You are on page 1of 8

Chekcat En&ceriq Seienre. Vol. 41, No. 6, pp. 1647-1654. 1986. OW9-X09/86 S3.00 + 0.

00
Printed in Great Britain. F%zrgamoa loum.3la Ltd.

PARAMETRIC SENSITIVITY OF A CSTR

R. M. CHEMBURKAR, M. MORBIDELLI and A. VARMA+


Department of Chemical Engineering, University of Notre Dame, Notre Dame, IN 46556, U.S.A.

(Received 5 December 1985)

Abstract-It is shown that the generalizedsensitivitycriterionrecentlydeveloped in the context of thermal


explosions and tubular reactors can be easily applied in the case of a CSTR as well. An illustrativeexample
concerningsensitivityanalysisof a singlenth order irreversibleexothermicreactionin a nonadiabaticCSTR
is presented.A generaked region of parametric sensitivity is identitkd where the reactor temperature is
parametrically sensitive simultaneously to al! the input parameters. Asymptotic analysis for large heats of
reaction is performed to investigate limiting hehaviour, which leads to the classical Semenov limit in the case
of large activation energies. It is shown that parametric sensitivity can occur even when unique steady states
exist for all Damkiihler number (Da) values.Furthermore,if operatingconditions are chosen so as to avoid
completelythe possibilityof parametricsensitivityfor all Da, then the possibility of steady-state multiplicity
is automaticallyavoided.

1. INTRODUCIION the classic work of Semenov (1928), several theoretical


It is well known that chemical reactors, in certain and experimental studies have been published in order
regions of operating conditions, may exhibit a para- to develop and verify criteria for ignition (Zeldovich
metrically sensitive behaviour, i.e. small changes in et al., 1985). The close connection between thermal
one or more of the reactor input parameters lead to explosion and chemical reactor theories is due to the
much larger changes in the output variables. This striking similarity between the pseudo-homogeneous
concept was first introduced in a pioneering study by models describing the temperature history of an
Bilous and Amundson (1956), who investigated the explosive medium and the temperature profile along
sensitivity behaviour of a pseudo-homogeneous tu- the tubular reactor axis. This allows direct application
bular reactor. of any criterion, originally developed in the context of
For exothermic reactions occurring in a tubular thermal explosions, to chemical reactors and vice versa.
reactor, the temperature profile along the reactor Earlier c1 priori criteria, which have recently been
length exhibits almost inevitably a maximum value reviewed and compared (Morbid&i and Vat-ma,
often referred to as a “hot spot”_ When such a reactor is 1985a), were all based on some geometric feature of the
operated in the parametrically sensitive region, small temperature-reactor length (or time) profile. In par-
changes in the inlet or other operating conditions may ticular, the criterion proposed by Adler and Enig
produce large increases in the hot-spot magnitude, (1964), which was recommended as the most reliable
thus seriously affecting reactor performance or cata- among them, defines runaway as the occurrence of a
lyst activity. Therefore in practical applications, it is region with a positive second-order derivative some-
usually desired to avoid such operating conditions where in the temperature-conversion profile, before
leading to parametric sensitivity, and it is particularly the hot spot.
beneficial to do so in the earlier stages of reactor Only recently, and in the context of thermal ex-
design. This necessitates the formulation of an a priori plosions, sensitivity criteria have been proposed based
criterion for parametric sensitivity, based upon which on the rigorous concept of sensitivity (Lacey, 1983;
any possible steady-state regime of the reactor can be Boddington et al., 1983). In particular, we have pro-
identified as either parametrically sensitive or non- posed (Morbidelli and Varma, 1985b) a generalized
sensitive. Using such a criterion, the regions of para- sensitivity criterion based on the sensitivity of the
metric sensitivity in the reactor parameter space can maximum temperature to any of the model input
be identitled in an a priori fashion, i.e. without parameters. It was shown that sensitivity exhibits a
reference to any specific reaction or reactor regime. sharp maximum when the heat of reaction parameter is
The practical importance of this problem has varied. The value of the heat of reaction parameter at
prompted a large number of investigations, all dealing which the maximum occurs is termed the critical value,
with the case of tubular reactors, where various criteria since it separates the sensitive and nonsensitive regions
for parametric sensitivity have been developed. of reactor behaviour. It was also found that the same
It is worth noting at this juncture that quite similar critical value is obtained regardless of which particular
problems arise in the context of thermal explosion input parameter is considered in the definition of
theory, where the occurrence of parametric sensitivity sensitivity. Thus this criterion identifies a generalized
is basically identitled with ignition phenomenon. Since region of parametric sensitivity, where the reactor
becomes simultaneously sensitive to small changes of
any of the input parameters. Moreover, the sensitivity
-0 whom correspondenceshould be addressed. boundaries obtained were in good agreement with
1647
1648 R. M. CHEMBURKAR et al.

those computed through the Adler and Enig (1964) adiabatic CSTR can be cast into the following compact
criterion mentioned above. This finding rigorously dimensionless form (Kaushus et al_, 1978):
supports the reliability of the latter criterion, where
F(v; Da, /?, y, n) = Da (1 +fl-0)”
sensitivity is defined without reference to any specific
model parameter-thus implicitly assuming the exist- x exp[y(l - l/v)j/(u - 1) -p-1) = 0 (1)
ence of a generalized sensitivity criterion.
(u-1)-/?(l -U) =o. (2)
In addition to its rigorous character, the great
advantage of the proposed criterion is that it can be All the distinguishing features of the system can be
extended to any reactor model and to any reactor determined by investigating eq. (1) alone, since eq. (2) is
characteristic, besides the hot-spot magnitude. This is merely a linear algebraic relation between the dimen-
not true of any of the previous criteria, which are based sionless concentration, u, and the dimensionless tem-
on some topological feature of the temperaturc- perature, 0.
reactor length profile, characteristic of tubular Let us define the absolute and the normalized
reactors. sensitivities as follows:
An example illustrating the generality of the pro-
posed criterion is given in the present work, where the ~(0; 4) = Idv/d+(
sensitivity behaviour of a single irreversible nth order and
exothermic reaction in a homogeneous nonadiabatic S(W+) = ld(lnWd(ln#I = IW/v)s(v;#l (3)
CSTR is examined. Even though this is the simplest of where 6 represents any of the four model input
chemical reactors, its detailed sensitivity analysis has, parameters, i.e. Da, 8, y and n. As discussed previously
to our knowledge, never been reported previously in in the context of thermal explosions (Morbidelli and
the literature. This may be because none of the Varma, 1985b), normalized sensitivity is a more con-
previous criteria can be applied in this case (in fact, venient quantity for sensitivity analysis since it has a
there is neither a temperature profile nor a hot spot), clearer physical meaning, which derives from the
and only the generalized criterion for parametric normalization procedure implicit in its definition.
sensitivity mentioned above can be used effectively. Hence it will be adhered to in this study. Sensitivities s
Specifically this allows an a priori identification of the or S, readily obtained by differentiating eq. (1) with
regions of generalized parametric sensitivity in the respect to q5, can be easily written in the following
reactor parameter space. form:
A particularly interesting aspect of this work is that, s(o; 6) = 1-(+/VI (=‘lW)lW_l~~)l (4)
due to the simplicity of the model studied, both
sensitivity and steady-state multiplicity behaviour of where the analytical expressions of all the partial
the reactor can be fully identified, This allows a better derivatives are reported in Table 1.
understanding of the connection between these two In Fig. 1, the steady-state value of the reactor outlet
phenomena, often regarded as separate issues, and by temperature, v, is shown as a function of the heat of
combining both these results, safe operating con- reaction parameter fi, together with the corresponding
ditions for a CSTR can be identified which avoid both normalized sensitivity of the outlet temperature to /3,
multiplicity as well as parametric sensitivity. i.e. S(v; B). The sensitivity displays a sharp maximum at
j3 = j?,, as fl is varied. At this point the outlet tempera-
2. SENSITIVITY ANALYSIS ture is most sensitive to changes in /?. For j3 values
The steady-state behaviour of a single irreversible larger than fi,, the reactor is in an ignited state where
nth order exothermic reaction occurring in a non- the reactor temperature has reached a much higher

Table 1. Exact analyticalformulae for various partial derivativesof F

dDa= (v- 1)

exP[Y(l-:>I
aF=D* n(1 +fi -vr-’ -(?I -1)8”_*
afi (v-1)
aF Da(l+p-vr
-=
a?

dF
--Da
an
Parametricsensitivityof a CSTR 1649
is also evident that sensitivity is u phenomenon in-
dependent of multiplicity; in fact, it may also occur in
the region of reactor steady-state uniqueness, as
shown, for example, in Fig. 1.
It is worth pointing out that in Fig. 2 the sensitivity
values refer to a reactor steady state on the low-
temperature branch as long as this branch is available.
When the temperature jumps to the high-temperature
branch at the bifurcation point /I*, the sensitivity
exhibits a discontinuity, dropping from an infinite to a
finite value. Throughout this study we concern our-
0.20 0.2t 024 026 028 0.30
selves mainly with the low-temperature steady state.
The case where the reactor is operating on the ignited
-6
brzinch is of less practical interest from the viewpoint
Fig. 1. Effectof the dimensionlessheatof reaction,S, on v and of defining temperature runaway and parametric sensi-
S(V; 8). Da = 0.11, y = 20, n = 1. tivity; nevertheless, comments about this case will be
made later in this study.
In Fig. 3 the sensitivity of the outlet temperature
with respect to various inlet parameters r$ is shown. It
value which is undesirable in most cases. Thus accord-
appears that the critical #Ivalue, i.e. the location of the
ing to the generalized sensitivity criterion, #I, cor-
maximum sensitivity, is the same to three significant
responds to the critical value for sensitivity, so that for
figures (#I, = 0.248) for any choice of 4. This clearly
/I < /I, the reactor is nonsensitive, while for fl > #i, it is
shows the generalized nature of the adopted criterion
parametrically sensitive.
because the reactor outlet temperature becomes sensi-
In this situation depicted in Fig. 1, the reactor
tive to all the model parameters simultaneously. Thus,
exhibits a unique steady state for all values of /l. It is
the parameter range characterized by /I > fl, can be
well known that nonisothermal CSTRs may exhibit
called the generalized region of parametric sensitivity.
steady-state multiplicity, leading to curves of classical
Thisclaim is further substantiated by the results shown
S-shape in the G-P plane under consideration. In this
in Table 2, where the critical values of + (defined as
case, as shown in Fig. 2, the temperature sensitivity
Da #) which maximize S(v; r#~) with respect to /I for any
increases for increasing /I values and becomes infinite
choice of the input parameter 4 are reported. The
at the bifurcation point /I = B*. In fact, at any bifur-
quantity + is defined to be similar to the Semenov
cation point aF /&I = 0, and therefore, from eq. (4),
number in explosion theory so as to facilitate com-
S(v, 4) + cl, for any 4. Since sensitivity attains its
parison between various reaction configurations and
maximum value at the bifurcation point /I*, it follows,
to tie in general concepts later in this paper. The
according to the adopted sensitivity criterion, that this
reported values of critical $ (i.e. $,) in Table 2 were
point also represents the critical condition for sensi-
obtained by computing the critical values of p(i.e. 8,)
tivity, i.e. /I, = j?*. This agrees with the results obtained
for any chosen y and Da, then using the definition of 1/1
by Barkelew (1984) about sensitivity behaviour of a
(i.e. +, = Dayj3,)_
CSTR. However, in that work, no sensitivity criterion
Two definite trends may be noted from Table 2:
was adopted, and the occurrence of sensitivity was
simply identified always with the occurrence of multi- (1) For smaller Da, and a fixed y, all the /3, values
plicity_ In the present work, such a coincidence follows become the same. However, the difference between
rigorously from the adopted sensitivity criterion, and it the /I, values found for different choices of 4
increases as Da increases.

1.3 -

I
>
1.2 - -
-4

_
^
cl_
._
m

I.1 -
8. = /3’- 0.340

IO* _ _ : . . .
0.20 0.25 0.30 0.35

Fig. 2. Effect of the dimensionlessheat of reaction, /3. on v


and S(u; B) in the multiplicity region. Da = 0.07, y = 20, n Fig. 3. Rehaviourof sensitivitiesS(u; I$)with the dimension-
= 1. less heat of reaction. #?.Da - 0.11, y - 20, n = 1.
1650 R. M. CHEMBURKAR et al.

(2) As y increases, the difference between the various PC


values decreases.
These results show that in the region of low heat of
reaction /I and activation energy y, the critical values
obtained from the analysis of S(u; 4) become de-
pendent on the particular choice of 4. Thus in this case
the validity of the concept of a generalized sensitivity
region is lost. The low overall energy available to the
system in these cases results in a mild nature of any
temperature-related phenomenon. This mild nature
can be evidenced by relatively low values of the
maximum S(I); 4) attained, as can be seen for the case of
large Da in Fig. 4. As Da decreases, the critical /I value
increases, and so does the value of sensitivity at
criticality. Thus for large B values, where thermal
sensitivity is a significant phenomenon, the generalized
sensitivity concept holds (see Table 2). For smaller /I
values (or larger Da values) where the concept does not
hold, the sensitivity in itself is a much less dramatic
phenomenon. In fact, for relatively large Da values,
there is no critical fi, value indicating that thermal
sensitivity no longer exists. Similar behaviour was
previously reported for tubular reactors using the
Adler and Enig criterion (Morbidelli and Varma,
1982).
From the values reported in Table 2 it may be seen
that the critical value of @, as Da + 0, approaches an
asymptotic value which depends on y. This can also be
proven analytically. For Da --* 0 (or, equivalently, for
large heats of reaction), the critical value PC can
invariably be obtained from one of the bifurcation
points of eq. (1) as long as y > 4. For n # 0 and large /i,
this bifurcation point corresponds to the smaller root
of
02 -yv+y = 0. (5)
Indicating this root by uC,the critical value of /l (i.e.
fi, = fl*) can be found by substituting v, into eq. (1).
Thus the asymptotic value of +, = Da yfi, is given by

Y (0, - 1) exp C - ~(1 - l/v31 (6)


which, in the limit of y + co, approaches e- ‘. For
n = 0, the bifurcation point is the smaller root of
eq. (5) for all values of /3_Hence, in the ease of n = 0, the
value of ti, given by eq. (6) is valid for all /I.

Fig. 4. Effect of Da on the behaviourof S(v; @ vs. 8. y = 20, n


= 1.
Parametric sensitivityof a CSTR 1651

This conclusion agrees with classical results pre- which S(V;~), irrespective of the choice of 4, is
viously reported in the context of thermal explosion maximized with respect to any of the involved input
theory (cf. Morbidelli and Varma, 1985b), which also parameters_
apply to pseudo-homogeneous plug-how reactors. In
particular, for large values of the heat of reaction
(equivalent to fi in the CSTR case; cf. Vat-ma and Aris,
1977), it was shown that the Semenov number, defined 3. REGIONS OF PARAMETRIC SENSITIVITY FOR
as the ratio between the rate of heat production by A CSTR

reaction and heat removal from the system (i.e. + After establishing the concept of a generalized
= Day/3 in the CSTR case), approaches a finite value region of parametric sensitivity for the case of a CSTR,
given precisely by eq. (6). Moreover, for y + co, this these regions are mapped by examining the behaviour
value approaches the classical Semenov limit e- ‘. This of S(v; Du) with respect to /YIfor various values of y (for
is a rather interesting finding, which, in view of the a first-order reaction, i.e. n = 1) and for different
different nature of the chemically reacting system reaction orders (for y = 20) in Fig. 5(a) and (b),
examined, indicates the intrinsic nature of the sensi- respectively. Each curve in these figures represents the
tivity concept. critical conditions for parametric sensitivity. Thus
Up to this point, S(v; 4) has been maximized with values of the operating parameters lying above the
respect to /3 only, in the process of finding 8,. We now curves indicate parametric sensitivity, while nonsensi-
examine the results when the maximization is done tive safe operation is obtained below the curves. These
with respect to some other parameter. In Table 3, half figures indicate that sensitivity is more likely as either
of the values are reproduced from Table 2 [where the activation energy y increases or the reaction order n
S(U; 4) was maximized with respect to /I], while the decreases.
others are obtained by maximizing S(u; 4) with respect The asymptotic behaviour in the limit of small Da,
to Da. Again the same conclusion as before is reached; which is not clearly seen in Fig. 5, is illustrated in Fig. 6
i.e. despite the discrepancies at large Da, the sensitivity obtained by transforming Fig. 5 from the &Da plane
criterion becomes general as Da is decreased. This to the $-Da plane. Figure 6 confirms attainment of the
reaRlrms the generalized character of the obtained limits given by eq. (6) for various y values, and the
sensitivity region, since it is bounded by a curve on Semenov limit +, = e-i as y becomes large.

Table 3. Critical values of 1,4= DONS; n = 1, y = 2Ot

The critical value found by maximizing S(q 4)


with respect to fi with respect to Da

d
DU Da B Y n Da B Y ”
l_
0.01 0.3992 0.3992 0.3992 0.3992 0.3992 0.3992 0.3992 0.3992
0.05 0.4488 0.4488 0.4488 0.4488 0.4488 0.4488 0.4488 0.4488

0.11 0.5454 0.5454 0.5454 0.5454 0.5452 0.5452 0.5458 0.5452


0.14 0.6076 0.6121 0.6129 0.6140 0603 1 0.6075 0.6106 0.6cr97
0.17 0.6640 0.68 10 0.6824 0.6851 0.6481 0.6638 0.6610 0.6698
0.20 0.7144 0.7504 0.7428 0.7504 0.6826 0.7159 0.7002 0.7254
0.30 0.8688 1.0332 0.9552 0.9618 0.7411 0.8709 0.7687 0.8824
0.40 0.9992 $ 1.1728 1.1752 0.7365 1.0091 0.7682 1.0040

+For entries above the solid line. the rector exhibits steady-state multiplicity; below the solid line, the steady
state is unique.
tThe curve S(Y; j?) does not exhibit any maximum in this case.

0.5 0.5
a b

E L!!sd
ynlo Il.1 y = 20
0.4 0.4

” -3
0.3 0.3

1;
0.2

0.1
50
20

100
‘i 0.2

0.1
2
I

0 0.5

0 0. I 0
0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
- Da - Da

Fig. 5. Effect of (a) dimensionless activation energy y and (b) reaction order non the parametrically sensitive
region in the B-Da plane.
1652 R. M. CHEMBURKAR et a!.

Fig. 6. Effect of (a) dimensionless activation energy y and (b) reaction order n on the parametrically sensitive
region in the $-Da plane.

4. COMPARISON OF THE REGIONS OF PARAMETRIC globally stable; any large perturbation in the operating
SENSITIVITY AND STEADY-STATE MULTIPLICITY conditions (or the transients during start-up or shut-
‘The multiplicity analysis of a CSTR, wherein an nth down) can be detrimental-thus global sensitivity is
order exothermic reaction occurs, has been the subject affected. In fact, using this reasoning, Bilous and
of numerous earlier studies. It is well known that for Amundson (1955) restricted themselves to multiplicity
y > 4, for every reaction order n, there exists a region in considerations while studying the sensitivity of
the &Da plane where multiple solutions are possible CSTRs. In the case of excursions like PBQa and P4Q4,
for eq. (1). the sensitivity exhibits a maximum when the boundary
Figure 7 represents typical behaviour in the /l--Ba of parametric sensitivity region is crossed, i.e. when the
plane for lixed values of y and n, where the region of excursion cuts across the solid curve in Fig. 7.
parametric sensitivity lies above the solid curve, and Consider now the case where the reactor is being
the multiplicity region exists between the dashed and operated in the ignited state; i.e. on the high-
the solid curves. For an existing unique nonsensitive temperature branch of Fig. 2, when Da K Da=_ in Fig.
steady state on the low-temperature branch, any 7. The boundary of the parametric sensitivity region in
change in the operating variables is considered safe as this case will be different from that found earlier by
long as these changes are restricted to the para- analysing the sensitivity of the low-temperature steady
metrically insensitive region, e.g. the excursion P,Q1 in state. It is self-explanatory from Fig. 2 that sensitivity
Fig. 7. The same holds for excursion PaQI, but during will now become infinite at /3 = /I,_ Hence in this case,
this excursion the multiplicity region is encountered. as the solid curve in Fig. 7 (loci of /?*) is crossed from
This type of excursion results when /l is increased from the top, the region of multiplicity is entered which will
a value less than fi* to a value lying between ~3. and /I*, affect global sensitivity; however, local sensitivity will
keeping Da constant as in Fig. 2. Thus it is evident from not be affected. In fact, local sensitivity is not affected
Fig. 2 that local sensitivity of the low-temperature as long as the operating conditions remain above the
steady state does not exhibit a maximum, and hence dashed curve in Fig. 7 (loci of fl*).
the excursion is safe from a local sensitivity point of In practical situations, the volumetric flow rate q is
view. However, the existing steady state is no longer the variable which can be subjected to change most
readily. Thus it is of significant interest to know a priori
whether for a given set of physicochemical par-
ameters, any multiplicity or parametric sensitivity is
O’O- encountered for any value of the flow rate q. In order to
conduct this analysis, it is necessary to cast the physico-
chemical parameters into dimensionless quantities
which are independent of the flow rate q_ Recently,
multiplicity analysis from this point of view was
reported by Balakotaiah and Luss (198 1). This involves
defining two new variables, a = 6/Da and B = yfl(l
+ a), which are completely independent of q for the
case of T, = T,. Since now only the Damkiihler
number involves q, the results obtained about multi-
plicity/sensitivity for some q can equivalently be said
for some Da.
Balakotaiah and Luss (1981) used singularity theory
to identify regions of all different types of multiplicities
for some Da in the a-8 plane. These results for the case
Fig. 7. Regions of multiplicity and parametric sensitivity in
the /?-Da plane. The region of parametric sensitivity lies above of T, = Tr are shown in Fig. 8. In Fig. 9, the a-B plane
the solid curve. The region of multiplicity is enclosed by the is shown divided into the region of multiplicity for
dashed and the solid curves. some Da, and the region where no multiplicity is
Parametric sensitivity of a CSTR 1653

4 some Da values, then parametric sensitivity also exists


m for some Da values; however, the converse is not true.
The same conclusion can also be reached from Fig. 7,
II
where for j3 > fl,, both multiplicity and parametric
iy
sensitivity occur for some Da values; however, for
I
c
#3 -z flFUsP
although multiplicity is not possible for any

F
CI

E
I Da, parametric sensitivity may occur for some Da.

P Ip

5. CONCLUDING REMARKS
In this study, the region of parametric sensitivity for
a CSTR has been identified. It is shown that a
generalized region of parametric sensitivity can be
defined, which is independent of the input parameter
Fig. 8. Identification of different types of multiplicity pat- chosen for defining the sensitivity. These findings are in
terns in the a-B plane (not to scale). (I) Unique steady state; complete agreement with results previously obtained
(ID isola; (III) mushroom: (IV) S pattern; (V) S+isola
in the case of thermal explosions and pseudo-
homogeneous tubular reactors (Morbidelli and
Varma, 1985b). Specifically, also for these reacting
20 I I I I systems, using the same sensitivity criterion, the exist-
ence of a generalized sensitivity region was proven. The
‘/ I
similarity among these systems is not only qualitative
but also quantitative. In the limit of large heat of
reaction parameter, the so-called Semenov number I&,
representing the ratio between the rate of heat produc-
tion and cooling, approaches the same asymptotic
value. Thus, using dimensionless parameters having
equivalent physical meaning (cf. Varma and Aris,
1977), the CSTR and the plug-flow reactor exhibit the
same sensitivity region, at least in the limit of large heat
-e of reaction_ This finding supports the intrinsic nature
of the adopted sensitivity criterion, which brings
Fig. 9. Comparison between the region of multiplicity and together seemingly different chemically reacting
the region of parametric sensitivity. Multiplicity exists for
some Da below curve a, while parametric sensitivity exists for
systems.
some Da below curve b. y = 100, T, = T,, n = 1. Another important conclusion of this work con-
cerns the comparison between the regions of par-
ametric sensitivity and steady-state multiplicity. This
possible for any Da by curve a Curve a is formed by the leads to an identification of the range of operating
hysteresis (S pattern) line for J!3< B, and the isola curve variables that will yield unique nonsensitive steady
for B > 3 (see Fig. 8); hence above curve a lies the states, required for safe operations. The results ob-
region where multiplicity is not possible for any Da, tained by analysing the a-B plane for the regions in
while multiplicity occurs for some Da below curve a. which steady-state multiplicity or parametric sensi-
Curve b in Fig. 9 is obtained using the following tivity is observed for some Da show that the former is
procedure. Every point on the solid curve in Fig. 7, say always completely enveloped by the latter. In practical
(Da,, #?J, represents the critical conditions for para- applications, since it is usually not desirable to operate
metric sensitivity. Each of these critical points is in the sensitive region, it follows that certainly the
translated into a critical line in the a-B plane by situations where the reactor exhibits multiplicity
varying S and computing a, = a/Da, and B, = ~$3,(1 would not be approached. Therefore, with respect to
+a). Each such line divides the a-B plane for that practical applications, it is of primary importance to
choice of /I, and Da, into two parts; below the line lies establish the boundary of the sensitivity region. Once
the region of parametric sensitivity, while nonsensitive this is avoided, multiplicity is automatically avoided.
operations occur above the line. Repeating this pro- This conclusion is also more general than what is
cedure for many such critical points (Da,, fi,) gives rise shown in this work for a homogeneous CSTR. In
to many such lines, and curve b is tangential to all these particular, Morbidelli and Varma (1986a, b) have
lines. Thus parametric sensitivity cannot occur for any examined a heterogeneous plug-flow reactor for both
Da above curve b, while it is possible for some Da sensitivity and multiplicity behaviour and reached
below curve b. similar results. With respect to reactor design, this
Figure 9 shows that the region of multiplicity for finding stresses the need for sensitivity analysis, which
some Da, is a subset of the region of parametric would automatically account for any possible steady-
sensitivity for some Da. Thus if multiplicity exists for state multiplicity as well.
1654 R. M. CHEMBURKAR etal.

Acknowledgement-One of us (A.V.) acknowledges with fond f feed condition


memory and great pleasure the period of his graduate studies * lower bifurcation point
under direction of the “Chief”.

Superscript
NOTATION * higher bifurcation point
a area of heat transfer
B YS(l +a)
B notation used in Figs 8 and 9 REFERENCES
C reactant concentration Adler, J. and Enig, J. W., 1964, The critical conditions in
specific heat thermal explosion theory with reactant consumption.
CP Comb. Flame 8,97-103.
Da Damkijhler number, V CT- ’ k( T,)/q
Balakotaiah, V. and Luss. D., 1981, Analysis of the multi-
E activation energy plicity patterns of a CSTR. Chem. Engw Commun. 13,
F (u; Da, 111-132.
B, Y. n) function defined by eq. (1) Barkelew, C. H., 1984, Stability of adiabatic reactors. Am.
rate constant as a function of tempera- Chem. Sot. Symp. Ser. 237, 337-359.
k(T)
Bilous, 0. and Amundson, N. R., 1955, Chemical reactor
ture, T
stability and sensitivity. A.I.Ch.E. J. 1, 513-521.
reaction order Bilous, 0. and Amundson. N. R, 1956, Chemical reactor
volumetric flow rate stabilityand sensitivity-II. Effect of parameterson sensi-
universal gas constant tivity of empty tubular reactors.A.I.Ch.E. J. 2, 117-126.
sensitivity of u with respect to 4 Boddington, T., Gray, P., Kordylewski, W. and Scott, S. K.,
1983, Thermal explosions with extensive reactant con-
normalized sensitivity of t, with respect sumption: a new criterion for criticality. Proc. R. Sot. A390,
to 4 13-30.
temperature of the CSTR Kaushus, W., Demont, J. and Hartmann, K., 1978, On the
coolant temperature steady states of continuous stirred tank reactors. Chem
Engng Sci. 33, 1283-1285.
mean temperature, (T,+ ST,)/(l + 6)
Lacey, A. A.. 1983, Critical behavior for homogeneous
dimensionless concentration, C/C, reacting systems with large activation energy. Int. J. En#ng
overall heat-transfer coefficient Sci. 23, 501-515.
dimensionless temperature, T/T, Morbidelli, M. and Varma, A., 1982, Parametric sensitivity
and runaway in tubular reactors. A.I.CLE. J. 28,705-713.
volume of the CSTR
Morbidelli, M. and Varma, A., 1985a, On parametric sensi-
tivity and runaway criteria of pseudohomogeneous tubular
Greek letters reactors. Chem. Engng Sci. 40, 2165-2168.
a S/Da Morbidelli, M. and Varma, k. 1985b. A generalized criterion
notation used in Figs 8 and 9 for parametric sensitivity: application to thermal explosion
z
theory. Proc. R. Sot.
B dimensionless heat of reaction, Morbidelli, M. and Varma, A., 1986a, Parametric sensitivity
(- AH) G/PC, T,(l + 6) in fixed-bed reactors: the role of interparticle transfer
6 Ua/PqcP resistances. A.I.Ch.E. J. 32, 297-306.
AH heat of reaction Morbidelli, M. and Varma, A., 1986b, Parametric sensitivity
an input parameter and runaway in catalytic reactors. Accepted for presen-
4 tation at ISCRE 9, Philadelphia, 18-21 May.
Y dimensionless activation energy, E/RT, Semenov, N. N., 1928, Zur theorie des verbrennungs-
P density prozesses. 2. Pbys. 48, 571-582.
# DayS Varma, A. and Aris, R., 1977, in Chemical Reactor Theory. A
Review (Edited by Lapidus, L. and Amundson, N. R.),
Chap. 2, pp. 79-155. Prentice-Hall, Englewood Cliffs, NJ.
Subscripts
Zeldovich, Ya. B.. Barenblatt, G. I., Librovich, V. B. and
C critical value Makhvjladze, G. M., 1985, The Mathematical Theory of
cusp cusp point Combustion and Explosions. Plenum Press, New York.

You might also like