles in bubbling fluidized beds (Ding & Gidaspow, 1990; Yuu,
∗ Corresponding author. Tel.: +86 10 62556951. Nishikawa, & Umekage, 2001). In the TFM (Jackson, 2000; Lu, Wang,
E-mail address: hzli@home.ipe.ac.cn (H. Li). & Li, 2009; van der Hoef, van Sint Annaland, Deen, & Kuipers, 2008;
1
Bold characters are for vectors or tensors. Xie, Battaglia, & Pannala, 2008), both phases can be considered

1674-2001/$ – see front matter © 2013 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.partic.2013.07.003
152 Y. Wang et al. / Particuology 15 (2014) 151–159

has been used more frequently to simulate commercial-scale flu-


Nomenclature idized beds. However, this model fails to account for the heteroge-
neous meso-scale structure, which is a key aspect of drag modeling
Latin letters
within CFD simulations; for a bubbling fluidized bed, the meso-
CDb0 drag coefficient of single bubble
scale structure is mainly related to the bubbles (Sundaresan, 2000).
CDe drag coefficient of multi-particle
In this work, we first partitioned a bubbling fluidized bed into
CDb drag coefficient of multi-bubble
three sub-systems and used seven parameters to describe the het-
D fluidized bed diameter (cm)
erogeneous meso-scale structure. A new model was proposed for
db bubble diameter (m)
calculating the drag between the gas and solid phases in bubbling
dp diameter of particle (m)
fluidized beds and a set of equations was built to calculate the seven
e elastic coefficient
parameters used in studies of Geldart B particles at low velocities.
fb volume ratio of gas in the bubble phase to that in
The fluidization process of Geldart-B particles in a bubbling flu-
total
idized bed was simulated using the Fluent software coupled with
g gravitational acceleration (m/s2 )
the new drag model. The simulation results showed better agree-
g0 radial distribution function
ment with experimental data than did results of the traditional drag
H height of the reactor (m)
model.
H0 initial packing height (m)
hd heterogeneous index
P pressure (pa) 2. Experimental
Re Reynolds number
t time step (s) 2.1. Materials
u real velocity (m/s)
U superficial velocity (m/s) The bed material in this experiment comprised quartz sand,
Ub superficial bubble velocity relative to bubble phase classified as group B particles within Geldart’s classification sys-
(m/s) tem (Geldart, 1973). The powder size was determined using a laser
Ue superficial velocity of the emulsion phase (m/s) diffraction particle size analyzer (LS13320, Beckman Coulter). The
Uge superficial gas velocity in the emulsion phase (m/s) particle size distribution of quartz sand particles is shown in Fig. 1;
Upe superficial particle velocity in the emulsion phase the mean and median sizes are 309.6 and 296.6 ␮m, respectively.
(m/s) Before each experiment, the powder was stored in the drying oven
Us average superficial slip velocity of the gas phase and (120 ◦ C) for at least 24 h, which could make them dried completely.
solid phase (m/s)
Usb superficial slip velocity between bubble and emul- 2.2. Apparatus
sion phase (m/s)
Use superficial slip velocity in the emulsion phase (m/s) Experiments were carried out in a three-dimensional fluidized
Ubr single bubble rise velocity (m/s) bed. As shown in Fig. 2, the fluidized bed comprised a 140 (inner
diameter) mm × 1000 (height) mm tube with a 3 mm-thick sin-
Greek symbols tered plate mounted at the bottom for gas distribution. The wind
ˇ drag coefficient (kg/m3 s) box was filled with glass beads to equalize the gas flow. There were
ε voidage a cyclone separator for gas–solid separation, and thereafter, a fur-
εs,max volume fraction of particles in close packing parti- ther separation was carried out in a bag filter. Separated particles
cles bed were eventually recycled to the bed bottom, so the total amount of
 density (kg/m3 ) bed material was kept nearly constant, even at high gas velocity.
 viscosity (Pa s) Air was used as the fluidizing gas, and the flow rate was
 stress tensor (Pa) accurately measured with a flowmeter, covering the range of
 granular temperature (m2 /s2 ) 0–417 L/min. A dehumidifier and an oil filter were also assembled
in-line with the gas feed.
Subscripts The hydrodynamics of the Geldart-B particles in fluidized beds
b bubble were investigated based on axial and radial solids hold-up data
e emulsion phase
g gas phase
mf minimum fluidization
p particle
s solid phase

fluid, and interpenetrating effects of each phase can be accounted


for using drag models. Applying a proper drag model is of vital
importance in TFM simulations (Beetstra, van der Hoef, & Kuipers,
2007; Behjat, Shahhosseini, & Hashemabadi, 2008; McKeen &
Pugsley, 2003; Zimmermann & Taghipour, 2005). Several drag
models have been developed to calculate the inter-phase momen-
tum exchange in gas–solid bubbling fluidized beds, including the
Wen–Yu (Wen & Yu, 1966), Syamlal–O’Brien (Syamlal & O’Brien,
1989), and Gidaspow (Gidaspow, 1994) drag models. Among them,
the Gidaspow drag model, also called the traditional drag model, Fig. 1. Size distribution of quartz sand particles.
Y. Wang et al. / Particuology 15 (2014) 151–159 153

Fig. 5. The axial solids hold-up distribution.

acquired by optical fiber probe (model PC-6M, produced by our


institute), under three superficial gas velocities of 0.1804, 0.2346,
and 0.2887 m/s. When experiments were conducted, the probe was
inserted into the fluidized bed at different heights. At each height,
the radial solids hold-up was recorded at 10 mm intervals between
Fig. 2. Experimental set-up and lateral measurement locations at cross-section the bed’s axial center and the wall, and two sets of measurement
(in mm). ports P1 and P2 were oriented perpendicularly, as shown in Fig. 2.
To ensure the validity and repeatability of sampled signals, the
sampling time was 60 s with a frequency of 1000 Hz. To prevent
particles adhering to the tip and blind zone, a glass cover was placed
over the probe tip.
The voltage signals of the optical fiber were correlated to the
solids hold-up computed from the measured pressure drops, and
the calibration curve is shown in Fig. 3.

2.3. Solids profiles

Fig. 4 shows the radial profiles for the solids hold-up at differ-
ent heights above the air distributor obtained at three different
gas velocities. It can be easily seen that the solids hold-up reduces
gradually with increasing bed height at the same superficial gas
velocity. And as the gas velocity is increased, the solids hold-up at
the same bed height decreases.
Fig. 3. Calibration curves of solids hold-up. Fig. 5 shows the axial solids hold-up profiles at different gas
velocities. It can be easily seen that the cross sectionally-averaged

Fig. 4. The influence of superficial gas velocity on radial profiles of time-averaged solids hold-up at three bed heights.
154 Y. Wang et al. / Particuology 15 (2014) 151–159

3.2. Hydrodynamic equations

In this section, we construct a system of seven equations, in


order to calculate the values of the seven unmeasurable parame-
ters introduced in the above section. Among the equations, five are
based on mass conservation or force balance, while the others are
an assumption and an empirical equation.

3.2.1. Superficial gas velocity in the emulsion phase


In this work, the experimental material is quartz sand, and the
operational gas velocity is relatively low. Accordingly, we assumed
the superficial gas velocity (Uge ) in the emulsion phase to be con-
stant and equal to the superficial incipient fluidization velocity of
the material (Umf ), i.e.:
Fig. 6. System partitioning for the bubbling fluidized bed.
Uge = Umf . (4)

solids hold-up at the same bed height decreases as the superficial 3.2.2. Force balance for particles in the emulsion per unit volume
gas velocity increases. The emulsion voidage is usually less than 0.8 and, consequently,
we can use the Ergun equation (Ergun, 1952) to calculate the drag
3. New drag model for Geldart B particles between particles and gas flows. The drag between a single particle
and gas flow is:
3.1. Partitioning of a bubbling fluidized bed
1 
FDe = CDe g dp2 Use
2
, (5)
2 4
To model bubbling fluidized beds, the overall system of a
bubbling fluidized bed is partitioned into three sub-systems: the where CDe is deduced from the Ergun equation, and is provided in
emulsion phase, the bubble phase, and the inter-phase, as shown Table 1.
in Fig. 6. Seven hydrodynamic parameters are needed to describe
the system, that is, the superficial gas velocity in the emulsion • The particle drag in the emulsion per unit volume, generated by
phase (Uge ), the superficial solids velocity in the emulsion phase gas in the emulsion phase, FDen
(Upe ), the volume fraction of bubbles (fb ), the relative gas velocity
with respect to the gas velocity in the bubble phase (Ugb ), the ris- (1 − fb )(1 − εe ) (1 − fb )(1 − εe ) ␲ 21 2
FDen = FDe = CDe d g Use
ing bubble velocity (Ub ), the bubble diameter (db ), and the voidage (␲/6)dp3 (␲/6)dp3 4 p2
of emulsion phase (εe ). As an approximation, the bubble phase is
assumed to consist of gas alone, omitting the presence of particles, 3 g 2
= CDe (1 − fb )(1 − εe )Use . (6)
i.e., εb = 1.0. 4 dp
The emulsion phase is assumed to be a homogeneous gas–solids
mixture. As a further approximation, it is viewed as a pseudo-fluid • The particle drag in the emulsion per unit volume, generated by
with mean density e , viscosity e , (Thomas, 1965) and superficial the bubbles, FDbn
velocity Ue , as described by:
fb p fb (1 − εe )p ␲ 21 2
FDbn = FDb (1 − εe ) = CDb d e Usb
e = p (1 − εe ) + g εe , (1) 3
(␲/6)db e (␲/6)db3 e 4 b2

3 p 2
= f (1 − εe )CDb U . (7)
e = g [1.0 + 2.5(1 − εe ) + 10.05(1 − εe )2 4 b db sb
+ 0.00273 exp(16.6(1 − εe ))], (2)
• The superficial gravity of the particles in the emulsion per unit
volume, Feg
g Uge + p Upe Feg = (1 − fb )(1 − εe )(p − g )g. (8)
Ue = . (3)
p (1 − εe ) + g εe

Table 1
Summary of the parameters and formulae of the new drag model.

Variables Emulsion phase Bubble phase Inter-phase

Superficial velocity Uge , Upe Ub , Ugb


εe
Superficial slip velocity Use = Uge − Upe 1−ε e
Usb = (Ub − Ue )(1 − fb )
g dp Use e db Usb
Reynolds number Ree = Rei =
 38Re−1.5
g e

i
0 < Rei ≤ 1.8
Drag coefficient for single particle or bubble CDb0 = 24
2.7 + Rei > 1.8
Rei
(1−εe )g
Effective drag coefficient with multi-particle/bubble correction CDe = 200 + 7
CDb = CDb0 (1 − fb )−0.5
ε3  d U
e g p se
3ε3
e
Y. Wang et al. / Particuology 15 (2014) 151–159 155

• Force balance for particles in the emulsion per unit volume, B particles was made to calculate Uge , and an empirical equation
FDen + FDbn = Feg , i.e. that applies only to the experimental conditions was selected to
calculate db . Therefore, this set of equations can only be used to
3 g 2 3 p 2
CDe (1 − fb )(1 − εe )Use + fb (1 − εe )CDb U determine the structural parameters for Geldart B particles under
4 dp 4 db sb
relatively low gas velocities.
= (1 − fb )(1 − εe )(p − g )g. (9)
3.3. Drag coefficients

Imitating the Gidaspow drag model (Gidaspow, 1994), we also


3.2.3. Force balance for bubbles adopt different formulas to calculate the drag between the gas
• The frictional drag between the bubble and emulsion phases, FDb
and solid phase under different situations. Meanwhile, the mean
␲ 21 2 voidage range is taken as the division standard.
FDb = CDb d e Usb . (10)
4 b2
• The buoyant force of the bubbles, Ffb 3.3.1. For the division εg ≤ εmf
Under such situation, the bubbling fluidized bed approximates
␲ 3
Ffb = d e g. (11) a fixed bed. Accordingly, the Ergun equation is chosen to calculate
6 b the drag between two phases:
• The gravitational force acting on the bubbles, Fwb
ε2s g g  
␲ ˇsg = 150 2
+ 1.75εs ug − up  , (18)
Fwb = db3 g g. (12) εg dp dp
6
• The force balance for bubbles where ug and up are the real velocities,

␲ 21 ␲ Ug Up Us
FDb = Ffb − Fwb , i.e., CDb 2
d e Usb = db3 (e − g )g. (13) ug − up = − = , (19)
4 b2 6 εg 1 − εg εg
and Us is the average superficial slip velocity of the gas and solid
3.2.4. Mass conservation of gas
phases.
Ug = Uge (1 − fb ) + Ub fb + Ugb fb = Uge (1 − fb ) + (Ub + Ugb )fb . (14)
   
Ug Up εg
3.2.5. Mass conservation of particles Us = − εg = Ug − Up . (20)
εg 1 − εg 1 − εg
Up = Upe (1 − fb ). (15)

As a further approximation, the amount of entrained particles 3.3.2. For the division εg ≥ 0.8
is assumed to be negligible, so we set Up = 0, and therefore Upe = 0. When the mean voidage is relatively high, the Wen–Yu formula
is adopted to calculate the drag between the gas and solid phases:
3.2.6. The empirical equation of the velocity of the bubbles
(1 − εg )εg  
According to Kunii and Levenspiel (1991), the bubble velocity ˇsg = 0.75 g ug − up  CD0 ε−2.65
g , (21)
Ub can be calculated using the following equations: dp
where CD0 is the standard drag coefficient,
For Geldart A particles, when the diameter of the bed, D ≤ 100 cm : ⎧
Ub = 0.34((U0 − Umf ) + 14.1(db + 0.5))(D/100)
0.33
+ Ubr , ⎨ = 0.44 Rep > 1000
CD0 = , (22)
For Geldart B particles, when the diameter of the bed, D ≤ 100 cm : ⎩ = 24 (1 + 0.15Rep0.687 ) Rep ≤ 1000
Ub = 0.0032((U0 − Umf ) + 11.3db0.5 )(D/100)
1.35
+ Ubr .
Rep

For db /D < 0.125 :


and Rep is the Reynolds number of the particles,
 g dp (ug − up )εg
Ubr = 0.711 gdb , Rep = . (23)
g
For 0.125 < db /D < 0.6 :
  
db
Ubr = (0.711 gdb )1.2 exp −1.49 ,
D

For db /D > 0.6, when the gas velocity is relatively high, slugging is more probable.
(16)

In other work, different empirical equations can be chosen


according to the material and the size of the fluidized bed. D is
the diameter of the cylindrical setup, so it is a constant; in our
experiment, D is 14 cm.

3.2.7. Mean voidage


εg = (1 − fb )εe + fb . (17)

In the simulation process, εg can be determined by the Fluent


software in advance, therefore, it can be taken as a known param-
eter.
Five of these equations are based on mass conservation or force
balance so they are universal. However, an assumption for Geldart Fig. 7. Schematic diagram of the simulated 2D bubbling fluidized bed.
156 Y. Wang et al. / Particuology 15 (2014) 151–159

Table 2 3.3.3. For the division εmf < εg < 0.8


Summary of the drag coefficients in different voidage divisions.
Under this situation, the Gidaspow drag model fails to account
Voidage ˇsg for the heterogeneous meso-scale structural effects. So, we pro-
division pose a new drag model based on the meso-scale structure of the
ε2 g  |Us | gas–solid bubbling fluidized bed. As shown in Fig. 7, the multi-
εg ≤ εmf 150 εsg 2 + 1.75εs d g εg
p
dp
phase heterogeneous structure of the gas–solid bubbling fluidized
0.75ε2g (1 −
εmf < εg < 0.8
 Use 2 Usb 2 
bed can be partitioned into three homogeneous structures.
 
εg ) Us g CDe (1 − fb ) d (1−εe) (1−εe )
+ CDb fb (1−ε
p
(1−ε )
p g Us g) g db Us

εg ≥ 0.8
(1−ε )
0.75 d g g
U  C −2.65
• The particle drag in the emulsion per unit volume, generated by
s D0 εg
p
the gas in emulsion phase FDen , as shown in Eq. (6).
• The particle drag in the emulsion per unit volume, generated by
Table 3
Summary of the parameters for solving drag model. the bubbles FDbn , as shown in Eq. (7).
• The overall drag between the gas and solid per unit volume FD
Parameter Value
(which is the same as Feg , shown by Eq. (9)).FD = FDen + FDbn ,
dp (␮m) 309.6
p (kg/m3 ) 2640 3 g 3 p 2
2
g (kg/m3 ) 1.225 FD = CDe (1 − fb )(1 − εe )Use + fb (1 − εe )CDb U . (24)
4 dp 4 db sb
g (m/s2 ) 9.81
Umf (m/s) 0.0928
g (Pa. s) 1.7894 × 10−5 • Mean drag coefficient CD
Ug (m/s) 0.1804, 0.2346, 0.2887 According to the definition,
D (cm) 14
(1 − εg ) 1  3 g 2
Table 4 FD = CD g dp2 Us2 = (1 − εg )CD U , (25)
(␲/6)dp3 2 4 4 dp s
Heterogeneous index at different gas velocities for Geldart-B particles.

Gas velocity Heterogeneous index


through comparison of Eqs. (24) and (25), an expression for CD is
(m/s)
⎧ obtained:
⎨ 1.0; (εg ≤2059.0ε
εmf )
− 196.2  U 2  U 2
0.1804 hd =
g
; (εmf < εg < 0.8) (1 − εe ) se (1 − εe ) p dp sb
⎩ εg + 720.8εg − 1706.0εg + 1006.0
3 2 CD = CDe (1 − fb ) + CDb fb .
1.0; (εg ≥0.8)
(1 − εg ) Us (1 − εg ) g db Us
⎧ (26)
⎨ 1.0; (εg ≤1756.0ε
εmf )
− 277.4
g
0.2346 hd = ; (εmf < εg < 0.8)
⎩ εg + 506.7εg − 1355.8εg + 869.1
3 2

1.0; (εg ≥0.8) • The drag coefficient between the gas and solid phases
⎧ Assuming the drag of every particle in the grid is the same, then
⎨ 1.0; (εg 1096.0ε
≤ εmf )
− 134.6
  
g
0.2887 hd = ; (εmf < εg < 0.8)
1 1
⎩ εg + 539.2εg − 1307.0εg + 791.7 g (ug − up ) ug − up  ε2g dp2 CD .
3 2
FD = (1 − εg ) (27)
1.0; (εg ≥0.8) (␲/6)dp3 2 4

Table 5
Summary of the governing equations for the TFM simulation.
∂(εg g )
Continuity equations of gas and solid ∂t
+ ∇ · (εg g ug ) = 0
+ ∇ · (εs s us ) = 0
∂(εs s )
∂t
∂(εg g ug )
Conservation of momentum of gas and solid ∂t
+ ∇ · (εg g ug ug ) = −εg ∇ pg + ∇ · (εg g ) + εg g g − ˇsg (ug − us )
+ ∇ · (εs s us us ) = −εs ∇ pg − ∇ ps + ∇ · (εs s ) + εs s g + ˇsg (ug −
∂(εs s us )
∂t
us )
Granular temperature equation 3 ∂(εs s s )
2
[ ∂t
+ ∇ · (εs s us s )] = (−ps I + s ) : (∇ us ) + ∇ · (
s ∇ s ) − s − 3ˇs
T
Gas phase stress g = g [∇ ug + (∇ ug ) ] + ( g − 2
 )∇
3 g
· ug
T
Solid phase stress s = s [∇ us + (∇ us ) ] + ( s − 2
 )∇
3 s
· us
 13 −1
εs
Radial distribution function g0 = 1− εs,max

Solid pressure ps = εs s s + 2 (1 + e) ε2s g0 s s


4
s 1/2
Bulk solid viscosity s = ε  d g (1
3 s s p 0
+ e) 

10s dp s 
 2 s 1/2
Shear viscosity of solid s = 96εs (1+e)εg g0
1+ 4
g ε (1
5 0 s
+ e) + 4
ε  d g (1
5 s s p 0
+ e) ␲
+ p√
s sin
2 I2D

150s dp (s ␲) 1/2  2 1/2


Granular conductivity of fluctuation energy ks = 1 + 65 εs g0 (1 + e) + 2ε2s s dp g0 (1 + e) ␲s
384(1+e)g0
 1/2 
4 s
Collisional energy dissipation s = 3ε2s s g0 s (1 − e2 ) dp 

⎪ ε2 g εs g  

⎪ 150.0 s 2 + 1.75 Us εg ≤ εmf

⎨
εg dp εg dp

2
ε g εs g  
Gas–solid drag coefficient ˇsg = 150.0 s 2 + 1.75 Us hd εmf < εg < 0.8


εg dp εg dp


⎩ 0.75 (1 − εg ) g CD0 Us  ε−2.65
g εg ≥0.8
dp
Y. Wang et al. / Particuology 15 (2014) 151–159 157

Fig. 8. Comparison of simulated time-averaged radial solids profiles obtained by new drag model and Gidaspow model with experimental data at different heights and
different superficial gas velocities of (a) 0.1804, (b) 0.2346 and (c) 0.2887 m/s.

According to the TFM drag coefficient definition, given by so,


  2
Gidaspow, it can be concluded that  
ˇsg = 0.75ε2g (1 − εg ) Us  CDe (1 − fb ) g (1 − εe ) Use
dp (1 − εg ) Us

(1 − εe ) p
U 2 
sb
ˇsg +CDb fb . (29)
FD = (ug − up ). (28) (1 − εg ) db Us
εg
158 Y. Wang et al. / Particuology 15 (2014) 151–159

Fig. 9. Comparison of time-averaged axial solids hold-up profiles.

The above derived drag coefficient formula in different voidage Table 6


Summary of parameters used in simulation.
divisions of the new drag model are summarized in Table 2.
Particle diameter (␮m) 309.6
3.4. Model implementation Particle density (kg/m3 ) 2640
Grid size x (mm) 5
Grid size y (mm) 5
To identify a discrepancy in the two drag coefficients yielded by Gas density (kg/m3 ) 1.225
the new drag model and Gidaspow drag model, a heterogeneous Initial bed height (m) 0.232
index, hd = (ˇnew model /ˇGidaspow ), is defined. It is obvious that when Superficial gas velocity (m/s) 0.1804, 0.2346, 0.2887
the mean voidage is less than the voidage at minimum fluidization Coefficient of restitution 0.9
Time step (s) 1.0 × 10−4
(εg ≤ εmf ), or is larger than 0.8 (εg ≥ 0.8), hd is equal to 1.0. Unsteady formulation First-order SIMPLE
The above derived Eqs. (4), (9), (13)–(17) were solved simul- Pressure-velocity coupling Phase coupled SIMPLE
taneously to determine the seven structural parameters Uge , Upe , Maximum number of 60
Ugb , Ub , db , fb , and εe , using the Maple software (Waterloo Maple, iterations per time step
Convergence criteria 1.0 × 10−4
Canada) based on the pre-specified operational and geometric con-
Under relaxation factors 0.5 for pressure, 0.2 for momentum and 0.4
ditions listed in Table 3. for volume and granular temperature
When the value of εg is between εmf and 0.8, the drag coefficient Maximum solid packing 0.5412
ˇsg is determined by the new drag model rather than Gidaspow volume fraction
model, yielding an hd not equal to unity. By fitting the calculated Acceleration of gravity (m/s2 ) 9.81
Viscosity of the air (Pa s) 1.7894
hd vs. εg , expressions of hd were obtained using the Matlab software
(MathWorks, USA), as a function of εg for different superficial gas
velocities, as listed Table 4. the results of the traditional model simulation display relatively
large deviations from experiment.
3.5. Simulation with the new drag model Fig. 9 shows comparisons between the simulated and
experimentally-measured axial solids profiles under different
The numerical simulation is based on the TFM within Fluent superficial gas velocities. It can be seen that the simulated result
6.3.26, with governing equations which are summarized in Table 5. of the new drag model at a superficial gas velocity of 0.1804 m/s
The solids stress in the momentum equations are closed with is slightly lower than the experimental result, while at velocities
the algebraic form of the kinetic theory of granular flow (KTGF) of 0.2346 and 0.2887 m/s, the simulated result is slightly higher.
(Goldschmidt, Beetstra, & Kuipers, 2002). The meso-structure- However, the results of the traditional drag model qualitatively dif-
dependent drag coefficient is thus incorporated into Fluent as a fer from the experimental data, while the results of the new drag
user-defined function (UDF). model are able to describe the experimentally measured data well
The simulation is conducted in the 2-D condition and detailed at all three gas velocities.
simulation parameters are summarized in Table 6. For the case of From the above results, it can be seen that the traditional drag
Geldart-B particles, as sketched in Fig. 7, the bed is 1.0 m in height model greatly underestimates the gas–solid drag at low superfi-
and 0.14 m in inner diameter. Gas enters the bed uniformly from cial gas velocities in bubbling fluidized beds and its simulation
the bottom inlet. The simulations ran for 20 s in physical time and results may, therefore, deviate from the experimental data con-
time-averaged observables were obtained over the last 10 s. siderably. By contrast, results of the new drag model show much
better agreement with the experimentally measured results.
4. Results and discussion
5. Conclusions
The simulated radial solids profiles obtained using both the
new drag model and the traditional model are compared to A new drag model for gas–solid bubbling fluidized bed was
the experimental data in Fig. 8 at three different bed heights proposed that incorporates the contribution of the heterogeneous
with superficial air velocities Ug of (a) 0.1804, (b) 0.2346 and (c) meso-scale structures which has not been accounted for in the tra-
0.2887 m/s. It can be easily seen that the results of the new drag ditional drag model. A set of equations was derived to determine
model show good agreement with the experimental data, while the structural parameters used in this new drag model. This set
Y. Wang et al. / Particuology 15 (2014) 151–159 159

of parameters was optimized exclusively for Geldart B particles Hosseini, S. H., Ahmadi, G., Rahimi, R., Zivdar, M., & Esfahany, M. N. (2010). CFD stud-
at relatively low gas velocities. The new model was incorporated ies of solids hold-up distribution and circulation patterns in gas–solid fluidized
beds. Powder Technology, 200, 202–215.
into the TFM to simulate the hydrodynamics of Geldart-B parti- Jackson, R. (2000). The dynamics of fluidized particles. Cambridge: Cambridge Uni-
cles in bubbling fluidized beds. Comparisons of results obtained by versity Press.
the new model and the Gidaspow model with experimental data Kunii, D., & Levenspiel, O. (1991). Fluidisation engineering (2nd ed.). Stoneham:
Butterworth-Heinemann.
were made, showing that the new model performed better than Lu, H. L., Wang, S. Y., Zhao, Y. H., Yang, L., Gidaspow, D., & Ding, J. M. (2005). Prediction
the traditional model. Because of the limited experimental data, of particle motion in a two-dimensional bubbling fluidized bed using discrete
we could not examine the utility of the new model at higher gas hard-sphere model. Chemical Engineering Science, 60, 3217–3231.
Lu, B., Wang, W., & Li, J. H. (2009). Searching for a mesh-independent sub-grid model
velocities. The question of whether our new model is suitable at
for CFD simulation of gas–solid riser flows. Chemical Engineering Science, 64,
higher superficial gas velocities will require further research. 3437–3447.
McKeen, T., & Pugsley, T. (2003). Simulation and experimental validation of a freely
bubbling bed of FCC catalyst. Powder Technology, 129, 139–152.
Acknowledgements
Rüdisüli, M., Schildhauer, T. J., Biollaz, S. M. A., & van Ommen, J. R. (2012).
Scale-up of bubbling fluidized bed reactors—A review. Powder Technology, 217,
The authors are grateful for supports from the State Key Devel- 21–38.
Sundaresan, S. (2000). Modeling the hydrodynamics of multiphase flow reactors:
opment Program for Basic Research of China (973 Program) under
Current status and challenges. AIChE Journal, 46, 1102–1105.
Grant Nos. 2009CB219904, 2013CB632603 and the National Sci- Syamlal, M., & O’Brien, T. J. (1989). Computer simulation of bubbles in a fluidized
ence and Technology Support Program of Ministry of Science bed. AIChE Symposium Series, 85, 22–31.
and Technology of the People’s Republic of China (Grant No. Thomas, D. G. (1965). Transport characteristics of suspension: VIII. A note on the
viscosity of Newtonian suspensions of uniform spherical particles. Journal of
2012BAB14B03). Colloid Science, 20, 267–277.
Tsuji, Y., Kawaguchi, T., & Tanaka, T. (1993). Discrete particle simulation of two-
References dimensional fluidized bed. Powder Technology, 77, 79–87.
van der Hoef, M. A., van Sint Annaland, M., Deen, N. G., & Kuipers, J. A. M. (2008).
Numerical simulation of dense gas–solid fluidized beds: A multiscale modeling
Beetstra, R., van der Hoef, M. A., & Kuipers, J. A. M. (2007). Numerical study of segre- strategy. Annual Review of Fluid Mechanics, 40, 47–70.
gation using a new drag force correlation for polydisperse systems derived from Wang, L. M., Zhou, G. F., Wang, X. W., Xiong, Q. G., & Ge, W. (2010). Direct numeri-
lattice-Boltzmann simulations. Chemical Engineering Science, 62, 246–255. cal simulation of particle–fluid systems by combining time-driven hard-sphere
Behjat, Y., Shahhosseini, S., & Hashemabadi, S. H. (2008). CFD modeling of model and lattice Boltzmann method. Particuology, 8, 379–382.
hydrodynamic and heat transfer in fluidized bed reactors. International Com- Wen, C. Y., & Yu, Y. H. (1966). Mechanics of fluidization. Chemical Engineering Progress
munications in Heat and Mass Transfer, 35, 357–368. Symposium Series, 62, 100–111.
Cundall, P. A., & Strack, O. D. L. (1979). A discrete numerical model for granular Xie, N., Battaglia, F., & Pannala, S. (2008). Effects of using two-versus three-
assemblies. Geotechnique, 29, 47–65. dimensional computational modeling of fluidized beds: Part I, hydrodynamics.
Ding, J., & Gidaspow, D. (1990). A bubbling fluidization model using kinetic theory Powder Technology, 182, 1–13.
of granular flow. AIChE Journal, 36, 523–538. Yuu, S., Nishikawa, H., & Umekage, T. (2001). Numerical simulation of air and particle
Ergun, S. (1952). Fluid flow through packed columns. Chemical Engineering Progress, motions in group-B particle turbulent fluidized bed. Powder Technology, 118,
48, 89–94. 32–44.
Geldart, D. (1973). Types of gas fluidization. Powder Technology, 7, 285–292. Zhu, H. P., Zhou, Z. Y., Yang, R. Y., & Yu, A. B. (2007). Discrete particle simulation
Gidaspow, D. (1994). Multiphase flow and fluidization: Continuum and kinetic theory of particulate systems: Theoretical developments. Chemical Engineering Science,
description. Boston: Academic Press. 62, 3378–3396.
Goldschmidt, M. J. V., Beetstra, R., & Kuipers, J. A. M. (2002). Hydrodynamic modeling Zimmermann, S., & Taghipour, F. (2005). CFD modeling of the hydrodynamics and
of dense gas-fluidized beds: Comparison of the kinetic theory of granular flow reaction kinetics of FCC fluidized-bed reactors. Industrial & Engineering Chem-
with 3D hard-sphere discrete particle simulations. Chemical Engineering Science, istry Research, 44, 9818–9827.
57, 2059–2075.

You might also like