Computerschem. EngngVol. 21, No. I, pp.

47--67, 1997
Copyright© 1996ElsevierScience Ltd
Pergamon Printed in Great Britain.All rights reserved
0098-1354(95)00254-5 0098-1354/96 $15.00+0.00

DESIGN AND CONTROL OF COUPLED REACTOR/COLUMN

SYSTEMS--PART 2. MORE COMPLEX COUPLED REACTOR/
COLUMN SYSTEMS

CHANG K. Yl* and WILLIAM L. LUYBENt

Chemical Process Modeling and Control Research Center and Department of Chemical Engineering
Lehigh University, Iacocca Hall, 1I I Research Drive, Bethlehem, Pennsylvania 18015 U.S.A.

(Received 8 November 1994; revised 25 August 1995;

Abstract-- This paper is the second in a series of papers that explore the challenging problems associated with
the design and control of coupled reactor/column systems. The steadystate designs and dynamics of three kinds
of coupled systems, a binary reactor/stripper system, a multicomponent reactor/rectifier system, and a
multicomponent reactor/rectifier/stripper system are studied in this paper. We concentrated on confirming that the
results from the simply binary reactor/rectifier system, considered in Part 1, can be applied to other
configurations and more complex systems. A design procedure and an optimum steadystate design for each
system are presented, and the effects of physical parameters on optimum design are described. Viable control
structures for various types of coupled systems are proposed. The control structure that worked well in a binary
system was found to require only minor modification to provide effective control of more complex systems.
Copyright © 1996 Elsevier Science Ltd

INTRODUCTION reactor, a stripper and two distillation columns as shown

in Fig. 1 M5.
Coupled reactor/column systems feature a reactor with a If the products boil intermediate to the two reactants,
distillation column sitting above the reactor (called a five units are needed: a reactor, a rectifier, a stripper and
coupled reactor/rectifier system) to remove low-boiling two distillation columns. The benefit of the M6 coupled
components or sitting below the reactor (called a system becomes questionable because one more unit is
coupled reactor/stripper system) to remove heavy com- required, compared with four units in a conventional
ponents. The advantages and the comparison with recycle system. The principle one derives from Fig. 1 is
reactive distillation were mentioned in part 1 (Yi and that one product should be the lightest or the heaviest in
Luyben, 1997). In the multicomponent reaction case (A coupled reactor/column systems. This is due to the fact
+ B ~ C + D), various configurations can be proposed. that the continuous separation of the lightest or heaviest
Fig. I shows different sequences of columns and product inhibits the backward reaction in a reversible
reactors in coupled reactor/column systems depending reaction.
on the volatilities of all components. Figure 1 also compares coupled systems with conven-
When the products are more volatile than the tional recycle systems. The stream from a reactor to a
reactants, a reactor/rectifier and one column are used as rectifier is vapor in the reactor/rectifier (M l) case, while
shown in Fig. 1 MI. If the reactants are more volatile the stream in the recycle system is usually liquid. The
than the products, a reactor/stripper and one column are difference may affect dynamic characteristics. Another
used as shown in Fig. 1 M2. If the reactants boil difference is that the column base becomes the reactor.
intermediate to the two products, a coupled reactor/ Therefore, the heat of reaction can be used as a heat
rectifier/stripper can be used as shown in Fig. I M3. If source in the reboiler in case of exothermic reactions.
the component volatilities are such that product C is the Part 1 studied a binary coupled reactor/rectifier
most volatile, reactant A is the next most volatile, system. In this paper, we will study three different kinds
product D has intermediate volatility and reactant B is of coupled systems: a binary reactor/stripper system, a
the heaviest component (C~c>C~A>ao>aa), the process multicomponent reactor/rectifier system, and a multi-
consists of a reactor with a rectifier to keep reactant B in component reactor/rectifier/stripper system. We concen-
the system and two distillation columns to achieve the trate on confirming that the results from the simple
removal of C and D and the recovery and recycle of binary reactor/rectifier system can be applied to other
reactant A (see Fig. 1 M4). When the volatilities are in configurations and more complex systems.
the order of ac>trA>ao>t~B, the process consists of a
BINARY COUPLED REACTOR/STRIPPER SYSTEM (B2)
* Current address: Korea Institute of Energy Research, 71-2
Jangdong, Yusung, Taejon 305-343, Korea. In the binary system with reaction A -~t B and heavy
* To whom correspondence should be addressed. product, one stripper and one reactor are used, Figure 2
47
 48 CHANGK. Yt and W. L. LUYBEN

M I : ( ~ c& oh) > (ok& ~.n) M2:( a A& ae) >(c~c & o~)) M3: o~> ( ~ & o~) > c~
~ C ~ C C

B
D D A+B

D D D

M4: 0h= > o~A > ~ D > Ot] M5 o~ >Oc >~B>c ~ M6: oq~ > ( ~c& o~) > ~

B ff"
D D
Fig. I. Variousconfigurationsof coupled reactor/columnsystems in multicomponentcase.

shows the comparison of alternative system configura- k,

fresh feed compositions zo, the specific reaction rates
tions for the binary system with reaction A ~ B and andk2,therelativevolatilityaSA,andtheproductpurity
heavy product. Figure 2 (a) is the general recycle system xo aregiven.Theoptimizationparameters ofthissystem
(Papadourakis et aL, 1987). Figure 2 (b) is the reactor/ are reactor holdup VRand numberof trays Nv The
stripper system with recycle (Luyben, 1993). Figure 2 followingdesignprocedurewasused:
(c) is the coupled reactor/stripper system, which is the SelectVRandN v
subject of this paper. For endothermic reactions, the load Calculate reactorconcentration z fromthe overallB
on the condenser can be smaller because some of the componentmaterialbalanceintheentiresystem.
heat of reaction can be provided by the heat of
condensation by, for example, circulating the liquid in
F,~. - BXB+ Vsk,
the reactor to the condenser by a pump (using a spray z= (l)
VR(k,+k9
condenser). Both the steady-state design and the devel-
opment of a control structure are discussed below.
Guess liquid flow rate in the stripper L = R.
Steady-state design for B2 case Calculate x, from a component material balance
Figure 3 shows the schematic diagram of a coupled around the reboiler and the equilibrium relationship.
reactor/stripper process in a binary system. The equa- Step up the column to get x,,÷~ from a component
tions are almost the same as those in the reactor/rectifier material balance at the n'h stage.
system, but the reactor equations and the overall balance I f y ~ is not equal to Xo, reguess L by interval halving
must be modified. Note that O/BAis less than unity since method and return to step 3.
component A is more volatile than component B. Vary VR and Nr until the minimum cost plant is
Design procedure of B2. Fresh feed flow rates Fo, the found.
 Coupled reactor/column systems--2 49
O ~ a l Recycle S~em (Papadoar~tis ¢t aL; t987)

'

ReaetorBtripp~withRecycle(Laybm;1993)

Fig. 2. Comparison of alternative system configurations for the binary reactor/stripper system (B2).

For the calculation of column and reactor sizes and for compared with other cases in order to see how design
economic analysis of the annual capital cost, the same variables affect the coupled reactor/stripper system. The
procedure used in Part 1 was employed. underlined numbers are the compared variables.
Optimum steady-state design. Table 1 gives results of The design variables of reactor holdup and number of
the steady-state design calculations for the B2 case in the trays have the same effects as Case B 1. The effect of
reactor/stripper system with F,, = 100 lb-mole/hr, zo = 0 reactor holdup can be compared by looking at Cases B2-
mole fraction B, kt = 0.35 hr -t, k., = 0.1 hr -~, B = 100 1, B2-2, and B2-3. As reactor holdup increases, the
Ib-mole/hr, asA = 0.5, A = 10000 Btu/Ib-mole (an concentration of product Xo in the reactor increases, and
endothermic reaction), and wh = 1/12 ft. the ratio of V/F,, decreases.
Case No B2-2 was chosen as the base case and was Comparing Case B2-2 with Cases B2-4 and B2-5, we
50 CHANGK. YI and W. L. LUYBEN

(b) A ~ B (heavy)

~ Qc

® Fo 7.0

Sm )per

Fig. 3. Schematic diagram of coupled reactor/stripper system in binary case (B2).

Table 1. Steady-state design in the reactor/strippersystem (B2)

Case B2-1 B2-2 B2-3 B2-4 B2-5 B2-6 B2-7

x,4m.f, of B) 0.99 0.99 0.99 0.99 0.99 0.95 0.99

V~ (Ib-mole) 1200 ~ ~ 1600 1600 1600 1600
Nr 1_88 18 18 L4 22 18 18
OlnA 0.5 0.j 0.5 0.5 0.5 0.5 0.333
.ro (m.f. of B) 0.594 0.640 0.668 0.640 0.640 0.646 0.640
VIF,, 2.372 2.138 2.013 2.303 2.087 1.808 1.306
D c (ft) 2.58 2.45 2.37 2.54 2.42 2.25 1.9 I
Dk(ft ) 9.14 10.06 10.84 10.06 10.06 10.06 10.06
QR ( 10~ Btu/h) 2.97 2.67 2.52 2.88 2.61 2.26 1.63
M, (Ib-moles) 0.73 0.66 0.62 0.71 0.64 0.55 0.39
MB (Ib-moles) 28.10 26.15 25.11 27.52 25.72 23.40 19.22
B (sec) 2.12 2.03 1.97 2.09 2.00 1.88 1.62
Heat exchanger areas Ift2)
reboiler 593 535 503 576 522 452 327
condenser 658 561 509 629 539 437 214
Capital costs ($1000)
cost reactor 247.5 296. I 340.2 296. I 296.1 296.1 296. I
cost column 128.0 121.1 117.2 103.0 140.4 110.7 93.1
cost trays 8. I 7.4 7, I 6.1 8.9 6.5 5. I
cost exchangers 242.9 222.8 211.7 237.1 218.3 194.5 140.1
Energy cost ($1000) 129.9 117.1 110.2 126.1 114.2 98.9 71.5
Total annual cost ($1000) 338.7 332.9 335.6 340. I 335.5 301.6 249.6

can see the effect of the number of trays. As the number volatility is shown by comparing Cases B2-2 and B2-7.
of trays increases, the V/F, ratio decreases. As relative volatility is reduced farther away from 1, the
The minimum cost condition for the B2 case is at V~ separation ability increases. The higher separation
= 1600, Nr = 18, V/F, = 2.138, which is Case B2-2 in ability makes the V/F, ratio decrease.
Table I. Table 2 provides parameter values and optimum
Effect of physical parameters on optimum design. We steady-state values of variables for the coupled reactor/
can see the effect of product impurity by comparing stripper system.
Cases B2-2 and B2-6 in Table 1. A lower V/F, ratio is Figure 4 shows how the physical parameters, which
needed with lower product purity. The effect of relative are relative volatility, reaction equilibrium constant, and
 Coupled reactor/column systems--2 51
Table 2. Parametervaluesand optimumsteady-statevaluesof variablesof the coupled
reactor/strippersystem(B2)
F,, 100.0 Ib-mole/hr z,, 0.0 mole fraction of B
R 313.8 Ib-mole/hr V 213.8 Ib-mole/hr
B 100.0 Ib-mole/hr N7 18 stages
Q~ 1.6854 x 106 Btu/hr V~ 1600.0 Ib-mole
M,, 0.668 Ib-mole MB 26.15 Ib-mole
.rn 0.99 mole fraction of B xo 0.640 mole fraction of B
TR 140 °F Tm 140.35 °F
I] 0.5697 x 10J hr
ct~ 0.5 A 10000.0 Btu/Ibmole
klo 0.2968 x 10 H hr L k2o 0.1930 x 107 hr ~
E~ 30000.0 Btu/lb-mole E, 20000.0 Btu/Ib-mole
blA -4068.8 b,~ 11.079
bin -4068.8 b,_n 10.386
M 50,0 Ib/lb-mole H,.,,p 250.0 Btu/Ib
C~, 0.75 Btu/Ib °F PI~ 50.0 psia
T~, 140.0 °F Tn 170.0 °F

distillate product composition, affect the optimum A composition measurement deadtime of 3 min was
points. included in the bottom composition control loop. The
Figure 4 (a) shows the effect of relative volatility on dynamic relationship between pressure in the reactor and
the optimum total annual cost. Optimum cost increases condenser heat removal was assumed to be represented
as relative volatility comes close to 1. The closer the by two I min first-order lags,
volatility of the product to the volatility of the reactant, Ziegler-Nichols tuning parameters in the non-linear
the more difficult the separation, making energy costs model were used. Figure 6 shows closed-loop responses
more significant. Figure 4 (b) shows that the total cost at of the system. The load disturbance (the solid line in Fig.
the optimum point decreases as the reaction equilibrium 6) is an increase in feed composition from z, = 0 to 0.1.
constant increases because optimum reactor holdup The structure provides good composition control. The
decreases due to the higher equilibrium conversion. For temperature in the reactor has a self-regulating charac-
the same value of the equilibrium constant, the optimum teristic, changing slowly to a new steady-state condition.
cost at k2 = 0.2 is smaller than at k2 = 0.1. Since the The dotted line in Fig. 6 shows closed-loop responses
difference between the forward and reverse reactions when the heat of reaction A is changed from 10,000 to
determines the net rate of reaction, the larger value of k2 9500 Btu/lb-mole and the forward activation energy E~ is
corresponds to a larger net rate of reaction, and this also changed from 30,000 to 29,500 Btu/Ib-mole. The
requires a smaller reactor. Figure 4 (c) shows the effect temperature and product composition show stable
of product impurity on the optimum point. As product dynamic responses.
purity is increased, the cost increases. We need to
increase the energy used in the column in order to purify
the product. M U L T I C O M P O N E N T R E A C T O R / R E C T I F I E R SYSTEM (MI)

Dynamics o f B2 The multicomponent case (M1 case) features products

Non-linear dynamic simulations were used to test that are both more volatile than reactants. The reversible
closed-loop performance. In Part I the best control reaction is A + B a= C + D. We use the notation that
structure for the reactor/rectifier system was found to be component A is j = I, component B is j = 2, component
(I) heat input to the reactor (QD controls the level in the C is j = 3 and component D is j = 4.
reactor, (2) distillate flow (D) controls the level in the Steady-state design o f M1
reflux drum, (3) reflux flow (R) controls the distillate
Design procedure o f M I. The optimization parame-
composition, and (4) fresh feed flow rate (F,,) is flow-
ters are reactor volume VR and number of stages NT, The
controlled. Reactor temperature was not controlled.
following design procedure was used given values of
Figure 5 shows a control structure that is the inverse of
fresh feed flow rate F,,, feed composition z,0, relative
the structure found to be the best for the reactor/rectifier
volatilities aj, and product purity XD~.
system. The structure uses reflux to control level in the
Select VRand N v
reactor, bottom flow to control bottom column level,
The following relations were obtained from overall
reboiler heat to control bottom product purity, and
component material balances.
condenser heat removal to control pressure. Reactor
temperature is not controlled. Note that circulation of Foz,,j - Dxoj + ~ VR(klzlz,_ - k2z3z4) = 0 (2)
reactor liquid through the condenser is used so that the
heat of condensation can provide some of the endo-
thermic heat of reaction. z,. -xD, = z,,2 -xo2=xo3 - z,,3= xo4 - z,~ (3)
52 CHANGK. Yt and W. L. LUYBEN
,~0

§4st

I .~
o73 o:, o'.~ o'6 0.7
Rebd/ve VolUllty

1100

1000

900

~ 7~
i

3OO

20~ I I I I
2 4 6 8 10
Roaotion Equit C o n ~ l , kl/k2

360 . . . . . . . . , . . . . . . . .

34O

~,o

26O

240 . . . . . . . . . . . . . . . . .
10 10.2 10"I
ProducX Irnpu~, I-XB

Fig. 4. Effects of physical parameters on the optimum cost (B2).

where vj is negative when j is 1 (A component) or 2 (B From component material balances at the n 'h stage,
component), and is positive when j is 3 (C component) y,,_ ~dat one lower stage is calculated.
or 4 (D component). If the impurity of one reactant is Repeat step 6 and 7 moving down the column.
fixed in the condenser, xoj can be calculated from If the zj~ do not satisfy the overall component
Equation 3 and the stoichiometry relationship from the balances, return to step 4 and reguess reflux ratio by
equimolar reaction. interval halving method. Note that the composition in
Given xoj, a rating problem is solved. Guess reflux the base of the column xaj is also the composition in the
ratio. reactor zj.
Start from top tray with component balance equations Vary VRand Nr to minimize total annual cost.
in the condenser. Optimum steadystate design. Table 3 gives results of
x,,~ is obtained from equilibrium relationships. the steady-state design calculations for the M I system
 Coupled reactor/column systems--2 53

Fo

Q ..... ~ l ~ ~ Q R
i
.......... ®.,

Fig. 5. Modified control structure schematic diagram for a coupled reactor/stripper system (B2).

Load disturbances solid: zo=0.1, dotted: =9500

1 ~.~_._.._L. , , ......

i ............... ............... i .............................

. . . . . . . .

~0.94 "'~"".":::::'......... i .................................................................................

":' i
0.92
1 4 5 9
Time, hr

155 ! !

150 ......... ;......... ~....... i ::::::::. . . . . . . . . . i:" ":-":i"": ' " ':: ' ::~::::::::"'::"::

~- iii: ..:::: i ............................. i....................................... i .........

140 ~; -- ~ ............... :- . . . . . . . . . . . . . . . . . . . . :. . . . . . . . . '~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

i : i
135 I I
1 6 9
Tim( hr
Fig. 6. Closed-loop responses to load disturbances z~ and A (B2).

with the conditions F,, = 100, zol = z02 = 0.5, Zo3 = z~ increases, the reflux ratio decreases. This is because as
= 0, k~ = 0.35, k2 = 0.1, D = 100. Series No. 2 was the concentration of products in the reactor increases, the
chosen as a base case and was compared with other cases separator requirements decrease to get the desired
in order to see how design variables affect the coupled product purity.
reactor/column system. The underlined numbers are the Comparing Case No. 4 with No. 2, we can see the
compared variables. effect of the number of trays. The reflux ratio increases
The effect of reactor size can be seen by a comparison as the number of trays decreases.
of Case Nos 1-3. As reactor size increases, the Figure 7 shows the tray liquid composition profiles of
concentrations of products in the reactor xa3 and x,4 the four components in the multicomponent reactor/
increase and the concentration of reactants xB~ and xa2 rectifier system for the optimum design conditions: aj =
decrease. This is because the concentration of reactants 0.8/1.0/1.875/1.5, xDj = 0.00510.00510.49510.495, VR =
in the reactor can be lower and still produce the same 7000, N r = 11 and RR = 4.2.
distillate products since VR is larger. As the reactor size Figure 8 shows the results of the steady state design
54 CHANG K. YI and W. L. LUYBEN

Table 3. Results of steady-state design in the M I case

Series No. No. I No.2 No.3 No.4 No.5 No.6 No.7 No.8
aA 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8
~A 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
ac 1.875 1.875 1.875 1.875 2.5 2.5 1.875 1.875
1.5 1.5 1.5 1.5 2.0 2.0 1.5 1.5
xo, 0.005 0.005 0.005 0.005 0.005 0.005 0.02~5 0.02,5
.r~), 0.005 0.005 0.005 0.005 0.005 0.005 0.025 0.025
,r,3 0.495 0.495 0.495 0.495 0.495 0.495 0.475 0.475
xDa 0.495 0.495 0.495 0.495 0.495 0.495 0.475 0.47;5
VR 4000 ~ ! 2000 7000 4000 7000 7000 7000
xn~ 0.7017 0.5747 0.5092 0.4275 0.5666 0.4787 0.4561 0.3483
xn, 0.0555 0.0507 0.0478 0.0752 0.0776 0.0711 0.0754 0.1093
xl~3 0.0743 0.1240 0.1503 0.1072 0.1389 0.1771 0.1804 0.1804
xBa 0.1685 0.2506 0.2926 0.3900 0.2169 0.2732 0.2881 0.362 I
RR 6.895 4.200 3.485 14.29 2.114 1.635 2.087 3.787
,v, I1 1_! 11 7 9 II 11 5
Dc (ft) 5.16 4.19 3.89 7.19 3.24 2.98 3.23 4.02
DR fit) 13.66 16.46 19.69 16.46 13.66 16.46 16.46 16.46
QR (106 Btu/h) 9.87 6.50 5.61 1.91 3.89 3.29 3.86 5.98
M o (Ib-moles) 65.79 43.33 37.37 127.05 25.95 21.96 25.73 39.89
M,, (Ib-moles) 1.77 1.17 1.01 3.42 0.98 0.81 0.97 1.64
13(sec) 8.03 6.08 5.50 12.47 3.36 3.28 3.36 3.69
Heat exchanger areas
reboiler (ft-') 1974 1300 1121 3823 779 659 772 1197
condenser (It-') 3290 2167 1869 6372 1298 1098 1286 1995
Capital costs (1000 $)
cost reactor 523.8 742.1 1038.1 742.1 523.8 742.1 742.1 742.1
cost column 180.9 144.7 133.7 179.0 93.7 100.6 109.6 73.6
cost trays 14.5 10.5 9.3 15.4 5.8 6.2 7.0 4.5
cost exchangers 485.0 369.7 335.8 745.5 264.9 237.7 263.4 350.4
Energy cost (1000 $) 432.3 284.7 245.5 837.3 170.5 144.3 169.0 262.1
Total annual cost (1000 $) 833.6 707.1 751.2 1398.0 466.6 506.3 543.1 652.3

0.9 ....... i ...................................................................................................

0.8 ....... i ...................................................................................................

ii:ii::ii:iiiii:iii:iiiiiiii:iii:iiiiiiiiiiiiiiiiiii::iii:ii
0,
0.6
i ........................ iiii:i:iiiiiii::i
-~ 0.5
~ 0.4 ..... i ...................

:..~"0.3
0.2
0.1-

zo xB 2 3 4 5 6 7 8 9 1'0 1'1 xD
Tray

I -'~" A ''t'- B "-I- C -B- D I

Fig. 7. Composition profiles in a multicomponent reactor/rectifier system ( M l).

2{: ~ -m-.
Nt-8
.-4.-.
| N~O
-IK-.
..1,

~ ...........
.=-...- ................................................................

Vi"~ "i=

.................... "' "~" FZ~'-"*'-'~=:::=::::;~ .............

~ ' ~ - - = _ ....... g-'-J:TZ"
-I
0 ~ i 5 6 7 8 9 1'0 1'1 1'2 1'3 14
Volume, VR

Fig. 8. Effect o f reactor volume on steady-state design in M I .

 Coupled reactor/column systems--2 55

1200
o.°

1100
vR=14oo6
I*

1000
,,.8
.•.'• v~-~doo
8
L)
"d 900

800

700

600 I I ! 1 I I

3 4 5 6 7 8 9 10

Reflux Ratio
Fig. 9. Effect of reflux ratio on total annual cost in M I.

calculations in the multicomponent reactor/rectifier sys- the reactor capital cost increases, the column capital cost
tem with the conditions XDt = XD., = 0.005, aj = decreases, and the energy cost decreases. The point for
0.8/1.0/1.875/1.5. The curves illustrate the relationship minimum cost is at Ve = 7000, Air = 11, and RR = 4. I.
among reflux ratio, reactor volume, and number of trays. The minimum cost point for each reactor volume is near
In order to achieve a desired separation, the reflux ratio the minimum reflux ratio.
decreases with an increase in reactor size. This is Effect of physical parameters on design. A compar-
because the product concentrations in the reactor ison of Case No. 2 and No. 6 in Table 3 shows the effect
increase as reactor size is increased, and the needed of relative volatility for the same size reactor. The more
reflux ratio decreases for the desired purity. We can see volatile the products, the easier the separation. The effect
here the interaction between reaction and separation. of relative volatility on the optimum cost is shown in
The maximum separation performance is required at the Cases No. 2 and No. 5 in Table 3. Case No. 5 is an new
minimum reactor volume condition. optimum design for the high relative volatility case. The
The minimum reactor volume. The minimum reflux more volatile the products, the easier the separation,
ratio and the minimum number of trays are of great making energy costs less significant. Note that the
practical importance in a conventional distillation col- optimum reactor volume is smaller.
umn. The coupled reactor/column system has a mini- Compared with Case No. 7, Case No. 2 with higher
mum reactor volume as an additional limiting case. purity has higher total annual cost due to higher column
There is a minimum value for reactor volume below capital cost and energy cost because reflux ratio is
which the process is not feasible for a given product increased as product purity is increased.
purity. For example, Ve,,i,, is about 2500 in Fig. 8. To
obtain the desired product purity, the reactor holdup Dynamics and control of Ml
must be greater than the minimum value. The reaction The base case for dynamic study was chosen as the
zone in a conventional the reactive distillation column is optimum steady-state design. Table 4 shows parameter
limited to the liquid holdup on the trays of column. By values and optimum steady-state values of variables of
using a large reactor that is greater than the minimum the coupled reactor/rectifier system for the multi-
reactor size, we can obtain almost pure products. component case.
Economic optimum analysis of MI. Figure 9 shows The best control structure for the binary process can
the economic optimization results from cost analysis be directly applied with only minor modification. Figure
with conditions xo~ = xo2 = 0.005, oti = 10 shows a control structure for the multicomponent
0.8/I.0/1.875/1.5 when Nr = I1. There are tradeoffs case of the reactor/rectifier system. Distillate flow rate D
among the reactor capital cost, the column capital cost controls the level in the reflux drum, and reboiler heat Qe
and operating energy cost. As the reactor size increases, controls the level in the reactor. Reflux rate R is used to
g.AC~ 2|-I-C
56 CHANGK. YI and W. L. LUYBEN
Table4. Parametervaluesand optimumsteady-state valuesof variablesof the coupledreactor/rectifiersystem (MI)
F,,4 50.0 Ib-mole/hr z,,j of F,,4 1.010.0/0.0/0.0 mole fraction
F,,n 50.0 lb-mole/hr z,,~of F,,B 0.0/1.0/0.0/0.0 mole fraction
R 389.3 lb-mole/hr V 489.3 Ib-mole/hr
D 100.0 Ib-mole/hr N7 12 stages
QR 5.1649 x 106 Btu/hr VR 7500.0 Ib-mole
M,, 1.423 lb-mole Mo 40.78 Ib-mole
xt~J 0.00510.005/0.49510.495 mole fraction x,~j 0.603/0.045/0.122/0.230 mole fraction
TR 140 °F T~ 137.37 °F
13 0.9069 x 10 ~ hr
o~j 0.8/1.0/1.875/ 1,5 A -20000.0 Btu/Ib-mole
k,) 0.1538 x 104 hr l k a) 0.8480 x 10 )° hf I
E~ I 0(X)0.0 Btu/Ib-mole E., 30000.0 Btu/lb-mole
b)~ -5006.8 b~ II ,954
b,R -5006.8 b,.n 12.177
bj c -5006.8 b2c 12,806
blo -5006.8 b,.o 12,583
M 50.0 lb/lb-mole H,,p 250.0 Btu/Ib
Ce 0.75 Btu/lb °F PD 50.0 psia
Te,, 140.0 °F To 110.0 OF
KMT 52278 lb/psia/hr

XD 1+XD2

FOA i - - ~)- -

Fig. 10. A modified control structure for multicomponent case of reactor/rectifier system (M 1).

control the impurity level, which is the sum of the Figure I1 (b) shows the load responses to feed
impurities in the product (xD~ + xo,_). The only change composition changes when zoAj is changed from 1.0/0/0/0
from the binary case is the control of one composition in to 0.9/0.1/0./0. Composition control is good, even
the reactor by manipulation of the flow rate of one feed though other flow rates and temperatures change very
component. This loop is used in order to precisely satisfy slowly. It takes over 50 hours to reach a new steadys-
the reaction stoichiometry. tate.
Figure ! 1 shows closed-loop responses to various load
disturbances. Figure I1 (a) shows load responses when
the feed flow rate of A (F,,A) is decreased 10% from 50 M U L T I C O M P O N E N T R E A C T O R / R E C T I F I E R / S T R I P P E R SYS.
to 45 lb-moles/hr. The product impurity is very well T E M (M3)
controlled, going to the steadystate within 1 hr. How-
ever, the flow rate of F,,n keeps changing to satisfy the In this section we consider a multicomponent reactor/
reaction stoichiometry for more than 50 hrs. Comparing rectifier/stripper system in which both reactants boil
reactor holdup of 7500 lb-moles with feed flow rate of intermediate between light and heavy products. The
100 Ib-moles/hr, we would expect this large time reactor/rectifier/stripper system shows unique tradeoffs
constant. among VR, NR, and N s. Effects of parameters on optimum
 Coupled reactor/column systems--2 57

o (a) Load disturbance: FoA=45

O

~'0.98 ................. : ......... : ......... ; ............................. : ................... : .........

X II i : : ! i i : i :
~0.96 . . . . . . . : ......... "-......... : ........ "~........ "......... : ......... "......... : ......... : .........
X 5 10 15 20 25 30 35 40 45
Time, hr

U. 1 4 0 ~ ..... !!' !!i i ! !

5 10 15 20 25 30 35 40 45
Time, hr

E ' ' " :

........ ......... i ......... i ......... i . . . . . . . . .

oi ' : i ::

5 10 15 20 25 30 35 40 45
Time, hr

o (b) Load disturbance: zoj of A =0.9/0.IIOJ0.

O

•..,., I .....
--:
[* , . ~ -t
:
. . . . . . . .
:
L . . . . . . . .
:
t . . . . . . . .
:
[ . . . . . . . J
: . . . . . .
:
1 . . . . . . . .
,:,
] .....
:
I ....
]

~ ~ i ! i i i i i i
~0.gs ...... i ......... ::.......... i.......... ::.......... i ......... i ......... i ......... i ........
~" I J :l i i i i i i i i J
~_. 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time, hr

0.5 1
i i i i ! i 1.5 2 2.5 3 3.5 4 4.5
,
Time, hr

0.5 1.5 2.5 3 3.5 4 4.5

Time, hr
Fig. 1I. Closed-loop responses to various load disturbances for MI case.

design are explored. The dynamics of the system were and D is the heaviest component. The equations
approximated by simple integrator/dead time transfer describing the system are given in the Appendix.
function models for the purpose of controller tuning. An
effective control structure was developed by drawing on
the experience gained in the study of more simple Steady-state design of M3
coupled reactor/column systems. Design procedure for M3. The optimum steadystate
Figure 12 shows the schematic diagram of a coupled design is found by performing a three-parameter search
reactor/rectifier/stripper process for a multicomponent on the number of rectifier stages N R, the number of
system. A reversible, second order reaction A + B = t C stripper stages Ns, and the reactor holdup VR.
+ D occurs only in the liquid phase in a continuous Guess values of R and B.
stirred tank reactor in which C is the lightest component Using the tridiagonal matrix method by Wang and
 58 CHANGK. Ymand W. L. LUYBEN

%
I-

T R D XDj

Vre LR

FoA ZoAj
Q VR
Fos ZoBj NF=NS+I

i •

NS Vst Ls

I -• QR

I ~B XBj
Fig. 12. Schematic diagram of a coupled reactor/rectifier/stripper process for a multicomponent system (M3).

Henke (1966) in which initial guesses of compositions

F(~,)
must be made, solve for every composition on all trays. ~÷,=Sk--- (7)
F'(80
Use a modified 0 method (Holland, 1975) to normal-
ize each of the calculated compositions. We defined 0 as
Check if these compositions satisfy the overall j
below, instead of the ratio of the new value to the old
component balance equations. If not, go to step 2.
value of the bottom product to the distillate composi-
Change R by the interval halving method if the
tions, because the overall component balances are
desired XD3 is not achieved, and repeat steps 1 through
strongly affected by the reaction terms composed of
4.
compositions in the reactor.
Change B by the interval halving method if the
desired xs4 is not achieved, and repeat steps I through
' ,4, 5.
zj/ - \zj/
Effect of optimization variables VR, NR, and N, on
design. Table 5 gives results of the steady-state design
The zj are calculated from overall component balance
calculations in the reactor/column system with the
equations. For example, j = I:
simulation conditions F,,A = 50 lb-mole/hr of pure A, F,,8
= 50 Ib-mole/hr of pure B, kl = 0.35, k2 = 0.1, aj =
~,.,.=F, Az,,AI + F,,aZoaI + VRk2z3z~ - Dxol (5)
/ \ ,,z,t 2121411, A = 0 Btu/lb-mole (no heat of reaction).
Case No. 1 is the minimum cost optimum steady-state
design. It was compared with other cases to see the
effect of design variables. The effect of reactor holdup is
To find the new O, use the Newton method.
shown with Case Nos l, 2 and 3 at constant number of
NC
F(0) = =Xo~"- 1 (6) trays. As reactor holdup is increased, the concentration
of products in the reactor increase, the concentration of
 Coupled reactor/column systems--2 59
Table5. Steady-statedesignin the reactor/rectifier/strippersystem
Case No. I No.2 No.3 No.4 No.5 No.6 No.7 No.8

.rD~(m.f,) 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.9.5

.rB4 (m.f.) 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.95
Vk (Ib-mole) ~ 2000 ~ 3000 3000 3000 4500 4250
N,~ 15 15 15 10 20 10 14 11
N~ 15 15 15 10 200 15 49 34
N7 31 31 31 21 41 26 64 46
or~ 2 2 2 2 2 2 1.25 1.25
a_~ 2 2 2 2 2 2 1.25 1,25
o~ 4 4 4 4 4 4 2.5 2.5
oq 1 I I I 1 I 1 1
z~ 0.2518 0.2872 0.2211 0.2521 0.2517 0.2511 0.2131 0.2153
z: 0.2535 0.2885 0.2238 0.2541 0.2534 0.2567 0.2135 0.2153
z~ 0.1946 0.1620 0.2247 0.2039 0.1926 0.2420 0.1053 0.1091
z4 0.3001 0.2623 0.3304 0.2900 0.3023 0.2503 0.4681 0.4603
D 50.00 50.00 50.00 50.00 50.00 50.00 50.01 50.10
B 50.00 50.00 50.00 50.00 50.00 50.00 49.99 49.90
V 258.8 309.3 225.8 354.9 250.4 312.9 458.7 426.5
R 208.8 259.3 175.8 304.9 200.4 262.9 408.7 376.4
D c fit) 2.69 2.95 2.52 3.16 2.65 2.97 3.57 3.45
D~ (ft) 12.41 10.84 14.71 12.41 12.41 12.41 14.20 13.93
QR (10' Btu/h) 3.23 3.87 2.82 4.44 3.13 3.91 5.73 5.33
M,, (Ib-moles) 0.70 0.87 0.60 1.02 0.68 0.88 1.35 1.25
MI~(Ib-moles) 21.56 25.77 18.82 29.58 20.86 26.08 38.22 35.54
MIj (lb-moles) 21.56 25.77 18.82 29.58 20.86 26.08 38.22 35.54
13( 10~ hr) 0.7328 0.7678 0.7087 0.7993 0.7251 0.7711 0.8515 0.8372
Heat exchanger areas (ffl)
reboiler 647 773 564 887 626 782 1147 1066
condenser 1078 1289 941 1479 1043 1304 1911 1777
Capital costs ($ 1000)
cost reactor 437.9 340.2 601.7 437.9 437.9 437.9 563.6 543.9
cost column 207.5 228.2 192.9 180.0 254.6 199.6 500.3 370.5
cost trays 14.9 17.1 13.4 12.9 19.1 14.5 47.5 32.4
cost exchangers 297.2 333.8 272.0 365.0 290.9 336.3 431.2 411.3
Energy cost ($1000) 141.7 169.3 123.6 194.3 137.1 171.3 251.1 233.5
Total annual cost ($1000) 460.8 475.7 483.7 526.3 471.2 500.8 765.4 686.2

reactants decrease, and reflux rate decreases because the coupled reactor/column system. The continuous removal
separation required in the distillation column is of products from the reactor mixture prevents the
decreased. The effect of number of trays can be seen by undesirable reverse reaction. Conversion is not limited
comparing Cases No. 1, No. 4, No. 5, and No. 6. As the by K~ in the coupled reactor/column system.
number of trays is decreased, the concentrations in the Table 7 shows the effects of the forward reaction rate
reactor are little affected, but the reflux rate increases. k, on the optimum steady-state design condition for
These results show that a tradeoff relationship exists several values of k,. As the forward reaction rate k~ is
among reflux rate, reactor holdup, and number of trays in increased, the optimum size of reactor decreases, the
the rectifier and stripper in the coupled reactor/rectifier/ optimum number of trays is the same. and the optimum
stripper. Figure 13 shows that the optimum point for the reflux ratio decreases. Larger kt's decrease the load on
minimum total annual cost is at N s = 15, NR = 15, and system.
VR = 3000. The reason why N s is equal to NR is that the When both k, and k2 are increased, the optimum VR is
relative volatility between the lightest component and decreased by the effect of k, and reflux ratio is
the intermediate components is equal to the relative decreased a little bit by the net effects of both k~ and k2.
volatility between the intermediate components and the
heaviest component. When the ratios are different, the Effect of volatilities on optimum design
optimum N s is different from the optimum NR. Table 8 shows the effects of relative volatilities on the
Effect of physical parameters on optimum design optimum steady-state design with the design conditions
Effect of reaction rates kl and k2 of xo3 = 0.99 and xB4 = 0.99, when k~ = 0.35 and k., =
Table 6 shows the effects of the backward reaction 0.1.
rate k2 on the optimum steady-state design with the The optimum number of trays in the rectifier increases
design conditions OfXD3 = 0.99 and xB4 = 0.99, when the as the relative volatility between the lightest component
a~ are 2/2/4/I. As the backward reaction rate is increased, (j = 3) and the intermediate components (j = 1, 2)
the optimum reactor holdup and optimum numbers of decreases. The optimum number of trays in the stripper
trays are little changed, but more reflux is needed. Even increases as the relative volatility between the heaviest
though the backward reaction rate is big, pure products component (j = 4) and the intermediate components (j =
are obtained. This illustrates the advantage of the 1, 2) decreases. It is expected intuitively that the required
60 CHAN~ K. Yt and W. L. LUVBE~

b.,

5000
400O
20 19 • , i 3OOO
lg 17 16 15 14 13 12 11 21100

VR
NR=NS

Fig. 13. Effects o f optimization parameters on m i n i m u m cost design (M3).

Table 6. Effect of reaction rate k., on optimum steady-state design (M3)

k~ k2 K,,~ VR N~ N~ RR TAC

0.35 0.1 3.5 3000 15 15 4.175 460.82

0.35 0.2 1.75 2750 15 15 4.901 487.38
0.35 0.35 1.0 2750 15 15 5.583 518.93
0.35 0.5 0.7 2750 15 15 6.156 545.05
0.35 0.7 0.5 2750 15 15 6.817 574.77
0.35 0.9 0.39 2750 15 15 7.398 600.63
0.35 1.0 0.35 2750 15 15 7.667 612.50

Table 7. Effect of reaction rates k~ and k2 on optimum steady-state

design (M3)

k, k, K,.q VR N,~ N~ RR TAC

0.35 0.1 3.5 3000 15 15 4.175 460.82
0.70 0.1 7.0 2000 15 15 3.387 390.14
0.70 0.2 3.5 2000 15 15 3.750 407.85

Table 8. Effect of volatilities on optimum steady-state design

Case otj V~ NR N~ RR TAC

1 111/3/0.33 2500 10 10 2.489 346.05

2 1/I/2J0.5 3000 15 15 4.175 460.82
3 11111.510.67 3750 26 26 7.234 674.80
4 1111210.8 4500 14 49 8.172 765.40
 Coupled reactor/column s y s t e m s - - 2 61
Table 9. Effect of product purities on optimum steady-state design

Case xo3 xD VR NR Ns RR TAC

I 0.95 0.95 4250 11 34 7.512 686.2

2 0.99 0.99 4500 14 49 8.172 765.4

Table 10. Parameter values and optimum steady-state values of variables of the coupled
reactor/rectifier/stripper system (M3 case)

F,~ 50.0 Ib-mole/hr z,,~ 1.0/0.010.010.0 Mole fraction

F,,~ 50.0 Ib-mole/hr z,,nj 0.0/1.010.010.0 Mole fraction
N~ 15 stages N,~ 15 stages
Nr 31 stages N~ 16'~' stage
1/,. 258.75 Ib-mole/hr V,, 258.75 lb-mole/hr
D 50.0001 Ib-mole/hr B 49.9999 Ib-mole/hr
R 208.75 Ib-mole/hr Qve -0.11753 x 10~ Btu/hr
Vk 3000.0 lb-mole M,, 0.7048 lb-mole
M, 21.56 Ib-mole Mo 21.56 Ib-mole
xoj 0.005/0.005/0.990/0.000 Mole fraction
xB~ 0.00510.00510.00010.990 Mole fraction
TR 140 °F TNF. j 136.0 °F
13 0.74283 x 10 ~ hr
c~j 2.012.014.011.0 ,~ 0.0 Btu/lb-mole
kho 0.1538 x 104 hr ~ k_,o 0.4393 x 10~ hr ~
E~ 10000.0 Btu/lb-mole E., 10000.0 Btu/Ib-mole
bu -7591.5 b.,~ 16.542
b L, ~7591.5 b_,., 16.542
b~ -7591.5 b_,3 17.236
bt~ -7591.5 b_,4 15.849
M 50.0 Ib/Ib-mole H,~, 250.0 Btu/lb
Ct, 0.75 Btu/Ib °F Po 50.0 psia
T~, 140.0 °F 7"/, 110.0 °F
K~n 22461 lb/psia/hr

trays and reflux ratio increase rapidly as either of these In the studies made to date, the heat of reaction has
relative volatility ratios approach unity. been assumed to be zero. The effects of the heat of
reaction on the optimum design in the coupled reactor/
Effect of product purity on optimum design rectifier/stripper system is an interesting topic of future
Table 9 shows effects of product purities on optimum work.
steady-state design with the design conditions of aj =
Dynamics and control of M3
1.25/!.25/2.5/I.0, k~ = 0.35 and k, = 0.1. As the purity
is increased, the optimum reactor holdup, trays and The base case for dynamic study was chosen as the
reflux ratio increase. The purity affects column size more economic optimum steadystate. Table 10 shows parame-
than reactor size. ter values and optimum steadystate values of variables of
the coupled reactor/rectifier/stripper system for the
Effect of heat of reaction multicomponent M3 case.
One of the advantages of the coupled reactor/column Control structure of M3. Previous studies of more
systems is the utilization of the heat of reaction. simple coupled systems produced effective control
Exothermic heat of reaction can be used as a heat source structures. The best control structure for a binary
for reboiler heat in the coupled reactor/rectifier system. coupled reactor/rectifier was one in which distillate
For endothermic reactions, the load on the condenser can composition was controlled by reflux, reactor level by
be smaller because some of the heat of reaction can be heat input, reflux drum level by distillate and pressure by
provided by the heat of condensation of the vapor from condenser cooling. For the reactor/stripper, the inverse
the stripper. of above structure was used. The structure used reflux to
In the coupled reactor/rectifier/stripper system, an control level in the reactor, bottom flow to control base
exothermic heat of reaction can be directly used to level, reboiler heat to control bottom product purity, and
provide more vapor to the rectifying section. This gives condenser heat removal to control pressure. For the
a larger vapor rate in the rectifying section than in the multicomponent reactor/rectifier case, the only change
stripping section. For an endothermic heat of reaction, was the flow rate of one feed component was used to
the vapor rate is higher in the stripping section than in control one composition in the reactor for the purpose of
the rectifying section. This possibility gives another precisely satisfying the reaction stoichometry.
design optimization parameter in addition to VR, Ne, and A combination of these control structures is used for
Ns. the reactor/rectifier/stripper system. We cannot claim the
 62 CHANGK. YI and W. L. LUYBEN

1 A
l

FoA I

. . . . . . . . . . i
61!

Fig. 14. A control structure for multicomponentcase of reactor/rectifier/strippersystem (M3).

structure developed is necessarily the best, but it works

well. Figure 14 shows the control structure used. It
gpe
Gu.)=--
- Ds

(8)
S
includes 1 pressure, 3 levels, I throughput, and 3
composition controllers. Pressure is controlled by con-
where Kt, is the slope of the ramp change in the
denser heat removal. The reflux drum level is controlled controlled variable for a step change in the manipulated
by distillate flow rate, bottom column level by bottom
variable (percent/hr) and D is the dead time (typically
flow rate, and reactor level by liquid flow from the
from a composition analyzer). Both Kp and D can be
reactor. The feed flow rate of A is flow controlled. The
calculated from the ultimate gain K,, and ultimate
feed flow rate of B is used to control the composition of
frequency co,, once they have been determined by an
component B in the reactor. The bottom product purity is ATV test (Tyreus and Luyben, 1992).
controlled by reboiler heat. The distillate product purity
is controlled by reflux flow rate. Notice that there is one
control valve not set: the valve on the heat exchanger in D =-- (9)
2o),,
the reactor. This controls the rate of heat addition or heat
removal. It can be used as an optimization variable (eg.
minimize energy cost) or to adjust the vapor loads in the
rectifying and stripping sections.
System identification of M3. This system is a multi- K,--E OAo
(10)
input multi-output (MIMO) process that has three
interacting composition loops once the inventory loops Friman and Waller proposed that for a MIMO process
have all been closed. To identify the nine transfer the two parameters in the off-diagonal elements can be
functions, we used the simple integrator/deadtime model determined in the same ATV test as the diagonal
and the relay feedback test as suggested by Friman and elements. Only three ATV tests are needed for a 3 × 3
Waller (1994). system. Using the integrator/deadtime model, the trans-
 Coupled reactor/column systems--2 63
Table I I. Ultimate gains and frequencies
from relay feedback test (M3)

K, to~

xoJR 0.994 12.083

xnJV,, 0.61 15.14
jF,, R 120.88 31.42

Table 12. PI tuning parameters of Z-N, T-L, and BLT tunings (M3)

ATV Test Transfer Function

Model
Z-N Tuning T-L Tuning BLT Tuning
K, % K, r, K~ r,
xo.~R 0.452 0.433 0.309 1.144 0.141 1.39
xJV,, 0.275 0.346 0.188 0.913 0.086 I. I I
::IF,,s 54.944 0.167 37.54 0.44 17.12 0.53

fer functions were identified. Table 11 gives the ultimate after 20 hours. T-L tuning gives reasonably good
gains and frequencies obtained. closed-loop performance.
The level loops were assumed to be perfectly
controlled. The signs of K;, constants were obtained from
Effect of reaction constants on controllability
step tests. The identified open-loop transfer function If the reaction constants are changed by adding
matrix is catalyst or increasing reactor temperature, both the
steady-state design and the dynamic controllability may
IjfxD33 be significantly changed. When both reaction constants
XB4 I = are doubled from k~ = 0.35, k_, = 0.1 to k I = 0.7, k2 =
NF.2J 0.2, the optimum reactor holdup is smaller and the reflux
rate is smaller (see Table 7). The dynamics of the two
-+ 12.155e-°'13°'~ _ 10.327e- 0.)35, _ 0.017e- 0.114, different systems were also compared, using the same
s S S
controller structure and T-L tuning parameters. Figure
17 shows the closed-loop load responses when the feed
- 23.121e -°l°z' +24.981e -°'"~ - 11.899e -°'°75 composition of A (zoAj) is changed from 1.0/0/0/0 to
0.7/0.3/0/0. The larger reaction constants case shows a
S S S
little faster load response in distillate product purity xo3
+0.036e -oJ)3), - 0 . 1 9 8 e -o.267~ +0.260e -o.oSn~ because the reactor holdup is smaller. The controllability
S S S
differences are small. We can conclude that the control
(11) structure and T-L tuning work well for a broad spectrum
of parameter values in coupled reactor/rectifier/stripper
Choice of tuning in M3. The transfer functions were systems.
used to choose an appropriate tuning method. Ziegler-
Nichols tuning, Tyreus-Luyben tuning (Tyreus and
CONCLUSION
Luyben, 1992) and BLT tuning were tested for this
system. Table 12 shows tuning parameter values of each The steady-state designs and dynamics of a binary
method. reactor/stripper system, a multicomponent reactor/recti-
Figure 15 shows that Z-N tuning gives a closedloop fier system, and a multicomponent reactor/rectifier/
unstable system: the Nyquist plot encircles the ( - 1,0) stripper system were studied.
point. Since BLT tuning gave more sluggish controller An optimum steady-state design for each system was
action than T - L tuning, we used T-L tuning parameters found as a base case for control study. The effects of
for this study. physical parameters such as reaction rates, relative
Closed-loop performance of M3. Figure 16 shows the volatilities, product purities on optimum design were
closed-loop responses when the feed composition of A described.
(ZOAj) is changed from I10/010 to 0.7/0.3/0/0 using three The best control structure from the simple binary
different tuning parameters. These responses are reactor/rectifier system was applied to these three
obtained from rigorous dynamic simulation. systems with minor modifications. A control structure
As predicted by the multivariable Nyquist plots, Z-N that was the inverse of the control structure found to be
tuning gives unstable closed-loop performance. BLT the best for the binary reactor/rectifier system was
tuning gives quite slow control action. The distillate successfully applied for the binary reactor/stripper
concentration does not even reach the set point value system (B2). For the multicomponent case (M l), which
64 CHANG K. Yl and W. L. LUYBEN

M u l t i v m a b l e Nyquist Plots:: m l i d : Z - N dmhed.-T-L d o n e d : B L T

'..-. i ! : : :

'. ~, i : ; ./. ............... ;. . . . . . . . . . . . .

........ :. ................ ! ................ !................ ! ................ i...............
". .%. : : : :
"'. : ,, T - L :

............... "..,: . . . . . .,,.< . . . . ; . . . . . . . . . . . . . . . . :. . . . . . . . . . . . . . . : ................ !. . . . . . . . . . . . . . .

• .. ~. . : • :
: "'.. ~... ! i :
• '- " ~ i

BLT "" . . . . . . . .

: ""-... : :...'"
• "..-.. ........ ..,.''.

I I I
-I
-5 -4 -3 -2 -1 0
REAL

Fig. 15. Multivariable Nyquist plots (M3).

mole)
features products that are both more volatile than
Fo:fresh feed rate to reactor (lb-mole/hr)
reactants, the best control structure for the binary FoA:fresh feed rate of A component 0b-mole/hr)
process was directly applied with only one modification: Foa:fresh feed rate of B component (lb-mole/hr)
H,~,l,:heat of vaporization (Btu/lb)
the flow rate of one feed component was used to control
K,.:gain controller
one composition in the reactor for the purpose of K,¢:reaction equilibrium constant, kt/k 2
precisely satisfying the reaction stoichiometry. An KMr:pseudo mass transfer coefficient (lb/psia/hr)
Kt,:constant of integrator/dead time transfer function,
effective control structure for the multicomponent reac-
(/hr)
tor/rectifier/stripper system (M3) was developed by K,,:ultimate gain
drawing on the experience gained in the study of more k:specific forward reaction rate (hr-))
kt0:pre-exponential factor of k t (hr -~)
simple coupled reactor/column systems.
k2:specific backward reaction rate (hr-~)
The chosen control structures handled various types of k,,:pre-exponential factor of k2 (hr-i)
disturbances well for systems having a broad range of L:liquid flow rate in the column (Ib-mole/hr)
L,:liquid flow rate at the n#' stage (lb-mole/hr)
physical parameters.
M:molecular weight (Ib/Ib-mole)
Ms:liquid holdup in the column base (lb-mole)
NOMENCLATURE
Mo:liquid holdup in the reflux drum (Ib-mole)
M,,:liquid holdup on the n 'h stage (lb-mole)
A:component A Nr:feed tray
B:component B NR:number of trays in rectifying section
B:bottom flow rate or recycle flow rate (Ib-mole/hr) Ns:number of trays in stripping section
b~:first Antoine constant ofj component Nr:total number of trays in column
b,..:second Antoine constant ofj component P:average pressure in the column (psia)
C:component C Po:pressure in the condenser (psia)
Cv:beat capacity of liquid (Btu/lb°F) /),,:pressure in the n#' stage column (psia)
D:component D P~:pressure in the reactor (psia)
D:distillate flow rate (Ib-mole/hr) P':vapor pressure of pure component j (psia)
Dc:diameter of column (ft) Qc:heat of condensation (Btu/hr)
DR:diameter of reactor (ft) Qa:heat input in reboiler (Btu/hr)
E~:activation energy of forward reaction (Btu/lb-mole) QvR:heat input in the reactor (Btu/hr)
Ez:activation energy of backward reaction (Btu/Ib- R:reflux flow rate (Ib-mole/hr)
 Coupled reactor/column systems--2 65

Ikdlml:Z-N SdldO'-L D d ~ : B L T
99.8 i i i i i i i i i

99.6 •. . . : : - - :.".-..-..-.~ .......... ~ ......... ~ ......... i ......... i ........ -i.-- ::

99.4

~ 99"2

~.4

~.~ ...................................
2 4 6
: .... ',.......... i'~ ...... ', .......... ? ........ :,...... "~i ........
$ I0 12 14 16 lit
1
20

m~i_...
Deled:Z4q ~lJlt'T-I. D I d s t : l I L T
210 -.. j i ~ J t e J ~ J

-..': ...... : ........ !.....

.. 7. :~.~ _.~:................ ~.. ........
:':..! .......... ::'.?
.........
:" ~:..........
: ~ - , . ! ?. . . . . . . . .
,'° ~:i .......... i...... :~ ..... ~:.:= ...... i ......... i .........
160 . . . . . . . . . . :. . . . . . . . . . ~-..'.'.'''.-'a . . . . "".'," ............. ". . . . . . . . . . . . . . . . . ~'.:.. ..... ~'~.''-C- .........

I I I i I I I I I
2 4 6 8 I0 12 14 16 18 20
Time, hr

I)elled~-N Sdid~T.L Dldled:llLT

~ z0:."
~ : .........
~ . : / %:..........
. i - ~ .i ..........
~ i : ~ "i . ..:..)
; i . : ~: - . ~ .: ":: •
. . . . . . . . . .

• '-::...'...
. . . . . .

: .~-~.i
. .

. ~.~.~:
. .

:.'.i ":: !'.~t

. . . . . . . . . . t
0 2 4 6 8 10 12 14 16 18 20

Fig. 16. Closed-loopresponses to load disturbances of zoAi=0.7/0.3/O/O(M3).

RR:reflux ratio yt~r:vapor phase composition of B component at the

TAC:total annual cost ($1000) final stage (mole fraction)
To:temperature in the condenser (°F) y,,/vapor composition ofj component at the n'~' stage
T~o:temperature of fresh feed (°F) (mole fraction)
T,,:temperature at the n'h tray (°F) z:composition in the reactor (mole fraction of B)
TR:temperature in the reactor (°F) zj:composition ofj component in the reactor (mole
V:vapor boilup (lb-mole/hr) fraction)
VR:reactor holdup (lb-moles) z,,:fresh feed composition of B component (mole
V~mi,,:minimumreactor holdup (Ib-moles) fraction)
V,,:vapor rate in the rectifying section (lb-moles/hr) z,,Aj:composition ofj component of fresh feed A (mole
Vs,:vapor rate in the stripping section (Ib-moles/hr) fraction)
wh:weir height fit) z,,~j:composition ofj component of fresh feed B (mole
xn:bottoms composition (mole fraction of B) fraction)
xBj:bottoms composition ofj component (mole z,,j:fresh feed composition ofj component to the
fraction) reactor (mole fraction)
xo:distillate composition (mole fraction of B)
xoj:distillate composition ofj component (mole Greek
fraction) a:relative volatility
x~:liquid phase composition at the n'h stage (mole %:relative volatility ofj component
fraction of B) ,8:liquid hydraulic time constant (hr)
x,,j:liquid phase composition ofj component at the n" A:heat of reaction (Btu/lb-mole)
stage (mole fraction) 0:constant in 0 method
66 CHANG K. Yt and W. L. LUYBEN

~ I ~03~t2d~.l l ) l l b l l t I,,O,?,lll-O3

99.6f . .~. . . . . i i i i i i ' S. . . . . .;. . ' . . . . . . . . . . i:. . . . . . . . . . f .......... : ......... i .......... ~. . . . . . . . .

,+.4
w
.'----...,..-...
~
~ . . - . . . . : ~ i " ....... ! .......... ? ......... ! . . . . . . . . . . : .........

99~ " • ........ ).......... i .......... ; .........

99.1 " ~

0 2 4 6 8 tO 12 14 16 18

~llddtl-0.35Jt2=O.l nMhd:tl~0.7 t2~03

21@ ! i u i ! u

~
lU
.... ' - _- . . . . . . . . . . . . ' . . . . . . . . . . !.......... .:. . . . . . . . . . i........... !. . . . . . . _:. . . . . . i. . . . . :
17~

......... i. . . . . ., i. ...... i. . . . . . . ! ......... i. . . . . . . . ! ......... ! ........ ! ......... ! ........

a~
IS©
I4C
......... ...... i........ i .~ :'i~.-. :. ,:!. . . . . . . i . . . . . . . . :. . . . . . . !. . . . . . . . . ~. . . . . . .
~zoi i i i i i" - - - i " - - -i-- - - _ L ' _
2 4 6 8 10 12 14 16 II 20
TiII~ lu

$Olfl:l I ~0.35~210.1 l ~ : l t 1~0.7,1tlm02

5( ! ! ! ! ! !

,: ..................... :: .......... !.......... i .......... :, ...................................................

! '°,,, i i i ii il

20 I I I I I £-- p,- a
0 2 4 6 8 I0 12 14 16 18 20
Time, ho'

Fig. 17. Comparison of effects of reaction constants on dynamic responses (M3).

r:controller reset (hr) reactor/column systems: Part l - - a binary coupled reactor/

oJ,:ultimate frequency (radians/hr) rectifier system. Computers chem. Engng 21, 25 (1997).

APPENDIX
REFERENCES

Friman M. and Waller K. V. Autotuning of multiloop control NONLINEAR DYNAMIC MODELS OF MULTICOMPONENT
systems. Ind. Eng. Chem. Res. 33, 1708 (1994).
Holland C. D. Fundamentals and Modeling of Separation REACTOR/RECTIFIER/STRIPPER SYSTEM (M3)
Processes. Prentice-Hall (1975).
Luyben W. L. Dynamics and control of recycle systems: Part The dynamic model of the multicomponent reactor/rectifier/
2-.--.comparison of alternative process designs, hu/. Eng. stripper system is provided. For other multicomponent systems,
Chem. Research, 32, 476 (1993). these equations can be modified easily. We make the following
Papadourakis A., Doherty M. E, and Douglas J. M. Relative assumptions: equimolal overflow, saturated liquid feeds and
gain array for units in plants with recycle. Ind. Eng. Chem. reflux, constant relative volatilities, constant pressure, theoret-
Research, 26, 1259 (1987). ical stages, partial reboiler, and total condenser. The assump-
Tyreus B. D. and Luyben W. L., Tuning PI controller for tions of equimolal overflow and negligible vapor holdup mean
integrator/deadtime processes. Ind. Eng. Chem. Res., 31, that the vapor rates through all trays of the column are the same.
2625 (I 992). The mathematical effect is that we do not need an energy
Wang J. C. and Henke G. E., Tridiagonal matrix for distillation. equation for the trays in the column. However, we do include
Hydrocarbon Processing, 45, 169 (1966). the energy balance for the reactor because its large holdup
Yi C. K. and Luyben W. L., Design and control of coupled represents significant thermal capacitance in the system.
 Coupled reactor/column systems--2 67
Reactor equations
d(Mmaurj)
Total continuiO': dt = Rx°~ - Lmxura + V,,y~_ ,j - V,,yur~ (21)

dVR=F,,A + Fon+Lur+I - LN¢+ V , , - V~, (12) Condenser and reflux drum:

dt
dM t~
.... V,,-R-D (22)
dt
Component material balance for component j:

d(V,zj) ~ .c . d(M~oj) - (R+D)xos (23)

- d t - - = r °az°aJ'r r <,~<,nj dt = v,~yurj -
Stripping section
+ LNF+IX:¢¢+ Ij - LNFZj+V~ryNF-Ij-- V,eyNFj
Reboiler and column base:
+ vjVRkloe - E'IR(T~+4~IIIZIZ2 ~ k2oe- E-umr"+460)z3z4 (13)
where v.+ is negative when j is 1 (A component) or 2 (B dMs
~=L, - V,, - B (24)
component), and is positive when j is 3 (C component) or 4 (D (1/
component).
Energy balance: d(M.xnj)
dt =L,x~j - V,,ynj - Bxn/ (25)
d(V.TR)
--=(Foa + F,,n)Tr,,+ LuF+iTuF+, - LN¢TNr n th tray."
dt
dM.
+ V.TNp_ +#+I,.,,_ V. 14,,,.
TNF+ ~ -
~-t = L,, +, - L,, (26)
• ' C,, + C,,

AVttt[ki . e- e'm<r++4t'°~Zi Z,. - k,oe- edtctrR++6mZ3Z4] d(M,,x,,s)

dt =L,,+~x,,+~j- L,er,,j+V,,Y,,-Ij- V,y,,j (27)
MC,
NS m tray."
QVR
+-- (14) dMm
MC,, dt = L u ¢ - Lus (28)
Pressure in the reactor." from bubble point calculation.
4
PR=Y~P')zj (15) d( M usxusj)
1=1 dt = Lurzj - Lusxusj + V+,ym- I u - V,,ymj (29)

In P+s=~blJ -+b_,j (16) Equilibrium relationship

TNp+460 -
Rate of vaporization: ot/c.j

y,,)= ~4k
=~10t~-X,,~ (30)
Kscr
v,,, = ~ . - . ( P . - P) (17)
M Liquid-hydraulic relationship:
Rectifying section g.-~t.
n th tray: L,,=L,,+. . . . (31)
/3
dM,
~=L,,+,-L,, (18) The parameter /3 is the hydraulic time constant which is
calculated from a linearized Francis weir formula.

d-(-M~'~"L)= L,,+ ~x,,+,j- L,er,,i+ V,,y,, _ ,j - V,,y,,j ([9) Temperature on each tray
An equation is obtained assuming Raoult's law and a vapor
N7~h tray: pressure of pure component j:

dMu-+-r= R - Lm
dt
(20) ]
b+ -460 (32)