You are on page 1of 3

Cbcmicnl Engineering Sc~enrc. Vol 43. No. 6. pp. 1419~1421. 1988 0009-2509188 S3.00 + 0.

00
Pnnted in Great Britam 0 1988 Pergamon Press plc

Scaling relationships for fluidized beds

(First received 3 March 1987; in revisedform 1 September 1987; acceptedfor publication 11 November 1987)

INTRODUCTION alternate reduced form of the dimensionless parameters can


The fluid dynamics ofa fluidized bed must be well understood be derived (Webb, 1984) for the viscous dominated subregion.
to properly design a l%dized-bed combustor or reactor. From Clicksman, (1984), conservation of mass for the fluid
There is a particular need to develop laboratory scale beds is
operating at standard conditions that will properly simulate
div (EU) = 0 (1)
the hydrodynamics of a commercial unit operating at elevated
pressure and temperature. and conservation of mass for the particle is
In a previous paper (Glicksman, 1984), a set of scaling laws
div [(l -E)u] = 0. (2)
was systematically developed that allow a bed operating at
ambient conditions to model the hydrodynamics of a bed at
In the equation of motion for the fluid, inter-particle forces
elevated pressure and temperature. In the genera1 case, both
are neglected. More precisely, the results which follow will
beds must be geometrically similar, and both beds must have
also apply to situations where inter-particle forces are
equal values of the Reynolds number, the Froude number, the
governed by particle inertia, gravity forces on the particle,
ratio of particle to fluid density, the dimensionless particle size
sphericity, and size distribution. Factors which are omitted
distribution. and sphericity. For small particles, a region was
include surface forces on particles due to static changes or van
identified where fluid viscous forces predominate the fluid
der Waals forces. Also the influence of the particle coefficient
inertia forces can be neglected. Similarly, for large particles, a
of restitution or the friction coefficient on inter-particle
region was identified where viscous forces can be neglected.
collisions is not included. Litka and Glicksman (1985)
In a companion (Glicksman and Nicastro, 1984), an
experimentally found that changes in the particle coefficient
experimental confirmation of the scaling laws was presented.
of restitution and coefficient of friction did not influence
Bubble frequency and amplitude spectra were measured in an
bubble characteristics and particle dispersion in a bubbling
atmospheric combustor. A scale model of the combustor was
fluid&d bed.
constructed that satisfied all of the scaling laws. The model
The equation of motion for the fluid can be written as
was fluidized with air at standard conditions; a consequence
of the scaling laws is that all of the hnear dimensions of the au
model are one quarter of the corresponding linear dimensions P,E Z+ti-gradU +~~,ge+gradp+B(ci--90.
[ 1
of the hot atmospheric combustor. For 677-micron mean
diameter particles in the combustor, and one quarter diameter (3)
steel particles in the model, the non-dimensional bubble The equation of motion for the particles becomes
frequency and amplitude characteristics in the two beds were
found to be virtually identical. The combustor was operated p,(l-s) E+ti.gradC +ip,g(l-s)-/3(S-_=O. (4)
at 1098 K with U, of 93 cm/s and U,, of 16 cm/s. [ 1
Recently, Newby and Keairns (1986) carried out a valid-
ation of the same scaling laws using a bed of 200-micron glass The boundary conditions for the bed at the side and
spheres fluidized with air at standard conditions and a bottom walls when there is no feed of particle into the bed are
geometrically similar bed of lOO-micron steel particles at x=O,D
fluidized with pressurized air. The beds had equal values of all and y = 0, Ho VN -- 0. (5a)
of the dimensionless scaling parameters in the scaling laws.
Newby and Keairns found close agreement between the where oN is the component of the particle velocity normal to
dimensionless bubble size and frequency for both beds. Horio the surface; the bed walls are at x = 0 and D. This boundary
er al. (1986) have recently proposed a set of scaling parameters condition also applies to an imaginary boundary high above
that has a different form from those mentioned above. They the expanded bed surface at H, if particles are not elutriated
carried out experiments on geometrically similar bubbling from the bed. For the gas velocity, at the bottom when a
beds of silica sand with mean diameters in the range of screen or fixed bed is used as a distributor, and also at a
10%180 microns. They showed agreement of mixing par- location well above the expanded bed surface:
ameters between beds scaled by their proposed set of
at y = 0, H, id = Y,. (5b)
parameters.
The purpose of this short paper is two-fold. First, the paper Along the bed walls:
demonstrates that, with proper non-dimensionalization of
the governing equations, two parameters previously ident- at x=O,D u = 0. w
ified do not appear independently, rather, they appear as a The boundary conditions for the pressure at the side walls
product in the viscous limit. Secondly, the reduced set of are
parameters for the viscous limit are shown to be identical to
those developed by Horio. x=O,D
ap
- = 0.
an
SCALING LAWS FOR THE VISCOUS LIMIT
which follows from eqs (5a) and (5~) combined with the
The equations of motion for the fluid and the particle resistance expression
continuum are used to develop the scaling laws for the viscous
limit, when fluid inertial effects are negligible. The develop- gradp = -B(U-6). (Se)
ment parallels that of the previous work (Glicksman, 1984).
Along the bottom, for a uniform distributor:
However, by scaling all of the length coordinates by the bed
dimension, L, rather than by the particle diameter, d,, an at y = 0 p = const. = pfmbrd + &.,,r&,f. (5f)

1419
1420 Shorter Communications

Along the top, above the expanded bed surface: Reynolds number. In the general case when both gas viscosity
and gas inertia are important, the more general gas to particle
Y = H, ? = ffreeboard (5g) interaction or drag term involves the Reynolds number and
The parameters will be nondimensional&d as d,/L as two independent parameters.
The resulting list of dimensionless parameters given in eq.
+z-V,QLLV,f=UD, (11) valid for the viscous limit is a more compact form of that
0 L given previously, Glicksman (1984) where it was not recog-
nized that Re and d,/L can be combined into a single
where the lerlgth scale is non-dimensionalized with respect to
parameter in the viscous limit. Both eq. (11) and the
the bed dimension, L. The proceeding continuity and momen-
previously derived parameter list apply to the viscous limits;
tum equations along with the boundary conditions can now
however, eq. (11) is less restrictive for modelling purposes
be written in non-dimensional form as because of the reduction by one of the number of independent
div (EU) = 0 (5) dimensionless parameters. It has been recognized previously
that proper non-dimensionalization of the governing equa-
div [ (1 - ~)t)] = 0 (6) tions can sometimes yield a more compact list of independent
7 dimensionless parameters (Kline, 1965).
(1 -c) g + (~.V)~ +$(I PL
-&)-PIUo(U-v) = 0 (7) Equation (11) can be written in a slightly different but
L I equivalent form, when it is noted in the viscous limit that V,,
du : LP, 57L is proportional to gp,d:/p so that eq. (10) can be rewritten as
Pf z + (Id. V)u
I 1 P. u.
(12)
+v -$ +~(d-d) = 0 (8)
( s 0 > P. %
The governing dimensionless parameters in the viscous limit
at x, = 0 and 1 [eq. (1 l)], can be equivalently written as
ON = 0 (9a) v_z u, L L,
u = 0 (9b) gLU,,D L2

sphericity, particle size distribution. (13)

Hario et al. presented a set of scaling relationships in terms


at y/D = 0
of a scaling parameter m:
(9d)
Q L, Lz
(9c)
m=-=-=_ (14)
DL L; L,

_*+&($)(+)
P (9f) v, -urn,- J&v.? - Umi)o (15)
p,u,2 AU.2
u mf= JmU;r (16)
at y = H,fD
L& = 0 (%) where the superscript o refers to the scale model bed.
Combining eqs (14) and (16) results in
u = 1 Ph)

~P = (17)
Pfraboard/Ps:~ (99
P.4
where g is added to both sides to nondimensionalize the
In the viscous dominated region, at small Reynolds Numbers, results.
terms in the fluid equation with coefficients of p//p. are Combining eqs (15) and (16):
negligible and the governing parameters are gL/V,f,
@L/p,V,,, L, jD and L, /L,, where L, , L, etc. are the relevant v v D
physical dimensions of the system. Note. U,/U,r which
appears in the pressure boundary condition [eq. (Sf)] will be
2=
v mf ( >
P
v mf .
shown below to be expressed in terms of the other governing The parameters given in eqs (14), (17) and (18) are identical to
parametkrs, it is not an additional independent parameter. those derived from the governing equations and given in eq.
From the Ergun equation, or equivalent, for the viscous (13).
dominated region: It must be borne in mind that eq. (13) is valid only when
fluid inertial effects are negligibre, i.e. they are a subset of the
W 150(1 -#pL lSO(1 -#L general relationships. In Glicksman (1984) the criteria for the
p= (10)
&sps&P,d; = E3ARed,
viscous limit is obtained when the ratio of the viscous forces
P34?
to fluid inertia forces in the Ergun equation is 10 or larger.
where Re is the Reynolds number based on d,. From Table 1 of that reference, for a bed of glass or sand
Therefore, in the viscous region, the independent governing fluidized with air at standard conditions, with Vo/Vmi of 3,
parameters can be expressed as the viscous limit occurs when particles are less than 200
microns. Horio et al. carried out their experiments with
Uz/gL, L/cd, Re), L/D, L, /L2, particle sphericity, particle
particles of mean diameter varying between 100 and 180
size distribution. (11)
microns.
Note in this case that d,/ L only appears in the parameter list The agreement that Horio found between geometrically
combined with the Reynolds number. This is a consequence similar beds serves to verify the viscous limit of the scaling law
of neglecting the gas inertia when gas viscosity dominates. for the range of conditions they considered. It further
The particle diameter is absent in the ratio of particle inertia indicates that surface forces due to electrostatic forces,
force to particle gravity force since both forces are pro- adhesion etc. are not important in this region since if they were
portional to the particle volume. The ratio of particle important, dynamic similarity would not be observed be-
diameter to bed dimension only appears in the ratio of tween particles with different ratios of surface area to volume.
particle inertia to gas velocity forces, i.e. combined with the Note that Horios scaling relationships are not equivalent
Shorter Communications 1421

to the full set of dimensionless parameters for the general REFERENCES


case; therefore, Horios scaling relationships can only be used Glicksman, L. R., 1984, Scaling relationships for fluidized
with confidence in the viscous limit. beds. Chem. Engng Sci. 39, 1373-1379.
Both the general form of the scaling relations and the Glicksman, L. R. and Nicastro, M. T., 1984, Experimental
simplified form for the viscous limit apply to all flow regimes verification of scaling relationships for fluidized beds.
of fluidization. However, for proper scaling high in the Chem. Engng Sci. 39, 1381-1391.
freeboard or in a dilute system such as a circulating bed the Horio, M., Takada, M., Ishida, M. and Naricika, T., 1986,
means to delineate the viscous and inertial limits must be Chem. Engng Sci. 39, 15 l-l 58.
reevaluated. For example, the Reynolds number of the fluid Kline, S. J., 1965, Similitude and Approximation Theory.
based on the bed diameter may be the proper parameter for McGraw-Hill, New York.
establishing the viscous limit of the dilute-phase systems Litka, T. and Clicksman, L. R., 1985, The inlluence of particle
rather than the Reynolds number based on the particle mechanical properties on bubble characteristics and solid
diameter. mixing in fluidized beds. Powder Technol. 42, 231-239.
Newby. R. and Keairns, D.. 1986, in Proceedings ojthe Fifth
Engineering Conference on Fluidization, Elsinore, Denmark
LEON R. GLICKSMAN (Edited by L. Ostergaard and A. Sorensen). Engineering
Department of Mechanical Engineering Foundation, New York.
Massachusetts Institute of Technology Webb, H. A., 1984, personal communication, Morgantown
Cambridge, MA 02 139, U.S.A. Energy Technology Center.

Chevnicol Engk~ermg Science. Vol. 43, NO. 6. pp. 1421-1424. 1988. 0009-2509;88 53.00 + 0.00
Printed in Grear Brium. C 1988 Pcrgamon Press pt

Simulation of Wei-Prater experiments for non-linear reaction systems in a


concentration-controlled recycle reactor

(Received 9 July 1987; accepted for publication 4 December 1987)

The method of Wei and Prater [l] IS an elegant method for For the simplest reaction system we considered the tri-
transforming highly coupled complex systems of monomol- angular network
ecular and pseudo-monomolecular reaction systems into
completely uncoupled hypothetical systems amenable to easy
solution. In this communication this method is extended to
apply to the determination of rate constants of complex
reaction systems obeying non-linear kinetics. This extension
is based upon the principle of the concentration-controlled
recycle reactor [2]. It enables the experimentalist to maintain (3)
the concentrations of selected components constant over a
whole reaction path. The selection must be made in such a
way that pseudo-linear kinetics results for each step of the
reaction network. The procedure just outlined was tested by
simulating appropriate experimental runs.
The experiments were evaluated by means of the
relationship and supposed that the rates are proportional to both the mole
fractions of Ai, ai, and the mole fraction of a homogenous
C? C9
a=X+- __
1__*XI+f_ji2X3+ . (1) (gaseous) catalyst B, us. For example, the rate of reaction
A, -+ AZ is
valid for a gradientless reactor, where a is the vector of the
mole fraction of the components, X is the characteristic vector
r, = k,,a,.a,. (4)
matrix, / is the exit molar flow rate per unit reaction volume WeiCPrater experiments were then performed by simu-
(or unit catalyst mass)> and C = X- f[fo is the vector of lation in aconcentmtion-controlled recycle reactor, where the
inlet molar flow rates per unit reaction volume (or unit mole fraction ofthe homogenouscatalyst B was held constant
catalyst mass)]. throughout the experiments. The steady-state mass balances
The matrix of eigenvectors, X, can be found in the usual for this system are
manner, i.e. by locating the straight-line reaction paths in the
simplex. The ratio of the characteristic roots, Al /AZ, can be jai-j:= 5f vikrk, i=l,...,4 (5a)
found from a curved reaction path using the relationship *=

(2)
(5b)
where b= X-la.

You might also like