Академический Документы
Профессиональный Документы
Культура Документы
SUBSCRIPTS
Combined bubble and foam fractionation operates successfully, with good stripping and enriching ratios obtainable.
However. the incremental advantage of the elongated bubble
section is small, a t least for the present system.
Increasing the reflux ratio increases the degree of enrichment. Increasing the rate of gas flow improves the degree of
stripping but decreases the degree of enrichment. Location of
the feed inlet can be important under certain circumstances.
Nomenclature
=
=
=
tops
feed
bottoms
literature Cited
PROCESS OPTIMIZATION BY
NONLINEAR PROGRAMMING
C A R L O S W .
D I
B E L L A A N D W I L L I A M
F. S T E V E N S
A recently proposed nonlinear programming technique i s used to optimize a continuous chemical process.
The problem i s to extremize a nonlinear criterion function, F(x,), subject to nonlinear constraints, G,(x,) = 0,
1 5 j 2 n; 1 5 i 5 rn
n. The method consists of finding a feasible solution, and then applying the
linear programming algorithm to the linearized mathematical model around successive feasible solutions
until an optimum i s reached. The example solved i s of sufficient complexity to demonstrate the value of
this method for chemical process optimization.
<
16
The impetus for the development of sequential search techniques in chemical engineering design is largely due to the increased availability and reliabilit). of knowledge concerning
physical and chemical phenomena. Growing familiarity with
the capabilities of digital computers has increased their use for
optimization calculations. Reliable chemical and physical
data permit the construction of mathematical models which are
sufficiently trustworthy in many cases. High-speed computers
enable sequential search iterations to be carried out in seconds.
Although both the mathematical model and the computer routine are formidable tasks. they must be knit together by suitable optimization techniques. Fortunately, such techniques
are becoming increasingly available. having been developed
mainly by mathematicians working in the areas of operations research and management science.
Optimization is onl>-one step in the process of arriving a t an
engineering decision. If the optimization is to be done
analytically. a t least two preliminary steps are necessary.
First: the engineer must develop a mathematical model of his
system. incorporating all that is known about the process into
the resulting expressions. Second: he must select a measure of
effectiveness, Lvhich is to be minimized or maximized by means
of the optimization procedure. Both steps require an intimate knokvledge of the system under study and its relation to
the rest of the company or the economy as a whole.
Mathematical
G,(xl",. . . .x,")
Let
aF
X?)
. . . .x,)
(1)
Subject to
Gr
(XI,
Lj
x2,
....x n )
xj
0
j
5; C;
1, 2, . . . .rn
w,
(11)
(X,
X,")
AYj
(12)
(2)
1, 2, . . . .n
. . . bl")
(9)
ax5
and
F(X1,
(XI",.
L'l, a constant
F - F" =
(3)
W3Axj
1-1
Subject to
and
.x3
2 0
(4)
and
s>
Subject to :
n
]
=1
Griffith and Stewart ( 7 ) have presented the following technique for solving the nonlinear programming problem as given
in Equations 1, 2. 3. a n d 4 by using a linear programming
technique sequentially. T h e problem is first linearized in the
region about some initial approximation of x j by expansion as a
Taylor's series. ignoring terms of higher order than linear.
T h e linearized problem then becomes
Maximize
= F(Xl0... . . . X n " )
Subject to
G , (.xl0,. . . ..no)
i
L,
and
x,
5 u,
l,..m
j=l,..n
(7)
T h e design problem optimized in this report was first proposed by Williams and O t t o ( 3 ) . T h e Williams plant contains
many of the characteristics of a typical chemical design problem
a n d is realistic enough so that a steady-state optimization has
meaning.
VOL. 4
NO. 1
JANUARY
1965
17
T,
MCHANGER
REACTOR
Fp,*
DISTILLATION
HEAT
g-
CSTR
cooling
water
&
DECANTER
COLI"
Lc
I
I
I
Icwiing
+B
+B
+C
-+
-+
C reaction coefficient, k l
P
E reaction coefficient, k?
G reaction coefficient, ka
kt
where
F,,
- 0.1 FRE -
Fp = 0
3.
4.
5.
= AIexp(--B,,'T)
A1
Gn
= 5.9755
F.4
B1 =
A? =
B2 =
AS =
BS =
6.
680
Rankine
0 (26)
(27)
(28)
F , L
FR - F p - F ,
~
7.
+ F,
(21)
Steady-State Balances
8.
FA
+ FB -
FO
- Fp
- F,
(22)
F,
0
(30)
% return
600 V p
Variable
(31)
FA
FB
FD
Fc
FRA
FRB
FRC
FRE
V
FR
FRP
T
7;return
Initial
Guess
10.000
40,000
20.000
5,000
6,000
35,000
10.000
15.000
150
100.000
6,000
61 0
3 x 109
Improzed
Initial Guess
Optimal
Solution
11,540
31 -230
36.010
13.546
31,523
36 6g7
3,609
18,187
60.915
3,331
60.542
60
157,391
10,817
656
100
72 75
2.010
8.820
39.910
2.360
31.660
223
92.640
7 890
610
1 2 x 104
=18
Units
Lb./hr
Lb./hr.
Lb./hr.
Lb./hr.
Lb. /hr.
Lb. /hr.
Lb /hr.
Lb./hr,
cu. ft.
Lb./hr.
Lb./hr.
O Rankine
c
/C
Method of Solution
G,
Find a set of
Y]
(XI
. . . . XI?)
1, . . . .9
(32)
such that
9
Gt2 =
:=
(33)
,=l
NO. 1
JANUARY
1965
19
produced intermediate nonoptimal solution vectors a n d corresponding values of the objective function. I t would be of
great value in both the design and operation of a chemical
plant. In design, variations in uncontrollable parameters
such as raw material costs, product price, etc., could be considered a n d a design with complete flexibility produced. When
the plant is actually operating and a change in one or more of
the parameters is encountered, it is necessary to change only
those values in the program and then reoptimize.
T h e method presented is simply another variation on the
many organized sequential search techniques proposed recently. As such, it is susceptible to the same shortcomings of
other methods-namely,
more than one extremum, concave
constraint space, convex objective function, etc. Fortunately, in most real systems, these problems are nonexistent or can be handled easily by multiple starting points.
T h e advantages of the method lie in the speed of solution, the
relative ease of programming, and, most importantly, its ability
to handle nonlinear equality constraints. Although many
nonlinear programming methods have been proposed in recent years. very little investigation has been made of their
relatiLe computational efficiency or practical worth. T h e
present paper is a step in the desired direction, presenting a
new combination of two known methods and indicating the
practical use of the combined method.
Nomenclature
A,
B,
= Arrhenius reaction
FA
reaction
rate exponential factor for ith
reaction
flow rate of reactant A to reactor, Ib./hr.
literature Cited
RECEIVED
for review Xovember 26, 1963
ACCEPTED
February 3, 1964
F U N D A M E N T A L S T U D I E S ON A N O D I C
PROTECT ION
Carbon Steel in Sulfuric Acid
Z
F0 R 0 U L I S
General Engineertng Diviszon, Esso Research and Engineering Co., 'Madison. .V. J .
The anodic polarization behavior of carbon steel in sulfuric acid was investigated b y potentiostatic techniques in acid concentrations from 0.5 to 96.4% H2S04 and in the temperature range - 1 .O" to 88" C. An
increase in acid concentration up to approximately 50 to 60% leads to a displacement of the polarization
curve toward more cathodic potentials and higher currents. A further increase up to 96.4% HzS04 displaces
the polarization curve toward more cathodic potentials and lower currents. An increase in temperature displaces the polarization curve toward higher currents, and decreases the passive potential range. Particular emphasis was placed on interpretation from an engineering viewpoint of the anodic polarization
parameters such as critical current, passive current, etc., as affected b y changing acid concentration and temperature.
20
Experimental