Вы находитесь на странице: 1из 12

The Chemical Engineering Journal, 16 (1978) 153 - 164

0 Elsevier Sequoia S.A., Lausanne -Printed in the Netherlands

Constrained Optimization of Cold-shot Converters

A. BURGHARDT and T. PATZEK


Polish Academy of Sciences, Research Centre of Chemical Engineeting, Gliwice (Poland)
(Received 24 February 1977)

Abstract maximize its output. Hence the closest


approximation to the optimal temperature
General conditions are derived for an optimal profile is desirable. Nowadays this is achieved
multi-bed adiabatic converter with coldshot in multistage adiabatic converters by inter-
cooling. The problem is solved with and with- stage cooling, either indirectly in heat
out inequality constraints on the temperature exchangers or directly by cold shots of the
of each bed. A numerical method is presented feed stream.
which quickly establishes the optimum A method of optimizing multistage
.working conditions of the converter. adiabatic converters with indirect cooling has
The compu tational problems arising in the been proposed by Horn and Kiichler [ 81.
optimization of ammonia quench converters Their algorithm maximizes converter output
and their solutions are described. Results for both a fixed number of beds and a exed
using both the modified Ben ton equation and inlet temperature. An optimization of cold-
the Temkin-Pyzhev equation are given. shot converters has been carried out by Lee
and Aris [9, lo]. The objective function
derived by Lee and Aris maximizes the total
1. INTRODUCTION net profit from a converter.
In this way the authors optimized an SOz
In the analysis of optimization problems with converter using a dynamic programming
temperature as the decision variable “ideal scheme. Unfortunately, their results differ
temperature control” is generally assumed. significantly from those published elsewhere.
Thus the possibilities for the practical realiza- The problem formulated by Lee and Aris was
tion of the best temperature profile are not also solved by Melange and Vincent [ll] who
investigated. The well-known Denbigh condi- used an identical objective function and
tion [l] for the optimum profile in a kinetic equation. The calculations by Melange
converter with a single reversible exothermic are in agreement with those by Hellinckx and
reaction is based on this assumption. van Rompay [12] , who applied the discrete
Since then, the optimal temperature profile maximum principle.
theory has been widely developed for several All the papers by Axis, Melange and
types of reactions, especially for the oxida- Hellinckx deal exclusively with the optimiza-
tion of SOa and for ammonia synthesis [2 - tion of cold-shot converters for SOz oxida-
61. The temperature profile obtained under tion. The optimization of other types of reac-
Denbigh’s conditions decreases monotonically tions is in most cases more complicated.
along the axis of the converter, and the A serious difficulty arises when the temp-
temperature of the feed - conversion being erature inside a converter must be less than
equal to zero -.- should be infinite. As shown some maximum value T*, owing to construc-
by Westerterp [7] , the industrial realization tional or technical considerations, e.g. because
of the optimal profile would be very cumber- of catalyst deactivation. If the optimal outlet
some, if not impossible. temperature of the first bed is higher than T*,
Nevertheless, knowledge of the temp- the volume of the bed will not be a minimum.
erature profile helps significantly in optimizing In this case, for indirect cooling, the optimum
heat exchange in a converter and thus to initial temperature of the second bed is given

153
154

by the condition derived by Horn [13] : 2 viAi= 0 (1)


i=l

‘Ilk 1 a(-rl) 1
a dq + .-?- - -=O
s 72
rl aT Assuming an ideal plug flow reactor model for
-1k ?2p
‘11P
each bed of the converter, we describe the
In contrast, it is known [14] that the optimal space-time of bed i as follows:
temperature in cold-shot converters may
‘7ik
exceed the maximum permissible value T* in (=
vi
--*=c,& ---
dr)

several beds.
(2)
vo s -ri (v, T)
7)ip
In the present work we have derived the
general conditions for an optimal cold-shot The relation between temperature and conver-
converter with inequality constraints imposed sion in each bed because of the adiabatic
on the temperature of each bed. A numerical nature of the process is
method is also presented, which rapidly 1
dv
-=_ CP
establishes the optimum working conditions (3)
dT (-Ah)w, = i-
of the converter. Furthermore, the computa-
tional problems of the optimization of am- We further assume that a is approximately
monia quench converters and their solutions constant. Integration of eqn. (3) gives imme-
are described. diately
T=h+aq (4)
2. MODEL OF A COLD-SHOT CONVERTER where
h= Tp --aqp (5)
Let us
consider a reactor with N adiabatic
beds. The reactant stream is divided into two Therefore, on the adiabatic path,
parts, a fraction X,G going to the preheater
Vik
where its temperature is raised from To to dq
T and the remainder (1 -- h1 )G serving as a
7: = Cohi (6)
s --ri(q, hi + aq 1
riip
b$ass cooling stream. The composition of
the stream is defined by the conversion degree We note that any parameter of the mixture at
Q (Fig. 1). the inlet to bed i is uniquely determined by
In bed i with space-time coordinate 7i vi-i, k and by the mass and heat balances of
the conversion and temperature increase from the ith mixing node as given in Fig. 2.
nip t0 Qik and from Tip t0 Tik, respectively. These two balances can be written as
After leaving bed i the mixture is cooled by
Ai-iGWi-i,k + (Xi - Ai_l)Gw,p = XiGwip (7)
mixing it with a fraction Xi+i -- hi of the
original feed, and the temperature and
conversion fall to Qi+i, p and Ti+l,p, respec- Ai-lG(Cp)i-l,kTi-l,k +
tively. Finally, the given COnVerSiOn nNk is + (hi -- hi_r)G(cp)oTa = hiG(cp)ip Tip (3)
achieved in bed N at rN.
IA us consider a single reversible reaction Using the definition of the degree of conver-
described by the stoichiometric equation sion we obtain from eqn. (7)

/ ( Ai++-AJG (‘Ai+2_li+)G (‘-‘H-~)G


_ i . g-i+1 r

rib a0
i
7 ik ‘lit,pA, ,,, G ‘lit+*
ro T T(I Tik T.
‘P Tip 1’CP Ti*lJt TP Tnh

Fig. 1. The reactor system.


155

hl = T1p
--aqlp
(Xi - ‘Ai-l)Gtwjp Furthermore we introduce the vector x charac-
To’ (cp). terizing the problem :

x = (r)Ik,~2k,...r~N-1,k,hl,h2,...,XN--1,hl)

(13)

Our constrained optimization problem can


now be formulated as follows. Given qNk and
ho = TO-- ag rp, minimize the space-time
Fig. 2. The ith mixing node.

T(X)=; Ti(X) (14)


9ip=‘f (17i-l,k-??lp)+qlp (9) i=l
I

Further, we assume that the specific heat of of the converter subject to the constraints
the mixture does not vary appreciably with i= 1,2 ,...,N-1
k%(x) 2 0 (15)
temperature and composition, i.e.
of the form
(cph-1,k w (Cp)ip * (Cp)O
T* -hhf
Then - --7)ik > 0 (16)
a
Ai-1 (Ti_1, k -
Tj:p= h_ TO) + TO (10) where T* is the maximum permissible temp-
I erature in the converter. Alternatively, we
Making use of relations (9) and (10) we can need to find a vector x’ such that
rewrite eqn. (5) as follows :
7(x’) = inn;T(X) (17)
hi = ho + F (h, - ho) (11)
1 where F is the set of feasible solutions. Let us
where define a function P,,(x) such that

0 forx E F
Consequently, we can give the reaction rate lim P,(x)= (13)
P+o
equation in the ith bed explicitly: +- forx$ F
(
ri = Pi 77, ?(h, -ho) +h, +aq\ (12) A new constrained objective function can now
i be defined as follows:

Fp(x) = T(X) +p,(x) (19)


3. THE OFTIMAL CONDITIONS The penalty function Pp(x) defined by rela-
tion (18) is determined as
An N-bed quench converter is completely de-
scribed by a set of 2lV independent variables P,(x) = f Nzl [min{O,gi(X)Il
2 (20)
or, alternatively, by the COnVeE3iOn qik (i = 1, i-
2, . . . , N) at the outlet of each bed, the fraction
hi (i= 1,2,..., N - 1) of the total feed flowing or using eqn. (16)
through each bed and the inlet temperature
2
T1, of the first bed. The remaining variables T* - hi
9 -- -qik (21)
such as the temperature or composition of the a
mixture in any cross section of the converter
are unique functions of the set of independent where p is an arbitrarily small positive number.
variables. Finally
For convenience we choose a new variable
156

F,(x) = lim F,(x) = T(X) + lim PP(3c) (22)


P-*0 P-*0

It is worth mentioning that if any constraint is violated the penalty function P,(x) forces F,(X)
to approach infinity.
It may also be shown [15] that any vector 3~’minimizing the objective function F,(x) in its
limiting form lies within or on the boundary of the permissible region F. The necessary condi-
tions for an optimum can be written as
lim aFp(x’)
-=_ aFo(x’)= o (23)
p.-~ ax ax
or
aF, aF, aFc aF, aF, =()
-.- ._ _- _.- _- (234
an&“’ a9N-1. k’ aAl ““’ ah,_,’ ah1
( 1
We obtain the limiting form of the objective function explicitly after inserting simultaneously
eqns. (9), (11) and (16) for nip, hi andgi, respectively:
r),k N qik
dv
- d7, -+
F, =X1 +x .Xi
s -rl(v, hl + W) i=2 s
hi-l(Vi-l,k -~lp)~hi+%P Xl
‘IIP -ri 7, x- (h, .-- ho) + ho + a?)
i

(24)

The conversion r)ik in the outlet stream from bed i appears as the upper limit of the ith integral,
the lower limit of the (i + 1)th integral and in the ith term of the penalty function P,(x). Thus

(25)

The first part of the alternative eqn. (25) corresponds to the unconstrained optimization
-rik = ‘-ri+i,p (26)
The second part gives

(27)

and in the limit ’


1 1

qik
or
=
P-+0
lim 1 i ---_P_
2
Ai -
Tik
-
-ri+
---
1, P
1
a
+T*_hi
a
1
(23)

T* -hi
qik = - (29)
a
It should be noted that for violation of the ith constraint the optimal solution lies on this very
constraint (Fig. 3).
The fraction Xi of the total feed flowing through the ith bed appears in the ith integral as a
multiplier, as a parameter and also in the lower limit of integration. hi appears also in the lower
limit of the (i + 1)th integral and in the ith term of the penalty function. Differentiating the ob-
jective function Fp with respect to hi (i = 2,3,. . . , N -- 1) we obtain
aFp_ 7)ip -7)lp __qik-qlp qik 1 a(--ri) %k dq
___- - + (hi -hoI -_--dq+ -+
ahi _- s r: aT s --ri
-rip ri+1, p Vp Qip
157

Fig. 3. Optimal temperature profile in a four-bed ammonia quench converter.

AI h,--ho
=0 (30)
91 a
The first part of eqn. (30) fulfils the condition for an unconstrained optimum :

tlip -'lllp ‘tlik -alp + qik 1 + hi -ho a t-c)


-_- - - dq=O (31)
--rip -ri+l, p sl -ri r” aT 1
‘lip

The second part, after inserting eqn. (27), takes the form

qip -?)lp _qik’-qlp 1 hi -ho a(-+ hi-ho 1 1


+ __+-- dq+~_------ 0 (32)
-rip -ri+l, P -ri rf aT t a 1-rik -ri+l.p

We note that AI appears as a multiplier of the first integral, in the lower limit of the second one
and in all integrals from 2 to N and constraints from 2 to N - 1.

aFP ‘)lk dq qlk -771p %k 1 ari


-= --
+ ; (hl - ho)
2 &q-
‘Xl s -3 -ra i=2 J
QlP QiP

2 N-l T* hi h, - ho =
-- x min 0, ----r)ik 0 (33)
P i=l f a a UXi

Again, using eqn. (27) leads to

aFP_ ‘ilk dq qlk.-- qlp %k 1 ijri


-- +; @r---h,) --dq-
ax,
VlP
J -rl
--- -r2p i=2 J
Tlip
rf aT

2 N-1 h, -ho
-- I: =0 (34)
Pi=2 a

Thus only the condition of optimality of the inlet temperature

lim aFP
- =0
/J-+0 ah1
158

is left. The variable hr = Tip - an iP appears as a parameter in all integrals and in all constraints;
therefore

T* --hi 1
dn - .? !? N$l min 0, (35)
P a i=l - a --nik X-.=
I O

or, inserting eqn. (27),

--dn-.?!Y! Nilmin =0
P a i=l

It should be noted that the use of the above condition in eqn. (34) leads to conditions (31) or
(32) for i = 1. Finally, the optimal conditions are of the form
for Tfk < T*

rik = ri+l,p i = 1,2,...,N- 1 (36)

for Tfk > T*


T* -- hi
qik = ~ i= 1,2 ,...,N- 1 (37)
a
and

Vip - rl lp hi - ho a (-ri)
--- - _+p -
--rip rf aT
i= 1,2 ,...,N-1 (38)

N
=
aIZ 0 (39)
i=l

As far as the objective function is concerned, erature Z+ , then nl;k has to be determined from
eqns. (36) - (39) express the most general eqn.(37)fori=l.
optimum conditions for an N-bed quench (b) The optimum inlet parameters nb,
converter. T& of the second bed can be found from eqn.
If the constraints are not exceeded, eqns. (36) if Tik < T*, or from eqn. (38) if Tik >
(36) - (39) reduce directly to the un- T*.
constrained optimization scheme of Melange (c) The procedure (a) and (b) is repeated
and Vincent [ 111, provided the costs of pre- for
heating are negligible. The optimum design
of an N-bed cold-shot reactor may be realized i= 2,3,..,N-1
by following a procedure which is based on Here it is worth mentioning that we pass from
eqns. (36) - (39). bed i to bed i + 1 along the mixing line, the
(a) Initially we assume that the inlet temp- equation of which is
erature TIP is known and that no constraint is
Tik - 7’0
approached, i.e. T=T,+ (7) -VIP) (42)
9r?r- Vlp
rlk = ?& (40)
(d) For i = N we require condition (39) to
‘I’he outlet conversion of the first bed can be fulfilled.
then be found from eqn. (38) for i = 1. If the (e) We compare the optimal outlet conver-
temperature at the outlet of the bed SiOn t7hk Of the rt?aCtOr obtained from eqn.
(39) with the given 1)Nk.
T;k = hl + a& (41) (f) If &k # QNk,we can change the initial
exceeds the maximum permissible temp- temperature TIP so as to obtain
159

6Vk = r)Nk (43) q 1 hi-he a(-ri)


+ -+ 2 - dr) (46)
JI -ri ri aT t
However, there is no need to do this, since Oip
according to eqns. (36) - (39) Thus the solution nik fulfils
proposed h?:k) = 0 (47)
TI, --, + dJk (44)
algorithm Acomputation of the integral in eqn. (46) by
is equivalent to the procedure the Gauss method allows us to replace this
integral by a continuous function of the
?Nk + Tip (45) upper limit of integration :
i.e. for every feasible value TIP, there exists
one and only one optimal value 7)&, and (43)
conversely. Of course, all our considerations
hold for
Thus eqn. (46) assumes the form
TNk< T*
i.e. the point (?)bk, T&k) belongs to the region Ip(~)=&!T! + r(q) (49)
of permissible solutions. -ri (17
)
(g) For a given nlIp, TIP and calculated where
values v;~, Ti’, Cj= 2,. . . , IV) we compute suc-
Ci = (77ip --r)lpM-cp)
cessively Xi using eqn . (11).
Finally let us denote the following: for an
Computing the proper root of eqn. (49) by
unconstrained optimization (ig&) > 0) eqn.
the Newton-Raphson method, or better by
(38) can be written as
the false position method [ 16 ] , gives the
&)‘VliP --1)lP __q --VIP + solution nfk. The same remark applies to
-rip -ri(V 1 computing &k from eqn. (39).

Fig. 4. Optimal outlet conversion qj,Tk(N = 1,2,3,4) VS.velocity 6 for the modified Benton equation.
160

M i j i ; i i * I; & & ir s s n * fg 20 H 22 28 Y .m M fl m3rmi

Fig. 5. Optimat outIet conversion r&k fN = 1,2,3,4) us. velocity s for the Temkin-P~z~ev equation.

Fig. 6. Optimal velocity s vs. inlet temperature Tip and optimal outlet conversion vhk (N = 1,2,3,4) for the
modified Benton equation.
161

TABLE 1
Optimal one-stage converter parameters from the modified Benton equation

TlP 615.0 625.0 635.0 645.0 655.0


*lk 793.0 793.0 793.0 793.0 793.0
%k 0.2754 0.2636 0.2517 0.2399 0.2280
71 0.343 0.250 0.184 0.137 0.102
s 10506 14393 19546 26363 35390
-

TABLE 2 9
Optimal two-stage converter parameters from the modified Benton equation

TlP 615.0 625.0 635.0 645.0 655.0 665.0


Stage I *lk 793.0' 793.0 793.0 793.0 793.0 793.0
qlk 0.2754 0.2636 0.2517 0.2399 0.2280 0.2162
71 0.246 0.184 0.139 0.105 0.080 0.061

Stage II , hl 0.719 0.737 0.755 0.772 0.786 0.806


T2P 652.4 661.6 670.5 679.2 685.9 695.9
%P 0.2161 0.2113 0.2059 0.2000 0.1930 0.1868
*2k 772.4 776.6 781.0 785.5 789.6 793.0
q2k 0.3611 0.3502 0.3393 0.3284 0.3183 0.3064
72 0.663 0.506 0.388 0.300 0.240 0.182
S 3960 5218 6826 8879 11244 14821

TABLE 3

Optimal three-stage converter parameters from the modified Benton equation

T1, 625.0 635.0 645.0 655.0 665.0 675.0 685.0 695.0


Stage I *a 793.0 793.0 793.0 793.0 793.0 793.0 793.0 793.0
r)lk 0.2630 0.2512 0.2394 0.2280 0.2162 0.2044 0.1925 0.1807
71 0.142 0.109 0.083 0.064 0.050 0.038 0.030 0.023
Stage II Al 0.573 0.594 0.615 0.629 0.653 0.675 0.706 0.735
T2P 661.7 670.6 679.3 685.9 695.9 704.0 711.8 719.5
q2P 0.2108 0.2054 0.1996 0.1930 0.1868 0.1795 0.1718 0.1637
*2k 772.4 776.4 780.7 784.7 789.6 793.0 793.0 793.0
q2k 0.3418 0.3307 0.3195 0.3101 0.2977 0.2849 0.2679 0.2507
72 0.321 0.249 0.194 0.158 0.119 0.091 0.066 0.048
Stage III h2 0.778 0.787 0.796 0.800 0.810 0.822 0.843 0.862
Tsp 665.8 673.5 681.0 686.8 695.4 703.8 714.4 723.8
1)3P 0.2801 0.2740 0.2674 0.2612 0.2535 0.2456 0.2360 0.2249
Tak 756.5 760.5 764.6 768.0 772.9 777.8 784.0 790.4
7)3k 0.3896 0.3790 0.3686 0.3593 0.3472 0.3350 0.3199 0.3053
73 0.787 0.617 0.486 0.400 0.307 0.235 0.169 0.124
S 2879 3694 4719 5787 7564 9866 13593 18460

4. COMPUTATIONAL EXAMPLE P tatm) To (X1 T*(X) l)lp YIP

300 293 793 0.065 0.1


We have performed an optimization of an
ammonia quench converter on the basis of the The optimization (Figs. 4 - 6) has been carried
algorithm given earlier. The technical para- out with the help of two kinetic equatiotis:
meters of the reactor are as follows. the Temkin-Pyzhev equation [17] and the
162

TABLE 4
Optimal four-stage converter parameters from the modified Benton equation

TlP 655.0 665.0 675.0 685.0 695.0 705.0 715.0

Stage Z Tlk 793.0 793.0 793.0 793.0 793.0 793.0 793.0


%k 0.2280 0.2162 0.2044 0.1925 0.1807 0.1689 0.1570
71 0.054 0.042 0.033 0.026 0.020 0.016 0.012

Stage ZZ X1 0.529 0.554 0.578 0.611 0.640 0.666 0.695


T2p 685.9 695.9 704.0 711.8 719.5 726.9 734.2
772P 0.1930 0.1868 0.1795 0.1718 0.1637 0.1551 0.1462
T2k 784.7 789.6 793.0 793.0 793.0 793.0 793.0
q2k 0.3101 0.2977 0.2849 0.2679 0.2507 0.2333 0.2157
72 0.133 0.101 0.078 0.057 0.042 0.030 0.022

Stage ZZZ A2 0.673 0.687 0.703 0.729 0.750 0.768 0.787


T3P 686.8 695.4 703.8 714.4 723.8 732.2 740.0
773P 0.2612 0.2535 0.2456 0.2359 0.2249 0.2128 0.1997
T3k 758.6 763.1 767.8 773.7 780.1 787.1 793.0
r)3k 0.3462 0.3337 0.3213 0.3061 0.2916 0.2777 0.2625
73 0.242 0.185 0.141 0.101 0.075 0.057 0.042
Stage IV X3 0.841 0.849 0.856 0.865 0.871 0.874 0.880
T4P 684.3 691.9 699.4 708.9 717.2 724.8 733.2
QP 0.3013 0.2929 0.2844 0.2735 0.2623 0.2508 0.2388
T4k 755.1 759.7 764.4 770.3 776.2 782.0 788.4
?4k 0.3867 0.3748 0.3628 0.3477 0.3335 0.3200 0.3055
74 0.578 0.449 0.348 0.253 0.189 0.144 0.107
S 3575 4634 5997 8235 11069 14637 19717

TABLE 5

An optimal one-stage converter according to the Temkin-Pyzhev equation

TlP 615.0 625.0 635.0 645.0 655.0 665.0


Tlk 793.0 793.0 793.0 793.0 793.0 793.0
qlk 0.2754 0.2636 0.2517 0.2399 0.2280 0.2162
71 0.291 0.235 0.192 0.157 0.128 0.103
s 12355 15288 18742 22943 28215 34792

TABLE 6
An optimal two-stage converter according to the Temkin-Pyzhev equation

TlP 615.0 625.0 635.0 645.0 655.0 665.0


Stage Z Tn 793.0 793.0 793.0 793.0 793.0 793.0
qlk 0.2754 0.2636 0.2617 0.2399 0.2280 0.2162
71 0.209 0.172 0.144 0.120 0.099 0.081
Stage ZZ Al 0.717 0.732 0.750 0.763 0.775 0.786
T2P 651.4 659.2 668.0 674.4 680.3 685.8
q2P 0.2157 0.2104 0.2050 0.1983 0.1912 0.1837
T2k 778.9 782.3 786.2 789.4 792.6 793.0
q2k 0.3666 0.3560 0.3449 0.3345 0.3241 0.3106
72 0.458 0.388 0.324 0.284 0.246 0.202
8 5395 6425 7688 8912 10430 12695

modified Benton equation [18] . A Hewlett- optimization which required about 10 min for
Packard 9830 computer was used for the a four-bed quench converter.
163

TABLE 7

An optimal three-stage converter according to the Temkin-Pyzhev equation

T,, 635.0 645.0 655.0 665.0 675.0 685.0 695.0 705.0 715.0
Stage I Tlk 793.0 793.0 793.0 793.0 793.0 793.0 793.0 793.0 793.0
%k 0.2517 0.2399 0.2280 0.2162 0.2044 0.1925 0.1807 0.1689 0.1570
71 0.112 0.094 0.078 0.064 0.053 0.044 0.036 0.030 0.024

Stage II X1 0.582 0.596 0.608 0.623 0.637 0.653 0.658 0.675 0.687
Tb 668.0 674.4 680.3 685.8 690.8 698.4 702.3 707.2 711.5
b 0.2050 0.1983 0.1912 0.1837 0.1758 0.1684 0.1597 0.1510 0.1420
T2k 777.5 779.9 782.9 784.8 786.6 789.4 792.5 793.0 793.0
q2k 0.3346 0.3231 0.3126 0.3009 0.2893 0.2761 0.2665 0.2525 0.2385
72 0.206 0.177 0.153 0.134 0.118 0.100 0.090 0.076 0.065
Stage III A2 0.776 0.782 0.784 0.793 0.801 0.805 0.804 0.814 0.821
Tsp 669.0 673.4 677.2 683.1 688.2 692.7 694.8 700.2 703.5
Q3P 0.2742 0.2666 0.2592 0.2521 0.2445 0.2349 0.2270 0.2177 0.2074
T3k 766.9 770.3 773.5 776.9 779.6 783.0 785.5 789.0 792.5
q3k 0.3900 0.3812 0.373 0.3631 0.3527 0.3418 0.3343 0.3228 0.3128
73 0.422 0.376 0.340 0.296 0.259 0.229 0.215 0.190 0.171

s 4868 5570 6307 7283 8358 9645 10558 12176 13808

TABLE 8

An optimal four-stage converter according to Temkin-Pyzhev equation

TIP 655.0 665.0 675.0 685.0 695.0 705.0 715.0 725.0 735.0

Stage z Tlk 793.0 793.0 793.0 793.0 793.0 793.0 793.0 793.0 793.0
qlk 0.2280 0.2162 0.2044 0.1925 0.1807 0.1689 0.1570 0.1452 0.1336
71 0.064 0.053 0.044 0.037 0.030 0.025 0.020 0.017 0.013

Stage II Al 0.498 0.513 0.526 0.542 0.547 0.562 0.571 0.582 0.591
T2p 680.3 685.8 690.8 698.4 702.3 707.2 711.5 716.3 718.8
V2P 0.1912 0.1837 0.1758 0.1684 0.1597 0.1510 0.1420 0.1329 0.1232
T2k 782.9 784.8 786.6 789.4 792.5 793.0 793.0 793.0 793.0
r)2k 0.3126 0.3009 0.2893 0.2761 0.2665 0.2525 0.2385 0.2236 0.2110
72 0.126 0.110 0.098 0.083 0.074 0.064 0.054 0.046 0.040

Stage III h2 0.643 0.645 0.661 0.669 0.668 0.678 0.682 0.687 0.694
T3P 677.2 683.1 688.2 692.7 694.8 700.2 703.5 706.8 709.5
‘)73P 0.2592 0.2521 0.2445 0.2349 0.2270 0.2177 0.2074 0.1963 0.1866
T3k 763.7 766.9 770.2 773.3 774.5 776.1 777.6 779.9 781.7
q3k 0.3615 0.3513 0.3415 0.3303 0.3213 0.3075 0.2951 0.2827 0.2721
73 0.221 0.191 0.168 0.148 0.134 0.113 0.099 0.088 0.079
Stage IV A3 0.820 0.824 0.826 0.831 0.831 0.833 0.830 0.830 0.834
T4p 679.1 683.5 687.2 691.9 693.1 695.3 695.4 697.3 700.3
94P 0.3082 0.3001 0.2934 0.2853 0.2779 0.2669 0.2560 0.2458 0.2375
T4k 759.6 762.8 765.9 769.5 771.8 775.2 778.0 781.1 783.8
q4k 0.4035 0.3947 0.3865 0.3771 0.3711 0.3615 0.3538 0.3449 0.3363
74 0.395 0.351 0.317 0.281 0.266 0.244 0.234 0.218 0.201
S 4471 5100 5738 6566 7129 8088 8831 9783 10772

The results of the computations for one-, 5. CONCLUSIONS


two-, three- and four-bed ammonia converters
are listed in Tables 1 - 8 and are illustrated in (1) The general conditions for an optimal
Figs. 3 - 5. Moreover, a nomogram has been N-bed quench converter have been derived on
obtained which facilitates the choice of the the basis of an ideal plug flow reactor model.
best converter inlet temperature for given Two schemes are presented, one for un-
technical conditions. constrained optimization and one for
164

optimization subject to N ‘-- 1 inequality Greek symbols


constraints on the temperature of each bed. 7) conversion of the reference component
(2) An algorithm which permits a rapid Ai fraction of the original mass flow rate
solution of the problem is also given. passing through the ith bed
(3) The general optimization requirements vi stoichiometric coefficient of
have been used for designing ammonia quench component i
converters. Computations have been carried 7 r’/q,, reduced space-time coordinate
out for both the Temkin-Pyzhev kinetic
equation and the modified Benton equation. Subscripts and superscripts
(4) The optimal ammonia converter designs 0 reference state (V = 0)
obtained with both equations do not differ 0 cooling stream
significantly. However, the Temkin-Pyzhev i bed index
equation allows faster computations owing to P bed inlet
a simpler form of the derivative a (--r)/a T. k bed outlet
However, the Benton equation is more conve- N number of beds
nient because of the constant activation * refers to the highest feasible temperature
energy. inabed
I
(5) The figures show that the differences in optimal solution
optimal outlet conversion between the three-
and four&age converters are small. Therefore
the use of a four-bed cold-shot ammonia
converter does not seem to be justified. REFERENCES
(6) Computations based on the reduced
1 K. G. Denbigh, Trans. Faraday Sot., 40 (1954)
(rik = ri+l,p ; i = 1,2,. ..,N - 1) scheme of
352.
eqns. (36) - (39) revealed that the temp-
2 A. Annable, Chem. Eng. Sci., 1 (1952) 145.
erature in consecutive beds exceeded the max- 3 P. H. Calderbank, Chem. Eng. Prog., 49 (1953)
imum permissible value T* = 520 “C consider- 585.
ably (see, for example, Fig. 3). This is in good 4 P. Mars and D. W. van Krevelen, Chem. Eng. Sci.,
agreement with the work of Kubee et al. [ 141. 3 (1954) 41.
5 C. van Heerden, Chem. Eng. Sci., 8 (1958) 133.
6 K. R. Westerterp, Chem. Eng. Sci., 73 (1961) 69.
7 K. R. Westerterp and W. Beek, Zngenieur (The
NOMENCLATURE Hague), 73 (1961) 15.
8 F. Horn and L. Kiichler, Chem. Zng. Tech., 31
(1951) 1.
AT,,, adiabatic temperature rise 9 K. Y. Lee and R. Aris, Znd. Eng. Chem. Process
ith component of reaction Des. Deu., 2 (1963) 301.
c concentration of the reference 10 K. Y. Lee and R. Aris,Znd. Eng. Chem. Process
component Des. Dev., 2 (1963) 307.
mean specific heat of the reacting 11 J. P. Melange and L. M. Vincent, Znd. Eng. Chem.
CP
Process Des. Dev., I1 (1972) 465.
mixture 12 L. G. Hellinckx and P. V. van Rompay, Znd. Eng.
F set of feasible solutions Chem. Process Des. Dev., 7 (1968) 595.
G mass flow rate 13 F. Horn, 2. Elektrochem., 65 (1961) 295.
h T - uq , variable characterizing an 14 J. Kube6, J. Burianove and Z. Burianec, Znt. Chem.
adiabatic bed Eng., 14 (4)(1974) 629.
15 W. I. Zangwill, Nonlinear Programming: A
--r reaction rate of the reference component
Unified Approach, Prentice-Hail, Englewood Cliffs,
S space velocity New Jersey, 1969.
vi volume of the ith bed 16 E. Isaacson, Analysis of Numerical Methods,
G volumetric flow rate at the reactor inlet New York, 1966.
W mass fraction of the reference 17 A. Nielsen, An Investigation on Promoted Iron
Catalysts for the Synthesis of Ammonia, Jul.
component
GjelIemps Vorlag Cop., 1956.
X vector of the optimization problem 18 A. Burghardt and M. Paiica, Chem. Stosow.,
YIP initial mole fraction of inerts VZZ2B (1970) 205.

Вам также может понравиться