Full Paper

Design Optimization of Crude Oil Distillation

By Jlong W. Seo, Min Oh,* and Tae H. Lee

In this work, the optimal design of a crude oil distillation unit was made. Since the method used here is based on the general
optimization framework, it is possible to design the process systematically. The results of this study gave better performances
than that of the existing process. The process in this study is currently in operation and has the capacity of 150,000 BPSD. Feed
locations, heat duties of pumparounds and operating conditions of the preheat train are selected as variables, and the energy
consumption, operating cost and annual cost are formulated as objective functions. The optimal feed locations of both the main
column and stabilizer are obtained by solving rigorous models and mixed integer nonlinear programming. Considering the
results of the optimal feed tray, the cost optimization is carried over to obtain the optimal annual cost, which considers both the
capital and operating cost, which is decreased from US$10,649,000/year to US$9,185,230/year.

1 Introduction development of process system engineering, extensive re-

search efforts have been made in the areas of simulation and
Notwithstanding the long history, the crude oil distillation optimal design.
unit (CDU) is probably the most widely used separation Generally speaking, optimization is one mathematical
process in chemical industry. Crude oil distillation operations methodology which can be defined as an activity to find the
are carried out in large equipment that normally consumes best solution to the process of interest within constraints. This
considerable energy. The major factor in making a crude oil methodology provides a process engineer with a powerful
distillation unit so complex is the multitude of products means to execute a process design task in the optimal sense.
ranging from liquefied gases for home cooking to asphalt for For the purpose of optimization, the objective function is
roofing and road construction [1]. In this respect, the introduced and it serves as a quantitative indicator of
optimization of design and operation is of great significance ªgoodnessº for a particular solution [4]. According to Edgar
and usually leads to complex equipment. Nonetheless, due to and Himmelblau [5], optimization problems for steady-state
its intrinsic complexities, the traditional process design of a distillation can generally be classified into categories with
crude oil distillation column is usually practiced based on increasing orders of complexity: (a) determine the optimal
heuristic approaches such as the engineers' experiences, operating conditions for an existing column, (b) determine not
design guidelines, and case studies from steady-state simula- only the values of the operating conditions, but also the
tions. The situation becomes even more complicated by the number of stages required for the separation, and (c)
fact that feedstock (crude oil) has an exceedingly complex determine the number of stages and the optimal locations
composition, consisting of many different types of hydro- for the feeds and side stream withdrawals.
carbons as well as inorganic and other organic compounds. One of the first efforts to solve design problems using an
The literature concerning the design and retrofit of CDU are optimization procedure was made by Srygley and Holland [6]
relatively scarce and generally fairly old. As Liebmann [2] using Hooke and Jeeves's [7] search procedure. The proposed
discussed, one of the reasons for its scarcity is that much of the procedure was limited to the determination of the minimum
knowledge in this area is kept in-house by the large refining number of stages required to give a specified separation at a
and contracting companies. It is argued, at this moment, that specified reflux ratio. Sargent and Gaminibandara [8]
the methodology based on the heuristic approach by no means considered the more general problem of the optimum
guarantees ªoptimal designº particularly in the sense of configuration of columns needed to effect a specified
energy and cost saving. separation. Ricker and Grens [9] proposed an iterative
Douglas [3] defined the process and plant design as a procedure based on the Naphtali and Sandholm [10]
creative activity whereby ideas were generated and then formulation for the minimization of the number of stages
translated into equipment and processes for producing new required to effect a specified separation at a given reflux ratio.
materials or for significantly upgrading the value of existing Al-Haj-Ali and Holland [11] determined the optimal operat-
materials. Current development of computer technology, both ing conditions required for an existing distillation column or
in hardware and software systems, permits the application of the design conditions for a new column. Fisher et al. [12]
various mathematical methods to perform a process design in proposed a general, shortcut solution based upon the
a systematic way. Furthermore, on account of the rapid simplified design and economic models with relatively few
outlet compositions fixed by marketing or environmental
± considerations. Kumar and Lucia [13] modified the successive
[*] J. W. Seo and T. H. Lee, Department of Chemical Engineering, Yonsei quadratic programming (SQP) approach of Wilson [14], Han
University, Shinchon-dong, Seodaemoon-gu, Seoul 120-749, Korea; [15], and Powell [16] in such a way that the Hessian matrix of
M. Oh, Department of Chemical Engineering, Taejon National Uni-
versity of Technology, Samsung-dong, Dong-gu, Taejon 300-717, Korea. the Lagrangian function correctly reflected the underlying
E-mail:minoh@hyunam.tnut.ac.kr thermodynamics of the separation processes.

Chem. Eng. Technol. 23 (2000) 2, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0157
0930-7516/00/0202-0157$$17.50+.50/0
17.50+.50/0 157
 Full Paper

Growing interest in energy conservation constitutes a

challenge to perform the optimization of operating conditions
of CDU. In the operation of a CDU, the quality of the products
has to be maintained to satisfy the specifications, and process
variables are of given ranges despite the change of operating
conditions, which allow the savings of operating and capital
costs. The importance of optimization of operating conditions
can also be justified by the fact that the results can be utilized for
existing plants even with revamping or retrofitting. Further-
more, the location of the feed tray has an influence on both the
mass and the energy balances of the column. Consequently, it
contributes to energy saving, on the one hand, and reducing the
capital costs, on the other. It should be, however, noted that the
location of the feed tray of the crude oil distillation column not
only involveswitheconomics butis also concerned withproduct
specifications. In an attempt to reduce its energy consumption,
extensive studies have been carried out since the 1970s and a
considerable number of novel distillation columns have been
proposed [17±20]. Hsie and McAvoy [21] performed simulation
and control of a crude column based on rigorous mathematical
models. However, all the works mentioned above did not
consider the heat exchanger network (HEN) of a preheat train
of a crude column, which has a great impact on energy saving.
In this work, the design of a crude oil distillation unit in
terms of annual costs was performed. Optimization was Figure 1. Schematic diagram of crude oil distillation unit.
involved both in the design of the equipment and in the
determination of the feed location and optimal operating vaporized to produce the vapor flow into the column. To make
conditions to minimize operating and capital costs. The this heating process as thermally efficient as possible, the
process being considered here is currently in operation and structure and operating conditions of the feed preheat train
has the capacity of 150,000 BPSD. The mathematical should be carefully considered. Since a large amount of heat
description of these problems is associated with a mixed set and a considerable number of streams are involved in the
of integer and continuous variables, which eventually leads to crude preheat train system (see Fig. 2), the structure is fairly
a mixed integer nonlinear programming (MINLP) problem. complex compared with other preheat systems. As a
New problem formulations and solution approaches were consequence, it is critical to design an efficient preheat train
based on the integration of classical frameworks of analysis in the pursuit of the energy saving distillation unit.
(e.g., control theory, optimization theory, numerical analysis).

2 Process Description

ThemajorfunctionalityoftheCDUistoseparatecrudeoilinto
a series of petroleum fractions characterized by boiling point
ranges. The plant studied in this paper is currently being
commercially operated in the South East Asia region and the
capacity of an atmospheric column is 150,000 BPSD. The crude Figure 2. Heat exchanger network.
unit that we examine in this study is composed of two main units:
an atmospheric distillation unit (ADU) and vacuum distillation
unit(VDU).TheADUisrelativelycomplexandconsistsofapre-
heatertrain,acrudechargeheater,acrudeoildistillationcolumn, 2.2 Atmospheric Distillation Column (ADC)
light naphtha (LN) stabilizer as well as 2 side-stripping columns.
The crude column separates the feed into an overhead light
naphtha cut, and side-draw cuts of heavy naphtha, kerosene,
2.1 Preheat Train and two gas oils, light and heavy. The boiling range and
quantity of these fractions are controlled by the column
The crude oil is heated by preheat train and a crude charge operation and can be varied to suit the processing require-
heater up to the point where a sufficient amount of the feed is ments of the refinery. The four side-cuts from the column pass

158 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0158 $ 17.50+.50/0 Chem. Eng. Technol. 23 (2000) 2
 Full Paper

to their respective strippers. Superheated stripping steam is zone, which is depressurized at 50 mmHg. On entering the
injected to remove any light components that may be vacuum column, the ATB feed is flashed so that the total vapor
absorbed in the heavy oil. flow up the column is sufficient enough to produce the
The major indicators of correct operation of the crude fractions taken from the column as side-cuts. There is an
column are the column temperature and pressure profile. A additional amount of vaporization known as overflash that
change in the temperature profile is made in two way,: i.e., heat will partly vaporize liquid on the lower trays. The vacuum
input through the crude heater and heat removal through the column separates the feed into side-draw cuts of LLVGO,
pumparound streams and product draw-off rates. The column LVGO, and HVGO. The remainder of the feed components
pressure is controlled at the crude column naphtha drum and leave the column as a vacuum residue (VR), after being steam
is set during the design at an optimum value based on stripped in the bottom trays. Superheated stripping steam is
condensing equilibrium conditions, cooling economics, and injected to remove any light components that may be
hydraulic requirements. Pressure is not a process variable absorbed in the residue.
from an operational viewpoint. Any significant change in the
column pressure profile will indicate instability in the tower,
i.e., dry or flooded trays. 3 Formulation of the Optimization Problem

A wide variety of optimization problems have a very similar

2.3 Pumparound structure. Indeed, it is this similarity that has enabled the
recent progress in optimization techniques. The optimization
The common way of attaining intermediate heat recovery is problem consists of an objective function to be maximized (or
known as pumparound reflux system [22]. As shown in Fig. 1 minimized) and constraints from design and operation
(see HN PA, Kero PA, LGO PA, and HGO PA), liquid is conditions. From the physical limitations of the process, some
withdrawn from the tray above the lower draw tray, cooled more constraints are usually enforced. For example, product
externally, and returned to a tray further up in the tower but rates, such as naphtha, kerosene, LGO, and HGO, must be
below the upper draw. The removal of heat in the side cooler kept within a certain range for specification. In this chapter,
results in a smaller requirement for the overhead reflux. This we formulate two types of optimization problems to obtain
technique is generally employed at both a low and an optimal feed tray and optimal operating conditions to
intermediate point in the column, which makes heat available minimize annual costs.
at two different temperature levels for exchange against crude
oil. So if we increase the heat recovery in the pumparound as
much as possible within operating constraints, we could save 3.1 Optimal Feed Tray
heat duty for crude oil, which in turn can reduce the heat load
of the overhead condensation unit and its size. An additional Until recently the location of the feed was usually
advantage of a pumparound in a crude column is to reduce the determined from the engineer's experiences or intuition.
diameter of the column by controlling the vapor and liquid However, since it has a great influence on both the mass and
traffic in particular zones. Thus, it is very useful for energy the energy balances of the column, determination of its
recovery to find the optimal feed tray location and the ªexactº location is of significance. The processing unit
pumparound conditions. In the processing scheme, each of the considered here consists of the main column (35 trays), light
pumparound reflux streams is controlled for both flow and naphtha stabilizer (30 trays), overhead condenser, and
temperature at the point where it reenters the column. reboiler. In addition, four pumparound and four side products
are included. To complete the formulation of the optimization
problem, objective function and some more constraints are
2.4 Light Naphtha Stabilizer considered.

The total overhead from the crude column contains a range

of components from light gases to light naphtha. The naphtha 3.1.1 Objective Function
stabilizer removes C2 and lighter gases for fuel gas and
separates the compressed liquids into an LPG stream and a In consideration of the highly energy consuming character-
light naphtha stream which are routed to their respective istics of a crude column, it would be sensible to choose an
storage system. energy recovery in the column as an objective function. The
heat recovery can be achieved with a pumparound and can
also be used to increase the temperature of the crude oil.
2.5 Vacuum Distillation Unit (VDU) Maximum heat recovery with pumparound reflux systems
reduces the heat duty of the condenser and the size of the
The atmospheric tower bottom (ATB) is heated again by overhead systems. Our main concern in this problem is the
the vacuum charge heater and fed to the vacuum column flash determination of the optimal feed location and the heat duties

Chem. Eng. Technol. 23 (2000) 2, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0159 $ 17.50+.50/0 159
 Full Paper

of the pumparound. For the purpose of optimization, we The column configuration and operating conditions re-
define the objective function in terms of heat duties of the quired to solve the optimization problem mentioned above
pumparound and condenser as follow1): are presented in Tabs. 1 and 2. The components are made up of
4 pure components and pseudo-components, numbering 12 in
P
minimize Qk ÿ Qcon (1) total. SRK and GS models are employed to calculate the
kˆ1 thermodynamic properties of the gases and liquids.
For the LN stabilizer the objective function is considered as
follows:
Table 1. Column configuration of the base case.

minimize Qreb (2)

3.1.2 Feed Tray Constraints

The candidate locations for feed are assumed to be 2  j 

TR ± 1 and the normal practice in industry indicates that the
number of feed stages is one. Each split rate should not exceed
the feed rate and the sum of all the split feed rates must be
equal to it.

fj  Fzj j 2 FT (3)

P
fi ˆ F (4)
i2TR

P Table 2. Operating conditions for the base case.

zi ˆ 1 (5)
i2TR

Eqs. (3ÿ5) mean only one tray is selected for feed.

Combined with other constraints as well as the objective
function, this leads to MINLP.

3.1.3 Pumparound Constraints

The flow rate of a side draw should not exceed the rate going
down to the stage, to prevent internal rates reaching zero.

Wj  Ljÿ1 8j 2 SP (6)

Furthermore, additional constraints such as upper- or lower-

bounds of variables, product rates, purity, and recovery are
imposed to give the physical meaning of the process.

3.1.4 Pressure Profile

When we apply a linear relationship between the top and

bottom of the column:

Pjÿ1 ÿ 2Pj ‡ Pj‡1 ˆ 0; 3  j  Nmax ÿ 2 (7)

where PNmax , PNmax ÿ1 , P2 , and P1 are of given values.

±
1) List of symbols used at the end of the paper.

160 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0160 $ 17.50+.50/0 Chem. Eng. Technol. 23 (2000) 2
 Full Paper

3.2 Optimal Annual Cost 4.1 Optimal Feed Tray

In this section, the optimization of annual cost comprising The solutions of optimization of the feed location are shown
the capital and operation costs is considered. The entire CDU in Tab. 3 and compared with other simulation results to verify
process including preheat train, the structure of which is the optimization. The optimal feed location obtained is tray
illustrated in Fig. 2, is considered here. The number of 34. But, in the base case of the original design, the feed
theoretical trays for vacuum distillation column is 12 and the location was 33 and the value of the objective function was
column includes 4 pumparounds. ATB is heated up to 378 C worse than the optimal case. For the cases of feed tray 32 and
by the heater and introduced on tray 11. Steam enters on tray feed tray 31, the values of the objective function were not
12. Capital costs are estimated [23,24] as the sum of the costs of improved. From this, it is clear that the value of the objective
the columns in the ADU and VDU (main columns, side function in the optimal case is the best.
strippers, and stabilizer), reboiler, condensers, and heat
exchangers. The costs for fuel, cooling water, and steam are Table 3. Optimal solutions for an atmospheric distillation unit.
considered as operating costs.

Capital cost
cost of column (main columns in ADU and VDU, side
strippers, stabilizer):

Ccol ˆ exp…acol  ln…Dcol  bcol † ‡ ccol †  NTRAY (8)

cost of reboiler:
Qreb
Creb ˆ areb  ‡ creb (9)
breb
cost of condenser:
The traffic in the internal column is represented in Fig. 3.
Qcond The flow rate in the optimal case is lower than that in the base
Ccond ˆ acond  ‡ ccond (10)
bcond case. Watkins (1969) reported that the flow rate profiles of
cost of heat exchanger (shell and tube): both liquid and vapor were reduced by the same percentage
when pumparound circuits were introduced. This result shows
Chx ˆ fd fm fp Cb that it is possible to reduce the size of the column. With a
2 reduction of the column size, a subsequent reduction in the
Cb ˆ exp…8:821 ÿ 0:30863…ln A† ‡ 0:0681…ln A† †
fd ˆ exp…ÿ1:1156 ‡ 0:0906…ln A††: type investment cost sfollows.
fm ˆ 0:8603 ‡ 0:23296…ln A†: material
fp ˆ 0:7771 ‡ 0:04981…ln A†: pressure (11)

Operating cost:
cost of cooling water: Ccw = 0.25 $/1000 gal
cost of fuel: Cfuel = 3.20 $/106 Btu
cost of steam: Cst = 4.31 $/106 Btu
Annual cost = capital cost  Crf + operating cost

…1‡i†k
Crf ˆ (12)
…1‡i†k ÿ1
where, interest rate of return, i, is assumed to be 9.5 % and
project life, k, is set to 25 years.

4 Results and Discussion Figure 3. Internal vapor and liquid flow in atmospheric distillation column.

To solve the proposed optimization problems, the GAMS Fig. 4 shows the column diameters of the base case and the
modeling package is used, which has MINLP solver DICOPT. optimized case, respectively. In the optimized case the
The optimal solutions were verified using the PRO/II steady- reduction of the column diameter is due to reduced internal
state simulator [25]. The optimal solutions are compared with flows in the column. Also the temperature profiles of the
other operating conditions column for both the base case and the optimized case are

Chem. Eng. Technol. 23 (2000) 2, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0161 $ 17.50+.50/0 161
 Full Paper

Figure 4. Diameter of atmospheric distillation column at each tray. Figure 6. Internal vapor and liquid flow in light naphtha stabilizer.

Figure 5. Temperature profiles of atmospheric distillation column. Figure 7. Diameter of light naphtha stabilizer at each tray.

drawn in Fig. 5, which are identical to each other. This means

that there are no changes in product specifications.
The optimization results of the stabilizer are summarized in 4.2 Optimal Annual Costs
Tab. 4 and compared with the simulation results of the base
case and two more cases. The optimal feed location for the Tab. 5 shows the result of optimization to minimize the
stabilizer is tray 14. From this comparison, the optimal design annual costs. Compared with the base case, the result
case is verified as the best result. decreased the annual costs from 10,649,000 to 9,185,230
US$/y. Temperature profile of the preheat train is shown in
Table 4. Optimal solutions for light naphtha stabilizer.
Fig. 8. The crude oil, initially at room temperature, is heated up
by passing through each heat exchanger. As one of the
constraints imposed is that the crude oil may not vaporize, the
temperature leaving the last heat exchanger is under the
boiling point. In the optimal case, the pressure of the preflash
pump was reduced to near the boiling point in the last heat
exchanger. Therefore, the operating cost of the preflash pump
The traffic in the stabilizer is represented in Fig. 6 which can be reduced under this operating condition. Fig. 9 shows the
shows lower flow rates in the optimal case than in the base heat composite curve of the cold stream, i.e., crude oil. Due to
case. This result is similar to those in the main column. the optimization procedure, energy conservation can be
Therefore, it is possible to reduce the size of the stabilizer realized, and the duty of the heater can be reduced by
(Fig. 7), which also reduces the investment costs. 20,000,000 kcal/h.

162 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0162 $ 17.50+.50/0 Chem. Eng. Technol. 23 (2000) 2
 Full Paper

Table 5. Results for the annual cost optimization.

Figure 9. Heat composite curve of the heat exchanger network.

l The performance of the existing CDU could be improved by

the optimization technique using MINLP.
l In the optimal design of the CDU, the significant decision
variables were found to be feed locations, heat duties of
pumparounds, and operating conditions of the preheat
train, which were determined to minimize annual costs.
l The energy recovery in pumparounds and preheat train led
to the optimal design of the CDU, resulting in the fact that
the annual costs of the CDU could be reduced to 86 % with a
savings of 20,000,000 kcal/h.
l The annual cost of the ADU was also reduced through
alternative operating conditions obtained by the optimiza-
tion of the flow rate of stripping steam and the yield.

Acknowledgments

The authors wish to thank STEPI, Korea, and LG

Engineering, Co., Ltd., Korea for the financial support for
this work.
Received: April 19, 1999 [CET 1104]

Symbols used

A [±] costing constant

Figure 8. Temperature profiles in the heat exchanger network. a, b, c [±] costing constant
D [±] costing constant
C [US$/y] cost
5 Concluding Remarks Crf [±] capital recovery factor
fd [±] type factor of heat exchanger
The optimal design of the CDU, which is currently in fj [kmol/h] molar flow rate of feed to stage j
operation, was studied using rigorous column models. fm [±] material factor of heat exchanger
Employing steady-state and dynamic simulations for the fp [±] pressure factor of heat exchanger
process, the optimal column configurations and the optimal F [kmol/h] feed molar flow rate
operating conditions were obtained by solving MINLP and FT [±] set of all possible feed trays
NLP. And finally, the results were verified by comparison with i [±] interest rate of return
the real process. The conclusions of this work are as follows: k [±] project life

Chem. Eng. Technol. 23 (2000) 2, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0163 $ 17.50+.50/0 163
 Full Paper

Lj [±] liquid molar flow rate leaving stage j References

N [±] number of trays
P [Pa] pressure [1] Hengstebeck, R. J., Petroleum Processing: Principles and Applications,
McGraw-Hill, New York 1959.
Q [kcal/h] heat duty
[2] Liebmann, K.; Dhole, V. R., Trans IchemE 76 (1998) Part A, pp. 335±347.
Qk [kcal/h] k th pumparound heat duty [3] Douglas, J. M., Conceptual Design of Chemical Processes. McGraw-Hill,
SP [±] set of side product trays New York 1988.
[4] Biegler, L. T.; Grossmann E. I.; Westerberg A. W., Systematic Methods of
TR [±] set of stages Chemical Process Design. Prentice-Hall, pp. 748±770, 1997.
Wj [±] side withdrawal molar flow rate on [5] Edgar, T. F.; Himmelblau D. M., Optimization of Chemical Processes.
stage j McGraw-Hill, New York 1988.
[6] Srygley, J. M.; Holland C. D., AIChE J 11 (1965) pp. 695±701.
zj [±] the binary variable associated with the [7] Hook, R.; Jeeves T. A., J. Assoc. Comp. Math. 8 (1961) pp. 212±229.
selection of stage j (1, if the feed enters [8] Sargent, R. W. H.; Gaminibandara K., Optimum Design of Plate
Distillation Columns. Optimization in Action, (L. D. W. Dixon, Ed.)
on stage j)
Academic Press, New York 1976.
[9] Ricker, N.L.; Grens E. A., AIChE J. 20 (1974) pp. 238±244. Naphtali, L.M.;
Sandholm. AIChE J. 17 (1971) pp. 148±153.
[10] Al-Haj-Ali, N.S.; Holland C. D., Hydrocarbon Processing, July (1979) pp.
Subscripts 165-175. Fisher, W.R.; Doherty M. F.; Douglas J. M., AIChE J. 31 (1985)
pp.1538±1547.
i index of stages [11] Kumar, A.; Lucia A., IMACS J. Appl. Numer. Math. 3 (1987) pp.409±425.
Wilson, R.B., Ph.D. Thesis, Harvard University, 1963.
j index of stages [12] Han S.P., Math. Program 11 (1976) pp.263±282.
k number of pumparound [13] Powell, M. J. D., A Fast Algorithm for Nonlinearly Constrained
Optimization Problems, in: Lecture Notes in Mathematics (G. A. Watson,
col column
Ed.), Springer-Verlag, Berlin 1978.
con condenser [14] Badhwar, R.K., Chem. Eng. Prog 66 (1970) March, pp. 56±61.
cw cooling water [15] Watkins, R. N., Petroleum Refinery Distillation, Gulf Publishing Company,
Houston 1973.
max maximum [16] Bagajewicz, M. J., Energy Savings Horizons for the Retrofit of Chemical
reb reboiler Processes. Application to Crude Fractionation Units. AIChE Spring
st steam Meeting, Mar., 1997a.
[17] Bagajewicz, M. J., On the Design Flexibility of Atomospheric Crude
Fractionation Units. AIChE Spring Meeting, Mar., 1997b.
[18] Hsie, W. H. L.; McAvoy T. J., Modeling, Simulation, and Control of Crude
Tower. AIChE annual meeting, Houston, April, 1990.
[19] Watkins, R.N., Hydrocarbon Processing, December (1969) pp. 93±106.
[20] Peters M. S.; Timmerhaus K. D., Plant Design and Economics for
Chemical Engineers. 4th ed., McGraw-Hill, New York 1991.
[21] Walas, Stanley M., Chemical Process Equipment, Butterworth-Heine-
mann, pp. 663±669, 1990.
[22] Simulation Science. PRO/II: Keyword Input Manual. March, 1996.

_______________________

164 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2000 0930-7516/00/0202-0164 $ 17.50+.50/0 Chem. Eng. Technol. 23 (2000) 2