Vous êtes sur la page 1sur 11

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/272265888

Modeling of Bubble-Structure-Dependent Drag


for Bubbling Fluidized Beds

Article in Industrial & Engineering Chemistry Research October 2014


DOI: 10.1021/ie502412g

CITATIONS READS

4 9

6 authors, including:

Wang Shuai Guodong Liu


Harbin Institute of Technology Harbin Institute of Technology
48 PUBLICATIONS 328 CITATIONS 53 PUBLICATIONS 271 CITATIONS

SEE PROFILE SEE PROFILE

Liyan Sun
Harbin Institute of Technology
33 PUBLICATIONS 108 CITATIONS

SEE PROFILE

Available from: Wang Shuai


Retrieved on: 13 October 2016
Article

pubs.acs.org/IECR

Modeling of Bubble-Structure-Dependent Drag for Bubbling


Fluidized Beds
Shuai Wang, Huilin Lu,* Qinghong Zhang, Guodong Liu, Feixiang Zhao, and Liyan Sun
School of Energy Science and Engineering, Harbin Institute of Technology, Harbin 150001, China
*
S Supporting Information

ABSTRACT: Considering the eect of bubble-emulsion structures in bubbling uidized beds, a bubble-structure-dependent
drag coecient model is developed. Accelerations in the bubble and emulsion phases are incorporated into the solution of the
drag coecient. Meanwhile, the inuence of solid pressure and bubble-induced added mass force is also taken into account. In
combination with the two-uid model, ow behaviors in two-dimensional and three-dimensional bubbling uidized beds are
simulated. The predictions by the present model with consideration of bubble eects are in more reasonable agreement with the
experimental results compared to the Gidaspow drag model. It is shown that the present model obtains a zonal distribution of the
drag coecient with solid concentration, which reveals that the drag coecient not only depends on the local solid concentration
but also is greatly inuenced by the local velocities.

1. INTRODUCTION In recent years, ltered TFMs have been developed to


Bubbling uidized beds have been widely applied to various account for the eects of unresolved structures for coarse-grid
industrial processes, such as coal gasication, uid-catalytic- simulations of gas-particle ows.1012 Constitutive models for
cracking regeneration, and polypropylene production.1,2 ltered TFMs are deduced through ltered results obtained
Fundamental knowledge of the hydrodynamic characteristics from highly resolved simulations. Parmentier and Simonin13
is essential for the design and scale-up of such reactors. modied the eective relative velocity appearing in the ltered
Computational uid dynamics (CFD) is a promising research drag through a functional subgrid drift velocity model. The bed
means to investigate the behavior of full-size systems and expansion was well predicted using coarse grids with the model.
complex phenomena in multiphase ow, which is dicult for Sarkar et al.14 constructed a subgrid drag model for cylinder
current measurement techniques. suspension. It was found that the presence of cylinders resulted
In the last 2 decades, substantial progress has been made to in a reduction in the ltered interphase drag force.
develop mathematical models to describe the behavior of Schneiderbauer and Pirker15 deduced two kinds of closures
bubbling-uidized-bed systems with acceptable accuracy.3,4 A for the unresolved terms of the solid stress and drag force in
two-uid model (TFM) has proven to be successful in the ltered TFMs: one derived from the ltered data and the other
prediction of the hydrodynamics of a bubbling uidized bed.5,6 on the basis of the assumption that the cluster formation led to
Reuge et al.7 pointed out that the solid stress model greatly heterogeneity inside the uidized beds. The simulated results
inuenced the simulation of bubbling uidized beds. The revealed that the prediction with subgrid modications using
prediction using the Princeton model was in better agreement coarse grids was in fairly good agreement with the highly
with measured data for bed characteristics compared to the resolved simulation.
Schaeer model. On the basis of multiscale analysis, Wang and The energy minimization multiscale (EMMS) approach was
Ge8 developed a mathematical model to characterize the kinetic widely used to deal with the impact of heterogeneous structures
feature of coarse particles. The results indicated that the on the drag force.16,17 Chalermsinsuwan et al.18 calculated the
unresolved subgrid structures aected the interphase drag drag coecient by employing a simplied EMMS model and
coecient and particle-phase stresses, which should be performed a simulation in a thin bubbling uidized bed with
considered explicitly. Geldart A particles. Wang and Liu19 revised the EMMS model
For bubbling uidization, bubbles have signicant eects on using an implicit cluster diameter expression and applied it to
physical mechanisms, which play roles similar to those of the simulation of bubbling uidized beds. These models derive
clusters in fast uidized beds. The gas tends to form dispersed from the calculation of cluster-based drag and are very
bubbles, which results in the reduction of gassolid interaction questionable when applied to dense bubbling uidized beds.
compared to the standard drag on the basis of a uniform Shi et al.20 proposed a bubble-based EMMS model and treated
particle distribution assumption. It was reported that very ne the bubble phase as the mesoscale structure. In this model, the
grids can capture local nonuniform patterns and the grid size accelerations for the emulsion and bubble phases were
should be on the order of 24 particle diameters in bubbling
uidized beds.9 Owing to the limitations of the computational Received: June 16, 2014
resource, coarse-grid simulation with appropriate subgrid Revised: August 29, 2014
models is required to reect the temporalspatial multiscale Accepted: September 5, 2014
structure for a large-scale industrial uidized bed. Published: September 5, 2014

2014 American Chemical Society 15776 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

Figure 1. Grid resolution of the gassolid bubbling uidized bed.

introduced. Compared with empirical relations and exper- T1-7, where g represents the gas viscosity and is assumed to be
imental results, the model agreed well with the available data. constant in the current work.
Wang et al.21 and Lv et al.22 established drag models on the At a high solid concentration, particles are closely packed and
basis of local structures, where local structural parameters are inuenced by sustained contact with multiple neighbors. The
obtained by solving the mass and momentum conservation frictional stress of sliding contacts is dominant. Here, the
equations and empirical correlations of the bubble velocity and friction stress model proposed by Srivastava and Sundaresan25
bubble diameter. The above models used the global hydro- is adopted to consider the frictional contribution, which
dynamic conditions to deduce the correlation between the drag modies the Savage26 model to take strain rate uctuations
coecient and voidage, which neglected the dependence of the into account. The kinetic theory is still used to describe the
drag coecient on the local information in the control volume. kinetic and collisional contribution. To provide a smooth
In this paper, a bubble-structure-dependent (BSD) drag transition between the shear ow and frictional regime, a
model is developed on the basis of the previous cluster- transitioning function is introduced.27 The kinetic and
structure-dependent (CSD) drag model.23 The heterogeneity frictional components of solid pressure are expressed by eqs
within a computational cell caused by bubbles is described by T1-9 and T1-10. The shear and frictional viscosities of the
resolving the overall system into subsystems. The pressure particle phase are given by eqs T1-14 and T1-15.
2.2. BSD Drag Coecient Model. In the previous study,
gradient due to particleparticle collision interaction and the
the CSD drag model was proposed to describe the eect of
bubble-induced added mass force are incorporated into
mesoscale structures in fast uidized beds.23 The local
calculation of the drag coecient. The accelerations in the
heterogeneous ow is resolved into three phases: the dense
emulsion and bubble phases are also taken into account. By phase in the form of clusters, the gas-rich dilute phase, and the
means of two-dimensional (2D) and three-dimensional (3D) interface. Bubbles, as typical mesoscale structures, lead to a
simulations of bubbling uidized beds, the bubble-based drag nonuniform drag solution in bubbling uidized beds.
model can give better predictions of the experimental data. Analogously, the local ow within the grid is separated into
three subsystems: the emulsion phase, the bubble phase, and
2. MATHEMATICAL MODEL the interphase, as displayed in Figure 1. Here, it is assumed that
An EulerianEulerian multiphase model is adopted in the the bubble phase only includes the gas phase and the
present study. The main assumptions of the model are as distribution of particles in the emulsion phase is uniform.
follows: (1) mass transfer between the phases is neglected Hence, the BSD drag force consists of two contributions: one
because of no reaction; (2) the diameter and density of the from the drag component in the emulsion and the other from
the bubble-induced drag component, which is expressed as
particles are uniform. The governing equations consist of the
follows:
conservation equations of mass and momentum. The kinetic
theory of granular ow is used to close the model, as reviewed gFgs g
by Gidaspow.24 Detailed information can be found in Table S1 BSD = = (neFde + nbFdb)
Uslip Uslip (1)
in the Supporting Information (SI).
2.l. Hydrodynamic Model. The continuity equations for where nb and ne are the number densities of bubbles and
the gas and solid phases are given by eqs T1-1 and T1-2 (see particles in the emulsion phase. Fde and Fdb denote drag
the SI). The momentum conservation equations of the gas and components in the emulsion phase and generated by bubbles,
solid phases are expressed by eqs T1-3 and T1-4, where and respectively. Detailed correlations are listed in Table S2 in the
denote the stress tensor and the gassolid drag coecient, SI.2831
respectively. To describe the uctuating energy of the particles, To solve the mutiscale drag coecient, six independent
granular temperature is introduced and dened as = C2/3, parameters are required, that is, the volume fraction of bubbles
where C represents the uctuating velocity of the particles. The (b), the voidage in the emulsion phase (e), the supercial gas
conservation equation of granular temperature is described by velocity in the emulsion phase (Uge), the supercial solid
eq T1-5. velocity in the emulsion phase (Use), the velocity of the bubble
Constitutive relations are used to close governing equations. (Ub), and the diameter of the bubble (db), which can be
The gas and solid stress tensors are expressed as eqs T1-6 and obtained by solving a set of nonlinear equations.
15777 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

2.2.1. Solid Momentum Equation of the Emulsion Phase. g = (1 b)e + b (6)


For simplication, it is assumed that the solid stress of the
emulsion phase is neglected. The solid momentum equation in 2.2.6. Stability Criterion. In this model, there are six
the emulsion phase along the ow direction is expressed as independent variables with ve hydrodynamic equations (eqs
follows: 26). Hence, the stability criterion is required to close the
equations. In the bubbling-uidized-bed system, particles tend
neFde + nbFdb = (1 b)(1 e)pg + (1 b)(1 e) to array themselves with minimal energy loss by drag and gas
tends to form bubbles and select an upward path with least
(s g )(g + as,e) + ps (2) resistance. The extremum condition of energy dissipation
consumed by the drag force is given as
(1 b)Us,e (1 b)Us,e 1
as,e = /z Ndf = (neFdeUg,e + nbFdbUb b) minimum
1 e 1 e (2a) (1 g)s
(7)
where as,e represents the accelerations of particles in the
emulsion phase. The solid pressure gradient takes the empirical For each grid cell, the local information (ug, us, g, and p/
particulate stress model and can be calculated by32 z) is obtained from the TFM and used to solve six
independent variables by employing ve hydrodynamic
ps = 108.76[(1 b)e] + 5.43[(1 b)(1 e)] (2b) equations (eqs 26) and a stability condition (eq 7). In this
way, the solution of the BSD drag coecient can be achieved.
2.2.2. Momentum Equations of Gas in the Emulsion Here, the BSD drag coecient is solvable in the range (mf,
Phase and Bubbles. For the ow of gas in the emulsion phase d). mf denotes the minimum uidizing gas volume fraction,
and bubbles, we assume that the stress of gas is neglected. From and d is the voidage where the ratio of the BSD and Gidaspow
the momentum equations of gas in the emulsion phase and drag coecients is 1. Once this interval is exceeded, the
bubbles at the steady state, we derive a pressure drop balance Gidaspow drag coecient model is still adopted. The bubble-
equation between gas in the emulsion phase and bubbles: based drag model is used to describe the eect of unresolved
b bubbles in the grid. When the bubble size exceeds the grid size,
neFde = nbFdb bg (ag,e ag,b) + pb the bubble can be directly solved by the grid.
(1 b)e (3) 2.3. Boundary Conditions and Numerical Solution
(1 b)Ug,e (1 b)Ug,e Procedure. Initially, the particles are loaded in the static bed.
ag,e = /z , The uniform gas velocity is set at the bottom inlet. The
e e atmospheric pressure is prescribed at the top outlet. For the
wall, there is no slip boundary condition for the gas phase and a
ag,b = ( bUb bUb)/z (3a) partial slip boundary condition for the solid phase is selected
with a specularity coecient of 0.5.34
where ag,e and ag,b are the accelerations of gas in the emulsion The simulations are conducted with a modied K-FIX code,
phase and bubbles. The added mass force can be written as which is a general purpose CFD code for modeling the
pb = C b(1 e) b[eg + (1 e)s ] hydrodynamics in uidsolid systems and has been successfully
used for the simulations of circulating uidized beds and
{ag,b [eag,e + (1 e)as,e]} spouted beds.35,36 The time step varies from 104 to 106 s, and
(3b)
the maximum convergence residual of 103 is set. Simulations
where the coecient Cb is calculated by 33
last for 40 s, and ow variables are time-averaged from 10 to 40
s when the quasi-steady state has been reached. It costs about 1
1 + 2 b week using the BSD drag model.
C b = 0.5
1 b (3c)
3. RESULTS AND DISCUSSION
2.2.3. Mass Conservation Equation of Gas. The gas ow
within the grid cell can be characterized by two-phase structures 3.1. Comparison with the Experiment of Laverman et
including gas in the emulsion phase and bubbles. Considering al. A 2D simulation is carried out on the basis of the
the mass balance of gas, the velocity of the gas phase in the grid experimental setup of Laverman et al.,37 as displayed in Figure
cell requires 2. In their experiments, a pseudo-2D uidization system with a
height of 0.7 m, a width of 0.30 m, and a depth of 0.015 m was
[(1 b)Ug,e + bUb] tested. Glass beads with a narrow particle size distribution of
ug = 400600 m were selected as uidized particles. Digital image
g (4) analysis and particle image velocimetry were used to obtain the
2.2.4. Mass Conservation Equation of Particles. The ow bubble information and instantaneous emulsion-phase velocity
of the particles in the grid is characterized by dispersed particles proles. Detailed operating conditions and physical parameters
in the emulsion phase. Hence, the solid velocity in the grid cell can be found in Table 1. From the study of Li et al.,38 the
requires inuence of the front and back walls was signicant. Here, the
frictional model proposed by Li and Zhang39 is applied to
(1 b) consider the wall eect, where the friction force and solid
us = Us,e
(1 g) (5)
uctuation energy supplied by the front and back walls are
accounted for.
2.2.5. Overall Gas Volume Fraction. According to the A mesh-independent investigation is previously conducted to
denition of voidage in a grid cell, we have evaluate the impact of the grid size on the predictions. Uniform
15778 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

grids with dierent sizes are adopted. Figure 3 displays the


time-averaged axial prole of a solid volume fraction under
dierent grid resolutions with two drag models. We can observe
that, although there is a similar axial distribution of the solid
concentration for the two drag models, the discrepancy
between them is apparent. The prediction by the Gidaspow
drag model with coarser grids diers from the others. When
ner grids are employed, a similar curve can be captured for the
two drag models, whereas the BSD drag model still obtains a
better prediction using a coarser grid resolution. This implies
that the BSD drag model can allow a coarser grid scheme
without losing accuracy.
Figure 4 shows a time sequence of the solid concentration
using two dierent drag models. Both models can predict the
movements of bubbles. The bubbles form near the inlet, pass
through the bed with coalescence, and erupt at the bed surface.
However, the dierence between them can be clearly
distinguished. In general, the BSD drag model captures a
larger bubble diameter than the Gidapsow drag model, which is
attributed to the fact that the interaction between the emulsion
phase and bubbles becomes weak with consideration of the
Figure 2. Schematic diagram of a simulated 2D bubbling uidized bed. bubble eects and the coalescence of bubbles occurs easily. The
gas tends to form a larger bubble to pass through the bed.
Table 1. Parameters for Numerical Simulation An instantaneous local solid concentration uctuation is
experiment (Laverman shown in Figure 5. A low solid concentration means a high gas
parameter et al., 2008) simulation unit volume fraction. When the solid concentration is lower than
particle density 2500 2500 kg/m3 0.2, the gas is thought to pass through the bed in the form of
particle diameter 485 485 m bubbles. The peak of the concentration reects a bubble
reactor height 0.7 0.7 m frequency. It can be found that there is an intense oscillation at
reactor diameter 0.3 0.3 m the center region. This indicates that the passing of bubbles
inlet gas velocity 0.45 0.45 m/s mainly appears at the center of the bed with the bursting of
gas viscosity 1.8 105 1.8 105 Pas bubbles.
gas density 1.2 1.2 kg/m3 Figure 6 displays the measured and simulated axial solid
initial static bed height 0.3 0.3 m velocity at two dierent heights. The prediction of the
initial concentration of the 0.6 Gidaspow drag model is also shown. Roughly, both models
particles can describe the nonuniform distribution of the solid velocity
grid size (Dx Dz) 1.0 1.0 cm along the radial direction. The negative velocity is captured
cm
restitution coecient of 0.95
close to the wall. By comparison, the simulated result by the
particleparticle present model is in better agreement with the measured result,
restitution coecient of 0.9 while the Gidaspow drag model overestimates the solid
particlewall velocity. With the application of the BSD drag model to
specularity coecient 0.5 consider the bubble eects, the experimental data can be more
accurately matched.
To investigate the motion of bubbles, contour plots of
voidage are converted to binary images. The software package
Image-Pro Plus40 is applied to identify the bubbles and their
coordinates. The gas volume fraction with a threshold value of
0.8 is used as a cuto point to demarcate bubble boundaries.
The bubble size is represented by the area of the bubble (Ab) in
the 2D bubbling uidized bed. Equivalent bubble diameters are
calculated from the measured void areas:

db,eq = 4Ab / (8)

Figure 7 presents the distribution of equivalent bubble


diameters with the bed height together with the experimental
results of Laverman et al.37 We can see that the BSD drag
model predicts a prole similar to that of the measured data.
Small bubbles are found at the bottom of bed, and with
Figure 3. Axial prole of the solid concentration with dierent grid
sizes. increasing bed height, the equivalent diameters of bubbles have
a rising trend. Compared to the Gidaspow drag model, the BSD
drag model gives larger bubble diameters, which are in better
agreement with the experimental data.
15779 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

Figure 4. Instantaneous concentrations of particles in bubbling uidized beds: (a) Gidaspow drag model; (b) BSD drag model.

Figure 5. Instantaneous concentrations of particles as a function of


time.

Usually, the bubble rising velocity mainly depends on the


diameter of the bubbles. Figure 8 shows variation of the bubble
rising velocity with the diameter of the bubbles. The proles of
the bubble rising velocity obtained from the correlation
proposed by Hilligardt and Werther41 and the experiments of
Laverman et al.37 are also plotted. It can be observed that, as
the bubble diameter increases, the bubble rising velocity is
promoted. Roughly, the trends are consistent. However, the
dierence is obvious. In this work, the bubble rising velocity is
obtained by means of tracking the movement of the individual
bubble center. The wall eects, coalescence, and breakup of
bubbles have a strong inuence on the bubble rising velocity,
even if the bubble sizes are similar. More detailed insight into
the bubble diameter and bubble rising velocity is still needed
for a deep understanding of the dynamics behavior in bubbling
uidized beds. Figure 6. Lateral proles of the time-averaged axial solid velocity: (a) z
= 0.105 m; (b) z = 0.245 m.
To evaluate the eects of solid pressure due to particle
particle collisional interaction and bubble-induced added mass
force, the variations of solid pressure and added mass force with between particles. While the bubble-induced added mass force
solid concentrations are displayed in Figure 9. It can be is weakened at a high solid concentration, it dominates at a low
observed that there is a clear dierence between them. With the solid volume fraction, where the movement of bubbles is
solid volume fraction increased, the contribution of solid intense.
pressure becomes signicant. This is attributed to the fact that a The accelerations in the dense and dilute phases have a great
higher solid concentration promotes the probability of collision inuence on the predicted results in the simulation of risers.42
15780 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

Figure 10. Prole of accelerations with the solid concentration.


Figure 7. Distribution of the bubble diameter with the bed height.

Figure 11. Proles of the solved bubble diameter and number density
of the bubbles with the solid concentration.
Figure 8. Variation of the bubble rising velocity with the bubble
diameter.

Figure 12. Prole of the drag component with the solid concentration.
Figure 9. Prole of the solid pressure gradient and added mass force
with the solid concentration. concentration increases, the accelerations in the emulsion and
bubble phases decrease and tend to zero. The acceleration of
Here, the accelerations in the emulsion and bubble phases with the bubble phase is higher than that of the emulsion phase. This
the solid concentration are shown in Figure 10. The implies that the rate of momentum change in the bubble phase
accelerations in the emulsion and bubble phases are directly is more obvious.
calculated from local independent variables. We can nd that Figure 11 shows the distributions of the solved bubble size
both accelerations display a similar trend. As the solid and number density of the bubbles with the solid
15781 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

Figure 13. Prole of the drag coecient with the solid concentration. Figure 16. Size distribution of quartz sand particles.

Table 2. Parameters for 3D Experiment and Numerical


Simulation
parameter experiment simulation unit
particle density 2650 2650 kg/m3
mean particle diameter 530 530 m
reactor height 1.2 1.2 m
reactor diameter 0.06 0.06 m
inlet gas velocity 0.55/0.75 0.55/0.75 m/s
gas viscosity 1.8 105 1.8 105 Pas
gas density 1.2 1.2 kg/m3
initial static bed height 0.2 0.2 m
initial concentration of the 0.6
particles
grid size (Dr D Dz) 0.25 0.05 cm rad
1.0 cm
Figure 14. Experimental system of a bubbling uidized bed mounted restitution coecient of 0.95
with an ECT sensor. particleparticle
restitution coecient of 0.8
particlewall
concentration. Here, the bubble density takes on a denition
specularity coecient 0.5
similar to that of the cluster density in the previous CSD drag
model. The bubble size is directly solved by the momentum
equation in the cell. It can be observed that, with increasing increases. On the contrary, the contribution from the bubbles
solid volume fraction, the bubble diameter is decreased and the decreases. This is due to the fact that there are small number
bubble density is improved. This implies that, at a high solid densities of bubbles at a high solid concentration. However,
volume fraction, more small bubbles appear in the emulsion when the solid volume fraction is lower, the bubble-induced
phase. The eect of the bubbles is gradually weakened. drag component is comparable to that in the emulsion phase,
There are two parts of contributions from the emulsion and which indicates that the bubble-induced drag component plays
bubble phases to the multiscale drag force in bubbling uidized an important role in determination of the total drag force.
beds. Figure 12 shows the distribution of the two contributions The distribution of the drag coecient with the solid
with solid concentrations. It can be seen that, with increasing concentration is shown in Figure 13. We can nd that both
solid concentration, the drag component in the emulsion phase drag coecients have a rising trend with increasing solid

Figure 15. Schematic diagram of an ECT system.

15782 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785


Industrial & Engineering Chemistry Research Article

Figure 17. Comparisons of simulated results (a and c) and measured


data (b and d) at dierent operating velocities (z = 0.2 m).

volume fraction. The BSD drag coecient is relatively low


compared to the Gidaspow drag coecient, which means that
the drag coecient becomes weak when the bubble eect is
considered. Meanwhile, the BSD drag model obtains a zonal
distribution of the drag coecient, which is dierent from that
by the Gidaspow drag model. This indicates that the drag
coecient not only depends on the local solid concentration
but also is greatly inuenced by other parameters such as local
velocities. Schneiderbauer et al.43 conducted a comparative
analysis of subgrid drag modications in bubbling uidized
beds. The results revealed that, although dierent dependencies Figure 18. Time-averaged proles of the solid concentrations along
on the void fraction and slip velocity led to the discrepancy in the lateral direction: (a) ug = 0.55 m/s; (b) ug = 0.75 m/s.
drag modications, the main features of bubbling uidized beds
can be well predicted by these subgrid modications. Quartz sand particles are selected as the bed material in this
Meanwhile, Milioli et al.44 pointed out that it was necessary experiment. The size distribution of the particles is displayed in
to consider the dependency of the drag coecient on the slip Figure 16. The mean diameter is 0.53 mm, which belongs to
velocity. Geldart B particles. The ow behaviors in bubbling uidized
3.2. Comparison with the 3D Experimental Results. beds are tested under two supercial gas velocities. The
3.2.1. Experiment Using an Electrical Capacitance Tomog- corresponding 3D simulations are also performed according to
raphy (ECT) Sensor. Experiments are conducted in a 3D experimental conditions. Detailed parameters and conditions in
bubbling uidized bed, as illustrated in Figure 14. The column the experiment and simulation are listed in Table 2.
is 0.06 m in inner diameter and 1.2 m in height with a static bed 3.2.2. Experimental and Simulated Results in a 3D
height of 0.2 m. Uniform uidizing air is injected into the Fluidized Bed. Figure 17 displays the contour plot of an
column through the distributor at the bottom. The ow rate is instantaneous solid volume fraction obtained by simulation and
measured using a owmeter. Particle concentrations are experiment. It can be observed that the model prediction can
measured by means of an ECT system. capture a distribution similar to that of the measured result.
Figure 15 displays a schematic diagram of an ECT system. It There is an accumulation of particles toward the walls owing to
consists of an electrical capacitance sensor with 12 measure- the wall friction. The motion of bubbles in the center leads to a
ment electrodes, a capacitance data acquisition system, and a lower solid volume fraction. In comparison to the solid
data postprocesser. The electrical capacitance sensor is placed distribution at low velocity, a high inlet velocity enhances the
around a measurement section of the ow eld. The lateral discrepancy of the solid concentration. The lateral
capacitance signals acquired from the sensors are transmitted proles of time-averaged solid holdup under dierent drag
to a capacitance data acquisition system having a capturing rate models are shown in Figure 18. We can nd that the solid
of 140 frames/s. The linear back-projection image reconstruc- concentration is low in the center regime and increases toward
tion method45 is employed to derive the image from the the wall. The BSD drag model can obtain a fair prediction with
capacitance data. experimental data. The traditional Gidaspow drag model
15783 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785
Industrial & Engineering Chemistry Research Article

underpredicts the solid concentration, which is related to the (4) Gao, X.; Wang, L. J.; Wu, C.; Cheng, Y. W.; Li, X. Novel Bubble-
overlook of the mesoscale structure eect in the grid. The Emulsion Hydrodynamic Model for GasSolid Bubbling Fluidized
prediction by the BSD drag model can agree reasonably with Beds. Ind. Eng. Chem. Res. 2013, 52, 1083510844.
the experimental data using coarse-grid resolution. From the (5) Asegehegn, T. W.; Schreiber, M.; Krautz, H. J. Influence of two-
and three-dimensional simulations on bubble behavior in gassolid
proles of solid holdup using two dierent inlet velocities, a
fluidized beds with and without immersed horizontal tubes. Powder
similar trend can be observed. A higher operating velocity Technol. 2012, 219, 919.
results in a signicant lateral discrepancy in the solid (6) Zou, Z.; Li, H. Z.; Zhu, Q. S.; Wang, Y. C. Experimental Study
distribution. and Numerical Simulation of Bubbling Fluidized Beds with Fine
Particles in Two and Three Dimensions. Ind. Eng. Chem. Res. 2013, 52,
4. CONCLUSION 1130211312.
A BSD drag coecient model is developed to account for the (7) Reuge, N.; Cadoret, L.; Coufort-Saudejaud, C.; Pannala, S.;
eect of the bubbles and incorporated into the TFM to Syamlal, M.; Caussat, B. Multifluid Eulerian modeling of dense gas
simulate the hydrodynamic characteristics in bubbling uidized solids fluidized bed hydrodynamics: Influence of the dissipation
parameters. Chem. Eng. Sci. . 2008, 63, 55405551.
beds. The eects of solid pressure due to particleparticle
(8) Wang, J. W.; Ge, W. Multi-scale analysis on particle-phase
collisional interaction and bubble-induced added mass force are stresses of coarse particles in bubbling fluidized beds. Chem. Eng. Sci. .
evaluated. At a high solid volume fraction, the contribution of 2006, 61, 27362741.
solid pressure is signicant, while the bubble-induced added (9) Wang, J.; van der Hoef, M. A.; Kuipers, J. A. M. Coarse grid
mass force dominates at a low solid volume fraction. A zonal simulation of bed expansion characteristics of industrial-scale gas
distribution of the drag coecient with the solid volume solid bubbling fluidized beds. Chem. Eng. Sci. 2010, 65, 21252131.
fraction is obtained by the BSD drag model, which implies that (10) Igci, Y.; Igci, Y.; Andrews, A. T., IV; Sundaresan, S.; Pannala, S.;
the drag coecient depends not only on the local solid OBrien, T. Filtered Two-Fluid Models for Fluidized GasParticle
concentration but also on other parameters such as local Suspensions. AIChE J. 2008, 54, 14311448.
velocities. (11) Igci, Y.; Sundaresan, S. Constitutive Models for Filtered Two-
To further verify the model, 3D simulations of bubbling Fluid Models of Fluidized GasParticle Flows. Ind. Eng. Chem. Res.
2011, 50, 1319013201.
uidized beds under dierent operating velocities are carried
(12) Holloway, W.; Sundaresan, S. Filtered models for bidisperse gas-
out. An ECT sensor system is employed to measure the lateral particle flows. Chem. Eng. Sci. 2014, 108, 6786.
distribution of the solid volume fraction in the bed. By (13) Parmentier, J. F.; Simonin, O. A Functional Subgrid Drift
comparisons with measured data, the BSD drag model gives a Velocity Model for Filtered Drag Prediction in Dense Fluidized Bed.
better prediction than the conventional drag model. AIChE J. 2012, 58, 10841098.
The drag force plays a vital role in the simulation of bubbling (14) Sarkar, A.; Sun, X.; Sundaresan, S. Sub-grid drag models for
uidized beds with Geldart A particles. In future work, the BSD horizontal cylinder arrays immersed in gasparticle multiphase flows.
drag model will be extended to the prediction of uidization Chem. Eng. Sci. 2013, 104, 399412.
with Geldart A particles in bubbling uidized beds. (15) Schneiderbauer, S.; Pirker, S. Filtered and Heterogeneity-Based

Subgrid Modifications for GasSolid Drag and Solid Stresses in


ASSOCIATED CONTENT Bubbling Fluidized Beds. AIChE J. 2014, 60, 839853.
(16) Yang, N.; Wang, W.; Ge, W.; Wang, L. N.; Li, J. Simulation of
* Supporting Information
S
heterogeneous structure in a circulating fluidized-bed riser by
Details of the equations used in the CFD modeling listed in combining the two-fluid model with the EMMS approach. Ind. Eng.
Tables S1 and S2 and a listing of the nomenclature used in this Chem. Res. 2004, 43, 55485561.
paper. This material is available free of charge via the Internet at (17) Shah, M. T.; Utikar, R. P.; Tade, M. O.; Pareek, V. K.
http://pubs.acs.org. Hydrodynamics of an FCC riser using energy minimization multiscale

AUTHOR INFORMATION
Corresponding Author
drag model. Chem. Eng. J. 2011, 168, 812821.
(18) Chalermsinsuwan, B.; Gidaspow, D.; Piumsomboon, P. Two-
and three-dimensional CFD modeling of Geldart A particles in a thin
bubbling fluidized bed: Comparison of turbulence and dispersion
*Tel.: +0451 8641 2258. Fax: +0451 8622 1048. E-mail:
coefficients. Chem. Eng. J. 2011, 171, 301313.
huilin@hit.edu.cn. (19) Wang, J. W.; Liu, Y. N. EMMS-based Eulerian simulation on the
Notes hydrodynamics of a bubbling fluidized bed with FCC particles. Powder
The authors declare no competing nancial interest. Technol. 2010, 197, 241246.

ACKNOWLEDGMENTS
This work was nancially supported by the National Natural
(20) Shi, Z. S.; Wang, W.; Li, J. H. A bubble-based EMMS model for
gassolid bubbling fluidization. Chem. Eng. Sci. 2011, 66, 55415555.
(21) Wang, Y.; Zou, Z.; Li, H.; Zhu, Q. A new drag model for TFM
simulation of gassolid bubbling fluidized beds with Geldart-B
Science Foundation of China (Grants 51390494 and particles. Particuology 2014, 15, 151159.
21276056).

(22) Lv, X. L.; Li, H. Z.; Zhu, Q. S. Simulation of gassolid flow in


2D/3D bubbling fluidized beds by combining the two-fluid model
REFERENCES with structure-based drag model. Chem. Eng. J. 2014, 236, 149157.
(1) Kunii, D.; Levenspiel, O. Fluidization Engineering; Butterworth- (23) Wang, S.; Zhao, G. B.; Liu, G. D.; Lu, H. L.; Zhao, F. X.; Zhang,
Heinemann: Boston, MA, 1991. T. Y. Hydrodynamics of gassolid risers using cluster structure-
(2) Yang, Y. R.; Yang, J. Q.; Chen, W.; Rong, S. X. Instability analysis dependent drag model. Powder Technol. 2014, 254, 214227.
of the fluidized bed for ethylene polymerization with condensed mode (24) Gidaspow, D. Multiphase Flow and Fluidization: Continuum and
operation. Ind. Eng. Chem. Res. 2002, 41, 25792584. Kinetic Theory Description; Academic Press: Boston, MA, 1994.
(3) Bokkers, G. A.; Laverman, J. A.; van Sint Annaland, M.; Kuipers, (25) Srivastava, A.; Sundaresan, S. Analysis of a frictionalkinetic
J. A. M. Modelling of large-scale dense gassolid bubbling fluidised model for gasparticle flow. Powder Technol. 2003, 129, 7285.
beds using a novel discrete bubble model. Chem. Eng. Sci. 2006, 61, (26) Savage, S. B. Analysis of slow high-concentration flows of
55905602. granular materials. J. Fluid Mech. 1998, 377, 126.

15784 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785


Industrial & Engineering Chemistry Research Article

(27) Wang, S.; Li, X.; Lu, H.; Yu, L.; Sun, D.; He, Y.; Ding, Y.
Numerical simulations of flow behavior of gas and particles in spouted
beds using frictionalkinetic stresses model. Powder Technol. 2009,
196, 184193.
(28) Ergun, S. Fluid flow through placed columns. Chem. Eng. Prog.
1952, 48, 9098.
(29) Zuber, N. On the dispersed two-phase flow in the laminar flow
regime. Chem. Eng. Sci. 1964, 19, 897917.
(30) Darton, R. C.; Harrison, D. The rise of single gas bubbles in
liquid fluidized bed. Trans. Inst. Chem. Eng. 1974, 52, 301304.
(31) Thomas, D. G. Transport characteristics of suspension: VIII. A
note on the viscosity of Newtonian suspensions of uniform spherical
particles. J. Colloid Interface Sci. 1965, 20, 267277.
(32) Gidaspow, D. Hydrodynamics of fluidization and heat transfer:
supercomputer modeling. Appl. Mech. Rev. 1986, 39, 123.
(33) Ishii, M.; Zuber, N. Drag coefficient and relative velocity in
bubbly, droplet or particulate flows. AlChE J. 1979, 25, 843855.
(34) Johnson, P. C.; Jackson, R. Frictional collisional constitutive
relations for antigranulocytes-materials, with application to plane
shearing. J. Fluid Mech. 1987, 176, 6793.
(35) Hao, Z. H.; Wang, S.; Lu, H. L.; Liu, G. D.; He, Y. R.; Xu, P. F.;
Wang, J. X. Numerical Simulation of Fluid Dynamics of a Riser:
Influence of Particle Rotation. Ind. Eng. Chem. Res. 2010, 49, 3585
3596.
(36) Yin, L. J.; Wang, S. Y.; Lu, H. L.; Wang, S.; Xu, P. F.; Wei, L. X.;
He, Y. R. Flow of gas and particles in a bubbling fluidized bed with a
filtered two-fluid model. Chem. Eng. Sci. 2010, 65, 26642679.
(37) Laverman, J. A.; Roghair, I.; Annaland, M. V.; Kuipers, H.
Investigation into the hydrodynamics of gassolid fluidized beds using
particle image velocimetry coupled with digital image analysis. Can. J.
Chem. Eng. 2008, 86, 523535.
(38) Li, T. W.; Grace, J.; Bi, X. T. Study of wall boundary condition
in numerical simulations of bubbling fluidized beds. Powder Technol.
2010, 203, 447457.
(39) Li, T. W.; Zhang, Y. M. A new model for two-dimensional
numerical simulation of pseudo-2D gassolids fluidized beds. Chem.
Eng. Sci. 2013, 102, 246256.
(40) Image Pro, version 5.1; Media Cybernetics: Silver Spring, MD,
2007.
(41) Hilligardt, K.; Werther, J. Local Bubble Gas Hold-Up and
Expansion of GasSolid Fluidized Beds. Ger. Chem. Eng. 1986, 9,
215221.
(42) Wang, S.; Liu, G. D.; Lu, H. L.; Xu, P. F.; Yang, Y. C.; Gidaspow,
D. A cluster structure-dependent drag coefficient model applied to
risers. Powder Technol. 2012, 225, 176189.
(43) Schneiderbauer, S.; Puttinger, S.; Pirker, S. Comparative analysis
of subgrid drag modifications for dense gasparticle flows in bubbling
fluidized beds. AIChE J. 2013, 59, 40774099.
(44) Milioli, C. C.; Milioli, F. E.; Holloway, W.; Agrawal, K.;
Sundaresan, S. Filtered Two-Fluid Models of Fluidized GasParticle
Flows: New Constitutive Relations. AIChE J. 2013, 59, 32653275.
(45) Xie, C. G.; Huang, S. M.; Hoyle, B. S.; Thorn, R.; Lenn, C.;
Snowden, D.; Beck, M. S. Electrical capacitance for flow imaging:
system model for development of image reconstruction algorithms and
design of primary sensors. IEE Proc., Part G: Electron. Circ. Syst. 1992,
139, 8998.

15785 dx.doi.org/10.1021/ie502412g | Ind. Eng. Chem. Res. 2014, 53, 1577615785

Vous aimerez peut-être aussi