Você está na página 1de 10

Computers and Chemical Engineering 27 (2003) 1143 /1152

www.elsevier.com/locate/compchemeng

MINLP synthesis of heat exchanger networks considering pressure


drop effects
S. Frausto-Hernandez a, V. Rico-Ramrez a,*, A. Jimenez-Gutierrez a,
S. Hernandez-Castro b
a

Departamento de Ingeniera Qumica, Instituto Tecnologico de Celaya, Ave. Tecnologico y Garca Cubas S/N, Celaya, Gto. C.P. 38010, Mexico
b
Facultad de Ciencias Qumicas, Universidad de Guanajuato, Col. Noria Alta S/N, Guanajuato, Gto. C.P. 36050, Mexico
Received 6 February 2003; accepted 7 February 2003

Abstract
Considerable research effort has been reported in both Pinch Technology and MINLP techniques for the synthesis of heat
exchanger networks. However, most of the design procedures assume constant stream heat transfer coefficients. A problem arises
because there is no guarantee that the values of the coefficients assumed during network synthesis are the same as those actually
achieved in detailed equipment design (Trans. IChemE 69 (1991) 445). It has been shown that network synthesis and detailed
exchanger design can be made consistent if network synthesis is based on allowable pressure drops rather than on assumed film
coefficients. Motivated by earlier applications in pinch technology, in this work we extend the simultaneous MINLP model for the
design of heat exchanger networks (Comput. Chem. Eng. 14 (1990) 1165) by removing the assumption of constant film heat transfer
coefficients and incorporating instead the effect of allowable pressure drop. Numerical results in two illustrative examples are shown
to demonstrate the potential benefits and scope of the proposed approach.
# 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Heat exchanger network; Pressure drop; MINLP synthesis

1. Introduction
Successful work has been done on heat exchanger
network synthesis (HEN) using both pinch technology
and MINLP techniques. Such developments have been
reviewed in a number of publications (Linhoff &
Flower, 1978; Linhoff & Ahmad, 1989; Papoulias &
Grossmann, 1983; Yee & Grossmann, 1990).
Recent approaches with both techniques have shown
to be capable of synthesizing near optimal networks for
real industrial problems. However, not much has been
done to ensure that the resulting networks are fully
consistent with what is finally achieved in terms of
industrial hardware. Although most of the current
synthesis techniques are based on the assumption of
constant film heat transfer coefficients, detailed heat

* Corresponding author. Tel.: /52-461-6117-575x156; fax: /52461-6117-744.


E-mail address: vicente@iqcelaya.itc.mx (V. Rico-Ramrez).

exchanger design is based on the satisfaction of three


major objectives (Polley & Panjeh Shahi, 1991):
1) Transfer of required heat duty.
2) Tube side pressure drop below a maximum allowed
value.
3) Shell side pressure drop below a maximum allowed
value.
Hence, since synthesis and detailed design are not
conducted on the same basis, there is no guarantee that
the values assumed for the heat transfer coefficients in
the synthesis stage are the same as those actually
achieved in equipment design.
Polley and Panjeh Shahi (1991) recognized that one
way of making consistent network synthesis and detailed exchanger design is to base network synthesis on
allowable stream pressure drop rather than on constant
film transfer coefficients. Since then, a number of
applications have been reported which suggest the
incorporation of pressure drop effects into the synthesis

0098-1354/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0098-1354(03)00042-5

1144

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

stage based on pinch technology (Serna, 1999). Further


applications of this approach have also been considered
for network retrofit based on MINLP techniques (Nie &
Zhu, 1999). In this paper, the simultaneous MINLP
model for heat exchanger design proposed by Yee and
Grossmann has been extended to incorporate pressure
drop considerations into the synthesis stage. So, the
network is synthesized based on allowable pressure drop
rather than on constant film heat transfer coefficients.
This paper is organized as follows. Section 2 briefly
describes the simultaneous MINLP technique for HEN
synthesis (Yee & Grossmann, 1990). Section 3 proposes
a strategy for incorporating pressure drop effects into
the MINLP approach. Section 4 summarizes the equations to be added to the MINLP formulation in the
extended version. Section 5 presents two illustrative
examples; it is shown that, although global optimum
cannot be guaranteed because of non-convexities in the
model, our results encourage further research on the
area. Finally, Section 6 concludes the paper.

2. MINLP synthesis of heat exchanger networks


The MINLP model proposed by Yee and Grossmann
(1990) is intended to provide an appropriate trade-off
between utility consumption, number of units and
exchanger areas. Such a model is based on a stagewise superstructure with the temperature driving forces
as optimization variables. The superstructure is constructed so that each cold stream can be potentially
matched with each hot stream in each of the stages (see
Fig. 1 for a two-stage superstructure). Also, alternative

parallel and series configurations are embedded as well


as possible rematching of the streams. The network
topology is determined by the vector of binary variables
whose values are obtained during the optimization
scheme.
The resulting MINLP formulation consists of the
following constraints: (i) overall heat balance for each
stream; (ii) heat balance at each stage; (iii) assignment of
superstructure inlet temperatures; (iv) feasibility of
temperatures; (v) hot and cold utility load; (vi) logical
constraints; (vii) calculation of approach temperatures;
and (viii) objective function. All the constraints but the
objective function are linear. The objective function
consists of the annual cost for the network. This annual
cost involves the combination of utility cost, the fixed
charges for the exchangers and the area cost of each
exchanger. The objective function is nonlinear because
of the heat exchanger area calculation. For the determination of the heat exchanger area, the expression
provided by Chen (1987) is used to approximate the
logarithmic mean of the temperature difference between
the streams of a given match. Also, the overall heat
transfer coefficients required by the area calculation are
obtained in terms of the constant (assumed as given)
film heat transfer coefficients of the streams.

3. Pressure drop considerations


The main issue here is that the heat exchanger
network to be synthesized must satisfy not only a
specified heat recovery target but also pressure drop
constraints. Hence, by removing the assumption of

Fig. 1. Two-stage network superstructure by Yee and Grossmann (1990).

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

constant heat transfer coefficients and adding information on stream pressure drops and physical properties,
the resulting network is made closer to industrial reality.
3.1. Pressure drop as a function of heat transfer
coefficients
Polley, Panjeh Shahi, and Jegede (1990) developed a
general relationship between frictional pressure drop
and convective film heat transfer coefficients as follows:
DP KAhm

(1)

where DP is the exchanger pressure drop, A is the heat


transfer area and h is the film heat transfer coefficient. It
has been shown that, although different exchanger
geometries gave relationships of this form, the value of
m varies from geometry to geometry. In this work we
assume that all the exchanger of the network are of the
same geometry and type (counter flow single-pass shell
single-pass tube exchangers). Serna (1999) developed
equations of the type of Eq. (1) based on the detailed
Bell /Delaware method. The equations derived by Serna
(1999) are used in this paper. Hence, for turbulent flow
in shell and tube exchangers, the equation for the tube
side is:
DPT KPT Ah3:5
T

(2)

and for the shell side is:


DPS KPS Ah5:109
S

(3)

where S and T stand for shell and tube, correspondingly.


The parameters KPT and KPS depend on the geometry
and the physical properties of the streams. Typical
expressions for those parameters are (Serna, 1999):




67:062f6:109
L  Dt
LD1:109
m1:297
j
e
j
KPTj 
(4)
gc
Dt
Fj rj kj3:406 Cp1:703
j
KPSi 

(0:023)2:5 gc Fi ri ki7=3 Cp7=6


i

non-convex and therefore global optimality cannot be


guaranteed.
It is important to make clear that relationships such
as Eqs. (2) and (3) apply only for those streams which do
not experiment change of phase. For that reason, in the
case of the hot utility (steam), it will be assumed that the
film heat transfer coefficient is still a given constant.
Furthermore, in this paper we also assume that the film
coefficient of cooling water is known.
3.2. Pressure drop calculations for heat exchanger
networks
When a process stream exchanges heat in more than
one unit, individual and overall pressure drops must be
calculated. We can add pressure drop only when units or
shells are located in series. For a parallel structure, the
pressure drop will be assumed as equal to the maximum
of the pressure drop in each of the branches (Nie & Zhu,
1999) calculated in terms of the flow rates of the stream
after splitting. Observe that this is an approximation; in
practice, the parallel flow rates will change until (at
steady-state) the pressure is the same for the exchangers
of every branch. Furthermore, in this work, we use a
simplified approach presented by Shenoy (1995) for the
calculation of the pressure drop in each of the stages of a
network superstructure. That approach assumes that the
pressure drop of a stream is linearly distributed according to the surface area of the exchange units. Hence, if
DPi is the pressure drop of the ith hot stream, the
pressure drop of the ith hot stream in stage k , DPik , is
given in terms of the total and individual surface area as
follows:
X
Aijk
j
DPik  X X
DPi
(6)
Aijk  Aicu
k

1=2 11=6

f4:5
i Di mi

(5)

where f is the a correction factor because of the


temperature dependence of the viscosity, m, r is the
density, gc is the gravity constant, k is the thermal
conductivity, Cp is the heat capacity, L is the length of
the tubes, Dt is the internal diameter of the tubes and De
is the external diameter of the tubes.
Hence, by providing the physical properties of the
streams, Eqs. (2) and (3) are used for the calculation of
the heat transfer coefficients in terms of the allowable
pressure drop for each of the streams. Notice that the
use of Eqs. (2) and (3) implies that not only the objective
function but also the constraints of the model will
include nonlinearities. The nonlinearities of the formulations will include both equations of type (2) and (3)
and the equation for the explicit calculation of heat
exchanger areas. Also notice that Eqs. (2) and (3) are

1145

where Aijk is the heat exchanger area for the match


between hot stream i and cold stream j in stage k , and
Ai cu is the area of the match between hot stream i and
the cold utility. A similar equation can be established for
the jth cold stream. This is of course an approximation,
but it greatly reduces the complexity of the model; the
reason is because it avoids the need of having expressions (2) and (3) for each stage of the model superstructure and, as a consequence, the nonlinearities of the
model are also reduced.
3.3. Calculation of cost of power
The power required in a heat exchanger can be
calculated as the product between the volumetric flow
rate and the pressure drop in each side of the exchanger
(shell and tubes). Hence, in order to incorporate the cost
of power to the objective function of the synthesis

1146

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

problem, we applied the expression proposed by Serna


(1999):
Cost of Power CWQDP

(7)

where CW is a cost coefficient (unit cost of power, $/kW


year), Q is the volumetric flow rate (given by the mass
flow rate, F , divided by the density, r ) and DP is the
pressure drop of the stream. Eq. (7) applies for both, hot
and cold streams.

4. HEN synthesis incorporating pressure drops effects


In this section we present the equations that have been
incorporated to the simultaneous MINLP formulation
of Yee and Grossmann (1990) in order to consider
pressure drop effects during the synthesis stage.

4.1. Notation
The following notation is used in the remainder of the
section. Hot streams are represented by index i and cold
streams by index j. Index k denotes the k th stage of the
superstructure. cu stands for cooling utility (cooling
water) and hu for hot utility. When used as an index, T
stands for total.
Sets
ST
HP
CP

stages of the superstructure [1. . .NS]


hot process streams [1. . .NH]
cold process streams [1. . .NC]

Variables
A
surface area
DP
pressure drop
DT
difference of temperature
T
temperature
h
film heat transfer coefficient
q
heat load
z
binary variable representing the existence/nonexistence of a match
Parameters
TIN inlet temperature
TOUT oulet temperature
CCU unit cost of cold utility
CHU unit cost of hot utility
CW unit cost of power
CF
fixed charge for exchangers
C
area cost coefficient
b
area cost exponent
Q
volumetric flow rate (determined as F /r )
DPP maximum allowable pressure drop

4.2. The extended model equations


Seeking concreteness, this section presents only the
equations that have been added to the simultaneous
MINLP formulation for HEN synthesis. The rest of the
formulation is the same as the one presented by Yee and
Grossmann (1990). It will be assumed that hot streams
flow in the shell side and the cold streams flow in the
tube side of the heat exchangers of the network.
Furthermore, since expressions (2) and (3) are not valid
for exchangers with change of phase, we will still assume
that the values of the heat transfer coefficients for the
utilities (cooling water and steam) are given constants.
The model equations are presented next.

4.2.1. Contact area for matches ij , icu and huj


Explicit equations for the calculation of heat exchanger area of each unit are required in the formulation. It
can be observed that, since the contact area directly
depends on the heat load, the contact area will be zero
for non-existing matches. As in previous approaches, the
Chen approximation for the logarithmic mean of the
temperature difference is used (Chen, 1987).

qijk
Aijk  


DTijk  DTijk1 1=3
(DTijk )(DTijk1 )
2

j  CP;

hi

i  HP;
(8)

hj

k  ST

Aicu  
(DTicu )(TOUT i  TIN




hi

qcui

DTicu  TOUT i  TIN
cu )
2

cu

1=3
(9)

hcu

i  HP
Ahuj  
(DThuj )(TIN


1
1


hhu hj

hu

 TOUT

qhuj


DThuj  TIN hu  TOUT j 1=3
j)
2

j  CP

(10)

4.2.2. Total contact area for hot and cold streams


Since specifications for pressure drop are given in
terms of the total allowable pressure drop for a stream,
evaluations of the total contact area for hot and cold
streams are necessary.
XX
AT i 
Aijk Aicu i  HP
(11)
AT j 

XX

Aijk Ahuj

j  CP

(12)

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

4.2.3. Overall pressure drop calculations


These equations relate the film heat transfer coefficients to the pressure drop for each hot and cold stream.
Besides, these values of pressure drop have to be less or
equal to the maximum allowable values for the pressure
drop of each stream (given).
DPi KP i AT i h5:109
i  HP
i
3:5
DPj  KP j AT j hj
j  CP
DPi 5DPP i i  HP
DPj 5 DPPj j  CP

(13)
(14)
(15)
(16)

Observe here that, as an approximation, we do not


consider different values for the heat transfer coefficient
at each stage, but we consider only lumped values as
function of the overall pressure drop of the stream.
4.2.4. Pressure drop for each stage
Pressure drop for hot and cold streams in each stage
of the superstructure is determined by using the Shenoy
(1995) approximation:
X
Aijk
DPik 

DPjk 

A
XT i
Aijk
i

AT

DPi

i  HP;

k  ST

(17)

DPj

j  CP;

k  ST

(18)

As mentioned before, this approximation allows the


calculation of the pressure drop at each stage without
having to use different heat transfer coefficients and
Eqs. (2) and (3) for each stage.
4.2.5. Objective function
As in previous approaches to HEN synthesis based on
MINLP techniques, the objective function consists of
the annualized cost of the network. However, compared
to the expression provided by Yee and Grossmann
(1990), there are two extra terms which are incorporated
to calculate the cost of power for hot and cold streams.
X
X
XXX
min
CCU qcui 
CHU qhuj 
CFij zijk
i

CFicu zcui 

CFjhu zhuj 

Cij (Aijk ) 

X
i

CWi* Qi DPi 

XX
i

Cicu (Aicu ) 
X

Chuj (Ahuj )b

CWj Qj DPj

(19)

1147

addition of Eqs. (8) /(18), and the substitution of the


original objective function by Eq. (19). We are very
much aware of the disadvantages that equations such as
Eqs. (13) and (14) represent with respect to the
numerical properties of the formulation, namely, convergence properties and the sufficient conditions for
optimality. Current research effort focuses in developing
a formulation with better convergence properties and
algorithms for global optimality (for fixed topology
problems).

5. Illustrative examples
This section presents two examples of MINLP synthesis of heat exchanger networks. It is assumed that all
the exchangers are single-pass tube single-pass shell heat
exchangers. Two cases are considered for each of the
examples:
a) Synthesis assuming constant heat transfer coefficients (pressure drop of each stream can then be
calculated for the resulting network).
b) Synthesis based on maximum allowable values for
pressure drop of the streams (actual pressure drop
and heat transfer coefficients are simultaneously
calculated during the synthesis stage).
In order to compare the results for each example, the
values of the coefficients for case (a) were proposed so
that the values of pressure drop of the streams of case a)
are equal to the values of the maximum allowable
pressure drop of case (b). Hence, one has to observe
that, although the values of case (a) are assumed
constant, a preliminary estimation of them has been
done in order to have a good basis for comparison. In
other words, the values for case (a) considered here for
the heat transfer coefficient are expected to be much
better values than a set of values arbitrary fixed, as it is
done in many of the literature examples. Also, in both
examples, a useful life of 5 years is considered for the
heat exchangers.
We want to emphasize again that, because of the nonconvexities presented in the formulation, global optimality cannot be guaranteed. However, preliminary
results in these small-sized examples encourage further
research in the area. Results are obtained by using
DICOPT// through the GAMS modeling environment (Brooke, Kendrick, Meeraus, & Raman, 1998).

5.1. Example 1
4.3. Summary
Our extension to the MINLP formulation given by
Yee and Grossmann (1990) consists therefore of the

This case was reported in the works of Shenoy (1995)


and Serna (1999). In this example the value of HRAT is
20 8C. The inlet and outlet temperature specifications
for each stream are given in Table 1. Table 2 provides

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

1148

Fig. 2. Resulting HEN for example 1 with assumed constant values of


heat transfer coefficients.

Table 1
Temperature specifications for the streams of example 1
Stream

Tin (8C)

Tout (8C)

F CP (kW/8C)

H1
H2
C1
C2
Steam
Cooling

175
125
20
40
180
15

45
65
155
112
179
25

10
40
20
15

Table 2
Assumed constant heat transfer coefficients of example 1
Stream

h (W/m2 8C)

H1
H2
C1
C2

2615
1333
917
166

the values of the heat transfer coefficients assumed as


constant for case a) of this example. The film heat
transfer coefficients for the utilities are assumed as 5000
W/m2 8C for steam and 2500 W/m2 8C for cooling
water. The physical properties and allowable pressure
drop for case b) are given in Table 3.

The resulting networks for both cases of these


example involve different topology but the same minimum utility consumption (605 and 525 kW for heating
and cooling, correspondingly) with a cost of $71 800/
year. Figs. 2 and 3 show the resulting network configurations obtained by assuming constant heat transfer
coefficients and by using maximum allowable pressure
drops, correspondingly. Table 4 summarizes the values
of heat load an exchange area for the case of constant
heat transfer coefficients.
For the purposes of comparison, Table 5 shows the
results obtained in this work as well as those of two
different designs obtained by Serna (1999) for the same
problem. Serna used different design strategies based on
pinch technology. Notice that, in Table 5, the values of
the annualized cost of the network considering power
cost are no reported for the designs of Serna (1999),
since the author did not calculate the cost of power in
his work. For the case of constant heat transfer
coefficients, the network consists of 7 units with a total
area of 423.26 m2 (investment cost is $75 553.75/year).
The cost of power is $10 883.76/year. Hence, the total
cost of the network is $158 237.51/year. For the case of
specified maximum allowable values of pressure drop,
the total area is 489.96 m2 (investment cost is $82 810.10/
year) and the cost of power is $2379.55/year. Hence, the
total cost of the network is $156 989.65/year. It is
important to observe that, even though the investment
cost is smaller for the case of assumed constant
coefficients, the total annualized cost of the networks
is smaller when the coefficients and the values of
pressure drop are simultaneously optimized. Table 6
shows the values of pressure drop for both of the cases.
Finally, Table 7 presents a comparison among the
values of the film heat transfer coefficients calculated
for case b) of this example with respect to those
calculated by Serna (1999). Table 7 also shows the
values of the contact area for each stream used in
pressure drop calculation. Note that the summation of
the column corresponding to the contact area is not
equal to the total area of the network. That is because
the exchange area of a given match is considered as
contact area for both the hot and the cold streams of the
match.

Table 3
Physical properties and allowable pressure drop for the streams of example 1
Stream

H1

H2

C1

C2

Allowable DPP (kPa)


Cp (J/kg 8C)
r (kg/m3)
m (kg/m s)
k (W/m 8C)

30
1658
716
0.24/10 3
1.1

25
2684
777
0.23 /10 3
0.24

10
2456
700
0.23/10 3
0.12

10
2270
680
0.23/10 3
0.011

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

1149

Fig. 3. Resulting HEN for example 1 incorporating pressure drop effects.

5.2. Example 2
We solved the synthesis problem introduced by Sagli,
Gundersen, and Yee (1990) and presented by Polley and
Panjeh Shahi (1991) for an HRAT of 20 8C. Temperature specifications of the problem are given in Table 8.
Table 9 provides the values of the constant coefficients
for case (a) and Table 10 summarizes the data about the
physical properties and allowable stream pressure drops
used for case (b). Figs. 4 and 5 show the configurations
obtained by assuming constant heat transfer coefficients
and by using maximum allowable pressure drops,
correspondingly. Table 11 summarizes the values of
Table 4
Results for example 1 assuming constant heat transfer coefficients
Exchanger

Area (m2)

Q (kW)

1
2
3
4
5
6
7

23.91
49.81
101.57
212.10
18.10
9.23
8.54

500
275
1700
700
500
105
525

heat load an exchange area for the case of constant heat


transfer coefficients. Tables 12 and 13 show the values of
heat exchanger area, annualized cost and pressure drop
for cases a) and b), which provide a basis for comparison.
Both of the resulting network involves minimum
utility consumption (1075 and 400 kW for heating and
cooling, correspondingly) with a cost of $122 250/year.
The network corresponding to constant heat transfer
coefficients consists of a total area of 562.78 m2
(investment cost is $54 672.01/year). The cost of power
is $11 739.02/year. Hence, the total cost of the network is
$188 661.03/year. For case b), the network consists of a
total area of 618.24 m2 (investment cost is $57 728.16/
year). The cost of power is $4556.78/year. Hence, the
Table 6
Pressure drop for streams of example 1
Stream

Constant h
DP (kPa)

Using maximum allowable DP


DP (kPa)

H1
H2
C1
C2

30
25
10
10

5.27
11.19
10
10

Table 5
A comparison for example 1
Computer time (s) Total area (m2) Network cost without power cost ($/year)
MINLP assuming
Constant h
0.4
Design 1 (Serna, 1999) /
Design 2 (Serna, 1999) /
This work
2.7

423.26
380.37
437.19
486.96

147 353.75
158 452.65
147 577.45
154 610.10

Network cost with power cost ($/year)

158 237.51
/
/
156 989.65

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

1150

Table 7
Calculated film heat transfer coefficients for example 1
Stream

Contact area (m2) (Serna, 1999)

h (W/m2 8C) (Serna, 1999)

Contact area (m2) (this work)

h (W/m2 8C) (this work)

H1
H2
ST
C1
C2
CW

122.52
278.18
21.02
181.02
228.69
12.01

2492
1358
5000
867
173
2500

57.16
409.19
20.61
140.50
337.14
9.32

2202
1124
5000
921
156
2500

Table 8
Temperature specifications for stream of example 2
Stream

Tin (8C)

Tout (8C)

F CP (kW/8C)

H1
H2
C1
C2
Steam
Cooling water

150
90
20
25
180
10

60
60
125
100
180
15

20
80
25
30

Table 9
Assumed constant heat transfer coefficients of example 2
Stream

h (W/m2 8C)

H1
H2
C1
C2

2615
1333
917
166

Fig. 4. Resulting HEN for example 2 with assumed constant values of


heat transfer coefficients.

Table 11
Results for example 2 assuming constant heat transfer coefficients

Table 10
Physical properties and allowable pressure drop for the streams of
example 2
Stream

H1

H2

C1

C2

DPP (kPa)
Cp (J/kg 8C)
r (kg/m3)
m (kg/m s)
k (W/m 8C)

20
2600
800
0.0005
0.12

30
2600
800
0.0005
0.12

10
2600
800
0.0005
0.12

60
2600
800
0.0005
0.12

total cost of the network is $184 534.94/year. As in


example 1, the total area and the investment cost for
case (b) are greater than those for case (a); however,
because of the power cost, the configuration of case (b)
is preferred since its total annualized cost is smaller.
Finally, Table 14 shows a comparison among the values
of the film heat transfer coefficients and contact area
obtained by Polley and Panjeh Shahi (1991) and the
ones calculated for case (b) in this work.

Exchanger

Area (m2)

Q (kW)

1
2
3
4
5
6
7

24.45
65.61
17.55
103.14
102.24
169.79
79.99

300
900
200
1050
1350
1075
400

5.3. Discussion
As shown in Tables 5 and 12, the difference in
computer time when using allowable pressure drop
compared to the one obtained with the constant heat
transfer coefficient approach is not really significant.
Moreover, the results show that, by incorporating
pressure drop effects and power cost to the formulation,
consistent values for the heat transfer coefficients and
optimal values for the pressure drop can be obtained.
However, the initialization procedure for obtaining
converge in the case of allowable pressure drop involves
to get an initial approximation for the exchange areas,
pressure drop and heat transfer coefficients. The non-

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

1151

Fig. 5. Resulting HEN for example 2 incorporating pressure drop effects.


Table 12
A comparison for example 2
Computer time (s)
MINLP assuming
Constant h 0.5
This work
3.8

Total area (m2)

Network cost without power cost ($/year)

Network cost with power cost ($/year)

562.78
618.24

176 922.01
179 978.16

188 661.03
184 534.94

Table 13
Pressure drop for streams of example 2
Stream

Constant h
DP (kPa)

Using maximum allowable DP


DP (kPa)

H1
H2
C1
C2

20
30
10
60

20
9.98
6.71
23.16

linear nature of the equations added to the formulation


makes this initialization procedure a difficult task. We
have not been able to get a systematic way to initialize
the variables and several runs have been performed to
obtain the order of magnitude of the new variables.
Finally, it should be clear that, although this work
improves the performance of the original model of Yee
and Grossmann (1990), further developments should be
aimed to remove the Shenoy approximation and to
calculate in a more reliable way the values of pressure
drop and heat transfer coefficients for each stage of the
model superstructure.

6. Conclusions
Most of the HEN design procedures based on
MINLP techniques reported to date do not consider

the pressure drop aspects of the problem. It is well


known, however, that detailed heat exchanger design is
based on the allowable pressure drop of the streams.
This works extends a previous approach to the HEN
synthesis problem based on mathematical programming
by incorporating the effect of pressure drop into the
synthesis stage. In this way, the optimization of the
problem minimizes capital cost but still satisfies both the
transfer of required heat duty and the pressure drop
constraints imposed by practical considerations.
The simultaneous MINLP formulation developed by
Yee and Grossmann (1990) has been extended by
incorporating equations that relate the overall pressure
drop of a stream to its film heat transfer coefficient
(Polley and Panjeh Shahi, 1991). Also, in order to
determine the pressure drop in each of the stages of the
configuration, the approximation provided by Shenoy
(1995) has been used. Finally, the cost of power has been
incorporated into the objective function of the optimization problem. Two small sized illustrative examples
have been used to demonstrate the scope of the
proposed approach. However, since we only considering
the case of single-pass tube single-pass shell heat
exchangers, we should recognize that our approach
may still produce suboptimal designs. Furthermore,
since the expressions which relate pressure drop and
heat transfer coefficients are not valid for exchangers
with change of phase, we are still assuming the film
coefficients for the utilities as given constants. Although

S. Frausto-Hernandez et al. / Computers and Chemical Engineering 27 (2003) 1143 /1152

1152

Table 14
Calculated film heat transfer coefficients for example 2
Stream Contact area (m2) (Polley & Panjeh Shahi,
1991)

h (W/m2 8C) (Polley & Panjeh Shahi,


1991)

Contact area (m2) (this


work)

h (W/m2 8C) (this


work)

H1
H2
C1
C2

937
1073
618
1109

154.71
293.52
326.39
200.08

721
680
565
814

157.66
149.534
257.1
192.76

the numerical disadvantages of the proposed approach


can also be easily identified, the results obtained so far
have shown the potential of the approach and encourage
further development in the area. As a matter of fact, a
formulation which deals with the non-convexities incorporated by equations like Eqs. (2) and (3) is now
under investigation.

Acknowledgements
The authors would like to thank the financial support
provided by the Mexican National Council for Science
and Technology (CONACYT) and by the Council for
Science and Technology of the State of Guanajuato
(CONCYTEG).

References
Chen, J. J. J. (1987). Letter to editors: comments on improvement on a
replacement for the logarithmic mean. Chemical Engineering
Science 42 , 2488 /2489.
Brooke, A., Kendrick, D., Meeraus, A., & Raman, R. (1998).
GAMS */a user guide . Washington, DC: GAMS Development
Corporation.

Linhoff, B., & Ahmad, S. (1989). Optimal synthesis of energy


management systems. Journal of Engineering Research and Technology 111 , 121 /130.
Linhoff, B., & Flower, J. R. (1978). Synthesis of heat exchanger
networks. American Institute of Chemical Engineering Journal 24 ,
633 /642.
Nie, X. R., & Zhu, X. X. (1999). Heat exchanger network retrofit
considering pressure drop and heat-transfer enhancement. American Institute of Chemical Engineering Journal 45 , 1239 /1249.
Papoulias, S. A., & Grossmann, I. E. (1983). A structural optimization
approach in process synthesis */II. Heat recovery networks.
Computers and Chemical Engineering 7 , 707 /721.
Polley, G. T., & Panjeh Shahi, M. H. (1991). Interfacing heat
exchanger network synthesis and detailed heat exchanger design.
Transactions of the Institute of Chemical Engineers 69 , 445 /457.
Polley, G. T., Panjeh Shahi, M. H., & Jegede, F. O. (1990). Pressure
drop considerations in the retrofit of heat exchanger networks.
Transactions of the Institute of Chemical Engineers 68 (Pt. A), 211 /
220.
Sagli, B., Gundersen, T., & Yee, T. (1990). Topology traps in
evolutionary strategies for heat exchanger network synthesis. In
H. Bussemaker & P. Iedema (Eds), Process Technology Proceedings 9, 51 /58.
Serna, G.M., (1999). Desarrollo de Algoritmos Rigurosos para la
Integracion Termica de Procesos. PhD thesis, Departamento de
Ingeniera Qumica, Instituto Tecnologico de Celaya, Mexico.
Shenoy, U.V., (1995). Heat exchanger network synthesis. process
optimization by energy and resource analysis, Gulf Publishing
Company.
Yee, T. F., & Grossmann, I. E. (1990). Simultaneous optimization
models for heat integration */II. Heat exchanger network synthesis. Computers and Chemical Engineering 14 , 1165 /1184.

Você também pode gostar