23rd ABCM International Congress of Mechanical Engineering

December 6-11, 2015, Rio de Janeiro, RJ, Brazil

NUMERICAL STUDY OF A BUBBLING FLUIDIZED BED USING A

COMBINED CFD-DEM CODE

Marcelo José Alba

Universidade Tecnológica Federal do Paraná / Universidade Estadual de Maringá
[email protected]

Flávio Clareth Colman

Universidade Estadual de Maringá
[email protected]

Alexandre Marconi de Souza da Costa

Universidade Estadual de Maringá
[email protected]

Paulo Roberto Paraíso

Universidade Estadual de Maringá
[email protected]

Luiz Mário de Matos Jorge

Universidade Estadual de Maringá
[email protected]

Abstract. The purpose of these instructions is to serve as a guide for formatting papers to be published in the Proceedings
of the XXII COBEM. The abstract should describe the objectives, the methodology and the main conclusions of the paper
in about 200 words. It should contain neither formulae nor reference to bibliography. The abstract will be included in a
printed volume to be distributed to the symposium participants, whilst the full paper will be published in the proceedings.
Fluidized beds are used in industries in the separation, mixing, drying and combustion operations. This work analyzed
the behavior of a fluidized bed with the open-source MFIX-DEM code developed by the National Energy Technology
Laboratory (NETL-USA). For the code, the gas and solid phases are treated in an Eulerian-Lagrangian framework. For
the gas-phase, continuity and momentum conservation equations was solved. For the particles, or the solid-phase, was
used the Discrete Element Method (DEM), which describes the contact and collisions phenomenon's from the spring-
dashpot-friction concepts, by means of an explicit numerical approach. A two-dimensional grid with gas injection was
used through a distributor at the bottom of the bed. The granular material analyzed was the soybeans. The gas velocity
was varied to determine the influence of pressure drop within the bed. The results obtained by the different drag
relationships were compared with the fluidization curves from the experimental data.

Keywords: fluidized bed, soybean, MFIX-DEM

1. INTRODUCTION

Fluidized bed systems are extensively employed in several industries such as: agroindustry, chemical and
pharmaceutical (KUNII and LEVENSPIEL, 1991). Generally, fluidized systems allow significant improvements in heat
and mass exchanges, with associated space and monetary revenues. Nowadays, a great portion gas-solids system studies
takes advantage of easily available computational resources to predict the heat and mass transfer behavior of fluidized
beds.
For the case of soybean grains, fluidized and spouted beds equipment are used to transfer heat through hot air injection.
The heat transfer is accompanied by a noticeable moisture reduction Soponronnarit et al. (2001). This process is important
for a prolonged storage of the product, avoiding microorganism wastage.
Spouted beds are the preferred fluidization routine for soybean grains, as the grains are classified in the Geldart D
group. In addition, fluidized bed drying is recognized as a fast drying technology, due to the large air to product contact
area achieved relative to a static bed caused by fluidization of the product, and the high air speed and high temperatures
used (SOPONRONNARIT et al., 2001). Soybean grains are particularly feasible for studies using the Discrete Element
Method (DEM) (TSUJI, et al., 1993). This feasibility resides in the circularity and sphericity properties, with great
uniformity in the particle diameter.
M.J. Alba, F.C. Colman, A.M.S. Costa, P.R. Paraíso and L.M.M. Jorge
Numerical Study of a Bubbling Fluidized Bed Using a Combined CFD-DEM Code

The CFD (Computational Fluid Dynamics) methodology when applied to research, development and optimization in
the process industries can lead to relevant cost reduction. This economies are associated to reducing the number of
physical prototyping. According to GIDASPOW, 1994, rarely are reported in terms of detailed economical figures.
Traditionally, two approaches are followed when study multiphase flow phenomena with CFD: Euler–Lagrange and
the Euler–Euler. The first, which involves the balance of forces at work upon the particle, requires considerable
computational effort. The Euler–Euler approach considers the dispersed phase as a continuous phase and is based on the
Navier–Stokes equations applied to each phase (WEN and YU, 1966; ANDERSON and JACKSON, 1967; SYAMLAL
and O’BRIEN, 1989; GIDASPOW, et al., 1992; TAGHIPOUR, et al., 2005; DUARTE, et al., 2009; ASEGEHEGN, et
al., 2011).
An important aspect which influences the accuracy of CFD results for fluidized beds is the methodology used to
extract the data. The data extraction methodology and the subsequent results, such as bed properties, bubble characteristics
and bed expansion, suffers from great variation between studies reported in the literature (ASEGEHEGN, et al., 2011).
Also, according to ASEGEHEGN, et al., 2011, data extraction can have as much influence as the use of different
constitutive relationships. For instance, HULME et al. (2005) has shown that different volumetric solids fraction at the
inlet led to different bubble average parameters. In the literature is also show different ways to define bed expansion ratio
(LOFSTRAND, et al., 1995; GELDART, 2004; TAGHIPOUR, et al., 2005).
Another important parameter that influences the simulation results is the time span used for extracting average
quantities. PATIL, et al., (2005), demonstrated that for different simulation time led to different bubble diameters. The
time span depends on parameters, such as bed geometry and superficial velocity, therefore becoming difficult provide a
general rule for all operating conditions. On the other hand, greater simulation times yields good results at greater
computational expenses.
One of most relevant parameters for fluidized systems is the minimum fluidization velocity (umf). According to Law
e Mujumbar (2006), fluidized bed operate normally at gas velocities to 2-4 times umf.
In this study we revisit the gas fluidization phenomena for a bed constituted from soybean grains. Three different bed
height configurations will be used with two different drag term correlations. The CFD code employed is the MFIX-DEM
(Multiphase Flow with Interphase eXchange coupled with Discrete Element Method). The results will be compared with
experimental data from Soponronnarit et al. (2001).

2. DESCRIPTION OF THE MODEL

MFIX (Multiphase Flow with Interphase eXchanges) is an open source CFD code developed at the National Energy
Technology Laboratory (NETL) for describing the hydrodynamics, heat transfer and chemical reactions in fluid-solids
systems. It has been used for describing bubbling and circulating fluidized beds, spouted beds and gasifiers. MFIX
calculations give transient data on the three-dimensional distribution of pressure, velocity, temperature, and species mass
fractions.
Figure 1 shows the boundary conditions for the numerical simulations. The pressure measurement points were located
at y = 0.0 m and e y = 1.0 m. The simulations were based on the experiments by Soponronnarit et al. (2001) for a soybean
fluidized bed, operating at three different initial heights : 0.08 m, 0.115 m e 0.15 m. The mesh was two-dimensional
cartesian with 20 x 100 cells. Uniformity in the inlet conditions were used for 13 different velocities. The particle diameter
was 0.007 m and its density 1200 kg/m3.

Figure 1. Malha do leito fluidizado baseado no modelo de Soponronnarit et al. (2001)

 23rd ABCM International Congress of Mechanical Engineering
December 6-11, 2015, Rio de Janeiro, RJ, Brazil

According to TSUJI (2007), particulate system flow can be classified in 3 types : (1) Collisionless ; (2) Collision
dominated; (3) prolonged contact dominated. The third is subdivided into : fluid absent, fluid present. In this way,
fluidized bed flows are classified as 3rd type with presence of fluid.
A second classification by Tsuji (2007) is based on the scales involved in the interaction of fluid and particles, namely:
micro, meso and macro. The task of identify boundaries between scales is left to the researcher, but in general, macro
models rely on empiricism, whereas micromodels have a sounding theoretical foundation. For this study was employed
the E method (micro-particle and meso-fluid) formulated by Tsuji (2007). In this method the particles are tracked
(Lagrangian approach) and the fluid is treated in terms of local averaging (Eulerian approach).

2.1 Discrete Element Method (DEM)

DEM, also known as distinct element method, was proposed by Cundall e Strack (1979) for disks and spheres. Based
on explicit numerical scheme, the method monitors the contact between particles and their resulting movement. This
method is also known as soft-sphere.
In their original work, Cundall and Strack (1979), compared the distinct element method with experiments. The
experiments were conducted with optically sensible materials, which allow an accurate determination of contact forces,
displacements and rotations, with a great computational cost. Their main conclusion was that the DEM is a valid tool for
granular systems. Actually, a great number of open and commercial codes refers to the DEM in describing the dynamics
of system with colliding particles, among them fluidized systems. In the DEM, collision and contacts are modeled as a
mechanical systems consisting of spring-slider and damper.
Here, material properties such as friction coefficient, spring and damper constant were selected based on the work of
Boac et al. (2009). Table 1 reports the granular and gas properties employed in the simulations.
The simulation were run for inlet gas velocities ranging from 0.45 m/s to 3.5 m/s with 13 uniform increments. For
each velocity the simulation time was 3 s.

Table 1. Physical properties, experimental conditions and initial and boundary conditions used in CFD-DEM
simulations.

Properties Literature values Choice values

Particle-particle 0.267; 0.55(1) 0.45
Static friction coefficient, μs
Particle-wall(4) 0.30(1) 0.30
Particle-particle 0.5; 0.7(1) 0.6
Particle Restitution Coefficient, e
Particle-wall(4) 0.6; 0.7(1) 0.6
Particle diameter, dp (x10-3 m) 7(3)
Particle Length (x10-3 m), l 7.0 – 8.2(1) --
Particle Width (x10-3 m), w 6.1 – 6.7(1) --
Particle Thickness (x10-3 m), t 5.5 – 5.9(1) --
Density of the solid, ρs, (kg/m3) 1130 – 1325.2(1) 1200
Bed height of particles, hbed (x10-2 m) 8.0; 11.5; 15.0(2) 8.0; 11.5; 15.0
Bed height (x10-2 m) 100(2) 100
Bed width (x10-2 m) 20(2) 20
Density of the gas (air), ρg (kg/m3) 1.205
Range of inlet air velocity, vair (m/s) 0.45 – 3.5(2) 0.45 – 3.5
(1) Boacet al. (2009) (3) Considered spherical
(2) Soponronnarit et al. (2001) (4) Acrylic

In the beginning of the runs the particles were randomly seeded in the bed interior without touching each other. When
the simulation started, under gravity force, the particles settled. This initial bed height corresponded to the heights reported
in the Soponronnarit et al. (2001) work.

2.2 Governing Equations: Gas phase

In MFIX–DEM, the gas-phase governing equations for mass and momentum conservation are similar to those in
traditional gas-phase CFD but with additional coupling terms due to drag from the solids-phase. The solids-phase is
modeled using discrete particles. The governing equations, implemented in MFIX (Syamlal, et al., 1993), for the gas-
phase continuity and momentum conservation, are:

   + ∇   
 = 0 (1)

  
M.J. Alba, F.C. Colman, A.M.S. Costa, P.R. Paraíso and L.M.M. Jorge
Numerical Study of a Bubbling Fluidized Bed Using a Combined CFD-DEM Code



 =  • 
     −  
 +    (2)
   


where,  is the gas-phase volume fraction,  is the density of gas-phase, 


 is the volume-averaged gas-phase velocity,
 is the momentum transfer term between the gas and the m solid phase, and 
th 
 is the stress tensor in gas-phase given
by:

̿ = −  ̿ + ̿ (3)

where  is the gas-phase pressure. The stress tensor in the viscous regime, ̿ , is assumed with Newtonian and given by:

" + # $
̿ = 2!  "  ̿ (4)

The tensor  ̿ can be identified like the deformation rate to the gas-phase 
" , given by:

1 '
 = &∇  + ∇   ( (5)
2

2.3 Governing Equations: Solid phase

In the DEM approach, the mth solid–phase is represented by Nm spherical particles with each particle having diameter
Dm and density ρsm. Solid phases are differentiated based according to radii and densities. For total of M solid phases, the
total number of particles is equal to
*

) =  ) (6)


These N particles are represented in a Lagrangian frame of reference at time t by:

+ ,-. ,., 0 ,-. ,., 1 ,-. ,., ,-. ,  ,-. 2 = 1 … . . ) (7)

where + ,-. ,. denotes the ith particle position, 0 ,-. ,. e 1 ,-. ,. denote its linear and angular velocities,  ,-. denotes
its diameter, and  ,-. represents its density.
The position, the angular and linear velocities of the ith particle evolve according to Newton’s laws as:

5+ ,-. ,.
= 0 ,-. ,. (8)
5

50 ,-. ,. ,-.

6 = 7' (9)
5
,-. ,- ∈:,.
7' = 6-  + 78 ,. + 7;,-. ,. (10)

5< ,-. ,.

- = =- (11)
5
,-
where  is the gravity acceleration, 7> ∈:,. is the total drag force on ith particle residing in kth cell and belonging to the
,-.
mth solid-phase, 7? is the net contact force acting as a result of contact with other particles, = - is the sum of all torques
,-.
acting on the ith particle, and 7@ is the net sum of all forces acting on the ith particle.
A advantage at DEM is the manner like it is solve the equations to the solid phase, taking to explicit treatment of
particle-particle collisions. To solid-gas flows at MFIX-DEM is employed the Cundall and Strack (1979) model, also
know like soft-sphere model.
To the drag force calculation, is used the follow model:
 23rd ABCM International Congress of Mechanical Engineering
December 6-11, 2015, Rio de Janeiro, RJ, Brazil

,D.
,- ∈:,. βF EF
7> = −∇PB ,xD .EF + JvB ,xD . − vIF ,xD .L (12)
εIF

where Pg(xk) and vg(xk) are the gas–phase mean pressure Pg and velocity vg fields at the kth particle location, E is the
,:.
particle volume, and M is the local gas-solid momentum transfer coefficient for kth cell.
,:.
The M represent the drag coefficient with different models developed and can be write by:
,∀- ∈:. ,D.
MF = βF = ρF , DF , QvIF ,xD . − vB ,xD .Q, ρB , !  (13)

where x: is the center of the kth cell, vB + ,-.  is the gas-phase mean velocity in the + ,-. position, where a cell-centered
value of vB is used. Similarly, vIF is a local cell average velocity of the mth solid-phase.
If considered a approximation of the constant drag force for all particles in a particular cell, the gas-solid momentum
,:.
transfer BF is estimated to the kth cell by

,:. ,D.
BF = εIF ∇PB ,xD . + βF JvB ,xD . − vIF ,xD .L (14)

3. RESULTS

The results for pressure loss were obtained from the post processing command line post_mfix utility available in the
MFIX code. Figures 1, 2 and 3, were generated with Paraview software (2015), and depict the volumetric fields and
particle positions for beds with different initial height.

(a) (b) (c) (d) (e)

Figure 3. First bubble formation after velocity change with hbed = 0.080 m: Gidaspow (above) vs. Syamlal (bellow). Gas
velocity: (a) 2.1 m/s, (b) 2.2 m/s, (c) 2.3 m/s, (d) 2.8 m/s and (e) 3.5 m/s

Concerning to the bed hydrodynamics, it can be seen that Gidaspow drag, corresponds to a slightly greater bubble
formation in the central region. Supplementally, the influence of turbulence is more highlighted for this drag model. This
can be noticed by the horizontal lining up in the regions where the gas voidage is in the 0.6 -0.9 range. These phenomena
are better depicted when comparing the figures for velocities near the minimum fluidization (a, b and c, Figures 1 and 2).
M.J. Alba, F.C. Colman, A.M.S. Costa, P.R. Paraíso and L.M.M. Jorge
Numerical Study of a Bubbling Fluidized Bed Using a Combined CFD-DEM Code

(a) (b) (c) (d) (e)

Figure 2. First bubble formation after velocity change with hbed = 0.115 m: Gidaspow (above) vs. Syamlal (bellow). Gas
velocity: (a) 2.1 m/s, (b) 2.2 m/s, (c) 2.3 m/s, (d) 2.8 m/s and (e) 3.5 m/s

From analysis of Fig 3d e Fig. 3e is noticed a greater dense bed height when the Gidaspow drag is employed.

(a) (b) (c) (d) (e)

Figure 3. First bubble formation after velocity change with hbed = 0.150 m: Gidaspow (above) vs. Syamlal (bellow). Gas
velocity: (a) 2.1 m/s, (b) 2.2 m/s, (c) 2.3 m/s, (d) 2.8 m/s and (e) 3.5 m/s

Figure 4 and Figure 5 show the bed pressure drop versus air superficial velocity. Gray dotted lines corresponds to
experimental data from Soponronnarit et al. (2001). Continuous lines corresponds to time averaged values extracted from
CFD-DEM simulations for each air superficial velocity. It was verified that an increase in bed height correlates with a
increase in pressure drop.
 23rd ABCM International Congress of Mechanical Engineering
December 6-11, 2015, Rio de Janeiro, RJ, Brazil

1.200,00 1.200,00
h=0.080 m - Numerical h=0.080 m - Numerical
h=0.115 m - Numerical h=0.115 m - Numerical
1.000,00 1.000,00
h=0.150 m - Numerical h=0.150 m - Numerical
h=0.080 m - Experimental h=0.080 m - Experimental
800,00 h=0.115 m - Experimental 800,00 h=0.115 m - Experimental
Pressure Drop (Pa)

Pressure Drop (Pa)

h=0.150 m - Experimental h=0.150 m - Experimental

600,00 600,00

400,00 400,00

200,00 200,00

0,00 0,00
0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 4,00 0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 4,00
Air velocity (m/s) Air velocity (m/s)

(a) (b)
Figure 4. Pressure drop versus inlet air velocity diagram: (a) Gidaspow and (b) Syamlal

Besides the noticeable influence of the drag models used in the CFD-DEM simulations on the particles dynamics
(Fig. 1, Fig. 2 and Fig. 3), it is verified that the pressure drop is not significantly affected (Fig. 5). For the bed with 0.150 m
height analyzed with Syamlal drag, the minimum fluidization velocity was 1.9 m/s. For all the other cases, the simulation
results predicted a minimum fluidization velocity of 2 m/s.
1.200,00
h=0.080 m - Syamlal
h=0.115 m - Syamlal
h=0.150 m - Syamlal
1.000,00 h=0.080 m - Gidaspow
h=0.115 m - Gidaspow
h=0.150 m - Gidaspow
800,00 h=0.080 m - Experimental
Pressure Drop (Pa)

h=0.115 m - Experimental
h=0.150 m - Experimental

600,00

400,00

200,00

0,00
0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 4,00
Air velocity (m/s)
Figure 5. Pressure drop versus inlet air velocity diagram: Gidaspow versus Syamlal

Figure 5 shows that only for the bed with 0.150 m height there is a deviation in the curves that shows the pressure
drop. Regarding the greater differences for pressure drop between numerical and experimental values for greater gas
velocities, this can attributed to a stronger interaction between particles, that can lead a more fluctuating pressure value.

4. CONCLUSIONS

The CFD-DEM coupled method yields good results for predicting the minimum fluidization velocity for the studied
cases, where numerical values are slightly higher compared to the Soponronnarit et al. (2001) experimental values, 2.0
m/s versus 1.9 m/s, respectively.
The differences for pressure loss between numerical and experimental values are more relevant; however, the curve
trends are analogous. A identified shortcoming resides in the several parameters of DEM were extracted from literature,
not being granted to represent the solids particles.
For future work, different drag models as well the assembly of a new experimental rig are being planned.

5. ACKNOWLEDGEMENTS

The authors would like to thank CAPES and Araucária Foundation for financial support and the National
Supercomputing Center (CESUP), of the Federal University of Rio Grande do Sul (UFRGS) for the availability of
resources and technology for the development of this research.
M.J. Alba, F.C. Colman, A.M.S. Costa, P.R. Paraíso and L.M.M. Jorge
Numerical Study of a Bubbling Fluidized Bed Using a Combined CFD-DEM Code

6. REFERENCES

Anderson, T.B., Jackson, R., 1967. A Fluid Mechanical Description of Fluidized Beds Equations of Motion, Industrial e
Engineering Chemistry Fundamentals, Vol. 6 (4), p. 527-539
Asegehegn, T. W., Krautz, H. J. Schreiber, M., 2011. Numerical Study of a Gás-Sólid Fluidized Beds Hydrodynamics:
Influence of Immersed Horizontal Tubes and Data Analysis. International Journal of Chemical Reactor Engineering,
Vol. 9 (1), p. 1-39.
Boac, J.M., Casada, M.E., Maghirang, R.G., Harner, J.P, 2009. Material and Interaction Properties of Selected Grains
and Oilseeds for Modeling Discrete Particles. ASABE Annual International Meeting, Paper Number 09-7166.
Duarte, C.R., Olazar, M., Murata, V.V., Barrozo, M.A.S. 2009. Numerical simulation and experimental study of fluid–
particle flows in a spouted bed. Powder Technology, Vol. 188, p. 195-205.
Geldart, D., 2004. Expansion of Gas Fluidized Beds. Industrial e Engineering Chemistry Research, Vol. 43 (18), p. 5802-
5809.
Gidaspow, D., 1994. Multiphase Flow and Fluidization: Continuum and Kinetic Theory Descriptions. Academic Press,
Boston.
Gidaspow, D., Bezburuah, R., Ding, J., 1992. Hydrodynamics of circulating fluidized beds, kinetic theory approach in
fluidization, Proceedings of the 7th Engineering Foundation Conference on Fluidization, p. 75-82.
Hulme, I., Clavelle, E., Van Der Lee, L., Kantzas, A., 2005. CFD Modeling and Validation of Bubble Properties for a
Bubbling Fluidized Bed. Industrial e Engineering Chemistry Research, Vol. 44 (12), p. 4254-4266.
Kunii, D., Levenspiel, O. 1991. Fluidization Engineering. John Wiley & Sons: New York, p. 491.
Law, C. L., Mujumdar, A. S., 2006. Fluid Bed Dryers. In Handbook of Industrial Drying. Taylor & Francis. Cap. 8, p
174-199.
Lofstrand, H., Almsted, A. E.,Anderson, S., 1995. Dimensionless Expansion Model for Bubbling Fluidized Beds with
and without Internal Heat Exchanger Tubes. Chemical Engineering Science, Vol. 50 (2), p. 245-253.
Soponronnarit, S., Swasdisevi, T., Wetchacama, S., Wutiwiwatchai, W., 2001. Fluidised bed drying of soybeans. Journal
of Stored Products Research, Vol. 37, p. 133-151.
Syamlal, M., O'Brien, T.J., 1989. Computer simulation of bubbles in a fluidized bed. AIChE Symp. Ser., Vol. 85, p. 22–
31.
Tsuji, Y., Kawaguchi, T. and Tanaka, T., 1993. Discrete particle simulation of two-dimensional fluidized bed. Powder
Technology, Vol. 77, p. 79-87.
Taghipour, F., Ellis, N., Wong, C., 2005, Experimental and computational study of gas–solid fluidized bed
hydrodynamics, Chem. Eng. Sci., Vol. 60, p. 6857-6867.
Wen, C.Y., Yu, Y.H., 1966. Mechanics of fluidization, Chem. Eng. Prog. Symp. Ser., Vol. 62, p. 100–111.

7. RESPONSIBILITY NOTICE

The authors are the only responsible for the printed material included in this paper.