This article was downloaded by: [Johns Hopkins University]

On: 01 January 2015, At: 14:21

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Heat Transfer Engineering

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uhte20

Application of a Genetic Algorithm for Thermal Design

of Fin-and-Tube Heat Exchangers
a a b
Gongnan Xie , Qiuwang Wang & Bengt Sunden
a
State Key Laboratory of Multiphase Flow in Power Engineering, Xi'an Jiaotong University ,
Xi'an, China
b
Division of Heat Transfer , Lund University , Lund, Sweden
Published online: 07 Oct 2011.

To cite this article: Gongnan Xie , Qiuwang Wang & Bengt Sunden (2008) Application of a Genetic Algorithm for Thermal
Design of Fin-and-Tube Heat Exchangers, Heat Transfer Engineering, 29:7, 597-607, DOI: 10.1080/01457630801922337

To link to this article: http://dx.doi.org/10.1080/01457630801922337

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Heat Transfer Engineering, 29(7):597–607, 2008
Copyright 
C Taylor and Francis Group, LLC
ISSN: 0145-7632 print / 1521-0537 online
DOI: 10.1080/01457630801922337

Application of a Genetic Algorithm

for Thermal Design of Fin-and-Tube
Heat Exchangers

GONGNAN XIE,1 QIUWANG WANG,1 and BENGT SUNDEN2

1
State Key Laboratory of Multiphase Flow in Power Engineering, Xi’an Jiaotong University, Xi’an, China
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

2
Division of Heat Transfer, Lund University, Lund, Sweden

Instead of the traditional trial-and-error process, a genetic algorithm (GA) is successfully applied to thermal design of
fin-and-tube heat exchangers (FTHEs). The design method uses a GA to search and optimize structure sizes of FTHEs. The
minimum total weight or total annual cost of FTHEs is taken as the objective function in the GA, respectively. Seven design
parameters are varied for the optimization objectives. The implementation of the design method consists of a GA routine
and a thermal design routine. In the GA routine, binary coding for tournament selection, uniform crossover, and one-point
mutation is adopted. In the thermal design routine, thermal design of the FTHE is carried out according to the conditions of
the structure sizes that the genetic algorithm generated, and the log-mean temperature difference method is used to determine
the heat transfer area under the combined structure sizes for a given heat duty. Optimization shows that it is possible to
achieve a great reduction in cost or weight, whenever such objectives have been chosen for minimization. The method is
universal and may be used for thermal design and optimization of FTHEs under different specified duties.

INTRODUCTION applied to industrial engineering applications. For gas-liquid

heat exchangers, fins can be used outside the tubes, or even inside
Compact heat exchangers (CHEs), including two types of tubes, based on different operating conditions and fluid types.
heat exchangers such as plate-fin types and fin-and-tube (tube- For gas-gas heat exchangers, fins can be employed on both sides.
fin) types, are widely used for gas-gas or gas-liquid applications. Heat exchanger design is a traditional and classical issue.
CHEs are compact, highly effective, small volume, and low in It can be based on previous design experience and industrial
weight and cost. Fin-and-tube heat exchangers (FTHEs) are em- requirements. The design of heat exchangers, in general,
ployed in many power engineering and chemical engineering including geometrical parameters and operating specifications,
applications, such as compressor intercoolers, air coolers, and cost estimation, and optimization, represents a complex process.
fan coils. Plate-fin heat exchangers (PFHEs) are widely used in The geometrical parameters are chosen first for a specified heat
gas-gas applications such as regenerators and recuperators in mi- duty under conditions of fouling. These parameters then are
croturbines [1, 2]. The FTHE (as shown in Figure 1) is one of the changed manually by a trial-and-error process to satisfy the
successful improvements of tubular heat exchangers. Because a condition on allowable pressure drop. Subsequently, the heat
relatively large thermal resistance is encountered on the gas side transfer coefficients (i.e., the heat duty) and friction factors (i.e.,
(generally, air side), fins are employed on the gas side to enlarge the pressure drop) are recalculated until both the heat duty and
the heat exchanger surface and increase the disturbance of the pressure drop meet specified requirements. Thus, the design
flow. In addition, if the operating pressure is high on the other task is a complex trial-and-error process and there is always
side, it generally is economical to use tubes. In the last 10 years, the possibility that the designed results are not the optimum.
many high-performance fin structures have been developed and In this sense, novel design and optimization methods as well
as optimization techniques for design of processes need to be
developed. In recent years, much literature has appeared on the
Address correspondence to Qiuwang Wang, State Key Laboratory of
Multiphase Flow in Power Engineering, School of Energy and Power
design and method of shell-and-tube exchangers [3–6], plate-
Engineering, Xi’an Jiaotong University, Xi’an, 710049, China. E-mail: fin/plate heat exchangers [7–10], and air-cooled heat exchangers
[email protected] [11–14].

597
 598 G. XIE ET AL.

Figure 2 Principal picture of economic optimization of heat exchangers.

Figure 1 Fin-and-tube heat exchanger.

cost) should be considered. As the material is selected, the

Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

In recent years, application of genetic algorithms (GAs) minimum area, to some degree, has to be considered relative
in thermal engineering has received much attention for to minimum weight, as in some applications the minimum
solving real-word problems [15]. For example, the fin-tube weight is important (e.g., in aerospace and air-refrigeration
heat exchanger performance was predicted using a GA [16]. applications). For this reason, in this work the minimum
Plate-fin heat exchangers were optimized by means of GAs total weight or annual cost of the designed heat exchangers,
[17–19]. A new design method was proposed to optimize a respectively, are considered as optimization objectives in GA in
shell-and-tube heat exchanger from an economic point of view order to observe the effects on the designed results, including
by a GA [20]. Optimization of the geometry of cross-wavy geometrical parameters and heat transfer area.
and cross-corrugated primary surface recuperators was studied
via GAs [21, 22]. The coefficients of heat transfer correlations THERMAL MODEL AND DESIGN PARAMETERS
for compact heat exchangers were obtained using GAs, based
on experimental data, and, in turn, these correlations were According to established knowledge, high effectiveness,
used to estimate their performance [23, 24]. These reports small volume and low cost are the common objectives in heat
suggested that GAs have a strong ability of search and combined exchanger design. In engineering practice, there are two design
optimization and successfully can optimize and predict thermal requirements to decrease volume/weight and annual cost of the
problems. Applications of GA in the field of thermal engineer- heat exchanger as much as possible under specified effective-
ing are new challenges. At this point, the GA technique may ness and allowable pressure drop; and to increase effectiveness
be used in the optimization design process in order to obtain as much as possible, given limitations in weight or dimensions.
optimal results under a specified heat duty within allowable Thus, before the optimal design is carried out, the design ob-
pressure drops. The aforementioned research draws much jective should be considered based on different requirements.
attention to design and optimization of shell-and-tube heat Without loss of generality, in the present work, the minimum
exchangers and plate/plate-fin heat exchangers. However, only total weight or total annual cost are considered. This means that
a few of the reported studies are concerned with thermal design the minimum total weight (TW) or total annual cost (TAC) are
and optimization of FTHEs by GAs. For this reason, the overall objective functions in GA, respectively.
objective of this study is to present a new design method of The appearance of GA and its many successful applications
FTHEs to find optimal design parameters within allowable pres- have shown that GA owns the ability of strong search and com-
sure drop, based on GA auto-search and optimization technique bined optimization. Thus, in the present study GA is used for the
without trial-and-error on the geometrical parameters. thermal design process instead of the traditional trial-and-error
On the other hand, for heat exchanger design, the balance process, and for optimization of structure sizes under specified
(trade-off) between heat transfer and pressure drop may be performance and design objectives.
considered (as shown in Figure 2). In general, a higher flow
velocity means a higher heat transfer coefficient and hence, a Fin-and-Tube Heat Exchanger
smaller heat transfer area and, correspondingly, a lower capital
cost. However, higher velocity generally will lead to higher In the present study an intercooler is considered. This inter-
pressure drop and higher power consumption and thus, a higher cooler is a so-called FTHE, as shown in Figure 3. For a FTHE,
power cost [25–28]. The heat exchanger area and the pressure hot air flows normal to a finned tube bundle, while cold water
drop mainly are associated with capital cost and operating cost, flows inside the smooth tubes. There are many design param-
respectively. Thus, before carrying out the thermal design, the eters, such as shape length, L, shape width, W , shape height,
design objective (e.g., minimum heat transfer area, minimum H , outside diameter of tube, Do , fin collar outside diameter, Dc ,
heat transfer engineering vol. 29 no. 7 2008
 G. XIE ET AL. 599

Table 1 Design specifications

Variable Unit Value

Air side Flow rate kg/s 58.2

Inlet temperature ◦C 104
Outlet temperature ◦C 51
Inlet pressure kPa 174.5
Allowable pressure drop Pa 5,200
Water side Flow rate kg/s 39.2
Inlet temperature ◦C 20
Inlet pressure kPa 174.5
Allowable pressure drop Pa 5,200
Heat duty kW 3115

GENETIC ALGORITHM
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

Simple Descriptions

The idea behind the genetic algorithm is to do what na-

ture does. The genetic algorithm originally was developed by
Holland [29], and was later comprehensively decrypted and im-
plemented into codes by Goldberg [30]. The genetic algorithm is
maintained by a population of parent individuals that represent
the latent solutions of a real-world problem. Figure 4 shows the
flow chart of the GA. Each individual is assigned a fitness based
on how well each individual fits a given environment and then is
evaluated by survival of the fitness. Fit individuals go through the
process of survival selection, crossover, and mutation, resulting
in creating the next generation, called child individuals. A new
Figure 3 Schematic view of fin-and-tube heat exchanger.

thickness of tube, longitudinal tube pitch, Pl , transverse tube

pitch, Pt , fin pitch, Fp , fin space, Fs , fin thickness, δf , and num-
ber of tube rows, N . In this work, the thicknesses of the tube
and fins are assumed to be constant, and thus are not to be op-
timized as they satisfy initial requirements, and almost do not
affect the performance of the heat exchanger. Note that plain fins
are selected on the outside of the tube bundle. However, other
fins such as plain wavy fins, slotted fins or louvered fins may be
adopted and the design method is straightforward.

Design Parameters

Table 1 lists the design specifications. The flow manner of

such a FTHE is cross-flow and both fluids are unmixed. The
material of the tubes is stainless steel with a thermal conductivity
of 15 W/(m · K), density of 7820 kg/m3 , while the fin material is
aluminum with a thermal conductivity of 170 W/(m · K), density
of 2790 kg/m3 . Seven design parameters, W , H , N , Do , Pt , Pl ,
and Fp are selected for optimization, as these parameters are
sufficient for the design of a FTHE. Note that the shape length,
L , can be determined by the number of tube rows, N , and the
longitudinal pitch of the tubes, Pl . The seven design parameters
are searched and combined in a genetic algorithm routine. Figure 4 Flow chart of genetic algorithm (GA).

heat transfer engineering vol. 29 no. 7 2008

 600 G. XIE ET AL.

Table 2 Search ranges and string length of design parameters

Variable Unit Search range Valid range Resolution String length

Diameter of tube, Do mm 7.0–13.0 6.9–13.6 0.01 10

Transverse pitch, Pt mm 20.5–30.5 20.4–31.8 0.01 10
Longitudinal pitch, Pl mm 13.0–32.0 12.7–32.0 0.01 11
Number of tube rows, N — 2–6 1–6 1 3
Fin pitch, Fp mm 1.0–8.5 1.0–8.7 0.01 10
Height of shape, H m 4.5–8.0 — 0.01 9
Width of shape, W m 3.0–5.0 — 0.01 8
Re in air side — 300–2∗ 104
Re in water side — 2300–106
W/Do — >=60

population therefore is formed by selection of good individuals It should be noted that in Table 2, the valid ranges refer to
from parent and child individuals. After some generations, the the ranges in which the empirical air-side correlations, based on
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

algorithm is converged in a best individual, which probably rep- 74 tested heat exchangers [32, 33], can be used with confidence.
resents the best solution of the given problem. More details about The search ranges in the table are selected based on the valid
description of GAs can be found in many books [30, 31]. The GA ranges so that the geometrical parameters and Reynolds numbers
is a probabilistic and robust algorithm, it aims at such complex can satisfy the valid ranges of the correlations. The detailed
problems, which belong to large-scale combinational optimiza- principles of the coding and decoding processes can be found in
tions and highly constrained engineering problems [31]. the abovementioned books [30–31].

Genetic Operators and Parameters

Variables Ranges
In the present study, tournament selection, uniform crossover,
Binary string is adopted for encoding the variables of a given and one-point mutation were selected, and niching (sharing) and
model, since in engineering applications it is easy to imple- elitism were adopted [34–36].
ment the way for coding and encoding. The search ranges and
binary string length of the seven design parameters are listed
in Table 2. The search range refers to the range for search in Tournament selection: Random pairs of individuals are selected
GA routine, while valid range refers to the application range from the population under a given probability, and the better
for the empirical correlations. Considering the ability of com- (with larger fitness) of each pair is allowed to mate. Thus one
puter handling bit operation and engineering applications, the child is created, which becomes some mix of the two parent
computational precision is set to two decimals (0.01). Although individuals.
some variables are discrete, they are considered as continuous Uniform crossover: It is possible to obtain any combination
for optimization purposes. To represent a set of geometrical pa- of two parents individuals (e.g., there exist two individuals,
rameters, an individual (one possible design) is denoted in the 011100 and 101011, thus the children could be 111010 and
form of a 61-character binary string. It can be seen from Table 001001).
2 that there are 670 choices of tube diameter (there exists 670 One-point mutation: There is a small probability that one or
values between 6.9–13.6 under the resolution of 0.01), 1,140 more of the children will be mutated (e.g., 111010 could be
choices for the transverse pitch, 1,930 choices for the longitudi- 101010).
nal pitch, 6 choices for the number of tube rows, 770 choices for Niching (sharing): The fitness of each individual is adjusted
the fin pitch, 350 choices for the height, and 200 choices for the according to similar degree, which is evaluated by a specified
width. Thus, here the design problem is considered as a large sharing function. The selection is conducted based on the
scale, discrete, and combined optimization problem. Based on new fitness. This technique ensures the variety of individuals
these choices of design parameters, the total number of possible (solutions) so that the global optimization solution will be
designs is 670 × 1140 × 1930 × 6 × 770 × 350 × 200 = obtained rather than the local optimization solution.
1015 . This number means that if the trial-and-error process for Elitism: The best parent is reproduced (copied) into the new
manual design is used, it is probably impossible to find the op- population. After the new population is generated, the GA
timal design as it will require enormous time and work until the checks if the best parent has been replicated.
global optimization design is reached. So the novel strategy or
technique of auto-search and combination will be necessary to The size of population and maximum evolution generation
present, and the GA might be the one among those optimization are set to 50 and 1,000, respectively. In the field of artificial
techniques that might do the job. intelligence, it is common to set the convergence criterion to be
heat transfer engineering vol. 29 no. 7 2008
 G. XIE ET AL. 601

reaching a maximum generation. Probability of crossover and It is worthwhile to note that the equivalent diameter in above
mutation are set to 0.5 and 0.005, respectively. The probability correlations is Dc . Then the heat transfer coefficient h can be
means that in the GA routine if the generated random number is determined by
less than a bias, the genetic operator is processed. The selection ρu c c p
of genetic parameters is a trial-and-error process, and with the h= j (2)
variation of these parameters, results are not exactly identical but Pr2/3
are very close to one another. On the other hand, in this study The friction factor, is calculated by the following correlation
the selection of genetic parameters and operators is based on the   f2   f3
previous knowledge, that is, on the Carroll’s recommendations f Pt Fp
f = 0.0267Re D1c (3)
[34, 35]. The task of thermal design is conducted in FORTRAN Pl Dc
language, with the main program’s job being to to run GA and
where
the subroutine program to carry out the thermal design process
under the specified heat duty. Pt Fp 0.00758
f 1 = −0.764 + 0.739 − 0.177 − (3a)
Pl Dc N

64.021
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

DESIGN METHOD AND DESIGN OBJECTIVE f 2 = −15.689 + (3b)

loge (Re Dc )
In this article, two objective functions are considered in the
15.695
design process. One is the minimum total weight (TW) and the f 3 = 1.696 − (3c)
other is the minimum total annual cost (TAC). The traditional de- loge (Re Dc )
sign method, log-mean temperature difference (LMTD) method, Then the pressure drop can be determined by
is used in the thermal design of FTHEs. The detailed expressions
  
and procedures are as follows. G 2 A ρm   ρ1
Ph = c f + 1 + σ2 −1 (4)
2ρ1 Ac ρ1 ρ2

Design Method For the water side, the friction factor, f, and Nusselt number,
Nu, can be obtained by the following correlations developed by
Once the geometrical (structure) sizes have been obtained by Gnielinski [37],
the GA-based individuals decoded, thermal design can be carried f = (1.82 log10 Re − 1.64)−2 (5)
out under the specified requirements. The LMTD method is used.
A step-by-step process is conducted to complete the design task. ( f /8)(Re − 1000) Pr
Details of the design process can be found in previous work [13, Nu = √ (6)
1 + 12.7 f /8(Pr2/3 −1)
14]. First the initial total heat transfer area Ain , the minimum
free-flow area Ac and Reynolds numbers Re on both sides are Then the heat transfer coefficient and pressure drop in water
calculated. Then the j-factors (Colburn factors) and f -factors side can be determined by
(friction factors) are calculated by the following correlations.
For the air side, when the number of tube rows is larger than h = Nu · k/Di (7)
one, the j-factors can be obtained [32, 33]
  j5   j6  −0.93 f · ρu 2i L
Fp Fp Fp Pc = (8)
j3
j = 0.086Re Dc N j4 (1) 2 Di
Dc Dh Pt
Note that whether the fluid flow and heat transfer is laminar or
where turbulent, it does not enter directly in the optimization process.
  0.41  In the thermal design process, only attention is drawn on the
0.042N Fp
j3 = −0.361 − + 0.158 ln N (1a) selection of correlations rather than the flow manner, and these
ln(Re Dc ) Dc
correlations are valid in wide ranges of the Reynolds number (as
listed in Table 2). On the other hand, the search ranges of the
0.076(Pl /Dh )1.42 design parameters in GA are restricted to the valid ranges of the
j4 = −1.224 − (1b)
ln(Re Dc ) correlations and whatever search variables being combined, the
obtained Re will be in the range of the correlations.
0.058N Figure 5 shows a flow chart of the design process of FTHEs.
j5 = −0.083 + (1c)
ln(Re Dc ) A step-by-step procedure for the heat exchanger design is listed
  in [13, 14]. In Figure 5, the reason for allowing solutions with
Re Dc area up to 20% higher is that in heat exchanger design other
j6 = −5.735 + 1.21 ln (1d)
N complex factors not considered will need additional area.
heat transfer engineering vol. 29 no. 7 2008
 602 G. XIE ET AL.

Total annual cost = annualized cost of heat transfer area +

operating cost of pump/compressor
TAC(xi ) = Cin + Cop (10)

Cin = CA ∗ An (10a)
 
P Vt P Vt
Cop = kel τ + kel τ (10b)
η h η c

here CA and kel are the price of per unit area and electrical en-
ergy, respectively, n and τ are the exponent of nonlinear increase
with area increase and the hours of operation per year, respec-
tively; P, Vt , and η are pressure drop, volumetric flow rate,
and pump/compressor efficiency, respectively.
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

On the other hand, when the geometrical sizes generated can-

not satisfy the specified performance, a penalty function has to

Figure 5 Flow chart of thermal design of heat exchanger.

Objective Functions

When the minimum TW is considered in the thermal design,

the objective function in GA is now set according to
Total weight = total weight of fins + total weight of tubes

T W = Af ∗ ρ f ∗ δ f + N ∗ Nt ∗ ρt ∗ W ∗ (Do2 − Di2 ) (9)

Constraints:

Ph < Ph,max , Pc < Pc,max
(9a)
xmin < xi < xmax
here the Ph,max , Pc,max are the maximum allowable pressure
drop on hot side and cold side, respectively. In order to compare
with previous works, two such values are set, namely 30 Pa
and 4500 Pa. x refers to the design variable to be optimized;
xmax , xmin refer to the upper bound and lower bound of the design
variables, respectively.
When the minimum TAC is considered in thermal de-
sign, the objective function in GA is now set [7, 8, 38], as
follows: Figure 6 Evolution process for minimum weight.

heat transfer engineering vol. 29 no. 7 2008

 G. XIE ET AL. 603

Table 3 Design results for different objectives

Variables Unit Reference[13] Minimum weight Minimum cost

Number of tube row / 5 2 2

Number of tubes / 1180 502 454
Longitudinal pitch of tube mm 22.0 27.26 27.26
Transverse pitch of tube mm 25.4 28.94 21.78
Inside diameter of tube mm 8 10.63 11.00
Outside diameter of tube mm 10 12.63 13.00
Number of fins / 2330 3318 2645
Fin pitch mm 2.0 1.23 2.87
Fin thickness mm 0.12 0.12 0.12
Width m 4.66 4.08 4.95
Height m 6 7.29 7.59
Length m 0.11 0.0545 0.0545
Volume m3 3.08 1.62 2.05
Heat transfer area m2 2,553.32 2,294.48 1,114.45
Reynolds number on air side / 1,790 2,365 2,535
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

Pressure drop on air side Pa 30 30 30

Reynolds number on water side / 6,590 11,000 12,470
Pressure drop on water side Pa 4,532 4,464 2,654
Total weight kg 2,103.76 1,335.05 1,388.52
Total annual cost $/year 11,397.14 10,873.74 7422.50

be conducted on the objective function. In this article, the step- 30 $/MWh, the pump efficiency is assumed to be 0.5. Note that
wise penalty factor is defined as follows the selections of cost parameters (e.g.,C A , n, kel ) depend on cur-
 rent practice experiences or recommended values [7, 8, 38]. The
1 1 < (Ain /A) < 1.2 and P < Pmax
pf = (11) real results will depend on these parameters but the proposed
0 others
procedure is general and is applicable for any selected values.
Thus, the fitness of the individuals should be adjusted into fol- The selected values thus represent a case study.
lowing equations: The evolution process for minimum weight and required heat
transfer area is shown in Figure 6. At the beginning of the evo-
For minimum TW: Fitness = p f · (T Wmax − T W ) (12)
lution process (about less than 50 generations), the individuals
with higher fitness are saved, and the individuals with small fit-
For minimum TAC: Fitness = p f · (TACmax − TAC) (13) ness are removed. After certain generation (about larger than 100
TWmax , TACmax are constants and assigned to 2,000 and 106 , generations), the differences between every individual are rela-
respectively. That is, the weight and annual cost of FTHE are tively large, in turn the variation of fitness for minimum weight
not larger than 2,000 kg and 106 $, respectively. The case of is small, finally a level off value is found. The designed results
invalid design with minimum fitness (is zero), will be removed under different design objectives and those in Xie et al. [13] are
from the possible combinations, and not be updated in the GA shown in Table 3, in which column 4 lists the designed results
evolution. for minimum weight. It is seen that the number of tube rows
decreases to two, however the area is almost similar because the
diameter of the tubes increases. The fin thickness decreases but
A CASE STUDY AND DESIGN RESULTS the number of fins is increased. The length of the shape decreases
by about 50%, and the volume decreases by about 50%. The to-
To demonstrate the described procedure, a case study is con- tal weight of the designed heat exchanger decreases by 35%. In
sidered as follows. this sense, in all of the possible designs, the results are optimum
Design a fin-and-tube heat exchanger with the following duty under the condition of minimum weight, and those presented in
(as listed in Table 1): 39.17 kg/s water enters tubes at 20◦ C to Xie et al. [13] are not the optimal.
cool 58.2 kg/s air from 104◦ C to 51◦ C. The inlet pressure is The number of valid designs of the heat exchanger for min-
174.5 kPa, and the heat duty for this case is 3155.45 kW. The imum weight is shown in Figure 7. The aforementioned design
allowable pressure drops on both sides are assigned about 3% procedure has shown that when the design is not valid, for ex-
(5,200 kPa) of inlet pressure. ample, heat transfer area does not meet the condition, or the cal-
For the economic calculations, the cost per unit area, C A , is culated pressure drop is not within the allowable pressure drop,
assigned to be 100 $/m2 , the exponent for the nonlinear increase the penalty function is added to the objectives (i.e., the fitness
with area increase, n, is assigned to be 0.6, the hours of oper- of individuals are assigned to zero). The individuals with lowest
ation is assumed to be 6,500 h/year, the price of electricity is fitness will be removed from the population, and the individuals
heat transfer engineering vol. 29 no. 7 2008
 604 G. XIE ET AL.

Figure 7 Number of valid designs for minimum weight.

with high fitness will possibly be generated through the genetic

Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

operators (selection, crossover, mutation). From Figure 7, the

number of individuals, which represents the valid design, are
larger than 70% of the population size. Thus, in every evolution
the best (optimal) design parameters, based on the desired per-
formance under the design objective, will be determined from
these individuals that possibly represent the best.
The evolution process for minimum annual cost and corre-
sponding weight is shown in Figure 8, where after a certain
generation (about larger than 400 generations), the differences
between every individual are relatively large, and as an effect,
the variation of fitness for minimum cost is small and finally
a level off value is reached. In Figure 8b, before about 200
generations, the corresponding weight of minimum cost at ev-
ery generation varies significantly, and after 400 generations the
difference between every generation is almost small. The de-
signed results for minimum annual cost compared to those in
Xie et al. [13], are listed in column 5 of Table 3. It is found that,
the number of tube rows decreases to two. The fin thickness in-
creases and in turn the number of fins is decreased. The length
Figure 8 Evolution process for minimum annual cost.
of shape decreases by about 50%, and the volume decreases by
about 30%. The total weight of the designed heat exchanger de-
as an objective in the design, the results are optimum compared
creases by 30%, and obviously, however, it is slightly larger than
to those in Xie et al. [13]. The results show that it is effective
that for minimum weight. The total annual cost also decreases
to use a GA technique to search and combine optimal parame-
by about 40%, meaning that in all of the possible designs, such a
ters for heat exchangers under different requirements. Making
design is optimum for the design objective of minimum annual
use of a GA provides strong ability of search and combined
cost.
optimization, which effectively might be better than the tradi-
The number of valid designs of heat exchanger for minimum
tional trial-and-error process for manual design of parameters,
weight is shown in Figure 9. Figure 9 shows that the number of
because there is always the possibility that the results from the
individuals representing the valid design is larger than 60% of
latter process are not optimal. As seen from the design proce-
the population size. Thus, in every evolution the best (optimal)
dure, the design results may depend on the selection and ap-
design parameters, based on the desired performance under the
plication of empirical correlations. Errors (deviations) between
design objective, will be determined from these individuals that
designed results and applied results (later through manufactur-
possibly represent the best.
ing and tests based on the design results) surely will be gener-
ated. The correlations used in the present work, have deviations
of about 8% on the airside [32] and about 20% on the water
DISCUSSION side.
In this work, the fin-and-tube heat exchanger is demonstrated,
The designed results in the present work are based on theo- however, other types of heat exchangers such as plate/plate-fin
retical design. According to the minimum weight or annual cost heat exchangers, primary surface recuperators, shell-and-tube
heat transfer engineering vol. 29 no. 7 2008
 G. XIE ET AL. 605

NOMENCLATURE

A Heat transfer area (m2 )

Ain Initial heat transfer area (m2 )
Ac Minimum free flow area (m2 )
CA Price per unit area ($/m2 )
Cin Annual cost of investment ($)
Cop Annual cost of operation ($)
cp Specific heat (kJ/(kg · K))
Dc Fin collar outside diameter (mm)
Dh Hydraulic diameter (mm)
Di Inside tube diameter (mm)
Figure 9 Number of valid designs for minimum annual cost.
Do Outside tube diameter (mm)
Fp Fin pitch (mm)
heat exchangers may be designed and optimized based on the Fs Fin space (mm)
proposed method. For example, when designing a plate-fin heat f
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

Friction factor
exchanger for a given duty, the design parameters, including fin Gc Volumetric velocity on air side (m3 /s)
pitch, fin height, shape sizes, number of passages, thickness of fin H Shape height (m)
and plate, and so on, can be searched and optimized via the GA h Heat transfer coefficient (W/(m2 · K))
integrated with a traditional design method. Besides, although k Thermal conductivity (W/(m · K))
plain fins are employed outside the tubes, other types of high- kel Price of electricity ($/MWh)
performing fins (e.g., plain wavy fins, slotted fins, louvered fins) j Colburn factor
may be considered in the design task. Thus, the GA may be L Shape length (m)
applied in heat exchanger problems to search and optimize the n Exponent in Eq. (10a)
complex thermal devices or networks. N Number of tube rows
Nu Nusselt number
pf Penalty factor
CONCLUSIONS Pl Longitudinal tube pitch (mm)
Pr Prandtl number
This article shows successful application of a genetic algo- Pt Transverse tube pitch (mm)
rithm for thermal design and optimization of fin-and-tube heat Re Reynolds number based on Dh
exchangers. A generalized procedure has been developed to Re Dc Reynolds number based on Dc
carry out the design and an optimization to find the minimum TAC Total annual cost ($)
weight or annual cost of the heat exchangers, respectively, based TW Total weight (kg)
on the LMTD and the genetic algorithm technique. A case study uc Minimum flow velocity (m/s)
has been presented to show the designed results by the proposed ui Velocity inside tube (m/s)
method. U Total heat transfer coefficient (W/(m2 · K))
It is concluded that the GA can provide a strong ability of Vt Volumetric flow rate (m3 /s)
auto-search and combined optimization in the thermal design of W Shape width (m)
heat exchangers without the trial-and-error process, and the heat x Optimization variable
exchanger configurations/structures can be optimized according
Greek Symbols
to different design objectives such as minimum surface area or
cost. The method can be extended for use in thermal design of
heat exchangers with different types of fins, such as plain wavy δf Fin thickness (mm)
fins, slotted fins, louvered fins, and so on. η Pump efficiency
σ Contraction ratio of cross-sectional area
ρ Fluid density (kg/m3 )
τ Hours of operation per year
ACKNOWLEDGMENTS P Pressure drop (Pa)
Pmax Allowable pressure drop (Pa)
This work was supported by National Natural Science Foun-
dation of China (Grant No. 50776068) and Program for New Subscripts
Century Excellent Talents in University of China (Grant No.
NCET-04-0938). The authors would like to acknowledge the 1,2 Inlet, outlet
reviewers for their constructive comments and suggestions. c Cold side
heat transfer engineering vol. 29 no. 7 2008
 606 G. XIE ET AL.

Dc Based on Dc ceedings of GT2006, ASME Turbo Expo 2006, May 8–11, 2006,
f Fin Barcelona, Spain. Paper No. GT2006-90260, 2006.
h Hot side [15] Sen, M., and Yang, K. T., Applications of Artificial Neural Net-
i Inside tube works and Genetic Algorithms in Thermal Engineering, In: Kre-
in Initial ith, F., (Eds), The CRC Handbook of Thermal Engineering, CRC
Press, Boca Raton, FL, pp. 620–661, 2000.
m Mean
[16] Pacheco-Vega, A., Sen, M.,Yang, K.T., and McClain, R. L., Ge-
max Maximum netic Algorithms-based Predictions of Fin-Tube Heat Exchanger
min Minimum Performance, Proceedings of 11th International Heat Transfer
t Tube Conference, August 23–28, Kyongju, Korea. Vol. 6, pp. 137–142,
1998.
[17] Ozkol, I., and Komurgoz, G., Determination of The Optimum
Geometry of the Heat Exchanger Body via a Genetic Algorithm,
REFERENCES Numerical Heat Transfer, Part A, vol. 48, pp. 283–296, 2005.
[18] Xie, G. N., and Wang, Q. W., Geometrical Optimization of Plate-
[1] McDonald, C. F, Low-Cost Compact Primary Surface Recuperator Fin Heat Exchanger using Genetic Algorithms, Proceedings of the
Concept for Microturbines, Applied Thermal Engineering, vol. 20, Chinese Society for Electical Engineering, vol. 26, no. 7, pp. 53–
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

pp. 471–497, 2000. 57, 2006. (in Chinese)

[2] McDonald, C. F, Recuperator Considerations for Future Higher [19] Mishra, M., Das, P. K., and Sarangi, S., Optimum Design of Cross-
Efficiency Microturbines, Applied Thermal Engineering, vol. 23, flow Plate-Fin Heat Exchangers through Genetic Algorithm, Inter-
pp. 1463–1487, 2003. national Journal of Heat Exchangers, vol. 5, no. 2, pp. 379–401,
[3] Poddar, T. K., and Polley, G. T., Optimize Shell-and-Tube Heat 2004
Exchanger Design, Chemical Engineering Progress, vol. 96, no. [20] Selbas, R., Kizilkan, O., and Reppich, M., A New Design Ap-
9, pp. 41–46, 2000. proach for Shell-and-Tube Heat Exchangers using Genetic Algo-
[4] Bell, K. J., Heat Exchanger Design for the Process Industries, rithms from Economic Point of View, Chemical Engineering and
ASME Journal of Heat Transfer, vol. 126, no. 6, pp. 877–885, Processing, vol. 45, no. 4, pp. 268–275, 2006.
2004. [21] Liang, H. X., Xie, G. N., Zeng, M., Wang, Q. W., and Feng, Z.
[5] Serna, M., and Jimenez, A., An Efficient Method for the Design P., Application Genetic Algorithm to Optimization Recuperator
of Shell and Tube Heat Exchangers, Heat Transfer Engineering, in Micro-Turbine, The 2nd International Symposium on Ther-
vol. 25, no. 2, pp. 5–16, 2004. mal Science and Engineering, October 23–25, Beijing, China,
[6] Unuvar, A., and Kargici, S., An Approach for the Optimum Design 2005.
of Heat Exchangers, International Journal of Energy Research, [22] Wang, Q. W., Liang, H. X., Xie, G. N., Zeng, M., Luo, L. Q., and
vol. 28, no. 15, pp. 1379–1392, 2004. Feng, Z. P., Genetic Algorithm Optimization for Primary Surfaces
[7] Jia, R., Sunden, B., and Xuan, Y., Design and Optimization of Recuperator of Microturbine, ASME Journal of Engineering for
Compact Heat Exchangers, 3rd International Conference on Com- Gas Turbines and Power, vol. 129, pp. 436–442, 2007.
pact Heat Exchangers and Enhancement Technology for the Pro- [23] Pacheco-Vega, A., Sen, M.,Yang, K. T., and McClain, R. L., Cor-
cess Industries, Eds: Shah, R. K., Davos, Switzerland, pp. 135– relations of Fin-Tube Heat Exchanger Performance Data Using
142, 2001. Genetic Algorithms Simulated Annealing and Interval Methods,
[8] Jia, R., and Sunden, B., Optimal Design of Compact Heat Ex- Proceedings of ASME the Heat Transfer Division, November 11–
changers by an Artificial Neural Network Method, Proceedings 16, New York, USA, Vol. 369–5, pp. 143–151, 2001.
of HT2003, ASME Summer Heat Transfer Conference, paper [24] Pacheco-Vega, A., Sen, M., and Yang, K. T., Simultaneous De-
no. HT2003-47141, 2003. termination of In-and-over-Tube Heat Transfer Correlations in
[9] Wang, L., and Sunden, B., Design Methodology for Multistream Heat Exchangers by Global Regression, International Journal
Plate-Fin Heat Exchangers in Heat Exchanger Networks, Heat of Heat and Mass Transfer, vol. 46, no. 6, pp. 1029–1040,
Transfer Engineering, vol. 22, no. 1, pp. 3–11, 2001. 2003.
[10] Wang, L., and Sunden, B., Optimal Design of Plate Heat Ex- [25] Jegede, F. O., and Polley, G. T., Optimum Heat Exchanger Design,
changers with and without Pressure Drop Specifications, Applied Trans. IChemE, vol. 70, pp. 133–141, 1992.
Thermal Engineering, vol. 23, no. 3, pp. 295–311, 2003 [26] Muralikrishna, K., and Shenoy, U. V., Heat Exchanger Design
[11] Mukherjee, R., Effectively Design Air-Cooled Heat Exchangers, Targets for Minimum Area and Cost, Trans. IChemE, vol. 78, Part
Chemical Engineering Progress, vol. 93, no. 1, pp. 26–47, 1997. A, pp. 161–167, 2000.
[12] Gonzalez, M. T., Petracci, N. C., and Urbicain, M. J., Air-cooled [27] Jafari Nasr, M. R., and Polley, G. T., An Algorithm for Cost Com-
Heat Exchanger Design Using Successive Quadratic Program- parison of Optimized Shell-and-Tube Heat Exchangers with Tube
ming, Heat Transfer Engineering, vol. 22, no. 1, pp. 11–16, Inserts and Plain Tubes, Chemical Engineering Technology, vol.
2001. 23, no. 3, pp. 267–272, 2000.
[13] Xie, G. N., Chen, Q. Y., Tang, L. H., Zeng, M., and Wang, Q. W., [28] Wang, L., Performance Analysis and Optimal Design of Heat Ex-
Thermal Design and Comparison of two Fin-And-Tube Heat Ex- changers and Heat Exchanger Networks, PhD thesis, Division of
changers, Proceedings of 2nd Chinese Heat Transfer Technology, Heat Transfer, Department of Heat and Power Engineering, Lund
November 8–11, 2005, Hainan, China, 2005. (in Chinese) Institute of Technology, 2001.
[14] Xie, G. N., Chen, Q. Y., Zeng, M., and Wang, Q. W., Thermal De- [29] Holland, J. H., Adaption in Nutrual and Aritifical Systems, Uni-
sign of Heat Exchanger with Fins inside and outside of Tubes, Pro- versity of Michigan Press, Ann Arbor, 1975.
heat transfer engineering vol. 29 no. 7 2008
 G. XIE ET AL. 607

[30] Goldberg, D. E., Genetic Algorithms in Search, Optimization and cation of computational intelligence in thermal engineering. He is currently a
Machine Learning, Addison-Wesley, Boston, USA, 1989. post doc at Lund University. He has published more than 10 articles in interna-
[31] Michalewicz, Z., and Fogel, D. B., How to Solve it: Modern tional journals or conferences, and has participated in several projects.
Heuristics. New York, Springer, 2000.
[32] Wang, C. C., Recent Progress on the Air-Side Performance of
Fin-and-Tube Heat Exchangers, International Journal of Heat Ex-
changers, vol. 2, pp. 57–84, 2000. Qiuwang Wang is a professor at the School of
[33] Wang, C. C., Heat Exchanger Design, Wunan Press, Taiwan, Energy and Power Engineering, Xi’an Jiaotong
China, Chap. 4, 2003. University, Xi’an, China. He received his Ph.D.
[34] Carroll, D. L., Chemical Laser Modeling with Genetic Algorithms, degree in Engineering Thermophysics from the
Xi’an Jiaotong University in 1996. He then joined
AIAA Journal, vol. 34, no. 2, pp. 338–346,1996.
the faculty of the university and took the professor
[35] Carroll, D. L., Genetic Algorithms and Optimizing Chemical
post in 2001. His main research interests include
Oxygen-Iodine Lasers, In: Wilson, H., Batra, R., Bert, C., Davis, computational fluid dynamics and numerical heat
A., Schapery, B., Stewart, B., and Swinson, F. (Eds). Develop- transfer, heat transfer enhancement, compact heat
ments in Theoretical and Applied Mechanics, Vol. XVIII, School exchangers, building energy saving and indoor air
of Engineering, The University of Alabama, pp. 411–424,1996. quality, etc. He has published more than 100 journal papers, half of which have
[36] Michalewicz, Z., Deb, K., Schmidt, M., and Stidsen, T., Test-case been published in international journals or conferences. He has also obtained 9
Downloaded by [Johns Hopkins University] at 14:21 01 January 2015

Generator for Constrained Parameter Optimization Techniques, Invent Patents of China.

IEEE Transactions on Evolutionary Computation, vol. 4, no. 3,
pp. 197–215, 2000.
[37] Gnielinski, V., New Equations for Heat and Mass Transfer in Tur-
bulent Pipe and Channel Flows, Ind Chem Eng, vol. 16, pp. 359–
368, 1976.
[38] Martin, H., Economic Optimization of Compact Exchangers, 1st Bengt Sunden received his M.S. and Ph.D. from
International Conference on Compact Heat Exchangers and En- Chalmers University, Göteborg, Sweden. He is
hancement Technology for the Process Industries, Eds: Shah, currently professor of heat transfer and depart-
R. K., Banff, Canada, pp. 75–80, 1999. ment head at Lund University, Sweden. His main
research interests are computational heat trans-
fer, heat exchangers, transport phenomena in
fuel cells, gas turbine heat transfer, combustion-
Gongnan Xie received his Ph.D in Power Engi- related heat transfer and enhanced heat transfer.
neering and Engineering Thermo Physics from Professor Sunden has published more than 350
the School of Energy and Power Engineering, articles in well-recognized journals, books and
Xi’an Jiaotong University, Xi’an, China. He re- proceedings. He has edited twenty books. He is the editor-in-chief of the In-
ceived his M.S. degree in Thermal and Power ternational Journal of Heat Exchangers, editor-in-chief for a book series Devel-
Engineering from Guangdong Ocean University opments in Heat Transfer. In addition, he is on the editorial board for another
in 2002, Zhanjiang, China. His research interests four journals. He is a fellow of ASME and serves as associate editor of journal
include computational fluid dynamics, numerical of Heat Transfer. He is an honorary professor of Xi’an Jiatong University, Xi’an,
heat transfer, compact heat exchangers and appli- China.

heat transfer engineering vol. 29 no. 7 2008