Beruflich Dokumente
Kultur Dokumente
G., Eds.; Sigthoff and Nordoff Amsterdam, The Netherlands, Moore, B. C. Principal Component Analysis in Linear Systems:
1980. Controllability, Observability and Model Reduction. IEEE
Garcia, C. E.; Prett, D. M. Advances in Industrial Model-Predictive Trans. Autom. Control 1981, AC-26, 17-32.
Control. In Proc. of Chemical Process Control III; Morari, M., Moore, C. Application of Singular Value Decomposition to the De-
McAvoy, T. J., Eds.; Elsevier: New York, 1986; pp 245-293. sign, Analysis, and Control of Industrial Processes. Proc. 2986
Gay, D.; Ray, W. H. Identification and Control of Linear Distributed Am. Control Conf. 1986,643-650.
Parameter Systems Through the Use of Experimentally Deter- Morari, M. Flexibility and Resiliency of Process Systems. Prep. Int.
mined Singular Functions. In Control of Distributed Parameter Symp. Process Systems Eng., Kyoto, Japan 1982.
Systems 1986; Rauch, H. E., Ed.; Pergamon Press: New York, Morari, M. Design of Resilient Processing Plants. IV. A General
1986. Framework for the Assessment of Dynamic Resilience. Chem.
Gay, D.; Ray, W. H. Application of Singular Value Methods for Eng. Sci. 1983,38, 1881-1891.
Identification and Model Based Control of Distributed Parameter Palazoglu, A. N.; Manousiouthakis, B.; Arkun, Y. Design of Robust
Systems. Prep. IFAC Workshop Model Based Process Control, Chemical Plants. A New Framework to Assess and Improve Ro-
Atlanta 1988. bust Dynamic Operability. Ind. Eng. Chem. Process Des. Deu.
Handman, S. E.; Leblanc, J. R. The Horizontal Ammonia Converter, 1985, 24, 802-813.
presented a t the AIChE Ammonia Safety Symposium, Los An- Rivera, D. E.; Morari, M.; Skogestad, S. Internal Model Control. 4.
geles; American Institute of Chemical Engineers: New York, 1982. PID Controller Design. Ind. Eng. Chem. Process Des. Deu. 1986,
Klema, V. C.; Laub, A. J. The Singular Value Decomposition: Its 25, 252-265.
Computation and Some Applications. IEEE Trans. Autom. Shinnar, R. Impact of Model Uncertainties and Nonlinearities on
Control 1980, AC-25, 164-176. Modern Controller Design-Present Status and Future Goals. In
Lau, H.; Jensen, K. Evaluation of Changeover Control Policies by Chemical Process Control-CPC III; Morari, M., McAvoy, T. J.,
Singular Value Analysis Effects of Scaling. AIChE J . 1985, 31, Eds.; Elsevier: New York, 1986; pp 53-93.
135-146. Stephens, A. D. Stability and Optimisation of a Methanol Converter.
Lau, H.; Alvarez, J.; Jensen, K. Synthesis of Control Structures by Chem. Eng. Sci. 1975, 30, 11-19.
Singular Value Analysis: Dynamic Measures of Sensitivity and Wong, C.; Bonvin, D.; Rinker, R. G.; Mellichamp, D. A. On Con-
Interaction. AlChE J . 1985, 31, 427-439. trolling an Autothermal Reactor at an Unstable Steady-State. IV
Litz, L. Order Reduction of Linear State-Space Models via Optimal Model Fitting and Control of the Laboratory Reactor. Chem.
Approximation. Proc. IFAC Symposium on Large Scale Systems; Eng. Sci. 1983, 38, 618-634.
Pergamon Press: New York, 1980; pp 195-202.
MacGregor, J. Continuous Emulsion Polymerization Reactor Con- Received for review March 5, 1990
trol. Proc. 1985 Am. Control Conf. 1985, 878-880. Accepted March 19, 1990
Part 1 presents reactor models for all regimes of behavior of dense bubbling fluidizing beds, and
the single parameter appearing throughout is the mean bubble size in the bed. The models are simple
in that they do not require numerical solution, and they reasonably fit the reported data for these
systems. Part 2 develops a freeboard entrainment reactor model to account for the extra conversion
taking place above the dense bubbling bed. This model extends naturally to fast fluidization and
other lean-phase contactors such as the circulating fluidized bed and suggests what type of information
is needed to make reliable reactor performance predictions in these systems.
Part 1. Dense Bubbling Fluidized Beds Davidson’s analysis of the flow of gas within and in the
To predict the behavior of a chemical reactor requires vicinity of a rising gas bubble and, secondly, Rowe’s finding
information on the stoichiometry, thermodynamics, heat that a rising bubble was accompanied by a wake of solids
and mass transfer, reaction rates, and lastly, flow or con- and that this was the main mechanism causing solid cir-
tacting pattern of materials in the reactor. For fluidized culation in a fluidized bed.
reactors it is the flow and contacting that cause particular These developmenta led to a new class of reactor model,
difficulties, and various modeling approaches have been the h y d r o d y n a m i c model, in which the bed behavior was
taken to account for this factor. In the early days, simple based on the characteristics of these rising bubbles. Over
contacting models such as plug flow, mixed flow, disper- a dozen of these models have been proposed but all for the
sion, and tanks in series were tried. Then, following the extreme of very fine particles in which the rising bubbles
lead of Toomey and Johnstone,’ 2-region models were are surrounded by very thin clouds of circulating gas. Why
proposed. The aim of these models was to account for the this extreme? Because it is this that is often of practical
observed nonhomogeneity of fluidized beds, in effect, the importance.
bubble and the dense phases. Over 30 distinctly different In this paper we first update one of the fine particle
models of this type have appeared in the literature. models, the K-L model. We then extend the analysis to
The 1960s saw two major advances in the understanding the other contacting regimes, thus to beds of intermediate
of gas/solid contacting in bubbling fluidized beds, firstly, particles in which bubbles are surrounded by thick over-
0888-588519012629-1226$02.50/ 0 0 1990 American Chemical Society
Ind. Eng. Chem. Res., Vol. 29, No. 7, 1990 1227
Small particles Large particles
fast bubbles Intermediate slow bubbles
Ub >>Umf/Emf Ub -(I to 5) u m,f / & m, f u b <Umfl&mf
. .
Cloud1
Very thin bubb
clouds
. . .
1 I
.. 1
('4 (e)
Figure 1. Many combinations of assumptions adopted in developing
g
a gas/solid contacting model for dense bubbling fluidized beds.
+ Ubr
I\
ub = uo - umf (4) 0.7 I I I I I 1 I
Recent experiments by Werther and Hilligardt3 in beds '- Rounded catalyst
I
i
of various diameters show that the rise velocity of bubbles
in large diameter beds of small particles can be very much O6 VYAcrylic
granules
larger than predicted by eq 4, due primarily to gulf
streaming. On the basis of these measurements, we pro-
pose the following correlations:
For Geldart A solids with d, 5 1 m,
ub = 1.551(~,- umf)+
14.l(db + 0.005))d>32 + Ubr [m/s] (5)
For Geldart B solids with d, I 1 m, 'k'I 0.3
0.3 ~ Irregular
j
II !
may be smaller than that calculated from eq 4. In these \
Magnesite
Crished
cases, always use the larger value. coal
powder
Next the fraction of bed consisting of bubbles is de- 0.1
termined by the flow regime; thus, for slow bubbles, or u b
Umfltmf,
0 I I 1 1
0 100 200 300 400 500 600
dp (w)
Figure 4. Wake fraction (f,)of three-dimensional bubbles a t am-
For intermediate bubbles with thick clouds that may well bient conditions, evaluated from X-ray photographs by Rowe and
overlap, or umf/tmf< ub < 5ud/emf, we have a difficult to Partridge.* V, = wake volume, v b = bubble volume.
represent regime. Interpolating between the slow and fast where the wake fraction (f,)is found from the measure-
regimes, we have, approximately, ments reported by Rowe and Partridge* shown in Figure
4.
The interchange rate coefficient between regions i and
J is defined by Ki,
= (m3of gas flowing from i to j and also
fromj to i)/(m3hubble)(s) [s-l]. Hence, between bubble
when and cloud
--1 - -- K , C A ,
dlvA
V, dt
K, = [ m3 Of gas
m3 of so1id.s
] (18)
eq 23 may have to be integrated numerically by using
bubble size profiles in the bed with changing Kh and K,
values corresponding to these changing bubble sizes.
The overall reactor efficiency when compared to plug
flow is then
where K, is the reaction rate constant. This measure is
independent of bed voidage and particle size if pore dif-
fusion effects do not intrude; hence, it is useful for linking 'bed
- ( catalyst needed in plug flow
- catalyst needed in a fluidized bed
reactor performance of fixed beds to fluidized beds.
For a feed rate u (m3/s) of reactant gas CAi(mol/m3) to
a catalyst bed containing solids of volume V, (m3),inte- effective first-order rate const in fluidized bed
gration of the performance expression gives the outlet true first-order rate const
concentration CAo, or the outlet fractional conversion XA, (25)
as
for plug flow 1 - X A = C A o / C A i = exp(-K,~) A comparison of eq 19 with eq 23 then gives
(19)
cAo 1
for mixed flow 1 - XA= - =-
C A ~ 1 + K,T
where the reactor ability measure is -
1 %I
= ,' [ m3 of catalyst
] (21)
+
-
Note that the efficiency is a function of the reaction rate
(26)
-
T
U m3 of feed/s constant. For slow reaction, 1;for very fast reaction,
'Ibed Y b d / ( l - tf)*
and where the dimensionless reaction rate group is Reactor Model for Bubbling Beds for Intermediate
Particles or ud/ed < u b < 5(id/ed). With fairly large
i = m, f, mf (22) particles, roughly Geldart I3 solids, the bubbling bed may
behave somewhere between the extremes of very fast and
For reversible reactions, replace all C A values by CA - very slow bubbles. Here bubbles surrounded by large
CA,Wuilib and X A by X A / X A , e q u i E b in eqs 19 and 20 and overlapping clouds rise faster, but not much faster, than
elsewhere, and the development throughout follows with- the emulsion gas. These overlapping clouds may well
out further change. Also, the performance expression comprise the whole of the emulsion phase.
developed here assumes isothermal flow with negligible A simple model to represent this situation is sketched
density change on reaction. When these assumptions are in Figure 2b. It views the bed as consisting of two regions,
not reasonable, Levenspie17shows how to account for this. bubble and emulsion, with just one interchange coefficient,
Next, consider the fluidized bed reactor. For the fine- Kbe,to represent the transfer of gas between regions. In
particle system with its five reaction and mass-transfer contrast to the fine particle model, here the upflow of gas
steps, the performance equation in its various forms is through the emulsion is not ignored.
For a first-order irreversible catalytic reaction, an ac-
counting of reactant gas A as it rises through these two
side-by-side regions gives
disappearance in bubble =
For vigorously bubbling conditions, or uo>> umf,we may (reaction in bubble) + (transfer to the emulsion)
use the approximation of eq 10. Thus, with eq 22, we get,
in slightly different form, disappearance in emulsion =
(reaction in emulsion) + (transfer to the bubble)
1 - X A = ex.( -K*) UO = exp[ - K 1f -L€f ] (23b) In symbols the above word equations become
1
1
YbKr + dCAe
1 -(1 - 6)Umf -dz
=
Kf=[ L
Kbc +
YcKr + 1 1 1 [S-'] (24)
(1- 6)(1 - t m f ) K r C A e - h K b e ( C A b - CAJ (28)
- +-
Kce ~eKr In these expressions, the rise velocity of bubble gas, not
The details of the derivation of these expressions are all just the bubble, is
given by Kunii and Levenspiel? and the basic parameter ub* = u b + 3umf (29)
needed to make these conversion predictions is the "mean" and d is given by eq 8. In addition, since the emulsion
bubble size in the bed. consists primarily of cloud gas, the interchange coefficient
For these fine-particle beds, bubbles quickly reach an between phases can be approximated by the bulk flow term
equilibrium size not far above the distributor. Thus, when of K b c of eq 16. Thus,
the reaction is not extremely fast and occurs significantly
throughout the bed, one can reasonably use an average Kbe 4*5(umf/db) (30)
1230 Ind. Eng. Chem. Res., Vol. 29, No. 7, 1990
Solving eqs 27 and 28 simultaneously subject to the from which the bubble fraction is
boundary condition
CAb = CAe = CAj at 2 = 0 (31) (7)
gives the overall conversion of reactant gas as
Also, the fraction of gas passing through the emulsion and
contacting solids is, with eqs 7 and 37,
or
and
Noting that tmf = 0.5 and u b Iud/cmfand that 6 cannot
reasonably exceed 0.25-0.35, eq 7 shows that this con-
tacting regime only exists when uo < 2 - 2 . 5 ~A~t higher
superficial gas velocities, channeling, slugging, and explo-
sive bubbling will replace ordinary bubbling behavior.
Geldan A panicles
05
7
x
- 02
01
0 05
--,\-
43; ( 0 )Calderbanke (perforated plate), 68,0.006,0.458,0.69,0.043;
(0) Van Swaaij,lo -70, 0.005, 0.23, -, 0.12; @) Van swaaij,lo -70, P = Pbed e-'' ~1 -17 =- ( I - q b~
ed ~ . . ~ ~
0 Pbed Ps 0 obed 1
Density, p Contact efficiency q
CAi
Li ( 1 -&i )
K, 7 =?,K
0
i = m, m f or f .
Figure 6. Conversion in beds of Geldart A/B and B catalyst. Lines
are calculated from eqs 23 and 24 of the fast bubble model with ud 0
L '&-o I 1 1 I I I 1
= 0.02 m/s. The following information is given in the order of 0 200 400 600 800
symbol, reference, d, (jtm), ud (m/s), dt (m), L, (m), u, (m/s). For
ozone decomposition: (0) Kobayashi,12 194, 0.021, 0.20, 0.1-1.0,
d, (w)
0.054.18; ( 0 )Potter,ls 117,0.017,0.23,0.1-0.4,0.06-0.08; (0) Cal- Figure 9. Decay constant for the exponential decrease of solid
derbank? 192,0.037,0.46,0.69,0.086; ( 0 )Van Swaaij?O --200,0.010, density in the freeboard 1,Chen et alJ1; 2, Bachovchin et al.=; 3,
0.30, -, 0.15. Hoggen et al.19; 4, Walsh et aL20; 5, Zhang et al."; 6, Nazemi et al.'*;
7. Lewis et al.16
ejected solids eventually return to the bed, Lewis et a1.I6 where p o is the density just above the surface of the dense
found that the density of solids in the freeboard falls off
exponentially with height with decay constant a, or bubbling bed, 9,prepared from reported literature
data, correlates this decay constant with particle size and
p = poeQz (41) gas velocity.
1232 Ind. Eng. Chem. Res., Vol. 29, No. 7, 1990
(43)
However, in the freeboard, the contacting efficiency is not
100% and the solid fraction is low. Accounting for these For the upper freeboard zone, simply apply the free-
two additional factors, the first-order expression at any board model, which leads to the conversion expression of
height z becomes eq 45 or eq 46. When used for fast fluidized beds, the
various physical quantities may fall in different ranges of
values. Thus,
(44)
In this paper, we present and demonstrate a numerical algorithm to detect and solve mixed systems
of nonlinear ordinary differential and algebraic equations that have an index problem. Symbolic
differentiation is used to generate the equations needed; symbolic elimination is not required as
in many other algorithms. We show that the so-called initialization and propagation problems are
in fact the same problem; they can therefore be handled by a common approach. Numerical as well
as structural singularities arising in index problems are accommodated. In principle, the algorithm
can be used with any integration scheme, including an explicit one. This algorithm is suitable for
solving index problems that previously could not be solved.