Ind. Eng. Chem. Process Des. Dev., Vol. 18, No.

2, 1979

333

Modeling of Shallow Fluidized Bed Combustion of Coal Particles

L. T. Fan,* K. Tojo, and C. C. Chang
Department of Chemical Engineering, Kansas State University, Manhattan, Kansas 66506

The dynamic and steady-state characteristics of a shallow fluidized bed coal combustor have been studied by using
a model in which the lateral solids mixing is taken into account. It has been found that the steady-state and
unsteady-state concentrations in the bed are influenced profoundly by the bubble size, and that the effects of excess
air rate, bed height, and particle density on the dynamic characteristics of the shallow fluidized bed coal combustor
are negligible compared with the effect of bubble size.

Introduction
I t is well known that one of the important advantages
of a fluidized bed is that mixing of fluidized particles in
it is intensive. This advantage is particularly appreciable
in a small fluidized bed. However, in a large shallow
fluidized bed, which is used to reduce blowing cost, lateral
solid mixing can be poor, leading to appreciable nonuniformity in lateral concentration profiles. This nonuniformity may impair an effectual property of the shallow
fluidized bed; that is, the bubble size remains small, thus
giving rise to a high transfer rate of gas between the bubble
and emulsion phases.
The present work is concerned with dynamic and
steady-state characteristics of a shallow fluidized bed coal
combustor which has so far received relatively little attention. A dynamic model in which the lateral solids
mixing is taken into account is developed. The effects of
some operating variables on the solids concentration
profiles are analyzed by means of the model in order to
determine an effective method for promoting lateral solids
mixing. Highley and Merrick (1971) studied the effect of
solid feed points on the lateral solids mixing in a large
fluidized bed reactor. Merry and Davidson (1973) have
proposed a method of Gulf Circulation which is generated by introducing uneven distribution of fluidizing gas
in order to promote the lateral solids mixing in a shallow
fluidized bed. However, none of these works is specifically
concerned with the shallow fluidized bed coal combustor.
Mathematical Formulation
The present model assumes that the reactor consists of
two phases, namely, the bubble and emulsion phases. The
assumptions of the model are as follows. (a) The voidage
of the emulsion phase remains constant and is equal to that
at the incipient state of fluidization. Thus the flow of gas
through the bed in excess of minimum fluidization flow
passes through the bed in the form of bubbles (see, e.g.,
Davidson and Harrison, 1963). (b) The emulsion phase
is well mixed in the axial direction. This is a valid assumption for a relatively shallow fluidized bed. (c) The
bubble size remains constant, and the flow of bubbles is
of the plug flow. This assumption is valid since there is
usually no sufficient time for the bubble to grow in a
shallow fluidized bed. (d) The overall rate of combustion
reaction represented as

c+0 2

- coz

is so high that the oxygen transfer into coal particles is the

rate-determining step (see, e.g., Avedesian and Davidson,
1973). (e) No elutriation occurs. (f) The bed is under
isothermal operation.
These assumptions give rise to the following governing
equations (see Appendix)

bubble phase:
aCab
- K(Cab - Cae)
aX

aCab
~

at
emulsion phase:
aCae

urn,

at

emf - = -(Cao

(1)

+r

Cae)

K(Cab -

k,
cae)d X - 6 CC,,

(2)

pd,

for oxygen, and

aC _
at

for carbon particles. The appropriate

initial and boundary
..
.
conditions-are: for t = 0
Cab = Cae = CaO
for t > 0
Cab = CaOat x = 0 (bed bottom)
aC/ar = aCab/ar = aCae/ar = 0
at r = 0 (center of the bed)
aC/ar = aCab/ar = aCae/ar = 0
at r = R (wall of the bed)
The feeding rate function, +,, is defined by

at 0 Ir 5 rf
arf2(1 - tb)L
=O
atrf<rIR
Method of Solution. Programming the numerical
solution of nonlinear parabolic partial differential equations is often a highly complicated, tedious, time-consuming, and unstable procedure. However, these difficulties have largely been eased by the available software
interface for solving the parabolic partial differential
equations implemented basically by the so-called method
of lines (Liskovets, 1965). The present study employs this
software interface for numerical calculations (Sincovec and
Madsen, 1975). Gears backward difference formulas
(1971) are used for the time integration. A modified
Newtons method with internally generated Jacobian
matrix is utilized to solve the nonlinear equations.
Results and Discussion
The functional relationships among variables and
nominal values of various parameters employed in numerical computation are listed in Tables I and 11, respectively. The average carbon concentration in the
emulsion phase is shown in Figure 1 as a function of time
with the bed height as the parameter. It can be seen that
+f

0019-7882/79/lii8-0333$01.00/00 1979 American Chemical Society

334

Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2, 1979

Table I. Variables Relationships

emf3

Urn*=

minimum fluidization velocity (Davidson and Harrison, 1971)

-.p - p g

5( 1 - E m f ) 6idp.11
bubble porosity (Davidson and Harrison, 1 9 6 3 )

Eb

uo - Umf

Ub

- Umf
- Umf

bubble velocity (Davidson and Harrison, 1 9 6 3 )

Ub

gas interchange coefficient (Kunii and Levenspiel, 1 9 6 9 )

K = K l K 2 / ( K l+ K , )

cm/s

Uo

0.711(gd~)'"

dB

1i s
D I/ 2gl/

Umf

K , = 4.5-

cm/s

5.85(-

K , = 6.78(mf +

d~ 5 ' 4

Dub

)'I2

dB3

dB
D, = 0 . 1 8 7 ~ b ~ ~ f
- bk m f

lateral dispersion coefficient of solids particles (Kunii, 1 9 6 6 )

Table 11. Numerical Values of Fixed Parameters
Used in Computations
dB = 5 , 1 0 , 1 5 cm
R = 4 0 cm
L = 20, 30, 5 0 cm
U, = 85.6-117.0 cm/s
(excess air = 10-50%)
temperature, 800 C
e m f = 0.5
(mol/cm3)
C, = 2.38 X
D = 1.74 c m 2 / s

Sh.D
Sh = 2.0 ( h g =-)
dP
P = 1.0, 0.75 g/cm3d , = 0.05 cm
D,, = 100 D o r c m z / s
Dab = D/lOO or 0 c m Z / s
feed rate
F = 6 g/s

the trend of concentration variation is almost independent

of the bed height. The steady-state carbon concentration
is approximately inversely proportional to the square of
the bed height. On the other hand, the effect of the rate
of excess air is negligible and the bed reaches the steady
state approximately a t 200 s when the bubble diameter is
5 cm. Figure 1also shows the well-known fact (Rengarajan
et al., 1977) that the steady-state concentration of carbon
particles is less than 1% by weight. A concentration of
1 g/cm3 roughly corresponds to 1% by weight in the
present system.
Figure 2 shows the effect of bubble size on the transient
carbon concentration. It can be seen that the concentration change is drastically influenced by the bubble size.
The average carbon concentration in the bed increases with
an increase in the bubble size. This implies that a small
bubble operation of the bed is more stable than a large
bubble operation because the bed with a low carbon
concentration can be easily controlled. When the bubble
diameter is large, e.g., 15 cm, the bed does not reach a
stable state because of the insufficient transfer of gas from
the bubble phase to the emulsion phase. On the other
hand, the small bubble operation reaches a stable state
easily.
The large bubble operation enhances the lateral solid
mixing and, thus, can minimize the possibility of generation of the extremely high concentration near the feeder.
Therefore, it is desirable to control the bubble size so that
it is neither too small nor too large.
Figure 3 shows the carbon concentration profiles a t
steady state with the rate of excess air and the bubble size

-3

d :$

110

cmZ/s

0 cm

= I . o g/cm3
r,= 2 0 Crn

L-

20 cm

"I
0

I U

.e

30 c m
50%

8 'I

I /

0
0.2

5 0 em

200

20

2000

zoo00

tim t ( ~ a c l

Figure 1. Effect of bed height and excess air rate on the transient
average carbon concentration.

cm

IS

!I'

6 .

0
02

5 cm

200

20
time

2000

20000

(I~c)

Figure 2. Effect of bubble diameter on the transient average

concentration.

Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2, 1979

ax+*

811.

dg

335

20 Y.

.S o c m

L -30sm
I, - 2 O m

"

02

20
tlrn I

10

20

30

40

roaioi

~)siIicn

x10
1sac1

zoo0

zoo00

Figure 4. Effect of solids density on the transient average carbon

Concentration.

lcmi

Figure 3. Effect of bubble size and excess air rate on the steady-state
carbon concentration profiles.

as the parameters. As can be seen from this figure, the

effect of the excess air rate is almost negligible in the range
of the excess air rate between 20% and 50%. Figure 3 also
shows that as the lateral mixing of the solids becomes
poorer, an appreciable concentration gradient is generated
along the radius, especially in a small bubble operation.
This phenomenon can sometimes be detrimental, especially for nonisothermal systems. As can be seen, a large
bubble operation drastically reduces the concentration
gradient.
In the present calculation, we have employed a correlation for the solids dispersion coefficient derived by Kunii
(1966) as listed in Table I. Note that the dispersion
coefficient is proportional to the bubble size. Hirama et
al. (1975) experimentally obtained a similar relationship
between the dispersion coefficient and bubble size. Solids
mixing in a fluidized bed in both radial and axial directions
is mainly induced by bubble motion (Toei et al., 1966),
which is obviously influenced by the bubble diameter. In
a fluidized bed where the bubble size is increased by
coalescence along the bed height, the axial dispersion
coefficient should be a function of the bubble size and
change along the bed height. Furthermore, the extent of
reactant conversion may be influenced considerably by the
bubble size. However, the detail of the relationship between the variable bubble size and the solids mixing is not
well known.
The effect of solids density on the transient average
concentration and on the steady state concentration are
shown in Figures 4 and 5 , respectively. In obtaining the
results, we have assumed that the solids density does not
change during the reaction. In real systems, however, the
density usually changes slightly with reaction time. As can
be seen in the figures, the density difference affects the
concentrations near the center and wall of the bed. In spite
of these effects, the trend of temporal concentration
variation and the extent of lateral mixing (concentration
profiles along the radius) are little influenced by the
density.
The effect of feeding area on the concentration profiles
in the steady and unsteady states is shown in Figures 6
and 7 , respectively. To generate a uniform concentration
profile in a large scale fluidized bed, a multipoints feeder
is usually used. As can be seen from Figure 7 , the con-

Figure 5. Effect of solids density on the steady-state carbon concentration profiles.

60m

02

20
hn l

200
(

2ooo

Figure 6. Effect of feeding area on the transient average carbon

Concentration.

centration gradient can be reduced considerably by enlarging the feeding area. The carbon concentration near
the center of the feeder is approximately proportional to
the feeding area with an exponent of 0.37.
The assumption of isothermal operation becomes less
valid if the solid concentration gradient in radial direction
becomes appreciable. Under such a condition, the energy
balance, in addition to the mass balance, must be carried

336 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 2 , 1979

WCDID

.20%
dg*SOcrn

L -3Ocrn

p .10em3

"

I.\/

This is eq 1 in the text.

Since the flow in the axial direction in the emulsion
phase is assumed to be of the complete mixing type, a mass
balance of oxygen over this phase is: (accumulation of
oxygen) = (rate of oxygen in by convection) - (rate of
oxygen out by convection) (rate of oxygen in by diffusion) - (rate of oxygen out by diffusion) + (rate of oxygen
in through gas exchange with bubble phase) - (rate of
disappearance by reaction) or

aCae
2arArL (1 - t b ) emf - = 2arAr(l - tb)Umf(CaOat
Cae) + 2arL(1 - t b ) NaeIr
- 2x(r + Ar)L(l L

'b) Naelr+Ar

r a M wlon r Ian1

Figure 7. Effect of feeding area on the steady-state carbon concentration profile.

to derive the governing equations. Furthermore, the

distribution in the size of coal particles may have to be
taken into account. Naturally, the degree of difficulty in
solving the resultant governing equations will be enhanced.
Our future efforts include considerations of the effects of
the temperature and particle size variation.
Concluding Remarks
The effects of operating variables on the steady-state
and unsteady-state carbon concentration in a shallow
fluidized bed combustor have been investigated by using
the two-phase model of a fluidized bed.
The steady-state and unsteady-state concentrations are
influenced profoundly by the bubble size. The time required to reach the steady state is controlled mainly by
the bubble size. The effect of the other parameters on the
concentrations is negligible when compared with the effect
of bubble size.
The change in the particle density with reaction time
has little effect on the steady-state and unsteady-state
concentrations. Therefore, it can be assumed that the
density remains constant in the bed.
Enlargement of the feeder area is an effective method
in reducing the lateral concentration distributions in the
bed. The maximum concentration a t the center of the
feeder is proportional to the feeder area with an exponent
of 0.37.
Appendix
Derivation of the Governing Equations. Consider
a cylindrical shell with a volume of 2arArL in the shallow
fluidized bed combuster. For simplicity, the bubble phase
and the emulsion phase in this volume element are lumped
separately.
Since the flow in the bubble is assumed to be of the plug
flow, a mass balance of oxygen over an incremental height
AX in this phase is: (accumulation of oxygen) = (rate of
oxygen in by convection) - (rate of oxygen out by convection) - (rate of oxygen through gas exchange with
emulsion phase) or

+ J 2arAr'?$(Cab

CaJ dX SarArL(1 - tb)Ra

Dividing this expression by 2arArL(1 - t b ) and letting Ar

0 gives

where N,, is the diffusional flux and is defined as

N ae = - D ae- aCae
(A-3)
ar
Ra is the reaction rate of oxygen per unit emulsion volume.
Based on the unreacted core model, the reaction rate for
a single coal particle is

Assuming that the controlling mechanism is gas film

diffusion, we obtain

r, = adplkgCae

(A-5)

Thus, the reaction rate per unit emulsion volume becomes

R, = ra.n
(A-6)
where n is the number of coal particles per unit emulsion
volume. n is related to the coal concentration, C, by
n=- C
64-71
where p is the carbon density of coal particles and is assumed constant. Substituting eq A-5 and A-7 into eq A-6
gives
k,
R, = 6 - CCae

(A-8)

Substituting eq A-3 and A-8 into eq A-2 gives

urn,

emf aCae = - (Cao- CaJ

at

+
r

(:

rDae$)

This is eq 2 in the text.

A mass balance of carbon over the emulsion phase is:
(accumulation of carbon) = (rate of carbon in from feeder)
+ (rate of carbon in by diffusion) - (rate of carbon out by

Ind. Eng. Chern. Process Des. Dev., Vol. 18, No. 2, 1979

D, = effective dispersion coefficient of solids, cm2/s

D,, = effective dispersion coefficient of oxygen in the emulsion

diffusion) - (rate of disappearance by reaction) or

aC
2.rrrArL (1 - q,) - = 27rrArFr+ 2xrL(1 - t h ) N,,, at

2 ~ ( r Ar)L(l - tb)NcJr+b
- 27rrArL(l -

tb)Rc

where

F'=

at 0 Ir Irf
atrf<rIR

=O

Dividing this expression by 27rrArL (1- t b ) and letting Ar

0 gives
aC
l
a
(A-10)
at = h - ; rNc) - R ,

(;

where N, is the diffusional flux and is defined as

aC
N =-D (A-11)
ar
and where
b'
$'F

at 0 Ir 4 rf

.rrrf2(1- tb)L

=O

atrf<rIR

R, is the reaction rate of carbon and is given by

R, = RaMc= 6

k,MC
-CCae
PdP

(A-12)

Substituting eq A-11 and A-12 into eq A-10 gives

aC

- =$p+;

at

( :)
--rDS--

337

6k,MC

- -CCae (A-13)
pd,

This is eq 3 in the text.

Nomenclature
C = carbon concentration in the emulsion phase, g/cm3
C, = oxygen concentration in the emulsion phase, mol/cm3
C a b = oxygen concentration in the bubble phase, mol cm3
CaO= initial oxygen concentration (feed gas), mol/cm

phase, cm'/s
Dab= effective dispersion coefficient of oxygen in the bubble
phase, cm'/s
D = gas diffusivity in the solid-gas boundary, cm2/s
d B =-bubble diameter, cm
d = particle diameter, cm
fl= solids feeding rate, g/s
g = gravitational constant, cm/s2
K = gas interchange coefficient, l / s
L = bed height, cm
R = radius of the bed, cm
r = radial distance from the bed center, cm
rf = radius of the feeder, cm
Sh = Sherwood number
t = time, s
U = superficial velocity of gas, cm/s
v",f = incipient fluidization velocity, cm/s
X = axial distance from the bed bottom, cm
t b = fraction of the bubble phase
p = gas viscosity, g/cm s or Pa-s
$f = carbon feed rate, g/cm3 s
p = particle density, g/cm3
Literature Cited
Avedesian, M. M., Davidson, J. F., Trans. Inst. Chem. Eng., 51, 121 (1973).
Davidson, J. F., Harrison, D., "Fluidized Particles", Cambridge University Press,
New York, N.Y.. 1963.
Davidson, J. F., Harrison, D., "Fluidization", Chapter 2, Academic Press, New
York, N.Y., 1971.
Gear, C. W., "Numerical Initial Value Problems in Ordinary Differential Equations",
Chapter 9, Prentice-Hall, Englewood Cliffs. N.J.. 1971.
Highley, J.. Merrick, D., A.I.Ch.E. Symp. Ser. No. 116, 67, 219 (1971).
Hirama, T., Ishida, M., Shirai, T., Kagaku Kogaku Rombur Syu., 1, 273 (1975).
Kunii, D., Levenspiel, O., J . Chem. Eng. Jpn., 2, 122 (1969).
Kunii, D., "Kagaku Kikai Gijutsu", Maruzen, No. 18, p 161, 1966.
Ucovets, 0. A., "The Method of Lines (Review)", "English Transhtiin in Difference
Equations", Vol. I, p 1308, 1965.
Merry, J. M. D., Davidson, J. F., AIChE J., 51, 361 (1973).
Rengarajan, P., Krishnan, R., Tseng, S.P., Wen, C. Y., A.1.Ch.E. 70th Annual
Meeting, New York, N.Y., 1977.
Sincovec, R. F., Madsen, N. K., ACM Trans. Math. Software, 1, 232 (1975).
Toei, R., Matsuno, R., "Kagaku Kikai Gijutsu", Maruzen, No. 18. p 135, 1966.

Received for review May 30, 1978

Accepted December 4,1978

This work was conducted under the sponsorship of the Engineering Experiment Station (Energy Study Project) of Kansas
State University.