Powder Technology, 77 (1993) 177-199 177

Simplified scaling relationships for fluidized beds

L.R. Glicksman, M. Hyre and K. Woloshun

Massachusetts Institute of Technology, Cambridge, A44 (USA)

(Received January 4, 1993; in revised form May 28, 1993)

Abstract

Simplifications to the full set of scaling parameters for dynamic similarity of fluidized beds, Glicksman, Chemical
Engineering Science, 39 (1984) 373, were explored. A new set of simplified scaling laws includes the Froude
number based on column height, the solid to gas density ratio, the ratio of superficial to minimum fluidization
velocity, bed geometric ratios, and particle sphericity and size distribution. When the gas to particle drag is
represented by either the Ergun equation or a single particle drag equation, the new simplified laws hold exactly
in both the viscous dominated and gas inertia dominated limits. For intermediate conditions, the gas to particle
drag is well approximated in models based on the simplified scaling laws. The simplified scaling laws allow very
small models to be constructed which properly simulate the hydrodynamics of a full size reactor or combustor.
Experimental confirmation of the new simplified scaling laws and the viscous limit scaling parameters, where
equality of the density ratio is omitted, were carried out in circulating fluidized beds. Within the viscous limit,
the solid to gas density ratio is an important modeling parameter when the slugging regime is approached. In
general, the solid to gas density must be matched to achieve good similarity. Using the new simplified scaling
laws, good agreement was observed even when the length scale of the air fluidized model was as small as l/16
that of an atmospheric combustor.

Introduction combustor at atmospheric pressure and a cold model.

The behavior of a fluidized bed is closely related to
the bed fluid dynamics. For design of large commercial
beds, such as fluidized bed combustors, it would be
very advantageous to simulate the hydrodynamics in a
small laboratory bed fluidized with air at ambient
conditions. It is crucial to ensure that the small scale
model accurately reflects the behavior of the larger
commercial bed.
In a previous paper, Glicksman [l], a set of scaling
relations was systematically developed which allowed
a bed operating at ambient conditions to model a bed
at elevated temperature and pressure. In the general
case both beds must have equal values of the Reynolds
numbers based on bed diameter and particle diameter,
equal Froude numbers and solid to gas density ratios.
In addition the beds must be geometrically similar and
have identical dimensionless particle size distributions
and sphericity. These conditions will be referred to the
full set of scaling parameters.
Several experimental confirmations of the full set of
scaling parameters have been carried out. Nicastro and
Glicksman [2] showed close agreement between the
time resolved pressure differences in a bubbling bed
With the full set of scaling parameters the scale model
had linear dimensions of the bed and the particles
which were one quarter of the combustor. Fitzgerald
et al. [3] showed qualitative agreement between two
beds fluidized with different gases; the disagreements
were probably due to static electric effects. Newby and
Keairns [4] carried out a validation of the full scaling
laws using a bed of 200 micron glass spheres fluidized
with air at standard conditions and a geometrically
similar bed of 100 micron steel particles fluidized with
pressurized air. They found close agreement between
the dimensionless bubble size and frequency for both
beds. Glicksman et al. [5] presented results for the
TVA atmospheric fluidized bed combustor; these results
were in close agreement with results obtained in a cold
scale model designed with the full set of scaling pa-
rameters. Chang and Louge [6] were able to simulate
combustors of different bed diameters with a single
experimental model by varying the composition of the
fluidizing gas. They also found that when the surface
of the particle was treated to yield artificially low friction
factors, the behavior of the bed was altered. They
conclude that the friction coefficient should not be an
additional scaling factor when ordinary particles, without

0032-5910/93/$6.00 0 1993 - Elsevier Sequoia. All rights reserved

178

the special treatment, are used in the full scale and TABLE 1. Runs conducted by Roy and Davidson [ll]
model beds. Glicksman et al. [7] carried out comparisons
between a small circulating bed combustor and a cold Runs at low and high temperatures
model constructed using the full set of scaling laws.
Condition/run A B C D E
For circulating beds, a dimensionless solids flow rate
must also be matched between the two beds. The two Agreement with run A Yes Yes No Yes
beds showed close agreement. The wall roughness was pS (kg mW3) 2650 7100 7100 7100 2650
found to exert an important influence on the bed dP (Xl@ m) 600 180 500 900 240
behavior; a dimensionless wall roughness must be in- kr (m s-) 0.15 0.09 0.64 1.25 0.07
P (bara) 1 1 1
cluded in the geometric similarity relationships for the
T (K) 1023 288 288 288 288
model. D (m) 0.135 0.045 0.045 0.045 0.045
When constructing a model fluidized with ambient U, (m s-t) 0.78 0.45 1.0 1.6 0.43
air, matching the full set of scaling parameters results Qnfl(gD)05 0.13 0.14 0.9 1.9 0.11
in a unique set of values for the particle density and Fr = (Vi, - U,r)/(gD).5 0.55 0.56 0.55 0.54 0.55
d,/D (x 103) 4.4 4 11 20 4.2
diameter and for the linear dimensions of the bed. To
Pfh (x 104) 1.5 1.7 1.7 1.7 4.9
model an atmospheric combustor operating at about Re = pf Uod,lp 4.1 5.5 33 105 7.4
800 C, the model has linear dimensions one quarter
Runs at low and high pressures
those of the combustor. By simplifying the set of scaling
relationships, it is possible to relax the constraint on Condition/run FGHI JK
the dimensions of the model relative to the full scale
bed. Glicksman [l] identified a viscous region, for small Agreement with run F - Yes Yes Yes No Yes
ps (kg me31 384 2650 2650 2650 2650 2650
particles and low velocities, where the gas inertial effects
d, (X106 m) 240 120 120 240 550 550
are negligible. Similarly, an inertial dominated region U, (m s-) 0.02 0.01 0.02 0.07 0.09 0.14
exists in beds of large particles at high velocities where P (bara) 1 6 1 6 1
the gas viscous effects should be minimal. Modified T (K) 288 288 288 288 288 288
criteria for the applicability of these scaling relationships D (m) 0.28 0.1 0.1 0.28 0.1 0.1
U. (m s-t) 0.27 0.16 0.17 0.32 0.24 0.29
were later suggested by Horio [8]. In both of these
vnflw)~s 0.01 0.01 0.02 0.03 0.09 0.14
regions, the simplified scaling laws permit some flexibility Fr = (U. - Umf)/gD)o~5 0.15 0.15 0.15 0.15 0.15 0.15
in model design. Horio et al. [9] proposed a set of d,/D (xl@) 0.9 1.2 1.2 0.9 5.5 5.5
scaling relationships which differed from those men- PflPs (X 1w 3.3 2.9 4.9 4.9 2.9 4.9
tioned above. Glicksman [lo] demonstrated that Horios Re = pf U,d,IcL 5 8 1.5 4 64 12
set was identical to the viscous limit for the full set
of scaling laws. Roy and Davidson [ll] examined the
limits of the viscous region as proposed by Glicksman
in bubbling beds at different temperatures and pressures the terminal velocity criteria. They built two scale models
(see Table 1). They compared measurements of the of a circulating bed combustor with linear dimensions
major frequency, maximum amplitude and standard which varied by a factor of four. The two models
deviation of the amplitude to determine similarity. Most exhibited reasonably good agreement with each other
of their tests were carried out at a particle Reynolds for vertical voidage distributions. The two models used
numbers of 8 or below. For tests at low Reynolds the same particle density and fluidizing gas conditions.
number they found that it was not necessary to match Although equality of gas to solid density ratio was
the gas to solid density ratio and the particle to bed mentioned as one possible scaling parameter, no attempt
diameter ratio. Two tests carried out at Reynolds num- was made to match this parameter between the model
bers of 64 and 105, respectively, did not agree with and the combustor. The solid to gas density ratio for
companion tests using the reduced set of parameters the models differ from that for the combustor by a
valid for the viscous limit. A single test carried out at factor of 5.5. Data was not presented for the combustor;
a Reynolds number of 33 did agree with companion thus, proper scaling between the cold models and the
tests; in this test the solid to gas density ratio was also combustor could not be verified. Ishii and Murakami
identical to the ratio for the companion tests. [13] compared two geometrically similar beds using the
Horio et al. [12] derived a scaling law for circulating same particle material and fluidizing gas properties,
beds from consideration of the core and annulus regions the scaling was based on the parameters given by Horio
in a circulating bed. The resulting scaling law matched et al. [12]. They found close agreement in measured
superficial gas velocity, particle velocity, and particle peak frequency, flow transition, and measurements made
terminal velocity to the square root of the linear di- with a light reflecting probe. Their peak particle Rey-
mensions. The particle size was determined to satisfy nolds number was about 4.5.
 Tsukada et al. [14] compare the behavior of a cir-
culating bed fluidized at pressure between 0.1 MPa
and 0.35 MPa. They maintained the particle diameter,
the solids flow rate and the gas velocity constant. They
found similar results between 0.1 MPa and 0.18 MPa
but found considerable differences at the highest pres-
sure which they attribute to exceeding an upper limit
in Reynolds number or to a change in gas bypassing,
their highest value of Reynolds number was approxi-
mately 5.
The purpose of this paper is to establish simplifications
to the full set of scaling laws. We will experimentally
explore the limits of the viscous approximation for
scaling circulating fluidized beds. A simplified form of
the scaling law will be developed which is valid
for both the viscous and inertial limit. The simplified
form allows the behavior of a large combustor to be
simulated by a much smaller model. The derivation of
the simplified form will be followed by experimental
confirmation.
Derivation of simplified scaling laws
The full set scaling relationships are obtained by
non-dimensionalizing the equations of motion for the
particles and the fluid in a fluidized bed along with
their boundary conditions. Using the bed dimension,
L as a typical length dimension, non-dimensionalization
of the equations yields the following dimensionless
parameters (Glicksman [lo]).
(1)
Factors omitted include surface forces on particles
due to static charge or Van der Waals forces. Also,
the influence of the particle coefficient of restitution
or friction coefficient on inter-particle forces is omitted.
Litka and Glicksman [15] showed that the friction
coefficient and coefficient of restitution have a negligible
influence on bubbling beds.
In eqn. (l), j3 is the coefficient of the fluid-to-particle
drag force per unit volume expressed as @(u-v) where
u and v are the fluid and particle velocities, respectively.
In the general case, flL/psuo is related to the viscous
and inertial forces of the fluid through the Ergun
equation or through the expression for drag on a single
sphere. These relationships indicate that pL/p,u,
is dependent on the Reynolds number based on
the particle diameter, d,/L and the dimensionless
particle size distribution. Substituting these para-
meters into eqn. (1) yields the full set of scaling
parameters,
We will explore the various forms of the drag re-
lationships to investigate simplifications to the full set
of scaling parameters.
When the particles are closely spaced in a bubbling
bed or possibly in the lower, dense portion, of a
circulating bed, the Ergun equation is appropriate for
the drag forces,
$ =&-v]=15op~ +--E)2 cllu-VI +I 75
.
P
x h(l-E) pf/u-v12
(3)
TX-
Rearranging,
PL _ I50 e(I-- e) CLL + 1.75
P&o 2 PsUoWd2
x w--E) I=--211
EL
-F--- T&G Ps
for the most general case if 4, d,/L, pJp,, the dimen-
sionless particle size distribution, and pfuodp/p are
matched between the full sized fluidized bed and the
model, the dimensionless drag coefficient given by eqn.
(3) will remain the same over all conditions. Substituting
these dimensionless parameters for /3L/psuo in the group
given in eqn. (1) leads to the full set of scaling re-
lationships. For a circulating bed, the dimensionless
solid circulation rate must also be added to the list of
governing parameters in eqn. (1) and resulting simplified
forms of the scaling parameters. Note that with the
bed geometry, particle properties, gas superficial ve-
locity, and solids recycle rate fixed, steady state con-
ditions in the fast bed or riser are fixed. This should
be independent of the specific geometry and inventory
of the recycle system provided that the recycle system
can deliver the desired solids recycle rate. Studies by
Rhodes and Laussmann [16] and Chang and Louge [6]
have shown that the solids holdup is, in fact, independent
of changes in the inventory of the solids in the recycle
system.
To obtain more flexibility in the modeling process,
simplifications to the full set of scaling laws must be
identified. We will investigate this by first exploring the
simplifications which hold for several limiting cases.
These limiting cases will span the range of operation
from incipiently fluidized beds to dilute circulating beds
or pneumatic transport. With such a broad range, the
limits of the simplification can be explored over a wide
range of fluidization conditions. In the present paper
180

experimental results will be limited to circulating or (Glicksman [lo]), the independent governing param-
fast fluidization. eters are

4 L Lx Gs
- - - 4, particle size distribution
Low Reynolds number gL (d,Re) L2 psuo

(11)
At low particle Reynolds numbers the Ergun expres-
sion can be simplified using only the first term in eqn. It was demonstrated that this can be rewritten as
(3). 4 uo L G
Thus, z , G, c , z, , #I, particle size distribution (12)

The new simplified scaling parameters, eqn. (lo),

contain all of the viscous limit parameters along with
(5) the particle to gas density ratio. The new simplified
scaling relationships should cover a wider range than
At the same limit, the minimum fluidization velocity
the viscous limit previously defined by the author, since
can be written as, the inertial term in the fluid momentum equation is
retained. Only the drag term is simplified.
$ = (p.-pr)g(l- lmr>= 150 (l -$f)2 E (6)
P

For gas fluidized beds where ps-pf can be replaced

High Reynolds numbers
by PI>
Consider the limit of high particle Reynolds numbers
(7)
where the inertial term in the Ergun equations dom-
inates.
In this limit,
Substituting eqn. (7) into eqn. (5),
(1-e)
-- pf Llu-?I]
-- g(l - l) f&f3L @-
P&o
-175
* E Ps wp
(13)
(8)
$0 - (1 - Glf&nf P&o
where u =u/u, and v =v/u, and u -v is a dimen-
and sionless slip velocity. The minimum fluidization velocity
uo (1 - E)2Emf3 can be expressed as
EFr= r (9)
s mf (l- QfF
F =(ps-pf)g(l-Emf)=1.75 9 p$f (14)
Thus, in the low particle Reynolds number limit, P

maintaining uoIu,f, Emf, and Fr identical between two rearranging and using ps in place of ps-pf,
fluidized bed guarantees that /3L/p,u, is also identical.
Although 4 and d, are eliminated between eqns. (5) EC - QS _ 1 75 (1-cnS Pf
and (7), in general particle spheric@ and dimensionless 2 (15)
krlf *TsiG
size distribution should be held constant in the scaling.
The use of 4 and a mean diameter in the Ergun Substituting this into eqn. (13) and multiplying by Fr,
expression only approximates the effects of these pa-
Fr PL _ 4 %f3Lb -vk(l-4
rameters. Note that if the two models display identical 2
dynamic characteristics then E is a dependent variable P&o SL %lf

whose distribution throughout the bed should be iden-

=- 4 E,:(l- e)Iu-v]
tical for both fluidized beds. In this limit, the governing E (16)
r&f2
parameters given in eqn. (1) can be expressed as,
At large particle Reynolds numbers, just as at low
4 Ps uo L, G, , . . . . Reynolds numbers, the dimensionless drag, /3L/psuo, is
z , pf , zf, ~z , PJU, , c$, particle stze distribution
identical when uo/u,f, e,,,f and Fr are identical. E, u
and v are dependent dimensionless variables which
(10) are identical for two similar fluidized beds. In this limit
where e,,,f will be a function of particle sphericity and the same set of governing dimensionless parameters
size distribution. In the viscous limit, derived earlier applies as in the viscous limit, given by eqn. (10).
 181

Low slip velocity numbers, it is reasonable to expect that they hold, at

Finally, consider the case when the magnitude of
the slip velocity between the particles and the gas is
close to L&E everywhere in the fluidized bed. With
the vertical pressure drop equal to the particle weight
the following holds for any value of the particle Reynolds
number,
least approximately, over the entire range of conditions
for which the drag coefficient can be determined by
the Ergun equation or an equation of similar form.
The validity of the simplified parameters can be checked
numerically for the intermediate range of values.
From eqn. (4),

= P&l -e) (17)

x ---
u-4 b--2)l Pf L
(21)
E 4% Ps
P=ps&#-4 ;-: (18)
where u -v can also be represented as u,./u@ Using
and the Ergun equation to determine umf,
1U
-+=u,(l-E) L for u--2)- - mf (19) 42
L
=pg(l--6,)
s hf e uo
Again, when u olu mf and Fr are identical for two beds -%)2 P&f ~ 1.75 (I- Gnf) %n,zp,
= 150 (l - - (22)
and the slip velocity is close to u,,& the dimensionless %f3 w,)2 %nf3 MP
drag coefficient is also identical for two beds.
For all three limiting cases identified above, similitude Dividing eqn. (21) by (22), and rearranging,
can be obtained by maintaining constant values for the 1.75

-1
dimensionless parameters, 1+ - - tpsRe EIU- 2)I
Fr( 1 - E,~)EZ 150 (1 - e)
iL\ L
=
\PsUo) 1 I 1.75 -- Me umf
3~~05
&P*Um*Lz
-G
P&o
, C#J,
dimensionless
size
particle
distribution
(20)
(1
Zf (1 - c)emt
150 (1 - Ed) uo -I

(23)
provided that the forces between the particles and gas
where Re = pfuod,,lp.
can be represented by the Ergun equation or an equiv-
alent expression. It is easy to verify by use of eqn. (23) the three
limits defined previously.
The advantage of the simplified set of scaling pa-
For the more general case, Fig. 1 shows the value
rameters over the full set, eqn. (2), is the increased
of p given by eqn. (23) relative to /I at low Re over
flexibility in the design of a model to simulate a com-
a range of conditions when u,/u, is 10 and 3, re-
bustor or chemical reactor. With the full set, after the
spectively, and Fr and A remain constant. When u,/
gas properties inthe model have been chosen, e.g. by
U and the slip velocity are high there is a larger
use of air at ambient conditions, there is only one
vl:ation of dimensionless drag coefficient with Reynolds
unique set of particle size and density, bed size, gas
velocity and solids circulation rate which can be used
in the model.
Using the simplified set of scaling parameters, the
choice of the fluidizing gas in the model fixes the solid
density. However, the model size can be altered, as
long as the Froude number is maintained constant by
j] . Uo/Umf = 3, Uslip/Uo = 1

altering z+,. The particle size is then set to maintain

u~/u,~ constant. This flexibility allows a model to simulate
much larger combustors or reactors than is possible
with the full set of scaling relationships.

General case Al

1 10 100 1000 10000

Re = pUod,,/p
Since the same simplified set of dimensionless pa- Fig. 1. Ratio of drag coefficient to low Reynolds number drag
rameters holds exactly at both high and low Reynolds coefficient, u&,,~-~O and 3.
182

number. Note that J3does not vary with particle Reynolds

number when the Reynolds number remains above
about lo3 or below about 10. Figure 2 illustrates the
PFBC
results when u,/u, is 1000, a condition approached
with very fine particle bubbling beds or circulating beds.
In the latter instance the use of the Ergun relationship
is questionable except for the dense lower part of the 1_1
,
bed. In this figure the slip velocity is 1/50u0 or less,
I I
which corresponds to 20~~ This limit on the slip
velocity is roughly twice the terminal velocity of large
particles. Note, if Fig. 2 had been extended to higher Exact Model

Reynolds numbers, p would approach a constant value,

as it did in Fig. 1.
Of particular concern is the error in the dimensionless
D
drag coefficient when a scale model is designed using
the simplified set of scaling rules, eqn. (20). The sim-
114 scale P T(C) L UO
plified scaling parameters allow small models to simulate Simplilied Model
PFBC 1 MPs 800 Lo uo
a given sized combustor. As the length scale is reduced
the superficial velocity must also be reduced to maintain
a constant Froude number. The particle diameter must
UY Exacl Model

114 kale
0.1 MPs 20 Lo uo

Simplified Model 0.1 MPs 20 Lo/4 uoI2

also be reduced to keep u&,~ constant.
1116 Scale 1116 Scale
To determine the validity of the simplified scaling Simplfied~Model Simplified Model 0.1 MPa 20 Lo/16 b/4

laws over a wide range of conditions we have used the UY

simplified scaling laws, eqn. (20), to design hypothetical
models whose linear dimensions are l/4 and l/16,
respectively, of the linear dimensions of a model de-
signed using the full set of scaling laws, eqn. (2). To
determine the validity of the smaller, simplified models,
the dimensionless drag coefficient /3L/psuo will be com-
pared between the simplified models and the model
using the full set of scaling laws. Figure 3 shows a
comparison of the exact model and the simplified models
for a pressurized fluidized combustor. Using the full
set of scaling laws the exact model, fluidized by ambient
air, is approximately the same size as the combustor.
The simplified models are reduced in size by their
respective assumed length scale. The other parameters
of the simplified model are then calculated to match
1 10 100 1000
Re = pUodp/p
Fig. 2. Ratio of drag coefficient to low Reynolds number
coefficient, u&,~= 1000.
Fig. 3. Exact and simplified models of a pressurized fluidized
bed combustor.
the simplified parameters. For example, when the length
scale is reduced to l/4 that of the exact model, the
velocity is reduced by l/2 to keep the Froude number
constant. The particle diameter is then reduced ap-
propriately to keep the ratio of uo/ud constant. These
calculations were carried out over a range of particle
Reynolds numbers, RepE, based on the full scaling law,
or exact, model. It was found that the particle Reynolds
number for the l/4 scale simplified model remained
roughly equal to 0.34Re, over a wide range of values
for Re, whereas the particle Reynolds number for the
1/16.scale model was roughly OXXe,. Tables 2 and
3 give the values for the exact and simplified scale
models of pressurized and atmospheric combustors,
respectively.
Using these Reynolds number scale factors, the errors
in the dimensionless drag coefficient /3L/p,u,, using the
simplified scaling models can be found from Figs. 1
and 2. These errors are shown on Figs. 4 and 5 for
u&,,,~ of 10 and 1000, respectively, plotted as a function
of Rep, based on parameters for the exact scaled bed.
For a particle Reynolds number of 1000 or less, which
corresponds to pressurized beds with particles of 1 mm
or less, the error in the drag coefficient with the
simplified scaling laws is 20% or less for a one quarter
10000
length scale model. The error is 40% or less for a
l/16 length scale model. At uo/u,r of 1000 and u~~/u,,,~
drag of l/50 the errors for the l/16 scale model are 20%
or less for RepE less than 103. For particles of 0.2 mm
 183

TABLE 2. Scaling pressurized beds; p = 1 MPa, T=800 C

umf LILC, uoluoexact Re(d,)

(m s-) Recxna

Exact 1 0.39 8.82 1 4.4E + 0.4 1 1 1

0.61 0.193 11.84 1.34 9.9E + 0.3 l/4 l/2 0.31
0.41 0.097 16.4 1.86 3.3E+0.3 l/16 l/4 0.103
Exact 0.7 0.24 10.7 1 1.5E + 0.4 1 1 1
0.47 0.124 14.6 1.36 4SE + 0.3 l/4 l/2 0.34
0.33 0.064 19.7 1.84 1.57E + 0.3 l/l6 l/4 0.12
Exact 0.2 0.024 29.1 1 350 1 1 1
0.14 0.012 36.2 1.24 120 l/4 l/2 0.35
0.10 0.006 42.2 1.45 43.8 l/16 l/4 0.125
Exact 0.1 0.006 42.2 1 43.8 1 1 1
0.07 0.003 46.9 1.11 15 l/4 112 0.35
0.049 0.0015 49.7 1.18 5.15 l/l6 l/4 0.123

The two rows following the exact case present the particle diameter, velocity, etc. for ti linear dimension decrease of l/4 and l/16,
respectively, maintaining constant Froude number, u:lgL and constant u&,,~ pr = 2 500kgIm3, 4 = 0.8, e,,,f= 0.5, Re = pru,-,d,lcL.

TABLE 3. Scaling atmospheric beds; p = 1 MPa, T= 800 C

Knf
(m s-l)
d4nf -4 /Jr= !5E!&!?f L/L, UOIUO
sxact Re(d,)
f4nf P2 kuct

- f4

( umf 1 exact

Exact 2 1.65 9.12 1 3.5E + 0.4 1 1 1

1.25 0.83 12.29 1.34 8.6E + 0.3 l/4 112 0.313
0.84 0.41 17.04 1.86 2.6E + 0.3 l/16 l/4 0.105
Exact 1 0.563 14.7 1 4.4E + 0.4 1 1 1
0.685 0.778 20.2 1.37 1.4E + 0.3 l/4 l/2 0.34
0.48 0.14 26.8 1.82 484 l/16 l/4 0.12
Exact 0.5 0.15 26.1 1 547 1 1 1
0.353 0.075 33.1 1.27 192 114 In 0.35
0.25 0.038 39.7 1.52 68 l/16 l/4 0.125
Exact 0.2 0.024 43.3 1 35 1 1 1
0.14 0.012 47.6 1.1 12 l/4 l/2 0.35
0.10 0.0061 49.9 1.15 4.4 l/16 l/4 0.125
Exact 0.1 0.0061 49.9 1 4.4 1 1 1
0.07 0.003 51.2 1.03 1.5 l/4 l/2 0.35
0.05 0.0015 51.7 1.04 0.55 1116 l/4 0.125

The two rows following the exact case present the particle diameter, velocity, etc. for a linear dimension decrease of l/4 and l/16,
respectively, maintaining constant Froude number, u&L. and constant u o/u ,+ pa= 2 500kgfm3, 4=0.8, l=0.5, Re =pfuodp/p.

or less, corresponding to a Reynolds number of 100
or less, the errors in drag coefficient are minimal. When
the Ergun equation applies for the drag coefficient, a
one quarter scale model based on the simplified scaling
laws should be valid for any conditions. A l/16 scale
model should be valid for diameters of about 0.2 mm
or less for a pressurized bubbling bed with u,Ju,r of
10 and u,,fuO of 0.3. At u&,,,~ of 1000 and u,r/u, of
l/50, the l/16 scale model should be valid for pressurized
beds with particles up to 1 mm in diameter. These
conclusions apply when the particle to fluid drag term
is given by the Ergun equation or similar relationships
and the scaled particles are not so small that inter-
particle surface forces come into play.
The particle to gas density ratio does not appear
explicitly in the expression for /3, eqn. (23). However,
184
1 10 100
Re- Uodp
Fig. 4. Error in drag coefficient, u,/~~=lO.
- empirical expression given by White [17],
II . ..
ULexact = l/4
ULexacl = 1/I 6
0.3 ,,.,, ,,,,,I
1 10 100
Re- Uod,
Fig. 5. Error in drag coefficient,
uti is a function of the density ratio. If the wrong solid
to gas density ratio is used in the simplified model,
the particle diameter in the simplified model required
to match ~&,,r changes. For example, if the simplified
model of a combustor pressurized at 1 MPa uses a
particle density of 2 500 kg me3 instead of the required
900 kg rnm3, the Reynolds number, ~&.,,&~, is reduced
by 40%. This increases the error in p between the
simplified and the exact model when /3 is derived from
the Et-gun equation.
For scaling beds with intermediate or large Reynolds
numbers the solid to gas density ratio is an important
parameter of the fluid momentum equation. It must
appear in the set of scaling parameters irrespective of
its influence on the drag coefficient.
In the freeboard of a bubbling bed or
portion of a circulating bed where particles
considered to act in clusters or groups,
amination of scaling of the gas to solid
made. Consider all of the particles grouped into clusters
with an effective diameter d, and the clusters occupying
a volume fraction lc of the bed volume. The cluster
to gas drag will be represented by the drag coefficient
for a solid sphere of diameter d,,
This can be rewritten as,
PL
-=-- 3 Pt
~u~---zIqcD 4 (l--E,) (25)
P&o 4 0 Ps c
1000 10000
If the diameter of individual particles does not in-
fluence the drag of a cluster of particles, then when
the solid to gas density ratio is held constant between
the combustor and the model, the dimensionless drag,
/3L/psuo, in eqn. (25) is a function of CD. We will use
the drag coefficient of solid spheres as a first approx-
imation for the relative change of CD with Reynolds
number based on cluster diameter. The drag coefficient
of a solid sphere can be closely represented by the
I c,= $ + 1+&G6
dc
+ 0.4
In the limit of large cluster Reynolds number, CD
,,,,I ,....I approaches a constant and pL/psuo is properly scaled
1000 10000
when the fluid to gas density ratio is held constant
between the model and the combustor. The ratio Ll
u,,/u~=~~. d,, and ecl should be the same between the properly
scaled model and the combustor. In this limit, uo/u,r
does not appear explicitly as one of the simplified
scaling parameters. However, if the cluster was formed
by bubble eruptions at the surface of a bubbling bed
u,/u, will certainly be important for the bed. Similarly,
in a fast bed, uo/umf will be important in the lower,
dense portion of the bed.
In the limit of small cluster Reynolds number, the
dimensionless drag pL/p,u, will not be the same for
the exact model and the simplified scale model unless
the cluster size scales with the particle diameter rather
than the bed dimensions.* However, as we will see
below, under any realistic conditions the cluster Rey-
nolds number will be near the upper limit.
If the reduced scale models faithfully reproduce the
dynamics of the exact case, then the cluster dimensions
should scale directly with the linear dimensions of the
bed. Thus, a one quarter linear scale model which has
*In the lower limit of Reynolds number, for clusters and particles,
C, is equal to 24/Re,, and u,~ is proportional to p.d,?p, eqn.
in the upper (7). It can be shown that @L/p,u, in eqn. (24) is proportional
are generally to l/Z+ L+,/u,,,~d,*/df. Complete similitude is not obtained unless
a similar ex- d, is also proportional to the bed length, resulting in the full
drag can be set of scaling laws.
 185

fr.
PL
-=-
PSUO 4
3
0
!?i Iu_-2)Ic
Ps D4
(1-e)
(27)

where C,, is the drag coefficient of a single particle.

- ULexact = l/16
This can be rewritten in terms of the single particle
Re: 0.1 MPa, 6 m/s, dc = 0.3m
terminal velocity which can be found from,

Psd3& _ 1 Pf+i c u2
6 24 Dt (W
Substituting eqn. (28) into eqn. (27) to eliminate CD,
one obtains,
I

10000
Re = pUodc/p (29)
Fig. 6.. Error in cluster drag coefficient for fixed u&,,,~ using C,
for a solid sphere.
Since uO/umfand Fr is held constant in the simplified
scaling process, we will examine the ratio uJumf to
determine if the drag coefficient, /3L/psuo, remains
a velocity one half that of the exact case will have a
constant.
cluster Reynolds number one eighth that of the exact
The Ergun equation, eqn. (4), can be solved to find
bed. From the relationship of CD with Re we can
U mb
determine the change of CD with model scale at a
given Reynolds number of the exact bed. Figure 6 &+&e=
shows the shift in ,C,, for length scale of l/4, l/8, and

1+
P
l/16, respectively, of the exact bed length as a function
of the cluster Reynolds number of the exact bed. Also - (150)(1- Q2 - (150)(1 - Q2 7(1 - %fY 43Ar
3
+
shown on the figure is the typical Reynolds number Gnf \ir %f3 Gnf
of an atmospheric combustor with a 0.3 m cluster
diameter, approximately 1.5X 104. In a bubbling bed, 3-w - %fM
3
the cluster diameter in the freeboard should be at least Glf
(30)
equal in size to the diameter of bubbles erupting at
the bed surface. For beds with horizontal tubes, the where Ar=pspfd3&p is the Archimedes number.
bubble diameter will be equal to or larger than the In the limit of low Ar, eqn. (30) becomes
3
horizontal tube spacing. In a bubbling bed without PfU-fde Gnf
Ar At-0 (31)
tubes, the bubbles and clusters can be much larger. In
CL = 150( 1 - lm*)
an open circulating bed the cluster diameter is more
difficult to determine. It is reasonable to assume that while for large Ar,
its diameter is proportional to the bed diameter, equal
in magnitude to the bed diameter or one order of (32)
magnitude smaller. From these considerations, the Rey-
nolds number based on the cluster diameter should be The terminal velocity for a particle can be obtained
104 or larger in an atmospheric combustor with a cluster as,
diameter of 0.2 m. The cluster Reynolds number should
be 105 or larger in a pressurized combustor. From Fig. D
6 it can be seen that a one quarter scale or an eight
scale model should have drag coefficients similar to Rearranging, and using eqn. (26) for CD,
the exact bed. For pressurized beds, the drag coefficients 4
should be very close in magnitude.

Terminal velocity of particles

If the drag coefficient, p, is influenced by the char- where
acteristics of individual particles, the detailed particle
pfutd
dynamics of the simplified scale models must be ex- Ret= -
amined. CL
In this case, For small values of Ar, eqn. (34) becomes
186

- WUmf

T=ECOC

dp = 0.2mm
0 0 A

I ,.,,,,a ,,,.,.I ,,,,I ,,,

I IO 100 1000 IOOiM lOOooO

Ar = SP,P~~~C~
0.1 1

Fig. 7. Ratio of terminal velocity to minimum fluidization velocity dp (cnmbustor), mm

vs. Archimedes number.

Fig. 8. Error in terminal velocity using simplified scaling at 1
MPa and 800 C.

Ret= $Ar Ar-0 (35)

while for large Ar

Re, = AAr Ar>l (36)

\i
At small and large values of Ar, the ratio of u&,,,~
approaches a constant value. In these two limits, the
simplified scaling laws will yield exact agreement of u,/
u,.,,~between the combustor and the simplified models.
In general, umf and u, should be a function of the
II
Archimedes number for a given value of 4 and E,,,~
0.1 I
Figure 7 illustrates the dependence of u,/u,~ on Ar. dp(combustor). mm
Also shown are the values of the Archimedes number Fig. 9. Error in terminal velocity using simplified scaling at 0.1
for a pressurized combustor at two particle diameters. MPa and 800 C.
Associated with the Archimedes number for the exact
model are the corresponding Archimedes number for ratio of uO/uti constant. To verify the new simplified
two models scaled using the simplified relationships at set of scaling parameters, tests were carried out com-
linear dimensions one quarter and one sixteenth the paring two geometrically similar beds, one four times
exact dimensions. larger than the other, fluidized with the same density
The errors in u,/u,f are shown on Figs. 8 and 9 for particles maintaining Froude number, u&,,,~, and the
simplified scale models at two different linear dimen- solid to gas density ratio constant. Further tests were
sions. Scaling a combustor with comparatively small then conducted to verity these parameters when scaling
particles, 0.2 mm or less, gives good agreement for u,l from hot combustors with dimensions 16 times that of
u,~ even at one sixteenth linear scale, while for large the model.
particles a linear scale of one fourth gives fair agreement
for u~/u,,,~Since z.+,/u,~is held constant in the simplified
scaling laws, close agreement of ut/umf also results in Experimental arrangement
close agreement of u,/uO.
Three cold and one hot circulating beds were used
during the course of the study:
Experimental (1) l/4 scale model of the UBC pilot scale combustor
Experiments were performed to test both the viscous (Fig. 10);
limit scaling, given by eqn. (12), and the new simplified (2) l/4 scale model of the 2.5 MW Studsvik combustor
scaling parameters, given by eqn. (10) in circulating (Fig. 11);
fluidized beds. For the viscous limit, experiments were (3) l/16 scale model of the 2.5 MW Studsvik com-
carried out in the same bed using particles of different bustor;
density while maintaining the Froude number and the (4) Studsvik 2.5 MW combustor.
 187

- Exhaust Air (To Filter)

terials were selected: steel, glass, and plastic. These

r
materials were readily available and allowed for the
-
evaluation of a wide range of density ratios. The powders
were prepared so that scaled size distributions and
fluidization properties matched those required for the
tests in which they would be used. Powder testing
included size analysis by sieving, measurement of umf,

_
4.
density measurement with a displacement pycnometer
4
or water displacement, and bulk density measurements.
Particle properties are given in Table 5.
Figures 12 through 15 show cumulative particle size
- Pressure Relief Line
(cm above prima distributions for particles used in a given scaling test
i
I series.
The required mean particle diameter needed in the
- Solid Flux Measurement Valve l/16 scale model of the Studsvik hot bed for using the
simplified scaling laws was calculated to be approxi-
mately 35 microns. Since the smallest sieve mean di-
47.15
ameter was 38 microns, it was not possible accurately
/- Retrn
Leg to predict the size distribution of the particles required.
In order to circumvent this problem, no attempt was
made accurately to measure the mean diameter of the
particles for use in the l/16 model, rather the particle
/- Air
L-Vave distribution was selected based on the minimum flui-
i
dization velocity. This is possible since the particle
diameter enters only indirectly through the minimum
Fig. 10. l/4 scale model of the UBC combustor. fluidization velocity.
In addition to the separation and recombination
cyclone Return process to achieve satisfactory size distributions, the
glass and plastic particles were also treated with anti-
static compounds in an attempt to reduce the level of
static electricity generated in the bed. The first com-
pound was Larostat 519, a fine powder which was mixed
in with the glass and plastic powders. Since the mass
fraction of Larostat in the mixtures was very small, it
was not anticipated that this would have any significant
effect on the hydrodynamics of the beds other than
through the reduction of static electricity. Copper ribbon
and rods were also used to aid in the discharge of
static in return lines and other equipment.
While the Larostat worked quite well in the reduction
of static, as more was added during operation, it began
severely to inhibit the flow characteristics of the glass
and plastic mixtures in the LValves and other con-
stricted areas. In order to further reduce bed static
levels, the glass and plastic particles and the insides
of the beds were spray coated with Anstac 2-M,an
Prim. Air I , alcohol based anti-static solution. After drying, this
a Scale(m) OS
virtually eliminated any remaining static effects.
Fig. 11. l/4 scale model of the Studsvik combustor. The solids circulation rate in the Studsvik combustor
was determined by measuring the velocity of an isotope
A summary of bed dimensions and the tests for which pill in the L-valve standpipe. This method indicated
they were used is provided in Table 4. considerable variations in measured solids flux for con-
In order to evaluate the applicability of the viscous secutive velocity measurements at the same condition.
limit scaling laws, and to simulate hot circulating beds In the l/16 scale Studsvik bed was measured by timing
operating at different pressures, three bed solid ma- pile-up of solids above a valve in the return leg. The
188

TABLE 4. Summary of beds used in tests

Bed Bed height A, Tests in which used

(m) (m)

l/4 scale UBC bed 1.62 l.l29E-3 Glass/steel viscous limit scaling
(no seconday air)
l/16 scale Stud&k bed 0.46 1.05E-3 glass/plastic viscous limit
(no boot or secondary air) scaling and simplified
scaling with constant pJpf
l/4 scale Studsvik bed 1.83 1.68E-2 scaling with constant pJpf
(no boot or secondary air)
l/16 scale Stud&k bed 0.46 1.58E-3 hot bed scaling with the
simplified scaling parameters
l/4 scale Studsvik bed 1.83 1.68E-2 hot bed scaling with the
full set of scaling parameters
Studsvik hot bed 8.0 0.43 hot bed scaling

TABLE 5. Particle characteristics

90 _ Glass Sample I
Density Mean Loose Sphericity U,,,, (Actual dp = 88.3)
80
(kg rnm3) diameter pack (c&s)
pm voidage

Glass I 2540 88.3 0.42 1.0 1.4

Glass II 2540 81.6 0.42 1.0 1.7
Glass III 2540 78.7 0.42 1.0 1.4
Glass IV 2540 112.3 0.42 1.0 2.8
Steel I 7250 49.5 0.65 0.6-0.8 1.4
Steel II 7250 -26 0.65 0.6-0.8 0.75
Steel III 7250 57.7 0.65 0.6-0.8 1.5
Plastic I 1400 99.5 0.55 0.6-0.8 0.85
Plastic II 1400 144.5 0.57 0.6-0.8 1.55
Studsvik hot 2700 243.0 0 IO 20 30 40 50 60 70 80 90 100
bed sand Equivalent Steel Size, microns

Fig. 13. Particle size distributions for viscous limit scaling laws.

80 80
l/16 Scale Studsvik Bed
l/4 Scale Studsvik Steel 70 _ Glass (Actual dp = 78.7)

160

0 20 40 60 80 100 120 140 160 180 200

Equivalent Hot Bed Particle Size, microns Equivalent l/4 Scale Bed Size, microns

Fig. 12. Particle size distributions for verification of the exact Fig. 14. Glass particle size distributions for simplified scaling.
scaling laws.

solid circulation rate in the l/4 scale UBC bed and by the visual contrast in solids. Both the valve pile-up
the l/4 scale Studsvik bed was measured by timing the and visual descent methods have been shown to give
rate of solid descent in the downcomer made possible acceptable results with agreement within 15% (West-
 189

Viscous limit scaling

1
The first set of simplified parameters evaluated were
l/16 Scale Studsvik Bed the viscous limit parameters developed previously
- Plastic (Actual dp = 99.5)
(Glicksman, [l]) and discussed above (see eqn. (12)).
In order to evaluate the validity of applying these scaling
parameters in the fast fluidization regime, two series
of tests were conducted. In these tests the values of
Fr, u,Iu, and bed geometric similarity were held
constant while the solid to gas density ratio was varied.
Table 6 provides a summary of the viscous limit tests.
The glass/steel tests were conducted in the l/4 scale
0 20 40 60 80 100 120 140 160 180 200
model of the UBC bed, and the glass/plastic tests were
Equivalent l/4 Scale Bed Size, microns conducted in the l/16 scale model of the Studsvik bed.
Fig. 15. Plastic particle size distributions for simplified scaling. Average solid fraction for several of the viscous limit
tests are provided in Figs. 16 through 19. Error bands
for these and all following solid fraction profiles are
based on deviations in the pressure and solids circulation
measurements and represent one standard deviation
phalen [18]). In the cold beds, differential pressure of the calculated average solid fraction.
transducers were used to measure time-mean and dy-
namic AP between adjacent pairs of pressure taps.
Pressure drop data were recorded with a computer
data acquisition system. In the Studsvik combustor, only loo
time averaged pressure drops were recorded. Solid M = 0.0029
fraction (SF) discussed in this paper is based on AP Fr = 0.57
measurements. Since SF only includes gravity effects,
this interpretation neglects particle acceleration and
shear stress an the column walls. The M measurements rp lo Resd = 9.8
i u, = 3.0 m/s
.=
yield an apparent solid fraction.
For all cold bed data, histograms of the solid fraction 8
9
data were created to assess whether pressure fluctuation A 1
amplitudes of scaled beds were similar. In addition,
fast fourier transforms (FFTs) of the pressure fluctuation
data were done to determine the similarity of pressure
fluctuation frequencies in scaled beds. The sampling
0.1
frequency used for the comparisons was 100 Hz. Earlier 0 10 20 30 40 50 60 70 80 90 100
studies have shown that this sampling rate is adequate Bed Height, %

for measurement of the cold bed fluctuations of interest Fig. 16. Solid fraction profiles, glass/steel viscous limit scaling,
(Westphalen, [l&3]). low velocity case.

TABLE 6. Viscous limit scaling

Glass vs. steel Glass vs. plastic

Glass 1 Glass 2 Steel Glass Plastic

u. (m s-l) 3.0-7.0 3.0-7.0 3.cb7.0 1.5-2.5 1.5-2.5

u,~ (cm s-l) 1.40 1.70 1.40 1.42 0.85
dp (pm) 88.3 81.6 49.5 78.7 99.5
Bed height (m) 1.62 1.62 1.62 0.46 0.46
Bed A,, (m) l.l29E-3 l.l29E-3 l.l29E-3 l.O5E-3 1.05E-3
Fr 0.57-3.1 0.57-3.1 0.57-3.1 0.50-1.4 0.50-1.4
Red, 17.4-40.9 16.2-37.8 9.8-22.9 7X-13.0 9.9-16.5
M (10-q 2.9 2.9 2.9 4.3-7.2 4.3-7.2
AIf% 2117 2117 6042 2117 1167
 When utilizing the viscous limit scaling laws to scale
M = 0.0029 between particles of different densities, the average
Fr= 1.90 solid fraction profiles were not in good agreement, even
below particle diameter based Reynolds numbers where
the viscous limit parameters should hold. Surprisingly,
the best agreement occurs as the superficial gas velocity
and particle Reynolds number is increased. Solid frac-
tion probability density functions (PDFs) and power
spectral densities (FFTs) were also in poor agreement.
The two different glass particle distributions, Glass I
and Glass II, in general showed good agreement, in-
dicating that changes in the size distribution did not
cause major differences in the bed hydrodynamics.
0 10 20 30 40 50 60 70 80 90 100 Figures 20-27 depict typical results for the probability
Bed Height, I
density functions and fourier transform plots. As the
Fig. 17. Solid fraction profiles, glass/steel viscous limit scaling, superficial velocity and particle Reynolds number was
high velocity case.
increased, the similarity in solid fraction profiles im-
proved. This is counterintuitive since the fluid inertial
effects should increase with velocity. However, addi-
tional parameters may be important near the boundaries
between flow regimes. For these experiments the flow
M = 0.0072
Fr = 0.50 may have been choked.
Rep, = 9.88
Reg, = 7.81 5 I
uo = 1.5 m/s
c
M = 0.0029
Fr = 0.57
= 17.4
Regll
= 16.2
= 9.8
uo = 3.0 m/s

0 lo 20 30 40 50 60 70 80 90
Bed Height, TO

Fig. 18. Solid fraction profiles, glass/plastic viscous limit scaling,

low velocity case.
1 3 5
Solid Fraction

Fig. 20. PDFs, glass/steel viscous limit scaling, low velocity case.
100 I I
M = 0.0043 2
Fr = 1.39 I -O- Glass 1
= 16.5 M = 0.0029
RepI
Reel = 13.0
I

I 20 30 40 50 60 70 80 90
Bed Height, %
Solid Fraction
Fig. 19. Solid fraction profiles, glass/plastic viscous limit scaling,
high velocity case. Fig. 21. PDFs, glass/steel viscous limit scaling, high velocity case
 191

50

M = 0.0029
Fr = 0.57 M = 0.0043
10
Fr = 1.39

&

1 1

b
5
9
0.1

0.01 0

Solid Fraction

Fig. 22. FFIs, glass/steel viscous limit scaling, low velocity case. Fig. 25. PDFs, glass/plastic viscous limit scaling, high velocity
case.

50
M = 0.0029 I

10

g l

0.1

I
0.01 ~~~~~~~~~~~~~~~~~.~.
0 10 20 30 40 50 0.01
Frequency 0 10 20 30 40 50
Frequency
Fig. 23. FFTs, glass/steel viscous limit scaling, high velocity case.
Fig. 26. FFTs, glass/plastic viscous limit scaling, low velocity case.

10
Rep, = 9.88
u, = 1.5 m/s

-25 0 25 50 75 100 0.01

Solid Fraction
0 10 20 30 40 50
Fig. 24. PDFs, glass/plastic viscous limit scaling, low velocity case. Frequency

Fig. 27. FFTs, glass/plastic viscous limit scaling, high velocity

Choking in CFBs depends on particle properties, gas
properties, and bed geometry. In more dilute systems
such as pneumatic transport, choking occurs when the
superficial velocity is decreased to the point where the
particle weight overcomes the gas shear forces and
case.
suspension begins to collapse. The solids are then
conveyed upwards in dense phase slugging flow (Satija
er al. [19]). Yang has proposed an empirical relation
192

for the onset of choking based on a study of dense they may have a wider range of validity in the bubbling
pneumatic systems (Yang [20]): regime.
2.2
2gD(g4.7-1)
=681x105 pf (37)
(%-~d2 * 0 Ps Simplified scaling with constant ZQ,/U,,,~
and ps/pf

along with an equation for the carrying capacity of the
gas
G=p, (l-6,)
Equations (37) and (38) allow for the calculation of
the flux and voidage at incipient choking given a su-
perficial velocity. Chang and Louge [6] compared Yangs
correlation with data they obtained in a cold circulating
fluidized bed for plastic and steel powders and found
it to be in excellent agreement.
Table 7 presents the non-dimensional loadings at
incipient choking predicted by the Yang correlation,
for the superficial velocities used in the tests of the
viscous limit scaling laws. In the runs at lower velocities,
Yangs correlation predicts that the beds were in or
near the choking regime. Because the Yang correlation
indicates that choking is a strong function of the solid
to gas density ratio, it may be that the viscous limit
scaling laws are unable to model bed hydrodynamics
near the boundary between different flow regimes. This
would explain the better agreement at the higher velocity
runs where the beds were not choked. These results
suggest that there are competing criteria which specify
the range of superficial velocity in which viscous limit
scaling laws apply for circulating fluidized beds. On
the one hand, superficial velocity must be kept low
enough to remain in the viscous limit. On the other
hand, the superficial velocity must be high enough to
prevent significant choking in the bed. While the viscous
limit scaling laws are limited in their applications to
fluidized beds operating in the fast fluidization regime,
The second set of simplified parameters which were
investigated were the simplified parameters based on
constant Froude number, uO/umf and pJpf, which have
been described above. In order to evaluate the validity
of these scaling parameters in the fast fluidization
regime, two series of tests were conducted. Table 8
provides a summary of the tests to evaluate the simplified
scaling parameters. Average solid fraction profiles for
several of the simplified scaling tests are provided in
Figs. 28 through 31.
When utilizing the simplified scaling laws to scale
between properly sized plastic or glass samples in two
geometrically similar beds the average solid fraction
profiles are in good agreement. For all superficial
velocities tested, there is close agreement even though
particle diameter based Reynolds numbers as well as
Reynolds numbers based on the bed diameters varied
significantly between corresponding runs made in the
two beds. Nearly all parts of the average solid fraction
profile curves were within confidence interval limits.
The probability density functions and power spectral
densities for these runs, several of which are depicted
in Figs. 32 through 39, also demonstrated good agree-
ment - especially in light of the poor agreement
obtained when the density ratio is not held constant.
Table 9 presents the non-dimensional loadings at
incipient choking predicted by the Yang correlation for
the superficialvelocities used in the tests of the simplified
scaling laws. Because Yangs correlation predicts that
these tests were conducted under conditions in which
the bed was choked, the above analysis indicates that
while the viscous limit scaling laws will not allow for
modeling bed hydrodynamics in regions approaching

TABLE 7. Choking predictions for the viscous limit tests - Yang correlation

Figure no. for Particle Superficial velocity Predicted M Measured M

corresponding (m s-) at incipient choking during tests
solid fraction profile (Yang correlation)

16 Glass I 3.0 4.2E-4 2.9E-3

16 Glass II 3.0 4.2E-4 2.9E-3
16 Steel 3.0 3.7E-3 2.9E-3
17 Glass I 5.5 0.133 2.9E-3
17 Glass II 5.5 0.136 2.9E-3
17 Steel 5.5 1.6E-2 2.9E-3
18 Glass 1.5 6.6E-3 7.2E-3
18 Plastic 1.5 3.8E-2 7.2E-3
19 Glass 2.5 2.6E-2 4.3E-3
19 Plastic 2.5 0.108 4.3E-3
 193

TABLE 8. Simplified scaling with constant pJpf 100

M = 0.0030
1/4 vs. 1/16 Glass 1/4 vs. 1/16 Plastic Fr = 0.50
Re1/4 = 26.7
1/4 Glass 1/16 Glass 1/4 Plastic 1/16 Plastic l \ ~%N. ' K Ret/t6 = 9.88
10
uo ( m s -I) 3.0-5.0 1.5-2.5 3.0-5.0 1.5-2.5 ~" ~ = 15m/s.
umt(era s -t) 2.88 1.42 1.50 0.85
dp (/zm) 112.3 78.7 144.5 99.5
Bed height 1.83 0.46 1.83 0.46
(m)
Bed A~ 1.68E-2 1.05E-3 1.68E-2 1.05E-3
(m~) O 114 Scale Studsvik Plastic
Fr 0.5-1.4 0.5-1.4 0.50-1.4 0.50-1.4 ~ 1/16 Scale 8tudsvik Plastic
Reap 22.2-39.0 7.8-13.0 26.7--47.8 9.9-16.5 0.1 , f I I I i , I , I , I
M (10 -3) 4.3-7.2 4.3-7.2 2.3-3.0 2.3-3.0 0 10 20 30 40 50 60 70 80 90
p,/~ 2117 2117 1167 1167 Bed Height, %
Fig. 30. Solid fraction profiles, plastic simplified scaling, low
velocity case.
100

M = 0.0072
Fr = 0.50 100
~ :1: Ret/4=22"2 M = 0.0023
Rel/t6 = 7.81 Fr = 1.39
10 .]. ~ I'%~,~ U01'4 =3"Om/s Ret/4 = 47.8
~ 10 Ret/16 = 16.5
Uot/4 = 5.0 m/s
.-~ Uol/t 6 = 2.5 m/s
r~ 1 O 1/4 Scale Studsvik Glass .
--41-- 1/16 Scale Studsvik Glass

I 1/4 Scale Studsvik Plastic

0.1 , I I I I I , I , I , I i
10 20 30 40 50 60 70 80 90 ~ 1/16 Scale Studsvik Plastic
Bed Height, %
0.I i ; i I , I , i I , I , I ,
Fig. 28. Solid f r a c t i o n profiles, g l a s s s i m p l i f i e d scaling, low velocity 10 20 30 40 50 60 70 80 90
case. Bed Height, %
Fig. 31. Solid f r a c t i o n profiles, plastic s i m p l i f i e d scaling, h i g h
velocity case.
100
M = 0.0043
Fr = 1.39
Rel/4 = 37.2
Ret/16 = 13.0 C h o k i n g in fluidized beds
10 "t~ Uol/4 = 5.0 m/s

Several correlations have been developed in attempts

!1 accurately to predict the choking velocity for various
particle types at a given mass flux (Leung et al. [21],
X 1/4 Scale 8tudsvik Glass Yousfi and Gau [22], Yang [20], Punwani et al., [23]
and Matsen [24]). Chong and Leung [25] have compared
.~. 1/16 Scale Studsvik Glass
choking gas velocity correlations in order to determine
0.1 i J J I , I , , I , I , I ,

10 20 30 40 50 60 70 80 90
which correlations are best suited for each class of
Bed Height, % gas-solid system using Geldart's [26] classification. The
Fig. 29. Solid f r a c t i o n profiles, g l a s s s i m p l i f i e d scaling, h i g h v e l o c i t y data used for comparison consisted of particles with
case. various sizes and densities. The root mean square
relative error of the Yang correlation is the lowest
when all particle groups are included. A brief dimen-
or fully within the choking regime, the simplified set sional analysis of the Yang correlation is presented
of scaling parameters will. below.
194

1.5

2
= Re,,d = 22.2
8 1
Re ,,,6 = 7.81
d
x Uo,,4 = 3.0 m/s
.=
Uo*,,6 = 1.5 m/s
B
a
55 Fr= 1.39
a 0.5
= 37.2
R%
P Re ,,t6 = 13.0
Uo,,4 = 5.0 m/s
Uo,,,6 = 2.5 m/s
0
-2 4
Solid Fraction

Fig. 32. PDFs, glass simplified scaling, low velocity case. Fig. 35. FFTs, glass simplified scaling, high velocity case.

M = 0.0030
Fr = 0.50
Re,,d = 26.7
u Re ,,,6 = 9.88
Uot,4 = 3.0 m/s
Uot,,6 = 1.5 m/s

0
0 I 2 3 -10 0 10 20 30 40 SO
Solid Fraction Solid Fraction

Fig. 33. PDFs, glass simplified scaling, high velocity case. Fig. 36. PDFs, plastic simplified scaling, low velocity case.

SO 0.25

& l/4 Scale Studsvik Glass

-__ l/16 Scale Studsvik Glass 0.2

8
i:
P
k 0.15 M = 0.0023
.6 Fr= 1.39

d
a3
8 .E 0.1
Fr = 0.50 f
9
Re,/, = 22.2
0.1
Re ,,,6 = 7.81 f
0.05
Uo,,4 = 3.0 m/s
Uot,,6 = 1.5 m/s
0.01 , 3 * - 1* 3 1. I 0
0 10 20 30 40 50 0 10 20 30
Frequency Solid Fraction

Fig. 34. FFTs, glass simplified scaling, low velocity case. Fig. 37. PDFs, plastic simplified scaling, high velocity case.

Non-dimensionalizing eqn. (37) as Rewriting the denominator of the left hand side,

2 (1u:
go (c4.7-1) 2 =6.81X105 0 f
2.2

( l-2 mf c1
 195

& 114 Scale Studwik Plastic

-II
- -- l/16 Scale Stud&k Plastic

Re ,,,6 = 16.5
u o,,4 = 5.0 m/s
Uo,,*6 = 2.5 m/s
0.01 0.01 * c n B
0 10 20 30 40 50 0 10 20 30 40 50
Frequency Frequency

Fig. 38. FFTs, plastic simplified scaling, low velocity case. Fig. 39. FFTs, plastic simplified scaling, high velocity case.

TABLE 9. Choking predictions for the tests of the simplified scaling laws - Yang correlation

Figure no. Particle Superficial velocity Predicted M at Measured hi

for corresponding (m s-r) incipient choking during tests
solid fraction protile (Yang correlation)

28 l/4 scale glass 3.0 7.3E-3 7.2E-3

28 l/16 scale glass 1.5 6.6E-3 7.2E-3
29 l/4 scale glass 5.0 2.7E-2 4.3E-3
29 l/16 scale glass 2.5 2.6E-2 4.3E-3
30 l/4 scale plastic 3.0 3.6E-2 3.OE-3
30 l/16 scale plastic 1.5 3.8E-2 3.OE-3
31 l/4 scale plastic 5.0 9.7E-2 2.38-3
31 l/16 scale plastic 2.5 0.108 7.2E-3

Equation (38) results in the following dimensionless
equation:
25. =(i _ lL5)(1_Ee)
P&c
Using eqns. (31), (32), (35), and (36) it was shown
that u,/u~ approaches a constant value for both small
and large Archimedes numbers. At intermediate values
of Ar, Figs. 8 and 9 show that the error in u,/u,
between beds scaled with the simplified scaling pa-
rameters is modest. Thus, throughout the entire range
of Archimedes numbers, Yangs choking expression is
dependent on the following parameters:
go
_.- wllf
and B which are identical to the simplified
2 7
UC UC Pf
scaling parameters.
The non-dimensional parameters which govern the
Yang correlation include a density ratio. This indicates
that the viscous limit scaling will not accurately predict
the choking behavior of circulating fluidized beds. This
may explain the poor agreement in solid fraction profiles
and probability density functions when scaling with
glass/steel and glass/plastic at low velocities. However,
when the density ratio is included, as it is with the
(41) simplified scaling laws, all the dimensionless parameters
which govern the Yang correlation are included in the
scaling laws.
Applying simplified scaling laws for scaling hot beds
In order to evaluate the validity of applying the
simplified scaling parameters to large scale hot beds
using length ratios on the order of l/16, solid fraction
profiles between the Studsvik 2.5 MW combustor and
a l/16 scale cold model were compared for six different
operating conditions. Table 10 provides a summary of
the tests to evaluate the simplified scaling parameters
when applied to a commercial sized hot bed. Figures
40 through 43 depict solid fraction profiles for several
of these runs. Included in these plots are results from
scaling the hot combustor. using the complete set of
scaling parameters obtained as part of an earlier study
(Westphalen, [27]).
196

TABLE M. Simplified scaling with constant p,/~+ - large hot 100

bed scaling

10
Studsvik l/4 scale l/16 scale
hot bed cold bed cold bed e Primary Air = 69.0
i
.g
u. (m s-l) 6-8 3-4 1.5-2.0 Eu 1

u,~ (cm s-) 1.40 0.75 9

dP (cun) 243 57.7 -26 ::
Bed height (m) 8.0 1.8 0.46 0.1
Bed A,, (m) 0.43 1.68E-2 1.58E-3 -El_ 114 Scale Cold Bed
Fr OS-O.9 OS-O.9 0.5-0.9
Red, 10.3-13.6 11.1-14.2 2.53-3.30 0.01
M (10-3) OS-l.8 0.5-1.8 0.5-1.8 0 10 20 30 40 50 60 70 80 90 100
PJPf 8400 6200 6200 Bed Height, %

Fig. 42. Solid fraction profiles, hot bed scaling with simplified
100
scaling laws, high velocity, 69% primary air.

10
Hot Bed Data:
ER I Pnmq Air = 49.0
Uo=6.16m/s
i
B
2e 1
% Primary Air = 69.0
B
0.1

0.01
0 10 20 30 40 50 60 70 80 - 90
Bed Height. %

Fig. 40. Solid fraction profiles, hot bed scaling with simplified I - I

scaling laws, low velocity, 49% primary air. 0.01 3 8

0 10 20 30 40 50 60 70 80 90
Bed Height. %

Fig. 43. Solid fraction profiles, hot bed scaling with simplified
scaling laws, low velocity, 69% primary air.

ae % Primary Air = 52.0

plifications of the full set of scaling parameters. Rather,
i
.s
there is a discrepancy between the hot bed and both
P 1 sets of scaling parameters which must be due to other
r;:

e
0 10 20 30 40 50 60 70
Bed Height %
Fig. 41. Solid fraction profiles, hot bed scaling with S,implified
scaling laws, high velocity, 52% primary air.
When utilizing the simplified scaling laws to scale
between a utility sized hot CFB and l/16 scale cold
model the average solid fraction profiles are in good
agreement for most of the conditions tested. Agreement
is excellent between the l/4 scale cold model which
utilized the full set of scaling parameters, and the l/
16 scale model which utilized the simplified set of
parameters. Thus, the disagreement between the hot
bed and the simplified model is not due to the sim-
factors.
Much of the disagreement between the hot and cold
beds may be due to the uncertainty in the measurement
of hot bed solids circulation rate where standard de-
_I viations in measured circulation rates were up to 50%
80 90 100 the average value. Agreement between the hot bed and
cold models is generally better near the top of the bed.
This may be explained in light of the choking phe-
nomenon described above. Since there was some error
in the solid to fluid density ratio when scaling between
the hot and cold beds, one would expect that there
would also be some error in the predicted choking
voidages. When choking ensues, the density in the
bottom is more significantly changed than in the top
of the bed. Comparison between the probability density
functions for the l/4 and l/16 scale cold models also
demonstrated excellent agreement as demonstrated in
Figs. 44 through 47. In most cases, the agreement in
histograms is better in the top of the bed. This is
 197

50

M,,4 = 1.45E-03 Rc,,~ = 13.9

2.5 M ,,,6 = 1.49B03 Re,,,6 = 3.17
10

f 1

.g

2
0.1
- l/4 Scale Cold Bed

0.01
0 1 2 3 4 5 0 10 20 30 40 50
Solid Fraction Frequency

Fig. 44. PDFs, comparison between l/16 and l/4 scale Studsvik Fig. 47. FFTs, comparison between l/16 and l/4 scale Studsvik
models, low velocity case. models, high velocity case.

Conclusions

A new set of simplified scaling laws has been proposed.

1.25
When the gas to particle drag can be represented by
6
._ the Ergun equation or similar expression, the new
2
LL
Fr,,l6 = 0.73
simplified scaling laws hold exactly in both the viscous
,x Rel,4 = 13.9 dominated and gas inertial dominated limits. When the

!o.75
2
2 0.5 Uo = 1.8 m/s
gas to particle drag is represented by drag on an isolated
particle, the simplified scaling laws also give the correct
limiting conditions for both the viscous and inertial
% limits.
0.25 For intermediate conditions, the gas to particle drag
is also well approximated in models using the simplified
0 scaling laws.
0 2 4 6 8 10
Solid Fraction Based on the experimental results of the viscous limit
Fig, 45. PDFs, comparison between l/16 and l/4 scale Studsvik scaling using different particle densities, the ratio of
models, high velocity case. particle to gas density cannot be ignored in order
properly to simulate a combustor or chemical reactor
outside the viscous limit. Within the viscous limit, the
I. I
density ratio must also be matched correctly to model

h
M,,4 = 6.07E-04 Re1,4 = 11.1
IN,,,6 = 6.23E.04 Re.,,16 = 2.53
a circulating bed near the slugging limit.
10
= 0.47 Uo,,4 = 2.9 m/s The simplified set of parameters which includes the
%4
Fr,,,6 =0.46 Uo,,,6 = 1.4 m/s solid to gas density ratio has been shown to give
acceptable results over a wide range of particle densities
and bed sixes, even when the length ratio is as small
as l/16. This has significant implications for the ability
to model large scale hot combustors with relatively
small cold laboratory models. Results of these hydro-
dynamic scaling experiments show good similarity when
using the simplified set of scaling laws and strongly
suggest that this new set of parameters is acceptable
0.011 3 * 3 1
0 10 20 30 40 50 over the range of conditions tested.
Frequency

Fig. 46. FFTs, comparison between l/16 and l/4 scale Studsvik
models, low velocity case. Acknowledgements

We are indebted to B. Zetharaeus and M. Morris

consistent with the agreement of the solid fraction of Studsvik AB, and Thomas Flynn of Babcock and
profiles. Wilcox, Alliance, Ohio for supplying us with data and
198

related information for the Studsvik fluidized bed com- voidage at incipient choking
bustor. This study was sponsored by the United States cluster voidage
Department of Energy, Morgantown Energy Technology voidage at minimum fluidization
Center. The data for the full scale model of the Studsvik sphericity
combustor was obtained by Detlef Westphalen of MIT gas viscosity
in a study sponsored by the Electric Power Research gas density
Institute. solid density
standard deviation
shear stress
List of symbols de1 operator

43 bed cross-sectional area Subscripts

Ar Archimedes number at choking
C, drag coefficient L gas
CP
specific heat of solid gl glass
D bed diameter mf at minimum fluidization
& cluster diameter Pl plastic
dP
particle diameter S solid
63 drag force stl steel
Fr Froude number based on bed length and su- t based on terminal velocity
perficial velocity
Fr, Froude number based on bed length and velocity
at incipient choking
solid particle friction factor References
fP

G mass flux of solid

gravitational acceleration L.R. Glicksman, Chem. Eng. Sci., 39 (1984) 1373.
g
M.T. Nicastro and L.R. Glicksman, C/rem. Eng. Sci., 39 (1984)
i, i, k unit vectors in x, y, and z
1381.
L bed length or height from distributor to exit T. Fitzgerald, D.B. Bushnell, S. Crane and Yeong-Cheng
A4 nondimensional solid circulation rate, G,/p,u, Shieh, Proc. 7th Int. Con& on Fluidized Bed Combustion, Vol.
Red, Reynolds number based on cluster diameter 2, 1983, p. 766.
and superficial velocity 4 R.A. Newby and D.L. Keaims, Proc. Fluidization V Conj,
Elsinore, Denma& 1986, pp. 31-38.
Red, Reynolds number based on particle diameter
5 L.R. Glicksman, T. Yule, R. Carson and R. Vincent, Froc.:
and superficial velocity 1988 Seminar on Fluidized Bed Combustion Technology for
RepE Reynolds number based on particle diameter Utility Applications, Feb. 1989, pp. l-20-1.
and superficial velocity using the full or exact 6 H. Chang and M. Louge, Powder Technol., 70 (1992) 259.
set of scaling laws 7 L.R. Glicksman, D. Westphalen, K. Woloshum, T. Ebert, K.
Roth, M. Lmts, C. Brereton and J. Grace, Proc. 11th Znt.
Re, Reynolds number based on particle diameter
Con& on Fluidized Bed Combustion, Montreal, 199I.
and particle terminal velocity 8 Ivk Horio, EPRI Rep. GS-6747, Feb. 1990.
SF solid fraction 9 M. Horio, A. Nonaka, T. Sawa and I. Muchi, AZChE J., 32
t time (1986) 1466.
U gas velocity 10 L.R. Glicksman, Chem. Eng. Sci., 43 (1988) 1419.
11 R. Roy and J.F. Davidson, Fluidization, 6 (1988) 293.
li gas velocity vector
12 M. Horio, H. Ishii, Y. Kobukai and N. Yamanishi, J. Chenr.
UC gas velocity at incipient choking Eng. Jpn., 22 (1989) 587.
U mf minimum fluidization velocity 13 H. Ishii and I. Murakami, in P. Basu, M. Horio and M.
uo superficial gas velocity Hasatani (eds.), Circulating Fluidized Bed Technology III, Per-
slip velocity gamon, Oxford, 1991.
USI
14 M. Tsukada, D. Nakanishi, Y. Takii, H. Ishii and M. Horio,
4 terminal velocity
Proc. Znt. Conf. Fluid. Bed Combustion, Vol. 2, ASME, New
V particle velocity York, 1991, pp. 829-834.
v particle velocity vector 15 T. Litka and L.R. Glicksman, Powder Technol., 42 (1985)
x9 Y, 2 Cartesian coordinates 159.
z axial bed location 16 MJ. Rhodes and P. Laussmann, Can. J. Chem. Eng., 70
(1992) 625.
17 F.M. White, II.rcous Fluid Flow, McGraw-Hill, New York,
Greek letters 1974.
P coefficient of interphase drag 18 D. Westphalen, M.S. Thesis, Massachusetts Institute of Tech-
E voidage nology, 1990.
19 S. Satija, J.B. Young and L.S. Fan, Powder TechnoL, 43 (1985) 23 D.V. Punwani, M.V. Modi and P.B. Taman, Proc. Znt. Powder
7<7
&,a,.
and Bulk Soli& Hand&f and Processing Conference. Powder
20 W. Yang; Powder Technol., 35 (1983) 143. Advisory Center, Chicano, IL, 1976. - _
24 John M. Matsen, Powder TechnoL, 32 (1982) 21.
21 L.S. Leung, R.J. Wiles and D.J. Nicklin, Znd. Eng. Chem. 25 Y.O. Chong and L.S. Leung, Powder Technol., 47 (1986) 43.
Process Des. Dev., 10 (1971) 183. 26 D. Geldart, Powder TechnoL, 7 (1973) 285.
22 Y. Yousfi and G. Gau, Chem. Eng. Sci., 29 (1974) 1939. 27 D. Westphalen, Ph.D. The& currently in progress, 1992.