Professional Documents
Culture Documents
www.elsevier.com/locate/ces
Abstract
The gas and particle motions in a bubbling &uidized bed both with and without chemical reactions are numerically simulated. The solid
phase is modelled as Discrete Element Method (DEM) and the gas phase is modelled as 2-D NavierStokes equations for 2-phase &ow
with &uid turbulence calculated by large Eddy simulation (LES), in which the e<ect of particles on subgrid scale gas &ow is taken into
account. The gas/particle &ow structure, the mean velocities and turbulent intensities can be predicted as a function of several operating
parameters (particle size, bed temperature, and inlet gas velocity). The lower the inlet gas velocity, the higher the ratio of particle collision.
The distributions of the particle anisotropic velocity show that the particles have no local equilibrium, and the distribution of gas kinetic
energy corresponds to the distribution of gas-particle coupling moment in the &uidized bed. An intensive particle turbulent region exists
near the wall, and the gas Reynolds stress is always much higher than the particle stress. The presence of the large reactive particles in
the &uidized bed may a<ect signi>cantly the gas and particle velocities and turbulent intensities. The e<ects of the bed temperature and
inlet gas velocity on the gas particle &ow structure, velocity, and turbulent intensity are also studied.
? 2003 Elsevier Ltd. All rights reserved.
Keywords: Mathematical modeling; Fluidized bed; Discrete element method; Large eddy simulation; Chemical reaction
1. Introduction
Fluidized bed combustion has been modeled from di<erent approaches, namely semi- empirical models and computational &uid dynamic (CFD) models. Semi-empirical
models are based on experimental >ndings and empirical
correlations obtained from measurements; they can account for complex &ow pattern in &uidized bed combustors
(Desroches-Ducame et al., 1998; Kulasekaran et al., 1999;
Knoebig et al., 1999; Adanez et al., 2001; Chen et al., 2001).
This type of approach is based to a large extent on measured
data, thus it is limited to the database it was developed from.
CFD models involve a full set of partial di<erential equations that describe the conservation of mass, momentum,
energy and chemical species (Peirano and Leckner, 1988;
Lathouwers and Bellan, 2001). CFD models are much more
complex than semi-empirical ones, and their computational
times are very long due to the unsteady character of the
numerous equations. However, detailed CFD models and
0009-2509/$ - see front matter ? 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2004.01.069
4194
useful and versatile tools to study the hydrodynamic behaviour of particulate &ows. They o<er a more natural way
to simulate a gassolid &ow with particles of di<erent size
and/or di<erent density, since each individual particle is
tracked in the simulation. Discrete particle models have
been combined with a Eulerian &uid model to simulate phenomena such as bubbling, slugging, and solid transport in
bubbling and circulating &uidized beds (Tsuji et al., 1993;
Hoomans et al., 1996; Helland et al., 2000; Yuu et al., 2001,
Zhou et al., 2002a,b). These models were properly validated
by comparison with experiments (Van Wachem et al., 2001).
The number of particles used in the numerical computation
is limited by the computer capacity. However, discrete particle models give an accurate description not only of the
particle motion (particle rotation, collision with other particles) but also of the chemical reactions, and heat and mass
transfer at the individual particle scale (Kaneko et al., 1999;
Rong and Horio, 1999; Zhou et al., 2002b). Kaneko et al.
(1999) used the discrete element method (DEM), in which
direct particle-to-particle heat transfer was rather illogically
neglected, to investigate the temperature e<ect on particles
and gas in a &uidized bed reactor producing polyole>n. Rong
and Horio (1999) simulated the thermal characteristics and
NOx emission of burning chars in &uidized bed. Their results
showed that the maximum char particle temperature is about
50 5 C higher than the average bed temperature, which is
much less than experimental measurements (Linjewile et al.,
1994; Joutsenoja et al., 1999). The heavy metal vaporization
dynamics was predicted in a &uidized bed reactor (Zhou
et al., 2002b) and a fair quantitative agreement was obtained
between the predicted vaporization rate and the measured
one. All models neglected the e<ect of turbulence on the
gas/particle motion, chemical species dissipation, and chemical reaction, although both experimental and theoretical
results veri>ed the strong intensity of turbulence in &uidized
bed (Peirano and Leckner, 1988; Zhou et al., 2000).
The general objective of this study is to develop a turbulent 2-phase gas/particle &ow model accounting for chemical reactions, which can predict the gasparticle turbulent
&ow structure, the thermal characteristics of burning coal
particles and the gaseous emissions in a coal combustion
&uidized bed. In Part I, the turbulent gas/particle &ow model
of Yuu et al. (2001) is extended in order to overcome their
assumptions: the magnitude of particle e<ects on subgrid
scale (SGS) &ow is obtained by solving simultaneously the
SGS kinetic energy equation and the >ltered mass and momentum conservation equations. In addition, the chemical
reactions in the turbulent &ow are taken into account. In this
model, the gas &ow is calculated by LES with two-way coupling, and the particle motion is treated by a Lagrangian approach in which the particleparticle interaction is modelled
as DEM. The e<ect of gasparticle interaction on subgrid
scale &ow is particularly clari>ed. Furthermore, the e<ects
of parameters associated with reactive particles: coal diameter, bed temperature, and inlet gas velocity on the turbulent
dense 2-phase &ow are examined.
@p
@
@
+
(
ij ) +
(ij )
@xi
@xj
@xj
+fg + + ;
(2)
where an overbar denotes application of the >ltering operation, is the porosity, f is the gas density, uf is the gas
n
R2
c
kst
c
kst
(3)
(u f; i vMp; i ) :
aTL + 1
@xi
4195
f C
(4)
n
p; k
n
k=1 p; k
M
nM =
; dp = k=1
;
; Mp =
Vgrid
n
n
R1
and
p; k
;
(6)
n
where n is the particle number in the computational cell,
vp is the particle velocity, dp is the particle diameter, p is
the particle density. Vgrid is the computational cell volume,
is equal to Nx Ny dM p , thus depending on the particles it
contains.
Therefore, Rep is de>ned as follows:
vMp =
k=1
f | u f v p |dM p
Rep =
:
f
(7)
vMpn = ( v pr n ) n ;
*
fcr = kst d - v p ;
*
v p
= v pr vpn ;
(8)
(9)
(10)
(11)
4196
(12)
fc = |fcn | t ;
(13)
*
(15)
Cd
where
is the e<ective drag coePcient.
The e<ective drag coePcient Cd depends strongly on the
local void fraction in the vicinity of the particle. According
to Wen and Yu (1966), it can be written as
Cd = Cd np ;
(16)
0:44
Rep 1000;
where Rep is the particle Reynolds number
Rep =
*
f | u f
f
*
v p |dp
(18)
*
vp
= F =mp + g
v
vp dmp
Vp
;
p +
mp
mp
dt
! = T =I;
(19)
(20)
*
rp
!p = !0 + !0 Nt;
(21)
= (F m + F l ):
*
(22)
n
Fd
=
;
Vgrid
(23)
i=1
*
Ft
Fd
*
v rel
**
n v
rel
Fd
*
v rel
n (uf vp );
(24)
t @nM 1 b2
;
@xi aTL + 1
(25)
initialization
void fraction
the particle
collide other particle
or the wall?
yes
no
t = t+t
start
repeat N times
(N = particle number)
energies (Donskoi and McElwain, 1999). Buring char particle evolves according to the shrinking core model. The
chemistry of the formation and destruction of NO and N2 O
is complex. So, the heterogeneous reactions of char with
CO, CO2 , NO and N2 O and the homogeneous reactions involving CO, O2 , NO and N2 O in the boundary layer are
simpli>ed according to the theory of Amand and Andersson
(1989), Johnsson (1994), and De Soete et al. (1999). The
colliding particle-particle heat transfers are modeled on the
basis of the analysis of the elastic deformation of colliding
spheres (Sun and Chen, 1988). The equations of conservation of species mass fraction and of energy for a &uid cell
describe the transportation of species (NO, CO, CO2 , H2 O,
NO, N2 O, and N2 ) and of energy in the gas, accounting for
the e<ect of chemical reactions on the gas particle &ow.
4197
end
4198
Table 1
Physical and numerical simulation parameters
t = 0.20 s
t = 0.50 s
t = 0.75 s
t = 0.90 s
t = 1.0 s
t = 1.30 s
t = 1.45 s
t = 1.70 s
t = 2.0 s
Value
Bed height, H
4:84 cm
Bed width, W
7:26 cm
Inlet gas (wall) temperature, 1123.15, 1173:15 K
Tf 0
Minimum &uidization velocity, umf = 7:01 104 d2p g(p f )=f
umf (Lin et al., 2002)
(for sand at 850:0 C, umf = 0:4 m s1 )
Inlet gas velocity, uf 0
0.4 and 0:6 m s1
Density,
Coal, pc
1100 kg m3
Sand, ps
2600 kg m3
Particle diameter, dp
Sand, ds
1:0 mm
Coal, dc
0.8, 1.5 and 2:0 mm
4.7
np value
Sti<ness, kSt
800 N m1
Friction coePcient,
0.3
Restitution coePcient, e
0.9
Poisson ration, /
Sand (walls)
0.30
Coal
0.37
Yound moduli, G
Sand (walls)
15:0 GPa
Coal
3:0 GPa
Number of grid cells
I J = 9 14
Total particle number
1480
Sand particles
1460
Coal particles
20
Time step, Nt
2:0 105 , s
50
40
6
30
Averaged value = 5.4 %
20
4
Averaged value = 0.38 %
10
0
0
0.0
0.5
1.0
Time / s
1.5
Properties
t = 0.00 s
2.0
Table 2
Parameters of the simulated cases
Case
Case
Case
Case
Case
Case
1
2
3
4
5
6
Coal number
0.6
0.6
0.6
0.6
0.6
0.4
850
900
850
850
850
850
0
20
20
20
20
20
Number of sand
particles
0.8
1.5
2.0
0.8
0.8
1480
1460
1460
1460
1460
1460
4199
2104
1104
0
0.00
0.075
0.050
0.01
Bed
0.02
width 0.03
/m
/m
50
3104
0.025
ht
100
4104
he
ig
Mean = 0.0
SD = 0.036
5104
0.04
(a)
0.000
0.05
Be
d
Particle number
150
2 0.5
(F 2+F ) )/N.m
Gas-particle moment (F= x y
-3
200
0
-0.3
-0.2 -0.1
0.0
0.1
0.2
0.3
Mean = -0.07
SD = 0.04
Mean = -0.003
SD = 0.05
80
2 -2
Turbulent energy / m .s
100
0.020
0.015
0.010
0.005
0.075
60
ht
0.025
ig
0.02
width 0.03
/m
he
Bed
40
/m
0.050
0.01
0.04
0.000
0.00
(b)
0.000
0.05
Fig. 6. Distributions of instantaneous gasparticle moment and gas turbulent energy in the &uidized bed at 1:45 s (case 1): (a) gasparticle
moment, (b) gas kinetic energy.
20
0
-0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4
(b)
0.025
Be
Particle number
0.030
(a)
the surface. The particle vertical turbulent intensity is maximum or medium near the walls whereas the gas intensity is
higher in the center inside the bed. The particle intensity is
almost uniformly distributed at the bed surface (0:031 m),
whereas the gas intensity is higher near the walls. The intensive particle surface &uctuation results in higher particle
intensity at the bed surface than in the bed. Fig. 9 plots the
distributions of gas and particle Reynolds stresses (de>ned
2
as ui; f(p) uj:f(p)
=uf0
). The particle stress is always much
smaller than the gas stress, the di<erence is up to 100 times.
4.2. E:ect of reactive particles on the 9ow structure
A binary mixture of sand and coal particles was considered, and the e<ects of bed temperature (case 2), coal
particle size (cases 35) and inlet velocity (case 6) with
chemical reactions on the turbulent gas/particle &ow were
studied. The higher the inlet gas velocity, the lower the ratio
4200
1.1
-0.005
-0.010
1.0
-0.015
0.9
-0.020
0.016 m
0.002
0.031 m
0.08
0.001
0.04
0.000
0.00
-0.001
-0.04
1.2
0.000
Bed height
0.005
0.12
0.003
1.3
Bed heights :
: 0.016 m
: 0.031 m
0.8
-0.025
0.00
0.01
0.02
0.03
Bed width / m
0.04
0.05
-0.08
0.05
-0.002
0.00
0.01
0.02
0.03
0.04
Bed width / m
Fig. 7. Gas and particle velocity distributions along the width in the bed
and at its surface (vertical component, case 1).
0.9
0.6
0.05
0.5
0.04
Bed height
0.016 m
0.03
0.3
0.031 m
0.00
0.01
0.02
0.03
0.04
0.05
0.30
0.80
0.27
0.75
0.12
0.11
0.10
0.09
0.08
0.65
Bed height
0.18
0.016 m
0.60
0.031 m
0.15
0.55
0.12
0.00
0.01
0.02
0.03
0.04
0.50
0.05
Bed width / m
0.02
0.03
0.04
0.05
0.04
0.05
Bed width / m
0.8
Fig. 8. Distributions of the gas and particle turbulent intensities along the
width in the bed and at its surface (case 1): (a) horizontal component,
(b) vertical component.
0.7
0.6
case 1
case 3
case 5
0.5
case 2
case 4
0.4
0.00
0.01
(a)
0.70
0.21
case 3
0.13
0.00
0.24
case 2
case 5
0.14
Bed width / m
(a)
0.4
case 1
case 4
0.15
Particle turbulent intensity
0.7
0.8
0.06
0.07
(b)
Fig. 9. Distributions of the gas and particle Reynolds stress along the
width in the bed and at its surface (case 1).
(b)
0.01
0.02
0.03
Bed width / m
Fig. 10. Comparisons of the gas and particle turbulent intensity distributions along the width in the bed (bed height = 0:016 m, vertical component): (a) particle turbulent intensity, (b) gas turbulent intensity.
the bed width (cases 15). For the particle phase, there exists an intensive turbulent intensity region near the walls.
At the bed center, the intensity is maximum or medium in
cases 1, 2, and 5 whereas it is minimum in cases 4 and
5. The particle turbulent intensity is always low, in particular at the bed center, in the presence of large particles.
The simulation shows that the presence of large reactive
particles in the &uidized bed may a<ect signi>cantly the
particle turbulent intensity. The distributions of the vertical gas turbulent intensity are similar. The intensities are
higher in the center. The bed temperature (case 2) has no
signi>cant e<ect on both the gas and particle intensities.
The gas Reynolds stress is always much higher than the
particle Reynolds stress in all cases. The gas and the particle velocities are not correlated, and the turbulent intensity
due to the particle collision plays a signi>cant role inside
the bed.
The particle turbulent intensity increases with the inlet
gas velocity, whereas the gas turbulent intensity and the
Reynolds stress decrease. However, the simulation shows
that the higher the inlet velocity, the higher the absolute
2
values of the gas and particle &uctuations (vf(p);
i ) and the
higher the Reynolds stress (ui; f(p) uj; f(p) ).
5. Conclusions
The turbulent gas and particle &ow both with and without
chemical reactions in a bubbling &uidized bed was studied.
The solid phase was modeled as discrete element method
(DEM) and the gas phase was modeled as 2-D Navier
Stokes equations for 2-phase &ow with &uid turbulence calculated by large Eddy simulation (LES), in which the e<ect
of particles on subgrid scale gas &ow is taken into account.
The following conclusions can be drawn based on the simulation of large, soft spheres and near minimum &uidization
velocities:
(1) The lower the inlet gas velocity, the higher the ratio
of particle collision. The initial ratio of colliding particles is up to 40%, and the mean ratios of particle
particle collisions and particlewall collisions are 5.4%
and 0.38%, respectively, when the inlet gas velocity is
1:5 umf .
(2) The vertical distribution of the particle velocity displays
two branches, each of them obeying a di<erent Gaussian
distribution. The anisotropic distributions indicate that
the particles in bubbling &uidized bed have no local
equilibrium.
(3) The presence of the particles increases the gas kinetic
energy and the kinetic energy distribution corresponds
to the gasparticle coupling moment distribution.
(4) An intensive particle turbulent region exists near the
wall, and the gas stress is always much higher than the
particle stress.
4201
(5) The presence of the large reactive particles in the &uidized bed may a<ect signi>cantly the particle &ow
structure, the gas and particle velocities and the turbulent intensities.
Notation
a
b
Cd
Cd
Ck ; C
dp
dn
d
e
F
Fc
fcn
fc
*
= 2k s , s
uf0
inlet gas velocity, m s1
uf
gas phase velocity, m s1
u f
>ltered gas phase velocity, m s1
uf
gas &uctuating velocity, m s1
vp
solid phase velocity, m s1
4202
vpr
vpn
vp
vp
vrel
vrel
v
Vp
Vgrid
x, y
relative velocity, m s1
normal relative velocity, m s1
tangential relative velocity, m s1
particle &uctuating velocity, m s1
relative velocity, m s1
relative &uctuating velocity, m s1
mean velocity of the mass &ow covering the particle
surface, m s1
particle volume, m3
volume of a computational cell, m3
x and y directions
Greek letters
/i
ij
NS
Nt
Nx
Ny
f
f
t
f
p
ij
ij
!
Poisson ratio
Kronecker delta
porosity in &uid cell
characteristic length, = (Nx Ny)1=2 , m
area of computational cell, NS = Nx Ny, m2
time step, s
x direction mesh spacing, m
x direction mesh spacing, m
coePcient of viscous dissipation
friction coePcient
gas viscosity, Pa s
kinematic gas viscosity, f = f =f , m2 s1
SGS kinetic gas viscosity, m2 s1
gas density, kg m3
particle density, kg m3
turbulent Schmidt number,
= 1:0
SGS stress,
ij = f vf (Sf; ij 2=3Sf; kk ij )
tangential unit vector
SGS stresses, ij = f uf; i uf; j
volumetric particlegas interaction, N m3
particle rotational velocity, s1
e<ect of mass transfer on gas moment, N m3
References
Adanez, J., Gayan, P., Grasa, G., de Diego, L.F., Armesto, L., Cabanillas,
A., 2001. Circulating &uidized bed combustion in the turbulent regime:
modelling of carbon combustion ePciency and sulphur retention. Fuel
80, 14051414.
Amand, L.E., Andersson, S., 1989. Emissions of nitrous oxide (N2 O)
from &uidized bed boilers. In: Proceedings of the 10th International
Conference on Fluidized Bed Combustion. ASME, pp. 4955.
Arastoopour, H., 2001. Numerical simulation and experimental analysis
of gas/particle systems: 1999 Fluor-Daniel Plenary lecture. Powder
Technology 119, 5967.
Chen, Z., Lin, M., Ignowski, J., Kelly, B., Linjewile, T.M., Agarwal, P.K.,
2001. Mathematical modelling of &uidized bed combustion. 4: N2 O
and NOx emissions from the combustion char. Fuel 80, 12591272.
Cundall, P.D., Strack, O.D.L., 1979. A discrete numerical model for
granular assemblies. Geotechnique 29, 4765.
De Soete, G.G., Croiset, E., Richard, J.R., 1999. Heterogeneous formation
of nitrous oxide from char bound nitrogen. Combustion and Flame
117, 140154.
4203