Paper No 1956
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Modeling fluidization in biomass gasification processes
Emmanuela Gavi*, Theodore J. Heindelt, Rodney O. Fox*
*Department of Chemical and Biological Engineering, Iowa State University, Ames, IA, USA 50011
Department of Mechanical Engineering, Iowa State University, Ames, IA, USA 50011
egavi@iiastate.edu, theindel@iiastate.edu, rofox@iiastate.edu
Keywords: fluidization, biomass gasification, segregation, binary particles, CFD
Abstract
Extensive validation of computational fluid dynamics (CFD) models is required when modeling biomass fluidization, because
several required model inputs are not known or not easily measured experimentally for biomass. In the present work, CFD fluidization
modeling of a biomass bed is validated by comparison with Xray computed tomography experimental data. A parametric study was carried
out by employing ground walnut shell or ground corncob as model biomass bed materials, and fluidization was performed at a gas velocity
twice the minimum fluidization velocity (Ug = 2Umf). An important result is the use of an "effective density" for biomass in the CFD model,
the use of which is necessary because the biomass particles can be characterized by an irregular shape and some degree of porosity, whereas
CFD models assume the biomass particles to be solid spheres. If the midvalue of the density range provided by the manufacturer is
employed, the bulk density of the solid phase in the bed is overestimated. It was observed that the bed height was well predicted with a
coefficient of restitution (COR) equal to 0.9 for a sphericity range of y = 0.81. For smaller y, the predicted bed height was higher,
consistent with the results of Deza et al. (2008). The results also suggest that the value of COR has a negligible effect on the predictions for
biomass systems.
Introduction
The number of studies on biomass fluidization has
increased in the last decade because of the interest in
biomass thermal conversion to energy through gasification
processes. However, the majority of these studies are of
experimental nature, and only a few modeling works are
available in literature. Among the few works present in
literature that explicitly deal with biomass modeling, only
one study by one of the authors of the present work (Deza
et al., 2008) is aimed at validating fluidization models with
experimental data. This validation step is of great
importance for the subsequent mixing and reaction
modeling, since the fluidization models available in the
literature have been developed and validated with only
standard particles, such as monodispersed dry particles
shaped as spheres, cylinders, discs and spheroids. In Deza
et al. (2008), a modeling study of the fluid dynamics of
biomass was conducted on a twodimensional bubbling
fluidized bed with the MFIX solver. The Gidaspow drag
model was used and the two coefficients that describe the
deviation of the biomass particle's behavior from standard
particles, namely the coefficient of restitution (COR) and
the sphericity coefficient, were varied. Deza et al. (2008)
found that the COR does not greatly affect the bed fluid
dynamics; however, the sphericity coefficient plays an
important role. Finally, for a bed composed of ground
walnut shell, they suggested a large COR (z0.85) and a
low sphericity (;0.6) is needed.
The aim of this study is to validate computational fluid
dynamics (CFD) simulations carried out with ANSYS
Fluent in a biomassfilled fluidized bed. A previous work
assessed the validity of the ANSYS Fluent solver and the
implemented drag models in a fluidized bed of glass beads
(Min et al., 2010). The present work takes a step forward
and studies the fluid dynamic behavior of a single
component biomass fluidized bed constituted by ground
walnut shell (GWS) or ground corncob (CCB), both
Geldart B particles. Experimental data for validation are
obtained by comparison with Xray computed tomography
(CT).
The paper is organized as follows. First a short review on
fluidization theory is presented. Then operating conditions
and numerical details are reported. Finally, results are
summarized and some conclusions are drawn.
Nomenclature
D Diameter of bed (m)
e,, Coefficient of restitution
F g Drag coefficient
h Bed height (m)
I Interaction term
M Mass (kg)
P Pressure (N m2)
S Stress tensor
u Velocity (m s')
V Volume (m3)
Greek letters
0 Sphericity coefficient
e Void fraction
p Density (kg m 3)
Paper No 1956
Subscripts
bed Bed
corr Correction
eff Effective
g Gas phase
IN Initial
sa Solid phase alpha
Fluidization Theory
In the EulerianEulerian multifluid model approach, the
gas and solid phases are treated as interpenetrating
continue as explained in Syamlal et al. (1993). The sum of
their volume fractions must sum to one:
N
6g +YeG =1
9=l
The subscript g stands for the gas phase, whereas the
subscript sl stands for the 2th solid phase.
The continuity equation for the gas phase is
S(,gpg)+ V (ggug) 0, (2)
whereas for each solid phase it is
a 0
v,(^ o,)+V(E (3)
in which pg and ps, are the densities, and ug and us, are the
velocities of the gas and ath solid phases, respectively.
The momentum balance equation for gas and ath solid
phase are the following:
a ( cgpgug)+V. (puu g)u = VS
at
N
Z Ig + g, pg,
(4a)
S(EpsaPsau ) + V (apsausa sa) V Sa +Iga
at
N
I ZI +SsaPsag,
(4b)
where Sg is the gas stress tensor, IgX is the term
representing the interaction between the gas and the 2th
solid phase, Ssa is the solids stress tensor, Ig is the term
representing the interaction of the ath solid phase with the
gas phase, and finally I, represents the interaction of the
ath solid phase with the 2th solid phase. In this work,
only one solid phase will be treated, therefore the term
Ia is not further defined here.
Table 1 summarizes the constitutive equations for the gas
and solid stress tensor for a single solid phase as they are
defined in ANSYS Fluent 6.3 (Fluent Inc., 2006). A simple
Newtonian closure is used for the gas stress tensor while
the kinetic theory of granular flows is employed to
calculate the solid stress tensor. Note that MFIX describes
the shearing granular flows by combining the viscous and
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
plastic flow regimes by introducing a switch at a critical
packing Sg* (Syamlal et al., 1993), whereas ANSYS Fluent
adopts the approach by Johnson and Jackson (1987), that
combines the two theories by adding the two formulas.
The interaction between gas and solid phase Ig, is defined
as
ga = CVP Pg F (us ug) (5)
in which the first term on the righthand side is the
buoyancy force and the second term is the drag force, with
the drag coefficient Fg,.
In the present work, the drag model derived by Gidaspow
(1994) is employed, in which for eg 0.8 the Ergun
equation is used, and for g > 0.8 the Wen and Yu
(1) equation is used:
ss(1 )/g sa glPg UUsa
150 s 2 +1.75
d2 d
a p pa
if <0.8
F =a
3D sagPgg Ug sa l 265
4D d P
pa
if s > 0.8
(6)
in which dpa is the size of the solid particle and CD is a
parameter defined as follows:
24(1+ 0.15Re0687)
C, = (7)
D Re
dpug us g
with Reg = 
V
g
The fluidization model requires two pieces of information
about the particle fluid dynamics that are not known nor
easily measurable experimentally for biomass particles: the
COR (ea,) and the sphericity coefficient 4. The COR
indicates the degree of elasticity of the collisions between
solid particles, so that e,, = 1 indicates perfectly elastic
collisions, and e,, = 0 indicates perfectly inelastic
collisions. The sphericity coefficient was introduced to
approximate biomass particle shape relative to a sphere. It
is defined as the ratio between the surface area of a sphere
with the same volume of the biomass particle, and the
surface area of the biomass particle:
S
sphere
Sbiomass
Sphericity in the range 0.8 < q <1 indicates a biomass
particle that is approximately isometrical. Sphericities
<<0.8 and q <0.5 indicate flat particles and
extremely flat particles, respectively (Cui and Grace,
2007).
Paper No 1956
Numerical Scheme
CFD simulations of single component biomass bed
fluidization were completed with the commercial software
ANSYS Fluent 6.3.26 on a Linux platform.
Twodimensional (2D) and threedimensional (3D)
simulations were run. The hexahedral cells size is 4 mm
and this leads to a grid size equal to 2888 for the 2D grid
and 59400 for the 3D grid. Timedependent simulations are
carried out with a time step equal to 104 s. The simulations
are solved for 10 s to allow for startup transients to die
down, and then the subsequent 60 s are used for time
averaging, by sampling every 10 time steps. The
computational time required to run a 70 s flow time
simulation is approximately 10 days for a 2D simulation on
8 processors and 48 days for a 3D simulation on 8
processors on the high performance computing (HPC)
machine at Iowa State University.
Ground walnut shell (GWS) or ground corncob (CCB)
provided easily obtainable model biomass systems and was
fluidized with a fluidization gas velocity Ug 2Umfand
no side gas injection. The details of the simulation set up,
in terms of chosen fluidization models, numerical scheme
details and case set up details are reported in Tables 2, 3
and 4, respectively.
This work focuses on developing an approach for the
biomass fluidization simulation. From this point on the
developed approach will be indicated as NEW. The
approach that is the current standard in fluidized bed
modeling assumes biomass particles to be spherical and
nonporous and will be identified in this work as STD.
The simulations that were performed are grouped into
three sets and summarized in Table 5.
Results and Discussion
Simulations were run on ANSYS Fluent with the standard
approach for the 3D grid for GWS and CCB beds with
different COR and sphericity (set 1). The results confirm
those found by Deza et al. (2008), that the COR does not
influence strongly the fluidization results, whereas the
sphericity coefficient plays an important role in the
fluidization. The results are reported for GWS, COR=0.9
and sphericity 0.6 and 0.9 in Fig. 1. Note the experimental
data in Fig. 1 and subsequent figures were obtained using
Xray computed tomography imagining of similar
coldflow fluidized beds. Details of these experiments can
be found in Min et al (2010), Franka and Heindel (2009),
and Drake and Heindel (2009).
In the STD approach, the density of the particles is set to
be equal to the nominal density of the biomass material
given by the manufacturer, and the solid packing limit is
set to 0.63. The initialization is performed with the
experimental observed packing (or bulk density) on the
theoretical bed height. It can be observed in Fig. 1 that
even though the bed height is well predicted with a
sphericity of 0.6, the void fraction is underestimated.
Because the drag models available in the literature were
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
derived for regular and solid particles, whereas biomass
particles can be irregularly shaped and porous, some
additional parameters need to be considered in the drag
model. The sphericity is not sufficient, as it only regulates
the transition of the particle shape from isometric to flat,
while the biomass particles considered here, though not
regularly shaped, are not generally flat but isometric. The
high drag and the low packing experimentally observed
can be caused by the porosity of the GWS and CCB
biomass particles employed. For these reasons the use of a
correction and an effective density is suggested here. The
effective density is computed for each material from the
experimental bed mass and volume, assuming the packing
limit of biomass particles to be equal to the experimental
packing limit of glass beads (e = 0.58):
Psa = Psa,eff = bed (9)
Vbed
The initialization of the simulations is performed by
employing a bulk density computed from the effective
density and a solid packing that is slightly lower than the
solid packing limit (Es, = 0.55) to facilitate the onset of
fluidization. The bed height is therefore raised in order to
introduce the correct amount of mass (h,A = 0.165m).
With the introduction of the effective density the correct
drag force is obtained; however, simulations still consider
the biomass particles as nonporous spheres.
In order to compare simulation results with experimental
data, it is necessary to apply a further correction, because
Xray CT experiments consider the gas phase present in
the particle pores and at the interstices between
nonspherical particles as bed void fraction. Therefore the
corrected simulated bed void fraction is calculated as
( Psanom
gg,corr 1 Psa,nom sa E (10)
Psa,eff )
The first set of simulations (set 1) is performed with the
NEW approach in 2D and is aimed at finding the best set
of parameters for GWS and CCB in terms of COR and
sphericity. The results are reported in Figs. 2 and 3. In the
graphs relative to GWS, for which two CORs were
employed, it can be observed that this parameter is not
determining for the simulation results.
The sphericity instead largely affects the results, both for
GWS and CCB. A good agreement with experiments in the
bulk of the bed is given by a sphericity of one. The
predicted bed expansion appears to be too large, however it
must be reminded that these are 2D simulations, and
therefore they have one degree of freedom less. This issue
is solved, as expected, in 3D simulations (set 2), the results
of which are shown in Figs. 4 and 5. Here the comparison
with the STD approach is also reported. With the NEW
approach a considerable improvement is obtained as
experiments and simulation results are in reasonable
agreement, both in terms of bed height and void fraction. It
can also be observed that small changes in the sphericity
coefficient do not strongly modify the results Therefore it
can be concluded that the best way to model biomass
fluidization is to employ the NEW approach with
Paper No 1956
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
COR=0.9 and sphericity 1.0.
Additional confirmation of these findings can be found in
the following validation of simulation results, carried out
along diameter lines collocated at different heights in the
reactor (h/D=0.25, 0.50, 0.75 and 1). These results are
shown in Figs. 6 and 7, where also the STD approach
results are reported. The NEW approach with coefficient of
restitution 0.9 and sphericity 1.0 gives results that are in
better agreement with experimental data, especially for
intermediate heights where the differences from the STD
approach are more evident.
A final comparison is reported in Figs. 8 and 9, in terms of
void fraction contour plots. The experimental
measurements, the results with the NEW approach with
coefficient of restitution 0.9 and sphericity 1.0, and the
results with the STD approach with coefficient of
restitution 0.9 and sphericity 0.6 and 1.0, are reported on
the two xz and xy planes. The bed appears defluidized with
the STD approach; even with sphericity 0.6 the correct bed
height is predicted. The best overall agreement is given by
the NEW approach, which is able to predict a void fraction
in reasonable agreement with experiments at most bed
heights and radial positions.
Conclusions
The modeling of the fluidization dynamics of a bed of
biomass material was treated, by employing the
commercial software ANSYS Fluent and a fluidization
model available in the literature. Experimental data were
used to validate the simulation results and it was found that
in order to predict the correct bed height and void fraction
in the bed, a correction was needed to account for the
irregularity and porosity of biomass particles (in this work
ground walnut shell and ground corncob). Therefore, an
effective density was used instead of the nominal density
of the biomass material, the initialization of the calculation
was performed in order to introduce the correct amount of
mass in the bed, and finally a correction term was used to
allow the simulation data to be compared with experiments.
With these adjustments to the model, a parametric study
was completed to identify the most appropriate coefficient
of restitution and sphericity, and it was found that good
agreement with experimental data is obtained with a
sphericity of 1.0 and COR=0.9.
0.6 0.7 0.8 0.9
void fraction, 
Figure 1: Time and planeaveraged void fraction as a
function of bed height for GWS (2D simulations with STD
approach). Symbols: experiments. Lines: simulations
(green 6 = 1.0, em = 0.9; cyan 6 = 0.9, e = 0.9).
0.5 0.6 0.7 0.8 0.9 1
[]
Figure 2: Time and planeaveraged void fraction as a
function of bed height for GWS (2D simulations with
NEW approach): symbols experiments; green 4 = 0.6, ea
= 0.9; cyan = 0.6, e< = 0.7; red = 1.0, e. = 0.9; purple
S= 1.0, e.= 0.7.
2.0
1.5
d 1.0o
..
Acknowledgements
This project was funded by the ConocoPhillips Company.
0.5 0.6 0.7 0.8 0.9 1
E,[]
Figure 3: Time and planeaveraged void fraction as a
function of bed height for CCB (2D simulations with NEW
approach): symbols experiments; green = 0.6, ea = 0.9;
red ) = 1.0, em = 0.9.
Paper No 1956
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
J0.
0.6
0.4'
It
0.8
JM
0.d
0.6 0.7 0.8 0.9
void fraction, 
h/D=0.25
P^i^
h/D=0.75
0.4'
5 0 5
position, cm
h/D=0.50
P" /
h/D=1.00
5 0 5
position, cm
Figure 4: Time and planeaveraged void fraction as a
function of bed height for GWS (3D simulations).
Experiments compared with simulations (COR=0.9):
symbols experiments; green STD approach =1.0; cyan
STD approach =0.6; red NEW approach =1.0; blue
NEW approach =0.9.
Figure 6: Average void fraction along diameter lines at
four different bed heights for GWS (3D simulations):
symbols experiments; red NEW approach 4 = 1.0, ea =
0.9; cyan STD approach 4 = 0.6, e, = 0.9; blue STD
approach 6 = 1.0; ea = 0.9.
0o.
o.e
0.4'
h/D=0.25
^^^~
h/D=0.50
0.8
0.6
0.4
void fraction, 
h/D=0.75
/^\
5 0 5
position, cm
h/D=1.00
5 0 5
position, cm
Figure 5: Time and planeaveraged void fraction as a
function of bed height for GWS (3D simulations).
Experiments compared with simulations (COR=0.9):
symbols experiments; green STD approach =1.0; cyan
STD approach =0.6; red NEW approach =1.0; blue
NEW approach =0.9.
Figure 7: Average void fraction along diameter lines at
four different bed heights for CCB (3D simulations):
symbols experiments; red NEW approach 4 = 1.0, e. =
0.9; cyan STD approach 4 = 0.6, em = 0.9; blue STD
approach 6 = 1.0; e. = 0.9.
Paper No 1956
o972105
0944211
0916316
0885421
0o 860526
0832632
080073?
0 776842
078947 E
0721053 0
0693158 I
0665I63
063736B
0609474
0 51579
0553684
0.525789
0 497995
047
5 0 5, >
x,cm
0972105
0944211
0916316
0888421
o 0860526
08 3632 1^i
0804737
0776842
0 78947
0721053
0693158 r
0665263
0637368
0609474
051579
0 53684
0525789 5
0 497895
047
p p I
y, an
Figure 8: Contour plots of void fraction for GWS in the xz (top row) and yz (bottom row) planes (3D simulations). From left
to right column: experiments, NEW approach q= 1.0; e, = 0.9, STD approach q= 0.6; e,= 0.9, STD approach q= 1.0; ea,
=0.9.
20
0972105
09 4211
0916316
0888421
0860526
0832632 15
0804737
o 7;842
0748947 5
0693158 X I
0637368
609474
o581579
0 55384
0525789 5
0 497895
047
I W2
0972105
094211
0916316
0888421
0860526
0832632 '5
0 77842
0o 74947
0721053
0693158 J
0665263
0 637368
0609474
0581579
055364
0 525789
0 497895
047
x, cm
y, cm
Figure 9: Contour plots of void fraction for CCB in the xz (top row) and yz (bottom row) planes (3D simulations). From left
to right column: experiments, NEW approach q= 1.0; e, = 0.9, STD approach q= 0.6; e,= 0.9, STD approach q= 1.0; ea,
=0.9.
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
I
j j
.....
Paper No 1956
References
Cui, H. & Grace, J.R. Fluidization of biomass particles: a
review of experimental multiphase flow aspects.
Chemical Engineering Science, 62, 4555 (2007)
Deza M., Battaglia F & Heindel T.J. A validation study for
the hydrodynamics of biomass in a fluidized bed.
FEDSM2008, 2008 ASME Fluids Engineering
Division Summer Conference, Jacksonville, FL:
ASME Press, Paper FEDSM200855158 (2008)
Drake, J.B., & Heindel, T.J., Repeatability of Gas Holdup
in a Fluidized Bed using Xray Computed
Tomography, FEDSM2009, 2009 ASME Fluids
Engineering Division Summer Meeting, Vail,
CO, ASME Press, Paper FEDSM200978041,
(2009).
Fluent Inc. Fluent 6.3 User's Guide (2006)
Franka, N.P., & Heindel, T.J., Local TimeAveraged Gas
Holdup in a Fluidized Bed with Side Air
Injection using Xray Computed Tomography,
Powder Technology, 193, 6978 (2009).
Gidaspow, D. Multiphase Flow and Fluidization.
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Continuum and Kinetic Theory Descriptions.
Academic Press. New York (1994)
Johnson, P.C. & Jackson, R. Frictionalcollisional
constitutive relations for granular materials with
application to plane shearing. Journal of Fluid
Mechanics, 176, 6793 (1987)
Joseph, G.G., Laboreiro, J., Hrenya, C.M. & Stevens, A.R.
Experimental segregation profiles in bubbling
gasfluidized beds. AIChE Journal, 53,
28042813 (2007)
Min, J., Drake, J.B., Heindel, T.J., & Fox, R.O.,
Experimental Validation of CFD Simulations of
a LabScale FluidizedBed Reactor with and
without SideGas Injection, AIChE Journal, To
Appear, 2010.
Syamlal, M., Rogers, W.A. & O'Brien, T.J. MFIX
Documentation Theory and Guide,
DOE/MC/213532373, NTIS/DE87006500
(1993)
Paper No 1956
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Table 1: Constitutive equations for gas and
(Fluent Inc., 2006).
solid stress tensor, for a single solid phase a, as defined in ANSYS Fluent 6.3
Gas stress tensor S, = P I+ p [VU + (VuY] g VugI
Solid stress tensor S Pa = P,,I + rs
Solid pressure Psa= saPsa0 + 2ps ( + ea s~ g0,
Radial distribution go,, = 1 1
Solid shear stresses zrs = sasaU[Vus +(Vu. )T] Esa s 2 s Vusal
Solid shear viscosity Psa = Psacol + P.sa,kn + safer
4 (2
Collisional viscosity p s,coi saPsd go,a e( a )
5 T)
Kinetic viscosity Psak = P + 2 ea )e s
6(3 ea) 5
Px sin q
Frictional viscosity Ps, fr si
12D
Solid bulk viscosity s) snPsYcdp goYO I + 1 +)
3 7z'
3 0 P ( saOa ( I+ ):Vu S r;", + ( g
Granular temperature (algebraic 12(1 e )2g
formulation) Yoa = da a S Psa a
>ga = 3KgaO
Paper No 1956
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Table 2: Numerical details in CFD simulations.
Description Value Comment
Pressurebased solver
Unsteady formulation Second order implicit
Time step 104 s Specified
Maximum number of iterations 100
Data sampling for time statistics 10
Operating pressure 1 atm
Gas density and viscosity 2.417 kg/m3, 1.8x10 Pa s
Inlet boundary conditions 0.362 m/s(GWS), 0.328 m/s(CCB) Ug=2Umf
Outlet boundary conditions Outflow Fully developed flow
Wall boundary for gas phase No slip Specified
Wall boundary for solidphase 0 Pa Specified
Convergence criteria 106 Specified
Pressurevelocity coupling SIMPLE Phasecoupled
Momentum discretization Secondorder upwind
Volume fraction discretization QUICK
Table 3: Chosen models for the solid phase.
Properties
Granular viscosity SyamlalO'Brien
Granular bulk viscosity Lun et al.
Frictional viscosity Shaeffer
Angle of internal friction constant=30
Frictional pressure basedktgf
Frictional modulus derived
Granular temperature algebraic
Solids pressure Lun et al.
Radial distribution SyamlalO'Brien
Elasticity modulus derived
Table 4: Bed material properties.
Properties GWS CCB
Particles diameter, gm 550 550
Particle nominal density, kg/m3 1300 1000
Particle effective density, kg/m3 985.61 687.8
Initial bed height (STD), mm 152 152
Initial solid packing (STD), 0.44 0.39
Solid packing limit (STD) 0.63 0.63
Initial bed height (NEW), mm 165 165
Initial solid packing (NEW), 0.55 0.55
Solid packing limit 0.58 0.58
Paper No 1956 7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Table 5: Description of the four sets of simulations that were performed, with the chosen parameters values
Simulations set Parameters
1. Fluent 3D parametric study with STD GWS: q=0.6, 0.8, 1.0; e,, =0.7, 0.8, 0.9
approach CCB: =0.6, 0.8, 1.0; e, = 0.9
2. Fluent 2D parametric study with NEW GWS: = 0.6, 1.0; eC = 0.7, 0.9
approach CCB: = 0.6, 1.0; e = 0.9
3. Fluent 3D final with the NEW GWS: = 0.9, 1.0; eaa= 0.9
approach CCB: = 0.9, 1.0; e.= 0.9
