Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Department of Chemical Engineering, Faculty of Engineering, University of Malaya, Kuala Lumpur 50603, Malaysia
Institute of Science & Technology for Advanced Studies & Research (ISTAR), Vallabh Vidyanagar, Anand, Gujarat 388120, India
Petroleum and Chemical Engineering Programme Area, Faculty of Engineering, Institut Teknologi Brunei, Tungku Gadong, P.O. Box 2909, Brunei Darussalam
a r t i c l e
i n f o
Article history:
Received 22 April 2013
Received in revised form 21 April 2014
Accepted 21 April 2014
Available online 26 April 2014
Keywords:
Outputinput multiplicity
CSTR
Open loop behavior
Parametric sensitivity
a b s t r a c t
The continuous stirred tank reactor (CSTR) presents challenging operational problems due to its complex
open-loop non-linear behavior such as input and output multiplicities, ignition/extinction, parametric sensitivity,
non-linear oscillations and perhaps even chaos. Many researchers have studied the non-linear dynamic behavior
of CSTR for several reaction schemes from a theoretical stand point. The present experimental studies are carried
out to identify the existence of input multiplicity in a non-isothermal CSTR. The sulfuric acid catalyzed hydrolysis
of acetic anhydride reaction system has been chosen and the regions of existence of input multiplicity are theoretically identied. The reactor showed the same steady state temperature for two input ow rates of reaction
mixtures into the CSTR reactor and thus conrming the existence of input multiplicity for the sulfuric acid
catalyzed hydrolysis of acetic anhydride reaction system. It was observed that the time taken by the reactor
was too long at a low feed ow rate of reaction mixture than at a higher feed ow rate reaction mixture. The
model simulations of steady state temperatures were in close agreement with experimental data for the above
feed ow rates of reaction mixture conditions.
2014 Elsevier B.V. All rights reserved.
1. Introduction
Non-linear phenomenon such as multiplicities and stability
problems are encountered in industrial chemical reactors, but apparently proprietary considerations have prevented the disclosure
of data and specic reaction. The occurrence of multiple steady
states under the same operating conditions of a reactor is known
as output multiplicity whereas input multiplicity means that more
than one set of input variables causes the same output variable.
Non-linearity is a necessary condition for input and output multiplicities [1]. Non-linear systems can show a variety of dynamic patterns ranging from stable operation at a unique stable steady state,
to unstable operation such as unstable steady state, oscillations,
and chaos. In general input multiplicities occur due to the presence
of competing effects in non-linear process or due to non-ideal
mixing with certain non-linear reaction kinetics or due to the recycle structure of non-linear processes. Constraints placed on the
http://dx.doi.org/10.1016/j.chemolab.2014.04.017
0169-7439/ 2014 Elsevier B.V. All rights reserved.
manipulated variable of non-linear process and increasing the number of controlled variables can eliminate input multiplicities [2].
However, one of the major difculties in dealing with non-linear
systems is the lack of unied mathematical theory for representing
various non-linear system characteristics. Most of the recent advances in the study and understanding of multiplicity in chemically
reacting systems have resulted from the applications of catastrophe
and singularity theories to situations which are relatively simple, i.e.
the number of intrinsic state variables involved are limited [3,4]. The
CSTR lumped reactor system with control has been reported by numerous researchers [512] investigating various aspects such as stabilization of CSTR, state and parameter estimation methods, design
and operability of the CSTR reactor and control of non-linear CSTR.
In physical terms, due to the existence of input multiplicity phenomena, there may be more than one set of manipulated variables
which can produce the same desired steady state in CSTR. Input
multiplicity can occur if at least one variable appears in a nonlinear fashion in the governing equation. The bounds on the values
of the manipulated variables have a strong impact on the number
of feasible inputs and a decrease in the range of values of the manipulated variable tends to decrease the number of possible inputs for a
specied output. Although there have been developments in the understanding of the dynamic behavior of CSTR including investigations on the existence of input multiplicity for several reaction
214
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
ST-2
ST-1
ST-3
CHT-2
CHT-1
R-1
CHT-3
R-2
R-3
T-2
T-1
C-2
C-1
TS
C-3
CSTR
schemes from a theoretical standpoint [14,1318,20], very few reaction systems have been tried experimentally [1820]. In the absence of a unied mathematical theory can represent various nonlinear system characteristics, it will be appropriate to study the
Table 1
Process variable combinations causing input multiplicity.
Tf
(C)
CAf
(mol/m3)
Ua W/C
q 106 m3/s
qc 106 m3/s
20.2
0.0027
38.34
0.2387
5000
20.2
0.0027
38.34
0.2865
6000
31.3
0.072
36.96
0.2300
5000
31.3
0.072
36.96
0.2760
6000
16.930
3.360
24.850
2.290
23.810
0.478
30.760
0.370
1.3110
1.3923
1.3110
1.4595
1.2131
1.3790
1.2131
1.4840
0.4367
0.0887
0.6700
0.0615
2.722
0.0546
3.510
0.0422
10.50
12.00
10.50
13.33
8.833
11.67
8.833
13.83
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
215
Table 2
The results of batch heat loss experiments to obtain heat transfer coefcient at different coolant ow rates.
Cooling water ow rate, qc 106 m3/s
Room temperature, C
1.333
1.667
3.667
5.167
7.167
9.167
10.500
12.083
16.667
80
80
80
80
80
80
80
80
80
34
34
34
34
34
34
34
34
34
0.5190
0.5743
11.73
13.68
15.85
17.71
18.83
20.05
23.05
Ua 2:290 10
qc
0:45
Table 3
List of dimensionless variables.
Items
Name
Vk0 expqE=RT
f
CAf
H
Cp T f
Dimensionless ow rate
p Vk0
Ua
expE=RT f
f
TT
Tf
X exp 1=
RTE f
216
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
manipulating the ow rate (q) and the coolant ow rate (qc), both the
manipulated variables q and qc should satisfy steady state mass and
energy balance equations for any feasible values of concentration (CA)
and absolute temperature (T), hence input multiplicity cannot occur
for any set of parameters.
The mathematical equations characterizing the dynamic behavior of
the non-isothermal CSTR are modeled using the unsteady state material
and energy balance and Arrhenius relation. The unsteady state material
and energy balance for a single CSTR are given as:
V
dCA
qCAf CA Vk CA
dt
Cp V
dT
qCp T f T UaTc T VH k CA :
dt
3. Reaction system
The acid catalyzed hydrolysis of acetic anhydride was chosen for
studying the dynamics of a CSTR reactor. The overall hydrolysis of acetic
anhydride reaction can be represented as:
H2 SO4 Catalyst
Acetic acid
The mechanisms of the hydrolysis of acetic anhydride are well documented. It is well known that the reaction proceeds via a substitution,
in which an attacking nucleophile replaces a substituent group on the
central carbon.
The reaction is of the rst order with respect to acetic anhydride. The
batch reaction kinetic equation for acid catalyzed hydrolysis of acetic
anhydride as given by [20] is:
10
rA 1:85 10 Cs exp11; 243:9=T CA
and
24.16
24.72
24.35
25.03
ETT f
CAf
e RTT f
1
CA
X
23.00
23.00
24.92
25.10
Steady state
predicted temp (C)
Expt. steady state
temperature (C)
Steady state predicted
temp (C)
Expt. steady state
temperature (C)
217
from Eqs. (6) and (7) and by introducing dimensionless variables as and
can be rewritten and shown in Table 3.
The mass balance in dimensionless can be written as:
1 ETTTT f
f
:
e
RTE
T f e
RTE
T f e
24.24
24.57
24.62
25.02
40.0
40.0
32.0
32.0
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
RTE
RTE
T e
T e
TC e
TC e
E
RT
T BT f e
T BT f e
E
RT
CA
0
CAf
10a
1
0
1 e
10b
RTE
f
T f e
T e
TC e
E
RT
BT e
0
T f
X
1
10c
23.00
23.00
24.92
25.10
RTE
f
BT eRT
0
T f
X
1
T f e
T e
TC e
T
T
BeRT
0
c 1
1
e X
Tf
BX
c 1
1
0
10d
10e
10f
BX
c
0
10g
BX
c
0
10h
BX
C
0
10i
BX
C
0
X
10j
10.50
12.00
10.50
13.33
0.4367
0.0937
0.6700
0.0615
5000
5000
6000
6000
q 106 m3/s
qc 106 m3/s
23.00
23.00
24.96
25.10
CAf mol/m3
Process variables
24.2
24.6
24.7
25.0
20.6
Steady state
predicted temp (C)
20.6
20.2
Feed temperature, Tf (C)
Table 4
Steady state temperature of CSTR under various process conditions (Tf = 20.2 C).
X
0:
X
11
A single steady state equation relating the dimensionless temperature T and the dimensionless ow rate is obtained by eliminating
the concentration term from the steady state mass and energy balance
equations for a perfectly mixed CSTR for a rst order exothermic reaction is as in Eq. (11).
218
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
Table 5
Steady state temperature of CSTR under various process conditions (Tf = 31.3 C).
Feed temperature, Tf (C)
31.3
32.0
Process variables
Expt. steady state temperature (C) Steady state predicted temp. Expt steady state temperature (C) Steady state predicted temp.
(C)
(C)
2.7210
0.0546
3.5160
0.0422
0.8333
11.6667
8.8333
13.8333
32.0
44.1
44.0
56.2
56.0
40.0
37.69
38.46
37.30
37.88
40.0
44.1
44.0
56.25
56.10
36.88
37.50
37.02
37.52
the experimental values of steady state temperature for the two feed
ow rates were also found to be in close agreement.
The experimental results showed that the effect of initial reactor
temperature was almost nil on the steady state temperature. For the
three different initial reactor temperatures, the value of steady state
temperature was almost the same. For feed temperature 20.2 C, the
experiments were carried out at three different initial reactor temperatures of 20.6 C, 32 C and 40 C. The steady state temperature for each
case was almost the same at a particular feed concentration as shown in
Tables 4 and 5. Similar results were obtained for the other feed temperatures. The time versus temperature prole of reaction for all the sets of
process variable combinations were also recorded. The transient time
versus temperature prole obtained for the experiments representing
different feed temperatures and feed concentrations are shown in
Figs. 5 and 6. For a particular process variable combination where
input multiplicity was observed, the time taken for reaching the steady
state temperature was much higher for the lower ow rate in comparison to the higher ow rate. For example, in case of Tf = 20.2 0C,
Ti = 40 C, CAf = 6000 mol/m3 (Fig. 5), the steady state temperature
for q = 0.0615 106 m3/s was obtained after 300 min whereas for
q = 0.67 106 m3/s, the steady state is reached before 50 min. The
same observation was obtained when the initial reactor temperature,
Ti = 32 C at feed temperature Tf = 24 C and feed concentration of
acetic anhydride CAf = 5000 mol/m3 conditions, as shown in Fig. 6.
The experimental time taken to steady state temperatures for feed temperatures, Tf, = 20.2 C, 24 C and 31.3 C are as shown in Tables 68.
It was observed that the time taken by the reactor was too long at a
low feed ow rate of reaction mixture than at a higher feed ow rate reaction mixture and is shown in Tables 68. It was found experimentally
that the reactor started with higher initial temperatures that reached
40
q = 0.670x 10-6m3/s
q = 0.062x 10-6m3/s
Temperature, T (C)
Temperature, T (C)
35
30
25
20
q = 0.889x 10-6m3/s
q = 0.072x 10-6m3/s
37
100
200
300
400
Time, t (min)
Fig. 5. Time versus temperature prole (Tf = 20.2 C, Ti = 40 C, CAf = 6000 mol/m3 and
q = 0.0615 106 m3/s and 0.670 106 m3/s).
32
27
22
50
100
150
200
250
300
Time, t (min)
Fig. 6. Time versus temperature prole (Tf = 24 C, Ti = 32 C, CAf = 5000 mol/m3 and
q = 0.0715 106 m3/s and 0.8880 106 m3/s).
Table 6
Experimental time taken to reach steady state reactor temperature (Tf = 20.2 C, CAf = 5000 mol/m3).
Feed temperature, Tf (C)
20.2
20.6
32.0
40.0
Process variables
80
166
62
380
34
112
22
126
20
86
18
98
CAf mol/m3
q 106 m3/s
qc 106 m3/s
5000
5000
6000
6000
0.4367
0.0937
0.6700
0.0615
10.50
12.00
10.50
13.33
Table 7
Experimental time taken to reach steady state reactor temperature (Tf = 24 C, CAf = 5000 mol/m3).
Feed temperature, Tf (C)
24
25
32.0
40.0
Process variables
51
294
42
372
20
82
14
200
18
130
10[p
136
CAf mol/m3
q 106 m3/s
qc 106 m3/s
5000
5000
6000
6000
0.8997
0.0715
1.21
0.0528
9.833
11.66
9.833
12.0
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
219
220
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
Table 8
Experimental time taken to reach steady state reactor temperature (Tf = 31.3 C, CAf = 5000 mol/m3).
Feed temperature, Tf (C)
31.3
32.0
40.0
30
348
12
412
12
100
10
112
Process variables
CAf mol/m3
q 106 m3/s
qc 106 m3/s
5000
5000
6000
6000
2.7210
0.0546
3.5160
0.0422
0.8333
11.6667
8.8333
13.8333
steady state temperature faster than that with the low initial reactor
temperature.
5.3. Parametric variation
CSTR parameters such as cooling water ow rate, feed ow rate of
reaction mixture, reactor feed temperature, and initial reactor temperature were varied in the region of input multiplicity by simulating the
dynamic equations for parameter sensitivity in CSTR.
The governing equations of the CSTR reactor system with i = qc as
parameter can be rewritten as:
dCA qCA0 CA
kCA F1
dt
V
12
qCp T f T
dT
HVkCA
UaTTc
F2
dt
VC w
VC w
VC w
p
13
d dT
F2 F2 F2 dCA dT dTc T
F2
dt dqc
CA T Tc dqc dqc dqc
qc
16
d dTc
F3 F3 F3 dCA dT dTc T
F3
dt dqc
CA T Tc dqc dqc dqc
qc
17
The above Eqs. (12)(17) were simulated and the results of simulations for parameter such as q, qc, Tf, Ti were shown in Fig. 7. The variation
of parameters such as q, qc and Ti within the range of 2% about the operating conditions do not show any signicant effect on both CSTR steady
state and transient reactor temperatures shown in Figs. 79. But however
the feed temperature seems to affect the transient reactor temperatures
and steady state temperature is greatly inuenced by a variation in
reactor feed temperature, Tf, as shown in Fig. 10.
6. Conclusions
dTc UaTTc qc
Tcin Tc F3
dt
Vc
Vc c Cpc
14
d dCA
F1 F1 F1 dCA dT dTc T
F1
dt dqc
CA T Tc dqc dqc dqc
qc
15
20.65
q= 0.088275e-6
20.6
q= 0.089167e-6
Temperature, T
20.55
q= 0.900583e-6
20.5
20.45
20.4
20.35
20.3
20.25
50
100
150
time, t
Fig. 7. Effect of ow rate of reaction mixture, q variation in the region of input multiplicity (CAf = 5000 mol/m3, Tf = 20.2 C, qc = 1.050e5 m3/s).
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
221
20.65
qc=1.0395e-5
20.6
qc=1.0500e-5
qc=1.6050e-5
Temperature, T
20.55
20.5
20.45
20.4
20.35
20.3
20.25
0
50
100
150
time, t
Fig. 8. Effect of cooling water ow rate, qc, variation in the region of input multiplicity (CAf = 5000 mol/m3, Tf = 20.2 C, q = 0.089167e6 m3/s).
where input multiplicity was observed, the time taken for reaching
the steady state temperature was much higher for the lower ow rate
in comparison to the higher ow rate. The simulation study on the
variation of parameters in the region of input multiplicity has shown
that the parameters such as feed ow rate of reaction mixture, cooling
water ow rate and initial temperature has no signicant effect affecting input multiplicity. The variation in reactor feed temperature, Tf, in
the range of 2% about operating feed temperature equal to 20.2 C has
shown to affect both t the steady state and transient temperature of
the reactor.
Nomenclature
a
area of heat exchange, m2
CP
specic heat capacity of the reaction mixture, J/(kg C)
E
activation energy, J/mol
feed concentration of acetic anhydride, mol/m3
CAf
CS
CA
(H)
k
k0
qA
qB
q
qC
R
rA
T
Tf
Ti
Tc
U
V
20.9
Ti=20.394
20.8
Ti=20.600
Ti=20.806
Temperature, T
20.7
20.6
20.5
20.4
20.3
50
100
150
time, t
Fig. 9. Effect of initial reactor temperature, Ti, variation in the region of input multiplicity (CAf = 5000 mol/m3, Tf = 20.2 C, q = 0.089167e6 m3/s, qc = 1.050e5 m3/s).
222
N.S. Jayakumar et al. / Chemometrics and Intelligent Laboratory Systems 135 (2014) 213222
20.65
Ti=19.998
Ti=20.200
Ti=20.402
20.6
Temperature, T
20.55
20.5
20.45
20.4
20.35
20.3
20.25
50
100
150
time, t
Fig. 10. Effect of feed temperature, Tf, variation in the region of input multiplicity (CAf = 5000 mol/m3, Ti = 20.6 C, q = 0.089167e6 m3/s, qc = 1.050e5 m3/s).
Greek notations
CAf
p Vk0
Ua
expE=RT f
f
dimensionless temperature, TT
Tf
Conict of interest
We don't have conict of interest.
Acknowledgment
This work was supported by the University of Malaya for fully
funding under HIR-MOHE (UM/MOHE HIR, Grant No. D000020-16001).
References
[1] M. Chidambaram, G.P. Reddy, Analysis of chemical reactors with input and output
multiplicities by Hammerstein systems, J. IETE 41 (5) (1995) 297.
[2] V. Balakotaiah, D. Luss, Input multiplicity in lumped parameter systems, Chem. Eng.
Commun. 39 (1985) 309.
[3] H.S. Chang, J.M. Calo, Exact criteria for uniqueness and multiplicity of an nth order
chemical reaction via catastrope theory approach, Chem. Eng. Sci. 34 (1979) 285.
[4] R. Aris, Chemical reactors and some bifurcation phenomena, Ann. N. Y. Acad. Sci. 1
(1979) 314.