Applied Thermal Engineering 29 (2009) 2954–2960

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Optimization design of shell-and-tube heat exchanger by entropy generation

minimization and genetic algorithm
Jiangfeng Guo, Lin Cheng, Mingtian Xu *
Institute of Thermal Science and Technology, Shandong University, Jinan 250061, China

a r t i c l e i n f o a b s t r a c t

Article history: In the present work, a new shell-and-tube heat exchanger optimization design approach is developed,
Received 25 March 2008 wherein the dimensionless entropy generation rate obtained by scaling the entropy generation on the
Accepted 3 March 2009 ratio of the heat transfer rate to the inlet temperature of cold fluid is employed as the objective function,
Available online 24 March 2009
some geometrical parameters of the shell-and-tube heat exchanger are taken as the design variables and
the genetic algorithm is applied to solve the associated optimization problem. It is shown that for the case
Keywords: that the heat duty is given, not only can the optimization design increase the heat exchanger effective-
Entropy generation number
ness significantly, but also decrease the pumping power dramatically. In the case that the heat transfer
Genetic algorithms
Shell-and-tube heat exchanger
area is fixed, the benefit from the increase of the heat exchanger effectiveness is much more than the
Optimization design increasing cost of the pumping power.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction tube bank by optimizing the global performance of the installation

Energy conservation is vital for the development of world econ-
omy. To use energy more efficiently is one of important measures
for saving energy. Heat exchangers are used to transfer thermal en-
ergy between two or more media and widely applied to power
engineering, petroleum refineries, chemical industries, food indus-
tries and so on. The shell-and-tube heat exchanger (STHE) shown
in Fig. 1 is the most common heat exchanger type. Therefore the
study of its optimization design is of great importance for conserv-
ing energy in heat exchange processes.
The available objective functions for heat exchanger optimiza-
tion designs may be classified into two groups, one is based on
the first law of thermodynamics and another is based on the com-
bination of the first and second law of thermodynamics. In recent
decades the second group has aroused widespread interest, which
includes the entropy and exergy. Based on the concept of entropy,
several heat exchanger performance criteria were developed; these
criteria have their own characteristics and constraints, but also are
interrelated as described in [1]. Among them the entropy genera-
tion number is the most frequently applied one proposed by Bejan
[2–4], and entropy generation minimization (EGM) has been
widely applied to the optimization design of heat exchangers.
Bejan [3] demonstrated that EGM may be used by itself in the
preliminary stages of design, in order to identify trends and the
existence of optimization opportunities. Vargas et al. [5] presented
an approach to determine the internal geometric configuration of a
* Corresponding author. Tel.: +86 531 993000 6503; fax: +86 531 88399598.
E-mail address: mingtian@sdu.edu.cn (M. Xu).
that uses the cross flow heat exchanger. Based on the EGM, Oğula-
tu et al. [6] analytically carried out an optimization design of a
cross flow plate heat exchanger, and examined the optimum result
by the experimental data.
From our knowledge the applications related to the EGM are
mainly based on the dimensionless entropy generation defined
by scaling the entropy generation rate on the heat capacity rate.
However it was found that the entropy generation number (EGN)
defined in such a manner exhibits the so-called ‘entropy genera-
tion paradox’ [7,8]. In order to avoid this paradox the ratio of heat
transfer rate to the cold fluid inlet temperature can be employed to
non-dimensionalise the entropy generation rate [7]. The obtained
dimensionless entropy generation number is called as the modified
entropy generation number in the following discussion and em-
ployed as the objective function in our heat exchanger optimiza-
tion design approach.
Usually the heat exchanger optimization design with multiple
design variables is more practicable, and its global optimum solu-
tion is more desirable. The genetic algorithm is a powerful tool to
address the multi-variable optimization problems. Recently the
application of genetic algorithm on thermal engineering has re-
ceived much attention [9–11].
In the present work, we attempt to put the entropy generation
minimization and genetic algorithm into the STHE optimization
design practice. In this method the modified entropy number is de-
fined as the objective function, the Bell–Delaware design method
for the STHE is utilized, the multiple design variables are assumed
and the genetic algorithm is used to solve the resulting optimiza-
tion problems. The influence of the selected design variables and

1359-4311/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2009.03.011
 J. Guo et al. / Applied Thermal Engineering 29 (2009) 2954–2960 2955

Nomenclature

Ao heat exchange area (m2) U heat transfer coefficient (W/m2 K)

Bs the ratio of the baffle spacing to shell diameter w tube side flow velocity (m/s)
cp specific heat (J/kg K) W_ overall pumping power (kW)
CER cost of energy recovery (€/kW h)
CE energy cost (€/kW h) Greek symbols
C* heat capacity rate ratio DCO increased annual operating cost (€/yr)
d tube diameter (m) DCS additional annual benefit from recycling (€/yr)
Ds shell inside diameter (m) DP pressure drop (Pa)
fi fanning friction coefficient DPbk pressure drop for an ideal cross flow section (Pa)
G mass flow velocity (kg/m2 s) DPwk pressure drop for the ideal window section (Pa)
H annual operating time (h/yr) d tube wall thickness (m)
jo heat transfer factor e heat exchanger effectiveness
L tube length (m) g overall pumping efficiency
m_ mass flow rate (kg/s) h central angle of baffle cut (rad)
n the number of tubes k thermal conductivity (W/m K)
Nb the number of baffles m kinematic viscosity (m2/s)
Nc number of effective tube rows in one cross flow section q fluid density (kg/m2)
Ncw number of effective tube rows crossed in the baffle win-
dow Subscripts
Ns Bejan’s entropy generation number 1 hot fluid
Ns1 modified entropy generation number 2 cold fluid
Ntu number of heat transfer units i inside
Pr Prandtl number o outside
Q actual heat transfer rate (W) s shell side
Qmax maximum possible heat transfer (W) t tube side
r fouling resistances (m2 K/W) w tube wall
Rb the correction factor for bypass flow
R1 the correction factor for baffle leakage Superscripts
Rs the correction factor for unequal baffle spacing at inlet i inlet
and/or outlet o outlet
s tube pitch (m)
S_ gen entropy generation rate (W/K)
T temperature (K)

other parameters on the entropy generation is explored by analyz-
ing the optimization results.
2. Optimization design method
2.1. Genetic algorithm
The basic principle of genetic algorithm was first proposed in
the 1970s by John Holland. The genetic algorithm is based on the
natural selection, which was found in biological evolution process.
In the optimization design application, before a genetic algorithm
can be put to work, a method is needed to encode potential solu-
tions to the optimization problem in a form that a computer can
process. One common approach is to encode solutions as binary
strings: sequences of 1’s and 0’s, where the digit at each position
represents the value of some aspect of the solution. A metric called
a fitness function that allows each potential solution (individual) to
be quantitatively evaluated. After a random initial population in
the ranges of design variables is generated, the algorithm creates
a sequence of new generations iteratively until the stopping crite-
rion is met. In this process, the selection of parents is based on
their fitness; children (next generation or population) are pro-
duced by making random changes to a single parent (mutation)
or by combining the vector entries of a pair of parents (crossover),

Fig. 1. Diagram of a typical STHE.

2956 J. Guo et al. / Applied Thermal Engineering 29 (2009) 2954–2960

start where v2,w is the kinematic viscosity of the tube side fluid at wall
temperature.
Fitness function In order to get the shell side pressure drop, it needs to calculate
the ideal pressure drops in cross flow section and window section,
Estimating ranges for varaibles
correct them by the effective correction factors, and finally sum all
Initial population the components to give the total shell pressure drop [13,14]. The
total pressure drop across the shell side is thus expressed as [13]:
 
Ncw
Evaluation / Fitness DP2 ¼ ½ðNb  1ÞDPbk Rb þ Nb DPwk R1 þ 2DPbk Rb 1 þ Rs ð6Þ
Crossover and Nc
computing mutation
More details about pressure drop in the shell side can be found in
Selection of parents [13,14]. The total pumping power is given as [16]:
Termination No  
criterion? _ ¼ 1 m_1 _2
m
W DP 1 þ DP 2 ð7Þ
1000g q1 q2
Yes where g is the overall pumping efficiency, and assumed as 80% in
Choose the best this paper.
individual
2.2.3. Entropy generation number
Stop
The irreversibility losses in heat exchanger are evaluated in
terms of entropy generation. The entropy generation rate caused
Fig. 2. Flow chart of genetic algorithm.
by finite temperature difference, S_ gen ; DT, can be written as follows
[15]:
and then replace the current population with the children to form Z  
o _ p dT
mc To To
the next generation. The algorithm selects individuals with better S_ gen ; DT ¼ _ p Þ1 ln 1 þ ðmc
¼ ðmc i
_ p Þ2 ln 2 ð8Þ
fitness values as parents, and eliminates the inferior. This guaran- i T 1;2 T1 T i2
tees the algorithm converges to a best individual, which probably For an incompressible fluid under non-adiabatic condition the en-
represents the best solution of the given problem [9–12]. The flow tropy generation rate S_ gen;DP caused by fluid friction is expressed
chart of a genetic algorithm is presented in Fig. 2. as follows:
0  1
2.2. STHE optimization design D P ln T o =T i
S_ gen;DP ¼ @m_ A
q To  Ti
2.2.1. Heat transfer calculation 1;2
There are many methods to calculate the shell inside diameter.  
o i
The following one is employed in the present work [13]: DP ln T 1 =T 1 DP lnðT o =T i Þ
¼m_1 1 þm _ 2 2 o 2 i2 ð9Þ
pffiffiffi q1 T o1  T i1 q2 T 2  T 2
Ds ¼ ð1:1  n  1Þ  s þ 3do ð1Þ
The total entropy generation rate in heat exchanger can be written
The tube side heat transfer coefficient ai can be computed from the
as:
Nusselt number correlation [13–15]:
 0:8 S_ gen ¼ S_ gen;DT þ S_ gen;DP
k1 w1 di
U i ¼ 0:023 Pr0:3 ð2Þ
di v 1 1 T o1 T o2 DP 1 lnðT o1 =T i1 Þ
_ p Þ1 ln
¼ ðmc _ p Þ2 ln
þ ðmc þ _1
m
T i1 T i2 q1 T o1  T i1
For the segmentally baffled shell-and-tube exchanger, the Bell–Del-
aware method is usually used in the STHE design [13–15]. In this DP 2 lnðT o2 =T i2 Þ
þ _2
m ð10Þ
method, the shell side heat transfer coefficient is determined by q2 T o2  T i2
correcting the ideal heat transfer coefficient through considering
the various leakage and bypass flow streams, which is given as fol- When the heat capacity rate of the hot fluid is larger than that of the
lows [13,14]: cold fluid, the outlet temperature of both fluids can be calculated as
 0:14 follows:
v2
U o ¼ jo G2 cp2 Pr22=3 ð3Þ
v 2;w T o1 ¼ T i1  eðT i1  T i2 ÞC  ð11Þ
where v2,w is the kinematic viscosity of the shell side fluid at wall T o2 ¼ T i2 þe ðT i1  T i2 Þ ð12Þ
temperature. The overall heat transfer coefficient U depends on
the tube side and shell side heat transfer coefficient and fouling where C  ¼ ðmc_ p Þmin =ðmc
_ p Þmax is the heat capacity rate ratio and
resistances, which is expressed as follows [13]: obviously C*6 1 [14], and for the expressions of the heat exchanger
      effectiveness e or various flow arrangements one may refer to Refs.
1 1 do do dw do 1 [14,15].
¼ þ rt þ þ rs þ ð4Þ
U U i di di kw di Uo Non-dimensionlising the total entropy generation rate by the
larger heat capacity rate ðmc _ p Þmax gives rise to the following entro-
2.2.2. Pressure drop calculation py generation number:
By neglecting the pressure loss in bend for the single tube pass,
S_ gen
the total tube side pressure drop is written as [13]: Ns ¼ ð13Þ
! _  cp Þmax
ðm
 0:14
L v1 q1 w21
DP 1 ¼ 4f i þ 1:5 ð5Þ This is called as the Bejan’s definition of EGN in the following dis-
di v 1;w 2
cussion. It was found that the Bejan’s definition of EGN would lead
 J. Guo et al. / Applied Thermal Engineering 29 (2009) 2954–2960 2957

to the ‘entropy generation paradox’. Therefore another non-dimen-
sionalising method for the entropy generation rate was proposed by
Hesselgreaves [7]:
S_ gen T i2
Ns1 ¼
Q coding is adopted. The normalized geometric selection, arithmetic
The dimensionless entropy generation rate defined by Eq. (14) is
called the modified EGN in the following discussion.
2.2.4. Optimization design of STHE
In the present section, we consider three optimization design
problems of the STHE with single tube pass by using the entropy
generation minimization and the genetic algorithm. The STHE un-
der consideration is shown in Fig. 1.
2.2.4.1. Example one. The known information of the STHE are
shown in Table 1, five design variables are selected and listed as
follows:
 Tube outer diameter do. The corresponding tube pitch s can be
determined based on this parameter [17].
 Number of tubes n. The range of n is 50–500 by experience
according to Table 1.
 Ratio of the baffle spacing to shell diameter Bs. Its range is 20–
100% according to the national standard of China [17].
 Central angle of baffle cut h. Baffle cut is recommended between
20% and 45% of the shell diameter, so the range of h is 1.8546–
2.9413 rad.
 Outlet temperature of the cold fluid in the heat exchanger T o2 .
The range of T o2 is 313.15–343.15 K (40–70 °C) by experience
according to Table 1.
The fitness function is defined as follows:
Fitness ¼ maxðN s1 Þ
where Ns1 is the modified EGN.
The constraint conditions are: length-diameter ratio is between
6 and 10; the baffle spacing is greater than 50 mm; the tube side
pressure drop is less than 5  104 Pa; the shell side pressure drop
is less than 5  104 Pa.
The optimization problem described above is solved by the ge-
netic algorithm with the genetic algorithm toolbox of MATLAB
[18]. According to the possible range of the design variables, a ran-
dom initial population of the design variables is created. Then the
iterative search process is carried out as shown in Fig. 2 until the
stop criteria are satisfied. During this process if any individual does
not satisfy one of the constraint conditions, it will be eliminated in
the subsequent process of evolution by the penalty function meth-
od (set Ns1 equal to 0.14). The maximum number of generations is
ð14Þ set at 500. Initial population includes 20 individuals; floating point
crossover and non-uniform mutation are employed. After 500 gen-
erations, the optimal solution of the STHE optimization design
problem under consideration is obtained and shown in Table 2.
The variation of the fitness function with respect to the number
of generations is depicted in Fig. 3. Obviously, the fitness value is
basically stable after 100 generations, and reaches the optimum
value when the termination condition is met.
For each generation in the evolution process, according to the
values of the fitness function of every individual, a best individual
is selected, the corresponding value of the heat exchanger effec-
tiveness, the heat capacity rate ratio, the number of exchanger heat
transfer units and the total pumping power of this individual are
thus obtained. The variations of the modified EGN with respect
to these parameters are illustrated in Fig. 4.
From Fig. 4a, one can see that as the heat exchanger effective-
ness increases, the modified EGN decreases. This shows an agree-
ment with the thermodynamic analysis based on the second law
of thermodynamics, which tells us that the undesirable thermody-
namic irreversibility quantified by the EGN diminishes the thermal
performance of the heat exchanger [19]. The heat capacity rate ra-
tio decreases with the decrease of the modified EGN as shown in
Fig. 4b, which indicates that as the outlet temperature of the cold
fluid in the shell side increases and the heat transfer rate keeps
constant, the shell side mass fluid rate decreases, thus if the ex-
changer is used to exchange heat between hot and cold water,
water saving can be achieved by the optimization process.
From Fig. 4c, one can see that when the number of transfer units
increases, the modified EGN decreases, which implies that the
reduction of the modified EGN is at the cost of increasing heat
transfer area when Cmin and the overall heat transfer coefficient
are fixed. It would be expected that the modified EGN caused by
finite temperature difference should achieve the minimum when
-0.11
-0.115
-0.12 change process of average value
change process of best value
-0.125
Fitness

Table 1
Known information for a STHE. -0.13
Fluid 1 (tube side) Fluid 2 (shell side)
-0.135
Inlet temperature Ti (K) 368.15 283.15
Outlet temperature To (K) 343.15 –
Mass flow rate m _ ðkg=sÞ 50 –
-0.14
Density q (kg/m3) 970 991.15
Constant pressure specific heat cp (J/kg K) 4200 4174
Kinematic viscosity m (m2/s) 3.36  107 6.96  107 0 100 200 300 400 500
Entrance pressure Pi (MPa) 6.5 5 Generations
Fouling resistance r (m2 K/W) 0.000086 0.00017
Prandtl number Pr 2.015 4.5878 Fig. 3. Variation of the fitness value of the best individual in each generation with
respect to the number of generations.

Table 2
The optimum parameters of the heat exchanger after the evolution of 500 generations.

do (m) n Bs h (rad) Ao (m2) U (W/m2 K) T o2 ðKÞ _ ðkWÞ

W e Ntu
0.016 526 0.70 2.33 99.37 1320.6 343.11 0.552 0.7054 1.5
2958 J. Guo et al. / Applied Thermal Engineering 29 (2009) 2954–2960

0.140 0.140
0.135 0.135
0.130 0.130
Ns1

Ns1
0.125 0.125
0.120 0.120

0.115 0.115

0.110 0.110
0.45 0.50 0.55 0.60 0.65 0.70 0.65 0.60 0.55
0.50 0.45 0.40
ε C*
(a) EGN versus the heat exchanger effectiveness (b) EGN versus the heat capacity rate ratio

0.140 0.140

0.135 0.135

0.130 0.130

Ns1
Ns1

0.125 0.125

0.120 0.120

0.115 0.115

0.110 0.110
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 2.1 1.8 1.5
1.2 0.9 0.6
Ntu W (kW)
(c) EGN versus the number of exchanger heat transfer units (d) EGN versus the total pumping power

Fig. 4. Variations of the modified EGN with respect to the heat exchanger effectiveness, the heat capacity rate ratio, the number of exchanger heat transfer units and the total
pumping power.

Ntu reaches the infinite, this agrees with the result depicted in
0.90
Fig. 4c. From Fig. 4d one can see that when the modified EGN de-
creases gradually, the total pumping power generally becomes 0.85
smaller and smaller due to the decrease of the mass flow rate of
2
Ns ×10

cold water. 0.80

For this example, if the Bejan’s definition of the EGN expressed
0.75
in Eq. (13) is employed in the definition of the fitness function, it
will yield the same result as presented above. The reason is that 0.70
in the optimization problem under consideration, the heat duty
is given, the comparison of Eq. (13) with Eq. (14) shows that the 0.65
two types of definitions of the EGN are equivalent. 0.60
0.30 0.35 0.40 0.45 0.50 0.55 0.60
2.2.4.2. Example two. The given data for this example is compiled in ε
Table 3. In this example instead of the outlet temperature of the Fig. 5. Variation of the Bejan’s EGN with respect to the heat exchanger
cold fluid, the tube length is selected as a design variable. And effectiveness.
the tube length ranges from 1 m to 12 m [17]. The other design
variables are the same as that in example one.
The Bejan’s definition of the EGN is first employed in the defini-
the decrease of the heat exchanger effectiveness, which is not
tion of the fitness function. After getting the optimal solution, the
the case as suggested by the thermodynamic analysis. Therefore
relation between EGN and the effectiveness is shown in Fig. 5.
the Bejan’s definition of the EGN is only valid when the heat duty
From this figure one can see that the decrease of EGN leads to
is fixed. Otherwise it will yield the controversial result.
If the modified EGN is employed in the definition of the fitness
Table 3
function, the variations of the modified EGN with respect to the
The given information of a STHE. heat exchanger effectiveness and the number of heat transfer units
are illustrated in Fig. 6a and b, respectively. From Fig. 6a and b, one
Fluid 1 (tube side) Fluid 2 (shell side)
can see that the decrease of the modified EGN leads to the increase
Inlet temperature Ti (K) 368.15 283.15 of the heat exchanger effectiveness and the number of heat trans-
Mass flow rate m _ ðkg=sÞ 50 20
fer units. Therefore the modified EGN avoids the ‘entropy genera-
Density q (kg/m3) 970 991.15
Constant pressure specific heat cp (J/kg K) 4200 4174 tion paradox’, thus demonstrates an obvious advantage.
Kinematic viscosity m (m2/s) 3.36  107 6.96  107
Entrance pressure Pi (MPa) 6.5 5 2.2.4.3. Example three. In order to further examine the optimization
Fouling resistance r (m2 K/W) 0.000086 0.00017
design of the STHE with the use of the modified EGN and the genet-
Prandtl number Pr 2.015 4.5878
ic algorithm. The following optimization problem is considered.
 J. Guo et al. / Applied Thermal Engineering 29 (2009) 2954–2960 2959

0.16 0.16

0.15 0.15

Ns1

Ns1
0.14 0.14

0.13 0.13

0.12 0.12
0.40 0.45 0.50
0.55 0.60 0.65 0.5 0.6 0.7 0.8
0.9 1.0 1.1 1.2
ε Ntu
(a) The modified EGN versus the heat exchanger effectiveness (b) The modified EGN versus Ntu

Fig. 6. Variations of the modified EGN with respect to the heat exchanger effectiveness and Ntu.

The tube outside diameter (do), the number of tubes (n), the ratio of
baffle spacing to shell inside diameter (Bs) and the central angle of
baffle cut (h) are selected as the design variables. The heat transfer
area is set to 60 m2 and kept constant. The modified EGN is taken
as the fitness function. After the evolution of 200 generations, the
final optimum parameters are shown in Table 4. Fig. 7 shows a se-
quence of snapshots of the modified EGN, the number of heat
transfer units, the heat exchanger effectiveness and the total
pumping power of the best individual in the evolution process.
From Fig. 7a and b, one can see that the number of heat transfer
units increases with decreasing the modified EGN, so does the heat
exchanger effectiveness. The variation of the total pumping power
with the entropy generation number is complicated as shown in
Fig. 7c, but the overall trend is that the pumping power increases
with decreasing the modified entropy generation number. There-
fore although the heat exchanger effectiveness is improved

Table 4
The optimum parameters of the heat exchanger after the evolution of 200 generations.

do (m) n Bs h (rad) Ko (W/m2 K) T o1 ðKÞ T o2 ðKÞ _ ðkWÞ

W e Ntu
0.016 318 0.36 2.09 1634.6 346.83 336.7 2.35 0.6308 1.175

0.1320 0.1320

0.1305 0.1305

0.1290 0.1290
Ns1

Ns1

0.1275 0.1275

0.1260 0.1260

0.1245 0.1245
0.1230 0.1230
1.00 1.04 1.081.12 1.16 0.58 0.59 0.60 0.61 0.62 0.63
Ntu ε
(a) The modified EGN versus Ntu (b) The modified EGN versus exchanger effectiveness

0.1320

0.1305

0.1290
Ns1

0.1275

0.1260

0.1245

0.1230
1.2 1.4 1.6 1.8 2.0 2.2 2.4
(kW)
W
(c) The modified EGN versus the pumping power

Fig. 7. Variations of the modified EGN with respect to the number of heat transfer units, the heat exchanger effectiveness and the total pumping power.
2960 J. Guo et al. / Applied Thermal Engineering 29 (2009) 2954–2960
through the optimization process, unfortunately the operating cost
rises. In such a situation it needs to examine whether the optimi-
zation process may lead to the higher performance of the heat ex-
changer. Firstly the benefit from the increase of the heat exchanger
effectiveness by the optimization process can be estimated as fol-
lows [20]:
_ p Þmin ðT i1  T i2 ÞDeH=1000
DC S ¼ C ER Q max DeH=1000 ¼ C ER ðmc
where CER = 0.0177 €/(kW h) [21], De is the increase of the heat ex-
changer effectiveness and equal to 0.0482, H is the annual operating
time and equal to 6000 h/yr. From Eq. (15), we obtain
DCS = 36322 €/yr which is the profit per year due to the improve-
ment of heat exchanger effectiveness. While the additional annual
operating cost due to the increase of the pumping power is esti-
mated as follows [16]:
_
DC O ¼ C E DWH
where CE = 0.12 €/kW h [16], DW _ is the increase of the pumping
power through the optimization process and equal to 1.01 kW.
From Eq. (16), we have DCO = 727.2 €/yr which is the price we have
to pay in order to get the higher heat exchanger effectiveness. The
comparison between DCS and DCO clearly demonstrates that the
optimization process can lead to significant benefit by enhancing
the heat transfer between the cold and hot water.
Note that in this example, the heat transfer area is fixed and the
design variables are only related to the geometry of the STHE.
Therefore in a STHE design process after the heat transfer area is
determined, the entropy generation minimization and the genetic
algorithm may be used to optimize the geometry of the STHE such
that the performance of the STHE can be improved significantly.
3. Concluding remarks
In summary, based on the second law of thermodynamics a new
heat exchanger optimization approach is developed. In this ap-
proach, the modified entropy generation number which can avoid
the entropy generation paradox is taken as the objective function,
the genetic algorithm is applied to solve the multi-variable optimi-
zation problems which not only yields the globe optimum solution
but also demonstrates the flexibility to select the design variables
and constraint conditions.
Three optimization design examples are presented. In the first
example five design variables are selected and the heat load is
given. It is shown that the optimization design process can remark-
ably increase the heat transfer effectiveness and decrease the
pressure drop.
The second example shows that the entropy generation number
defined by Bejan suffers from the ‘entropy generation paradox’,
while the modified entropy generation number can avoid such a
paradox. Therefore the modified entropy generation number is
preferable in the heat exchanger optimization design applications.
From our knowledge, the present work is the first attempt to use
the modified entropy generation number as the objective function
in the heat exchanger optimization design.
In the third example, the heat transfer area is fixed which may
be obtained by the traditional heat exchanger design method, then
the tube and baffle geometrical parameters are selected as the de-
sign variables and thus are allowed to change in the optimization
design process. It is found that the optimization design enables
us to improve the heat exchanger performance significantly, and
is profitable from economic point of view.
In the traditional optimization design of the STHE based on the
first law of thermodynamics, the thermodynamic irreversible
ð15Þ losses in a heat exchanger has not been accounted for. Furthermore
it is found that the traditional optimization design with the total
cost as the objective function suffers from decreasing the heat ex-
changer effectiveness [20]. In this point the optimization approach
developed in the present work is more advantageous.
Acknowledgements
The support of our research program by National Basic Research
ð16Þ Program of China (Project No. 2007CB206900) is greatly
appreciated.
References
[1] M. Yilmaz, O.N. Sara, S. Karsli, Performance evaluation criteria for heat
exchangers based on second law analysis, Exergy, An International Journal 1
(4) (2001) 278–294.
[2] A. Bejan, Entropy Generation through Heat and Fluid Flow, Wiley, New York,
1982.
[3] A. Bejan, Entropy Generation Minimization, CRC Press, New York, 1995.
[4] A. Bejan, A study of entropy generation in fundamental convective heat
transfer, ASME Journal of Heat Transfer 101 (1979) 718–725.
[5] J.V.C. Vargas, A. Bejan, D.L. Siems, Integrative thermodynamic optimization of
the crossflow heat exchanger for an aircraft environmental control system,
Transactions of the ASME 123 (August) (2001) 760–769.
[6] U. Oğulatu, F. Doba, T. Yilmaz, Irreversibility analysis of cross flow heat
exchangers, Energy Conversion and Management 41 (2000) 1585–1599.
[7] J.E. Hesselgreaves, Rationalisation of second law analysis of heat exchangers,
International Journal of Heat and Mass Transfer 43 (2000) 4189–4204.
[8] A. Bejan, Second law analysis in heat transfer, Energy 5 (1980) 721–732.
[9] G.N. Xie, B. Sunden, Q.W. Wang, Optimization of compact heat exchangers by a
genetic algorithm, Applied Thermal Engineering 28 (2008) 895–906.
[10] Resat Selbas, Önder Kızılkan, Marcus Reppich, A new design approach for
shell-and-tube heat exchangers using genetic algorithms from economic point
of view, Chemical Engineering and Processing 45 (2006) 268–275.
[11] M. Mishra, P.K. Das, S. Saranqi, Optimum design of crossflow plate-fin heat
exchangers through genetic algorithm, International Journal of Heat
Exchangers 5 (2004) 379–401.
[12] M. Zhou, S. Sun, Genetic Algorithms: Theory and Application, National Defense
Industry Press, Beijing, 1999 (in Chinese).
[13] M.Z. Shi, Z.Z. Wang, Principles and Design of Heat Exchanger Device, Southeast
university press, Nanjing, 1996 (in Chinese).
[14] T. Kuppan, Heat Exchanger Design Handbook, Marcel Dekker Inc., New York,
2000.
[15] R.K. Shah, D.P. Sekulić, Fundamentals of Heat Exchanger Designing, John Wiley
& Sons, Inc., USA, 2003. pp. 735–804.
[16] Antonio C. Caputo, Pacifico M. Pelagagge, Paolo Salini, Heat exchanger design
based on economic optimization, Applied thermal engineering 28 (10) (2008)
1151–1159.
[17] State Bureau of Quality and Technical Supervision. Tubular Heat Exchangers,
GB151-1999, Standards Press of China, Beijing, 1999. (in Chinese).
[18] C.R. Houck, J. Joines, M. Key, A genetic algorithm for function optimization: a
Matlab implementation, ACM Transactions on Mathematical Software (1996).
[19] A. Bejan, Advanced Engineering Thermodynamics, Wiley, New York, 1988.
[20] Jiangfeng Guo, Mingtian Xu, Lin Cheng, The application of field synergy
number in shell-and-tube heat exchanger, Applied Energy (2009),
doi:10.1016/j.apenergy.2009.01.013.
[21] M.S. Söylemez, On the thermo economical optimization of single stage
refrigeration systems, Energy and Buildings 36 (9) (2004) 965–968.