Você está na página 1de 13

A LBMDEM solver for fast discrete

particle simulation of particlefluid flows

Qingang Xiong, Ehsan Madadi-Kandjani


& Giulio Lorenzini

Continuum Mechanics and


Thermodynamics
ISSN 0935-1175
Volume 26
Number 6
Continuum Mech. Thermodyn. (2014)
26:907-917
DOI 10.1007/s00161-014-0351-z

1 23

Your article is protected by copyright and


all rights are held exclusively by SpringerVerlag Berlin Heidelberg. This e-offprint is
for personal use only and shall not be selfarchived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com.

1 23

Author's personal copy


Continuum Mech. Thermodyn. (2014) 26:907917
DOI 10.1007/s00161-014-0351-z

O R I G I NA L A RT I C L E

Qingang Xiong Ehsan Madadi-Kandjani Giulio Lorenzini

A LBMDEM solver for fast discrete particle simulation


of particlefluid flows

Received: 11 March 2014 / Accepted: 15 March 2014 / Published online: 3 April 2014
Springer-Verlag Berlin Heidelberg 2014

Abstract The lattice Boltzmann method (LBM) for simulating fluid phases was coupled with the discrete
element method (DEM) for studying solid phases to formulate a novel solver for fast discrete particle simulation
(DPS) of particlefluid flows. The fluid hydrodynamics was obtained by solving LBM equations instead
of solving the NavierStokes equation by the finite volume method (FVM). Interparticle and particlewall
collisions were determined by DEM. The new DPS solver was validated by simulating a three-dimensional gas
solid bubbling fluidized bed. The new solver was found to yield results faster than its FVMDEM counterpart,
with the increase in the domain-averaged gas volume fraction. Additionally, the scalability of the LBMDEM
DPS solver was superior to that of the FVMDEM DPS solver in parallel computing. Thus, the LBMDEM
DPS solver is highly suitable for use in simulating dilute and large-scale particlefluid flows.
Keywords Lattice Boltzmann method Discrete element method Discrete particle simulation
Particlefluid flows

1 Introduction
The past two decades have witnessed a rapid development of the EulerLagrange discrete particle simulation
(DPS) for simulating particlefluid flows [16]. In traditional DPS, the fluid volume fraction incorporated
NavierStokes (NS) equation is solved by traditional grid-based computational fluid dynamics (CFD) methods, such as the finite volume method (FVM) [79] or meshfree CFD approaches such as the smoothed particle
hydrodynamics (SPH) [10]. The solid phase is represented by discrete particles, each of which is tracked by
the Newtons second law to give much more detailed information compared with the EulerEuler two-fluid
model (TFM) [1113] in which the solid phase is modeled as a continuum. In most cases of DPS, the grid
size in FVM or the smooth length in SPH for discretization of the fluid NS equation is at least an order of
magnitude larger than the particle diameter. In this situation, computational effort is mostly devoted to particle
dynamics when the domain-averaged concentration of the solid phase is far from dilute. As a result, developing
a fast fluid hydrodynamics solver is of little importance when the mean particle loading is not low. However,
in many natural processes and engineering applications, e.g., air pollutant dispersion and low concentration
pneumatic conveying systems, the domain-averaged volume fraction of the solid phase is relatively low and
Communicated by Andreas chsner.
Q. Xiong E. Madadi-Kandjani
Department of Mechanical Engineering, Iowa State University, Ames, IA 50011, USA
E-mail: xiong@iastate.edu
G. Lorenzini (B)
Department of Industrial Engineering, University of Parma, 43124 Parma, Italy
E-mail: giulio.lorenzini@unipr.it

Author's personal copy


908

Q. Xiong et al.

the majority of the computational demand is shifted from the solid phase to the fluid phase. Additionally,
if the simulation domain is relatively large and parallel computation is required, traditional grid-based CFD
methods will face great challenges for good performance when many processors are used. Although meshfree
approaches are free from the problem of parallel scalability, its modeling accuracy and speed are relatively
lower than traditional grid-based CFD methods. Therefore, in dilute or large-scale systems, solving the fluid
phase with an accurate, fast, and highly scalable scheme is extremely important to DPS.
In recent years, the lattice Boltzmann method (LBM) [14] of simulating fluid mechanics has attracted great
interest because of its explicit nature and high parallelism while maintaining excellent modeling accuracy, and
it has already been viewed as an efficient tool to replace traditional grid-based CFD methods in a variety of
fluid flow applications [15,16]. LBM in computer simulations of particlefluid flows has also been applied,
especially in direct numerical simulations [1721]. Recently, LBM has also been largely applied in TFM
[2225], accelerating modeling speed significantly compared with traditional grid-based CFD approaches.
However, developing a LBM-based fluid hydrodynamics solver in DPS is much less common than TFM.
Sungkorn et al. [26] first attempted to use LBM in DPS for dilute particleladen suspensions, an approach
followed by Song et al. [27]. Nonetheless, it can be seen in these pioneering papers that the derivation of
the LBM-based fluid hydrodynamics solver which is rather complex, seems to be difficult for engineers to
implement. Additionally, the focus of these studies is more on numerical validations and applications than on
modeling efficiency. As already discussed, development of a LBM-based fluid hydrodynamics solver in DPS
is largely motivated by the attractive modeling speed and parallel efficiency of such a solver. Thus, developing
a fast and engineer-friendly LBM-based fluid hydrodynamics solver with a rigorous theoretical foundation is
unquestionably important.
In this article, a LBM-based solver for fluid phases was developed, which can accurately recover the
fluid continuity and momentum equations at low Mach number. The discrete element method (DEM) [28]
was applied to the interparticle and particlewall collisions. Several critical points in this novel approach are
discussed. A three-dimensional (3-D) gassolid bubbling fluidized bed was simulated by this LBMDEM DPS
solver, and the results were in good agreement with the literature. The modeling efficiency of this LBMDEM
DPS solver was demonstrated.
2 Derivation of the LBMDEM DPS solver
2.1 Governing equations for fluid and solid phases
For incompressible fluid laden with solid particles, the governing equations are expressed as
f f
+ (f f u) = 0
t
(1)
f f u
s
+ (f f uu) = f p + (f f ) + M + f g,
t
where f , u, and f , are the volume fraction, velocity vector, and stress tensor of the fluid in the grid. f is the
fluid material density, which is constant. M s is the sum of forces exerted by all solid particles within the fluid
grid.
Solid particles are tracked individually by Newtons second law. Forces exerted on each solid particle
include buoyancy from the fluid pressure gradient, interparticle and particlewall collisions, particlefluid
interactions, and gravity. Therefore, each solid particle evolves as follows
m

dU
= V p + Fc + Ff + mg,
dt

(2)

where m, U and V are the mass, velocity, and volume of a solid particle, respectively. Fc and Ff refer to the
interparticle/particlewall collision force and forces exerted by the fluid phase. p is the hydrodynamic pressure
that comes from the fluid, and g is the gravitational acceleration. Because of the Newtons third law, it can be
obtained that for a computational cell
Ms =

Nc
1 
Ffk ,
Vcell
k=1

where Nc is the number of solid particles in this cell, Vcell is the volume of this cell.

(3)

Author's personal copy


A LBMDEM solver

909

Fig. 1 LBM discretization of the simulation domain and density function distribution (for simplicity, only two-dimension is
shown here)

2.2 Lattice Boltzmann method for fluid dynamics


Although the lattice Boltzmann method is also a grid-based method, it is purely explicit and directly derived
from the microscopic Boltzmann equation. In LBM, the simulation domain is discretized into a regular mesh
and an instantaneous density distribution function f q (x, t) resides at each lattice. x is the position of the lattice,
and q is the direction of discrete velocity. The schematic diagram of the domain discretization and f q (x, t) in a
lattice is shown in Fig. 1 using the D2Q9 model as an example. The fluid dynamics is governed by the evolution
of f q (x, t), which consists of the so-called collision and propagation steps. In this study, the widely used single
relaxation BhatnagarGrossKrook (BGK) scheme [29] was employed to obtain the collisionpropagation
equation of f q (x, t)

1  eq
(4)
f q (, u) f q (x, t) + q ,

where is the dimensionless relaxation parameter related to the fluid kinematic viscosity , and is the fluid
eq
density. cq is the lattice speed for f q (x, t) to propagate to adjacent lattices. f q (, u) is the equilibrium density
distribution function. q comes from the body forces such as particlefluid interactions and gravity. Therefore,
and u can be obtained as


=
f q , u =
cq f q .
(5)
f q (x + cq dt, t + dt) = f q (x, t) +

q
eq

In order to recover the NS equation with second-order accuracy, f q (, u) is formulated as




cq u 9 (cq u)2
3 u2
eq

,
f q (, u) = Aq 1 + 3 2 +
c
2 c4
2 c2

(6)

eq

where Aq is the prefactor in f q (, u). Equation (6) is valid when the Mach number of the simulation Ma is
reasonably small. In this study, we chose the values of dx and dt to ensure that
|umax |
0.05,
c
is the maximum velocity in the simulation domain. The fluid pressure in LBM reads
Ma =

where umax

p=

c2
.
3

Using the following formulation for q [30]




1
Aq [3(cq u) + 9(cq u) cq ] (M s + f g),
q = 1
2

(7)

(8)

(9)

Author's personal copy


910

Q. Xiong et al.

Fig. 2 DEM collision model for inter-particle collision

and the relation between f and as


= f f ,

(10)

f f
+ (f f u) = 0
t
f f u
+ (f f uu) = (f p) + (f f ) + M s + f g.
t

(11)

Equation (4) is physically equivalent to

It is worth noting that in the LBM expression of the volume-averaged NS equation, p is proportional to the
fluid material density f , not the fluid superficial density . Thus, p is obtained as
p=

c2
.
3 f

(12)

It can be found that Eq. (11) differs from Eq. (1) in the pressure gradient term. Since (f p) = pf + f p,
Eq. (1) can be recovered by adding a term pf to the right-hand side of Eq. (11). Then,


1
q = 1
2


Aq [3(cq u) + 9(cq u) cq ] ( pf + M s + f g).

(13)

With this modification, Eq. (1) is completely approximated by Eq. (4). f can be calculated using the LBM
scheme as
f (x) =

f (x + cq dt) f (x cq dt)
3
Aq cq
+ O[(dx)2 ].
2 q
c2 dt

(14)

p in Eq. (2) can be obtained in the same way.

2.3 Discrete element method for particleparticle/wall collision


The discrete element method was employed to simulate interparticle and particlewall collisions analogously
to the combination of a spring and a dash pot, as shown in Fig. 2. The normal force between solid particle i
and j was computed as

Author's personal copy


A LBMDEM solver

911

Fig. 3 Schematic diagram of lattice category, control volume, and the interpolation process (for simplicity, only two-dimension
is presented here)

Fcji,n = kn ni j n U i j,n
x j xi

ni j = 
x j xi 
U i j,n = [(U i U j ) ni j ]ni j
 mm
i j
kn
2 ln e m i +m
j
n =
,
2 + (ln e)2

(15)

where kn , , and n are the normal spring stiffness, interparticle overlap, and normal damping coefficient,
respectively. e is the coefficient of restitution for normal collision. For the tangential force, it reads





kt t t U i j,t if Fcji,t  f Fcji,n 








Fcji,t =
 c


 c

Fc
f
ji,n  t i j if F ji,t  > f F ji,n 
U i j,t = U i j U i j,n
U i j,t

ti j = 
U i j,t 
t
t = t0 + U i j,t dt
t0

t =

2 ln e

mi m j
m i +m j kt

2 + (ln e)2

(16)

where kt , n , t , and f are tangential spring stiffness, tangential displacement, tangential damping coefficient,
and friction coefficient, respectively. t0 is the initial time of this collision. Due to the torque exerted by the
tangential force, the angular motion of the particle i is expressed as

Author's personal copy


912

Q. Xiong et al.

Fig. 4 Flowchart of the LBMDEM DPS solver

Ii

 di
di
=
ni j Fcji,t ,
dt
2

(17)

where d and I are the particles diameter and moment of inertia. For detailed information and formulation
about DEM, it can be referred to [8].

2.4 Solution procedure of the LBMDEM DPS solver


The solving procedure of the foregoing derived LBMDEM DPS solver is similar to that of the conventional
FVMDEM DPS solver. In DPS, the critical point is the coupling between the solid and fluid phases. At the
beginning of each time step, the fluid volume fraction f is computed by the number density of solid particles
within the control volume of the respective fluid lattice. The concept of control volume of a fluid lattice and
the interpolation of the fluid volume fraction and velocity to the position of the solid particle for calculation of
Ff are shown in Fig. 3. The reason for introducing control volume to each fluid lattice is to make computation
of the fluid volume fraction near boundaries more conveniently. At the same time, fluid superficial density ,
velocity u, and pressure p at each lattice are computed by Eqs. (5) and (12). Then, the fluid volume fraction

Author's personal copy


A LBMDEM solver

913

Fig. 5 The configuration and boundary conditions of the bubbling fluidized bed

and velocity are interpolated to all of the solid particles to calculate Ff for each solid particle. In this study,
the interpolation formulation for f (xk ) for solid particle k is
f (xk ) =

W1 f (x) + W2 f (x + c2 dt) + W3 f (x + c1 dt) + W4 f (x + c5 dt)


,
W1 + W2 + W3 + W4

(18)

where W is the area of the rectangle as shown in Fig. 3. u(xk ) is calculated in the same way. Thus, Fkf is the
function of f (xk ), u(xk ), and U k as
Ffk = Ffk (f (xk ), u(xk ), U k ).

(19)

With Eq. (3), M s is obtained. This is followed by calculation of the fluid volume fraction gradient by Eq. (14).
With the obtained quantities, Eqs. (6) and (13) are used to execute fluid lattice evolution governed by Eq. (4).
To be consistent with the conventional FVMDEM DPS solver, the fluid pressure is recalculated using the
updated fluid superficial density, after the execution of Eq. (4). Finally, the particle integration is launched
using Eqs. (2), (15), (16), and (17). The flowchart of the LBMDEM DPS solver is illustrated in Fig. 4.

3 Numerical validation
To validate the LBMDEM DPS solver, a 3-D gassolid bubbling fluidized bed [31] was simulated with the
D3Q27 LBM scheme. The configuration and boundary conditions of the bubbling fluidized bed are shown in
Fig. 5. The gas was supplied from the bottom with velocity u 0 , and an atmosphere pressure p0 was assigned to
the top outlet. The left and right walls are given no-slip boundary condition while the front and back faces are
periodic. The solid and gas physical properties are listed in Table 1. Only the gassolid drag force is considered
in Ff , since drag force is widely viewed to play the dominant role in gassolid interactions. The Gidaspow
type of drag coefficient [32] was chosen in this study. Thus, Ff for solid particle k reads

Author's personal copy


914

Q. Xiong et al.

Table 1 Physical and simulation parameters of the bubbling fluidized bed


Parameters

Value

Domain-averaged gas volume fractions, f


Density of solid particle, s
Diameter of solid particle, d
Normal spring stiffness, kn
Tangential spring stiffness, kt
Restitution coefficient, e
Density of gas, g
Viscosity of gas, g
Bed depth, L
Bed width, W
Bed height, H
Inlet gas velocity, u 0
Atmosphere pressure, p0
Grid spacing of LBM, dx
Time step, dt
Number of CPU cores

0.7
1495 kg/m3
7.5 105 m
7 N/m
2 N/m
0.9
1.2 kg/m3
1.8 105 Pa s
4 104 m
2 102 m
4 102 m
8.2 103 m/s
105 Pa
2 104 m
1 106 s
8

Fig. 6 Distribution of solid volume fraction with the axial position of the bed

Ffk =
k =

Cd =

Vk k
[u(xk ) U k ]
1 f (xk )
3 f (xk )[1f (xk )]
f |u(xk ) U k | [f (xk )]2.65
4 Cd
dk

f (xk ) > 0.8

150 [1f (xk )]2 f f + 7 [1 f (xk )] f |u(xk ) U k | f (xk ) 0.8


4
dk
f (xk )dk2



24
0.687
Rek 1000
Rek 1 + 0.15Rek

(20)

0.44
Rek > 1000
f (xk ) |u(xk ) U k | dk
Rek =
.
f
This bubbling fluidized bed was also simulated by the FVMDEM DPS solver, which was carried out on the
open-source software OpenFOAM , for comparisons of the modeling accuracy, speed, and parallel scalability.
The axial distributions of time-averaged solid volume fraction when the system reached statistically steady
state are shown in Fig. 6. It can be seen that the curves between these two DPS solvers are in excellent agreement
and also match the results in the literature [31] very well. The slightly larger difference between these two DPS
solvers at the bottom of the bed is, possibly, due to a different way of computing the f . In the LBMDEM DPS

Author's personal copy


A LBMDEM solver

915

Fig. 7 Speedup of the LBMDEM DPS solver to the FVMDEM DPS solver with different f

solver, the f at the inlet was less than unity due to the control volume, while in the FVMDEM DPS solver,
the f at the inlet was assigned to be unity. Based on Eq. (20), it can be found that k near the inlet calculated
in the LBMDEM DPS solver was larger than that from the FVMDEM DPS solver, resulting in a larger drag
forces for solid particles near the inlet to move upward. Thus, it can be concluded that the LBMDEM DPS
solver is established with a solid theoretical foundation.

4 Numerical performance
The speedups of the LBMDEM DPS solver relative to the FVMDEM DPS solver, along with different
domain-averaged gas volume fractions f , are shown in Fig. 7. It can be clearly observed that, with increased
f , the LBMDEM DPS solver is increasingly faster than the FVMDEM DPS solver. This confirms that the
LBMDEM DPS solver is much more suitable for simulating dilute particlefluid flows.
The parallel scalabilities of the LBMDEM and FVMDEM DPS solvers are shown in Fig. 8. Each CPU
core simulated the same domain size as that for one core in Sect. 3. Obviously, the parallel scalability of the
LBMDEM DPS solver is increasingly higher than that of the FVMDEM DPS solver as more CPU cores are
used. Nearly, linear scalability of the LBMDEM DPS solver is achieved in the entire range, opposite to the
performance of the FVMDEM DPS solver. The superiority of the LBMDEM DPS solver to the FVMDEM
DPS solver on multicore scalability is largely attributed to the inherent parallelism of LBM. Therefore, the
LBMDEM DPS solver is a promising alternative to the conventional FVMDEM DPS solver when a large
number of processors are required for large-scale applications, even in dense particlefluid flows.

5 Conclusion
A new DPS solver that uses LBM to solve fluid dynamics equations was developed. Through the addition
of a pressure termed as body force, the LBM BGK scheme was proved to be mathematically equivalent to
the volume fraction averaged NS equation. In this study, DEM was used to model interparticle/particlewall
interactions. The implementation process of this LBMDEM DPS solver was discussed in detail. This solver
was validated by the FVMDEM DPS solver for simulating a 3-D gassolid bubbling fluidized bed. The
numerical validation and the discussion of modeling speed indicate that our proposed LBMDEM DPS solver
is a promising platform for simulating dilute and large-scale particlefluid flows.
Although DEM was employed to simulate the solid phase in this study, our proposed philosophy can be
extended to other discrete approaches. Only the computational part of the solid phase needs to be reformulated,
while the overall framework remains unchanged. Thus, the LBM-based DPS solver is expected to have great

Author's personal copy


916

Q. Xiong et al.

Fig. 8 Parallel scalabilities of the LBM- and FVMDEM DPS solvers

flexibility and to be suitable for application to a variety of problems in simulating dilute and large-scale
particlefluid systems.

References
1. Xu, Y., Padding, J., van der Hoef, M., Kuipers, J.: Detailed numerical simulation of an intruder impacting on a granular bed
using a hybrid discrete particle and immersed boundary (DPIB) method. Chem. Eng. Sci. 104, 201207 (2013)
2. Pepiot, P., Desjardins, O.: Numerical analysis of the dynamics of two- and three-dimensional fluidized bed reactors using
an EulerLagrange approach. Powder Technol. 220, 104121 (2012)
3. Zhu, H., Zhou, Z., Yang, R., Yu, A.: Discrete particle simulation of particulate systems: theoretical developments. Chem.
Eng. Sci. 62, 33783396 (2007)
4. Zhu, H., Zhou, Z., Yang, R., Yu, A.: Discrete particle simulation of particulate systems: a review of major applications and
findings. Chem. Eng. Sci. 63, 57285770 (2008)
5. Li, T., Gopalakrishnan, P., Garg, R., Shahnam, M.: CFDDEM study of effect of bed thickness for bubbling fluidized
beds. Particuology 10, 532541 (2012)
6. Wang, S., Shen, Y., Ma, Y., Gao, J., Lan, X., Dong, Q.: Study of hydrodynamic characteristics of particles in liquidsolid
fluidized bed with uniform transverse magnetic field. Powder Technol. 245, 314323 (2013)
7. Capecelatro, J., Desjardins, O.: An EulerLagrange strategy for simulating particleladen flows. J. Comput. Phys. 238,
131 (2013)
8. Zhou, Z., Kuang, S., Chu, K., Yu, A.: Discrete particle simulation of particlefluid flow: model formulations and their
applicability. J. Fluid Mech. 661, 482510 (2010)
9. Sakai, M., Abe, M., Shigeto, Y., Mizutani, S., Takahashi, H., Vir, A., Percival, J.R., Xiang, J., Pain, C.C.: Verification and
validation of a coarse grain model of the DEM in a bubbling fluidized bed. Chem. Eng. J. 244, 3343 (2014)
10. Sun, X., Sakai, M., Yamada, Y.: Three-dimensional simulation of a solidliquid flow by the DEM-SPH method. J. Comput.
Phys. 248, 147176 (2013)
11. Igci, Y., Andrews, A.T., Sundaresan, S., Pannala, S., OBrien, T.: Filtered two-fluid models for fluidized gas-particle suspensions. AIChE J. 54, 14311448 (2008)
12. Zhang, Y., Bo, Y., Wu, Y., Wu, X., Huang, Z., Zhou, J., Cen, K.: Flow behavior of high-temperature flue gas in the heat
transfer chamber of a pilot-scale coal-water slurry combustion furnace. Particuology (2013) (in press)
13. Wang, S., Liu, G., Lu, H., Yang, Y., Xu, P., Sun, L.: Computational fluid dynamics of riser using kinetic theory of rough
spheres. Powder Technol. 228, 5668 (2012)
14. Chen, S., Doolen, G.D.: Lattice Boltzmann method for fluid flows. Annu. Rev. Fluid Mech. 30, 329364 (1998)
15. Zhou, H., Mo, G., Wu, F., Zhao, J., Rui, M., Cen, K.: GPU implementation of lattice Boltzmann method for flows with
curved boundaries. Comput. Methods Appl. Mech. Eng. 225, 6573 (2012)
16. Janen, C., Krafczyk, M.: Free surface flow simulations on GPGPUs using the LBM. Comput. Math. Appl. 61, 3549
3563 (2011)
17. Yin, X., Zenk, J.R., Mitrano, P.P., Hrenya, C.M.: Impact of collisional versus viscous dissipation on flow instabilities in
gassolid systems. J. Fluid Mech. 727, R2 (2013)
18. Xiong, Q., Li, B., Zhou, G., Fang, X., Xu, J., Wang, J., He, X., Wang, X., Wang, L., Ge, W., Li, J.: Large-scale DNS of
gassolid flows on Mole-8.5. Chem. Eng. Sci. 71, 422430 (2012)
19. Xiong, Q., Li, B., Xu, J., Wang, X., Wang, L., Ge, W.: Efficient 3D DNS of gassolid flows on Fermi GPGPU. Comput.
Fluids 70, 8694 (2012)

Author's personal copy


A LBMDEM solver

917

20. Ladd, A.J.: Numerical simulations of particulate suspensions via a discretized Boltzmann equation. Part 1. Theoretical
foundation. J. Fluid Mech. 271, 285309 (1994)
21. Ladd, A.J.: Numerical simulations of particulate suspensions via a discretized Boltzmann equation. Part 2. Numerical
results. J. Fluid Mech. 271, 311339 (1994)
22. Wang, T., Wang, J.: Two-fluid model based on the lattice Boltzmann equation. Phys. Rev. E 71, 045301 (2005)
23. Sankaranarayanan, K., Sundaresan, S.: Lattice Boltzmann simulation of two-fluid model equations. Ind. Eng. Chem.
Res. 47, 91659173 (2008)
24. Luo, L.-S., Girimaji, S.S.: Lattice Boltzmann model for binary mixtures. Phys. Rev. E 66, 035301 (2002)
25. Luo, L.-S., Girimaji, S.S.: Theory of the lattice Boltzmann method: two-fluid model for binary mixtures. Phys. Rev.
E 67, 036302 (2003)
26. Sungkorn, R., Derksen, J.J.: Simulations of dilute sedimenting suspensions at finite-particle Reynolds numbers. Phys.
Fluids 24, 123303 (2012)
27. Song, F., Wang, W., Li, J.: A lattice Boltzmann method for particlefluid two-phase flow. Chem. Eng. Sci. 102, 442
450 (2013)
28. Cundall, P.A., Strack, O.D.: A discrete numerical model for granular assemblies. Geotechnique 29, 4765 (1979)
29. Qian, Y., dHumires, D., Lallemand, P.: Lattice BGK models for NavierStokes equation. Europhys. Lett. 17, 479
484 (1992)
30. Guo, Z., Zheng, C., Shi, B.: Discrete lattice effects on the forcing term in the lattice Boltzmann method. Phys. Rev.
E 65, 046308 (2002)
31. Wang, J., van der Hoef, M.A., Kuipers, J.A.M.: CFD study of the minimum bubbling velocity of Geldart A particles in
gas-fluidized beds. Chem. Eng. Sci. 65, 37723785 (2010)
32. Gidaspow, D.: Multiphase Flow and Fluidization: Continuum and Kinetic Theory Descriptions. Elsevier, Amsterdam (1994)

Você também pode gostar