Professional Documents
Culture Documents
00
Printed in Great Britain. 8 1987. Pergamon Journals Ltd.
Abstract-A detailed population balance model is presented for a duidized bed reactor incorporating: the
formation of bubbles at the grid plate, their rise with velocities governed by their sizes, random coalescence
between bubbIes, gas exchange between bubbles and the dense phase, and a first order chemical reaction in
the dense phase under well-mixed conditions. Reaction conversion is calculated as a function of
dimensionless parameters relating the rates of various competing processes such as coalescence, dense phase
mixing, mass exchange between bubbles and dense phase and reaction rate. Comparison ofconversions with
those of Davidson et al. (1977) show significant variations indicating that the dynamics of bubble size
distributions could have non-trivial erects on the extent of reaction. Fluctuations in bubble populations did
not seem strong enough to translate to strong fluctuations in reaction conversion.
341
342 IAN R. SWEET et al.
becomes inappropriate. It is of interest, therefore, to Gas exit
Shah et al. (1977a) have argued that bubble popu- reflects the fact that all entering bubbles contain gas of
lations may become so small that fluctuations could, in reactant concentration c0 _
some circumstances, play a considerable role. Thus the
more general stochastic framework presented by The population balance equation
Ramkrishna and co-workers (1973, 1976) may be The population balance model is now complete and
necessary for such analysis. In this connection, the the relevant equation for the expected number density
simulation technique of Shah et al. (1977b) based on fi (u, .z, c) is readily identified. Under steady-state con-
the method of the interval of quiescence, automatically ditions, we have
accounts for any fluctuations that may be important.
In this paper we will use the coalescence model of $c V(U,Z)f 1(~TG41+~[u(o)f, (u,z,c)l
Argyriou et al. (1971), slightly modified by Shah ef al.
(1977a) (to account for the fact that coalescable
bubble-pairs may not necessarily be exactly at the same
bed height), to analyse the effect of bubble coalescence
and interactions with dense phase on reaction conver-
sion. Thus the coalescence frequency is taken to be
where
S(U - Urn,)
(7)
N =
0 9 ug = u(uo).
uouo
The condition that the entering bubbles are all of
uniform volume u0 translates to
c” = (=$!+bH+(+)Cp. (16)
complete the formulation of the problem. The quantitative analysis of the model consists in
solving eq. (9) simultaneously with eq. (13). Before we
3. DENSE PHASE MODEL address this issue it is instructive to see how the present
The balance equation (9) [or its dimensionless form model degenerates into the more elementary models of
(1 I)] is written without regard to the state of mixing in the past. Suppose, for example, the entering bubbles
the dense phase and to this extent eq. (9) is quite leave without any coalescence between bubbles.
346 IAN R. SWEET et al.
Assume further that the bubble volume does not to the population balance framework has been dis-
change as it ascends through the bed. Then the present cussed by Ramkrishna (198 1). The simulation not only
model must simplify to the one presented by Davidson produces average values but also computes fluctu-
et al. (1977) which assumes that the dense phase is well ations about the average values.
mixed and that the bubble phase consists of identical Shah et al. (1977a) have employed their simulation
non-coalescing bubbles. When no coalescence occurs technique to predict bubble size distributions in a
we must set o = 0 in eq. (II) to obtain fluidized bed without chemical reaction. The inclusion
of chemical reaction is conveniently made in the
foregoing simulations. The technique, being essentially
transient in nature, requires an initial condition stipu-
For constant bubble volume, q = 0. Further if 4 is lating the state of the entire population (that is the
independent of < (but dependent on v) then the above bubble trivariate distribution in size, location and
equation becomes reactant concentration in this case) and evolves to the
steady-state distribution. With chemical reaction to be
accounted for, since the concentration change in each
bubble depends upon the dense phase concentration
(not at first available because of the coupling between
The foregoing equation is readily solved subject to
the population balance and the dense phase equations),
condition (12) yielding
the bubble population can be simulated only for
assumed values of the dense phase concentrations, c~.
9, (S,rl.i) = exp [ “‘y&I i 1-t6 5, - (<, - 0 For each such assumed value of cp, the simulation
(18)
yields the number density fi, which on substitution
xexp
[ rli(rl)P
1>
“(‘1!(1-, -1 b(q- 1)
into eq. (14) will produce a new value of c,, generally at
variance with the assumed
used in a successive
one. While
approximation
this could be
scheme for the
where we have set 5, E c,/cO and p = (U - U,,)/U.
assumed and calculated dense phase concentrations to
Substitution of (18) into the dimensionless version of
agree within specified limits, another approach, which
(14) gives after some algebra
has some computational advantages, is as follows.
1 --fieeX Because the bubble coalescence rate has been assumed
tl, = (19)
to be independent of reactant concentration (and thus
1 +k’-flepX
where the dense phase concentration) the dynamics of the
x ~ K(l)(l -8) 4 bubble size distribution in the bed is unaffected by the
reactant concentration in the bubbles. Thus the mul-
B
tiple simulations do not have to be repeated for each
the right-hand side of which is readily shown to be the assumed cp if, in the simulations, the updating of
same as (4 + KGa)H/u,u(v,), an expression used by bubble concentrations during a quiescence interval is
Davidson et al. (1977). When eq. (18) is plugged into performed for each cp. The calculated values of cp for
the dimensionless form of (15), denoting cbH/cO by <bH different assumed values of it are obtained from a
we get single set of sample paths. By plotting the calculated cp
<,(l -e-x)+e-x. vs the assumed value the intersection of the curve with
<bH = (20)
the line representing equal values of the ordinate and
When (19) and (20) are employed in (17) abscissa the correct steady-state value of c,, is obtained.
This value of the dense phase concentration may again
l-z++z
c’=Peb,,+(l-B)&,= 1_z+-t, , Z=PepX be used for a simulation as a second check and to
obtain the actual number density function fi. Once the
which is the expression derived by Davidson et al. number density function is calculated the fraction
(1977). Thus we have shown that the coalescence model unconverted in the gas leaving the bed, c’, is obtained
here simplifies appropriately in the limit of vanishing from eq. (17).
coalescence. Computations were made on a Cyber-205 at the
Purdue University facility. The simulation times re-
quired varied from 30 s to 9 min of CPU time to
4. COMPUTATIONS compile and run. Naturally, smaller coalescence rates
The numerical solution of eqs (11) and the dimen- needed the greatest amount of computer time since
sionless form of (13) is a very formidable task because larger numbers of bubbles are encountered in this case.
of the trivariate nature of 4, and the possibility of
statistical correlation between bubble states (which
presents a closure problem). However, a very effective 5. RESULTS AND DISCUSSION
alternative is available in a simulative approach due to At the level of analysis in this paper we are
Shah et al. (1977b) which consists in artificially realiz- examining effects not quantitatively studied in exper-
ing the process by Monte Carlo simulations. The imental works on fluidized bed reactors. Hence the
mathematical relationship of the simulation technique computations made in this paper have generally been
Population balance modelling of bubbling fluidized bed reactors-I 347
with a view towards establishing the major directions Table 2. Heights and thicknesses of
in which the reactor behaviour changes with the bed slices
phenomena for which account has been made. In this
Slice no. Height (in.)
connection the experimental data of Argyriou er 02.
(1971) provide a useful starting point for our compu- 1 o-1.54
tations. Table 1 lists the dimensions of the bed, 2 1.54-3.08
operating conditions, the feed bubble size, and values 3 3.084.62
4 4.62-6.16
of the parameters in the coalescence model. Table 2 5 &l&7.70
shows a discrete partition of the fluidized bed with 20 6 7.70-9.24
slices within which the events occurring in the bed will 7 9.24-10.78
be viewed. The calculations include not only average 8 10.78-12.32
values of various quantities but also fluctuations about 9 12.32-13.86
10 13.8~L15.40
the mean. 11 15.4CLl6.94
The main parameters whose effects are to be 12 16.94-18.48
evaluated are the dimensionless coalescence rate w, the 13 18.48S20.02
dimensionless reaction rate k’, and the dimensionless 14 20.02-21.56
15 21.5C23.10
constant Z which contains the exchange rate of 16 23.10-24.64
reactant between the dense phase and the bubbles. 17 24.626.18
First, we examine the bubble number distribution in 18 26.18-27.72
the bed. Since the model does not account for any 19 27.72 29.26
20 29.2C30.8
dependence of coalescence rate upon either reaction or
gas exchange the only relevant parameter which can
alter the bubble number and size distribution in the increased coalescence rates. For the lowest coalescence
bed is the coalescence parameter. Figure 2 shows the rate the standard deviation about the average number
mean value of the cumulative number of bubbles and of bubbles increases progressively throughout the bed.
its fluctuations as a function of bed height for three This follows from the fact that at the entrance of the
different coalescence rates. The figure clearly reflects bed the number of bubbles is assumed to be known
the drastic reduction in the number of bubbles with exactly (i.e. zero standard deviation) and thereafter
random coalescence leads to an increasing standard
Table 1. Values of fluidized bed parameters used in the deviation about a smaller average. If the bed were high
simulations enough this standard deviation would eventually drop
since the mathematical limit of the process is a single
u mf 0.042 in. s ’ bubble with zero standard deviation. Although this
u 0.08652 in. s 1
14.0 in.“’ s- ’ does not occur with the low coalescence rate the drop
k,
kz 0.364, 1.5, 3.64 in.-‘s-l in standard deviation at higher rates is apparent from
ks 0.125 Fig. 2. Figure 3 illustrates the number of bubbles in
H 30.8 in. each slice and the standard deviation for the low
DR 11.5 in.
coalescence rate. Table 3 reinforces the above
VO 0.037 in.3
conclusions.
I
0.:0,87 0.186410
I 0.322634
I 0.458857
I 0.5930W
I 0.731303
I 0.867527
I I. 00373
I
Fig. 2. Cumulative average number of bubbles vs dimensionless reactor height for all three coalescence rates.
348 IAN R. SWEET ef al.
24.0
z
I’
* Number of bubbles
2
12.0
0 2 x standard deviation
z t
Fig. 3. Averagenumber of bubbles and twice the standard deviation about the averages vs dimensionless bed
height for w = 291.3.
Figure 4 shows the bubble volume distribution on a Next, we examine the model predictions for re-
cumulative basis at three chosen bed heights, one at the action conversion. The conversion (or c’ which is the
bottom region, another in the middle and the third at result of deducting the conversion from unity) depends
the top. At the bottom of the bed the cumulative upon the parameters, o, k’and Z. Figure 5 shows c’, the
number fraction rises quickly to unity for small bubble unreacted gas fraction, as a function of k’ for various
volumes (since the bubbles there are small). For the coalescence rates for a fixed Z (representing a fixed gas
middle section the rise to unity is gradual as one should exchange rate). Alongside is also plotted the result
expect. Near the top of the bed the existence of large from the model of Davidson et al. (1977). Each of the
sized bubbles leads to a steep rise of the cumulative curves in Fig. 5 shows the obvious increase in conver-
distribution to unity for a large bubble volume as it sion at higher reaction rates while higher coalescence
should. rates show diminished conversion as a result of by-
0 I I I I
1.0 5.0 9.0 l3..0 17.0 21.0
T)
Fig. 4. Cumulativefraction of bubbles withindifferentvolume ranges at the lowest coalescence rate in the
bottom, middle and top regions of the bed.
e
Population balance modelling of bubbling fluid&d bed reactors-1 349
I .o
A w = 291.3
I I I I 1
8.0 16.0 24.0 32.0 400
k’
k’
Fig. 5. Unreacted gas fraction vs dimensionless first-order
reaction rate constant for three coalescence rates and Fig. 6. Unreacted gas fraction vs dimensionless first-order
D = 0.02232 in. s-‘. reaction rate constant at w = 2913.0 with varying transport.
passing of the reactant with bubbles with inadequate preach that of Davidson et al. (1977) in the infinite
exchange between the dense phase and the bubbles. As limit. Table 4 gives the numerical results of the
the coalescence rate drops the conversion curve ap- computations represented in Figs 5 and 6. In par-
proaches that of Davidson et al. (1977), a result that ticular, fluctuations in conversion were found to be
was established also analytically earlier in the paper. negligible even at the highest coalescence rate used.
Figure 6 shows the effect of exchange rate as Figure 7 is a graph of the cumulative number
measured by 2 on the conversion. Lower exchange fraction of bubbles vs the dimensionless concentration,
rates reduce conversion significantly because the by- 5, for the top, middle and bottom regions of the bed.
passed reactant transfers less efficiently to the dense Bubbles near the grid plate have concentrations dis-
phase to which the reaction is confined. The higher tributed almost evenly from 65 to 100% of the feed
exchange rates produce conversion curves which ap- concentration. However, by the time bubbles reach the
w Z k C’ 5, hi C’ (Davidson) SD C’ x lo3
Fig. 7. Cumulative fraction of bubbles within different concentration ranges in the top, middle and bottom
regions of the bed: well-stirred dense phase model.
middle of the bed a vast majority have concentrations INT-8206424 which made this researchpossible. One of us
less than 30% of the feed concentration. As might be (D. Ramkrishna)is also grateful to the Indian Instituteof
expected, at high transport rates and low coalescence Science, Bangalore where part of this research was ac-
complished for a Visiting Professorship sponsored by the
rates bubble phase concentrations tend to decrease. University Grants Commission of India.
6. CONCLUSiONS
We have been able to incorporate the population NOTATION
balance framework to account for the effect of by- a interfacial area of bubble
passing of reactant with bubbles which coalesce ran- C concentration of reactant in bubble
domly and individually exchange gas with the dense C’ dimensionless exit concentration of reactant
phase on the reaction conversion. The conversions in the efhuent gas phase
predicted can vary substantially from those of =H exit concentration of reactant in the gas
Davidson et al. (1977) depending upon the values of phase
the coalescence rate and the exchange rate parameters. ‘hi average reactant concentration in the bubble
The fluctuations in conversion were negligible for the phase at reactor exit
most part, at least in the range of parameter values CO reactant concentration in feed gas
included in this work. The superficial velocities used in CP reactant concentration in well-mixed dense
this work were relatively low, being twice the minimum phase
fluidization velocity. It is likely that for higher fluidiz- c rate of change of reactant concentration in
ation velocities the volume fraction of reactants in the bubble
bubble phase may be substantial enough to lead to H height of fluid bed
fluctuations in the conversion. Such computations Hmf height of bed at incipient fluidization
were not pursued in this work because of limited k rate constant of reaction
availability of data to serve as a guide. It is essential k dimensionless reaction rate
that efforts be made to determine coalescence rates k2 constant in coalescence frequency governing
before a sound quantitative understanding of fluidized coalescence rate
bed reactors can emerge. k, shape factor for non-spherical bubbles
We have emphasized that one of the objectives of KG mass transfer coefficient for bubble for ex-
this work is to show that the facility now exists to change of reactant with dense phase
analyze rather complex features of fluidized bed fl average number density function (product
reactors and their effects on conversion. The demon- density of order 1); average number of pairs
stration was confined to a first-order isothermal reac- of bubbles (product density of order 2)
tion and in the absence of bubble break-up. It is NO number of bubbles per unit volume in feed
possible to incorporate bubble break-up into the Pa pressure at bed top
analysis without any special strain on the computation 4 gas exchange rate between bubble and dense
provided, however, quantitative information is phase
available. Non-isothermal reactions and complicated Q coalescence frequency
kinetics introduce no conceptual problems although S cross-sectional area of fluidized bed
the computational burden is increased enormously. U bubble velocity
u superficial velocity of fluidization
Acknowledgements-The authorsgratefullyacknowledgethe u mf superfTcia1velocity at incipient fluidization
National Science Foundation for an InternationalGrant 0 bubble volume
Population balance modelling of bubbling fluid&d bed reactors-1 351
v.max maximum bubble volume Hulburt, H. M. and Katz, S. L., 1964, Some problems in
V rate of change of bubble volume particle technology. A statistical mechanical formulation.
X dimensionless parameter defined below eq. Chem. Engng Sci. 19, 555574.
Kumar, R. and Kuloor, N. R.. 1970, The formation of bubbles
(19) and drops, Adv. Chem. Engng 8, 255-368.
Z height measured from grid plate Lee, W. and Weekman, V. W., 1976, Advanced control
Z dimensionless parameter defined below eq. practice in the chemical process industry: a view from
industry, A.I.Ch.E. J. 22, 27-38.
(20)
Matsen, J. M., 1973, Evidence of maximum stable bubble size
in a fluidized bed A.Z.Ch.E. Symp. Ser. 69, No. 128, 3&33.
Greek letters Matsuura, A. and Fan, L. S., 1984, Distribution of bubble
B dimensionless parameter defined below eq. properties in a gas-liquid-solid fluidized bed. A.Z.Ch.E. J.
(18) 30, 894903.
6 Dirac a-function Mori, S., Miho, S., Yamada, M., Hiraoka, S. and Yamada, I.,
1984, Radial distribution of bubbles in the cylindrical
41 dimensionless number density function fluidized bed. A.I.Ch.E. Symp. Ser. 80, No. 241, l&16.
PP
dimensionless coalescence rate parameter Narasimhan, G., Nejfeit, G. and Ramkrishna, D., 1984,
w dimensionless exchange rate for bubble Breakage functions for droplets in agitated liquid-liquid
K dimensionless exchange rate for bubble dispersions. A.Z.Ch.E. J. 30, 457467.
Nguyen, T. H., Johnson, J. E., Clift, R. and Grace, J. R., 1976,
r dimensionless concentration of reactant in
Prediction of bubble distributions in freely-bubbling three-
bubble dimensional fluidized beds. In Fluidization Technology
5,
dimensionless dense phase concentration of (Edited by D. L. Keairns), pp. 2Os-214.
reactant Orcutt, J. C., Davidson, J. F. and Pigford, R. L., 1962,
dimesionless version of c Reaction time distributions in fluidized catalytic reactors.
4
CEP Symp. Ser., 38, l-15.
dimensionless bubble volume Potter, 0. E., 1971, Mixing. In Fluidization (Edited by
T dimensionless bubble velocity Davidson, J. F. and Harrison, D.), Chap. 7. Academic Press,
r dimensionless height of bubble location London.
Ramkrishna, D., 1978, The prospects of population balances.
Chem. Engng Educat. 14, Winter.
Ramkrishna, D., 1981, Analysis of population balance-IV.
REFERENCES The precise connection between Monte Carlo simulation
Argyriou, D. T., List, H. L. and Shinnar, R., 1971, Bubble and population balance. Chem. Engng Sci. 36, 1203-1209.
growth in coalescence in gas fluidized beds. A.I.Ch.E. J. 17, Ramkrishna, D., and Borwanker, J. D., 1973, A puristic
122-130. analysis of population balances-I. Chem. Engng Sci. 28,
Bukur, D. and Amundson, N. R., 1975, Mathematical mod- 1423-1435.
eling of fluidized bed reactors-III. Axial dispersion model. Ramkrishna, D., Shah, B. H. and Borwanker, J. D., 1976,
Chem. Engng Sci. 30, 1159-l 167. Analysis of population balances-III. Agglomerating
Calderbank, P. H. and Toor, F. D., 1971, Fluidized beds as populations. Chem. Engng Sci. 31, 43-2.
catalytic reactors. In Fluidization (Edited by Davidson, J. F. Rice, W. J. and Wilhelm, R. W., 1958, Surface dynamics of
and Harrison, D.), Chap. 8. Academic Press, London. fluidized beds and quality of fluidization. A.Z.Ch.E. J. 4,
Caram, H. and Amundson, N. R., 1979, Fluidized bed 423429.
gasification reactor modeling. 1. Mode1 description and Rowe, P. N., 1971, Experimental properties of bubbles. In
numerical results for a single bed. Ind. Engng Chem. Proc. Fluidization (Edited by Davidson, J. F. and Harrison, D.),
Des. Dev. 18, 80-96. Chap. 4. Academic Press, London.
Davidson, J. F. and Harrison, D., 1963, Fluidized Particles. Shah, B. H., Ramkrishna, D. and Borwanker, J., 1977a,
Cambridge University Press, Cambridge. Simulation of bubble populations in a gas fluidized bed.
Davidson, J. F. and Harrison, D., 1971, Fluidizarion. Chem. Engng Sci. 32, 1419-1425.
Academic Press, London. Shah, B. H., Ramkrishna, D. and Borwanker, J., 1977b,
Davidson, J. F., Harrison, D., Darton, R. C. and LaNauze, Simulation of particulate systems using the concept of the
R. D., 1977, The two-phase theory of fluidization and its interval of quiescence. A.1.Ch.E. J. 23, 897-904.
application to chemical reactors. In Chemical Reactor Yates, J. G., Rowe, P. N. and Cheesman, D. J., 1984, Gas entry
Theory (Edited by Lapidus, L. and Amundson, N. R.), effects in fluidized bed reactors. A.1.Ch.E. J. 30, 89&893.
Chap. 10. Prentice-Hall, Englewood Cliffs, NJ. Yoshida, K., Nakajima, K., Hamatani, N. and Shimizu, F.,
Harrison, D. and Leung, L. S., 1961, Bubble formation at an 1978, Size distribution of bubbles in gas Auidized beds. In
orifice in a fluidized bed. Trans. Inst. Chem. Engrs 39, Fluidizarion (Edited by Davidson, J. F. and Harrison, D.),
409414. pp. 13-18. Cambridge University Press.