Académique Documents
Professionnel Documents
Culture Documents
PII: S1359-4311(13)00846-6
DOI: 10.1016/j.applthermaleng.2013.11.042
Reference: ATE 5187
Please cite this article as: H. Zarea, F.M. Kashkooli , A.M. Mehryan, M.R. Saffarian, E.N. Beherghani,
Optimal design of plate-fin heat exchangers by a Bees Algorithm, Applied Thermal Engineering (2013),
doi: 10.1016/j.applthermaleng.2013.11.042.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
HIGHLIGHTS
We analyzed a plate-fin heat exchanger with offset strip fins.
This is the first application of a Bees Algorithm for plate-fin heat exchanger design.
Preliminarily design and effectiveness of PFHE was improved by minimizing the entropy generation units.
The results show the superiority of this method over GA, PSO and ICA and preliminary design.
PT
RI
U SC
AN
M
D
TE
C EP
AC
1
ACCEPTED MANUSCRIPT
Hossein Zarea1, Farshad Moradi Kashkooli1*, Abdollah Mansuri Mehryan1, Mohammad Reza
PT
1
Department of Mechanical Engineering, Shahid Chamran University of Ahvaz, Ahvaz, Iran.
RI
In this study, the application of Bees Algorithm (BA) in the optimum design of a cross flow plate-fin heat exchanger
with offset strip fin is investigated. First, heat exchanger is optimized and designed according to the effectiveness
SC
optimization. Then, an analysis based on the second law of thermodynamics and minimizations of entropy
generation units is performed. Specific heat duty, space restriction and permitted pressure drop are considered as the
U
constraints for the number of entropy generation unit's minimization. Hot and cold flow length of the heat
AN
exchangers, number of fin layers, fin frequency, fin height, fin strip length and fin thickness are introduced as
optimization variables. The effectiveness and accuracy of the suggested algorithm are compared with literatures.
The results have shown that BA can find optimum configuration with higher accuracy in comparison with Genetic
M
Algorithm (GA), Particle Swarm Optimization (PSO), Imperialist Competitive Algorithm (ICA) and preliminary
D
design.
TE
Key words: Plate-fin heat exchanger, offset strip fin, Bees Algorithm, optimization.
1. Introduction
EP
One of the important types of heat exchangers is compact heat exchanger. The advantage of these
C
exchangers is high heat transfer area in volume unit which leads to decrement of space, weight,
AC
improvement of energy efficiency, required cost and improvement of design process in comparison with
ordinary heat exchangers [1]. Plate fin heat exchanger (PFHE) is one of the important types of compact
heat exchangers. Due to low volume and weight, high effectiveness and multi-current capability, these
exchangers are widely used in gas to gas application including: micro-turbines, cryogenics (cooling in
*
Corresponding author. Tel. :( +98) 917-7360841; Fax :( +98) 711-6123864; E-mail: Farshadmoradikashkooli@ymail.com.
1
ACCEPTED MANUSCRIPT
very low temperature), petroleum, petrochemical, chemical and aerospace industries [2]. The design of a
PFHE is a complex process based on trial-and-error method in which geometrical and operational
parameters are selected to satisfy specified requirements such as outlet temperature, heat duty and
pressure drop. Some of commonly used fins in these exchangers are plain, wavy, louver, perforated, offset
PT
strip and pin fins. Since the average boundary-layer thickness decreases significantly when offset-strip
fins are used, the convection coefficient increases. Therefore, these fins have higher heat transfer
RI
effectiveness than plain flat fins [3]. Generally, high effectiveness of compact heat exchangers is related
SC
to the high pressure drop. Thus, finding a relation between heat transfer increment and consumption
power due to high pressure drop is necessary and the analysis based on the second law of
U
thermodynamics is used for this aim and could be the best method to evaluate this situation [4-6].
Minimizing entropy generation method (EGM) was described as a modeling and optimization method by
AN
Bejan [7]. Many works have been dedicated to the optimization of heat exchangers using traditional
mathematical techniques [8– 12]. Recently, application of evolutionary algorithms has gained much
M
attention in design of heat exchangers. In the early efforts, GA was successfully used for optimization of
D
shell-and-tube heat exchanger [13, 14]. Recently, Genetic Algorithm based on the random search is
TE
widely used in designing and optimization of compact heat exchangers. Ahmadi et al [15] minimized the
number of entropy generation units and total annual cost as two objective functions in a plate fin heat
EP
exchanger with offset strip fins. Sanaye and Hajabdollahi [3] used GA to obtain the maximum
effectiveness and the minimum total annual cost as two objective functions in a plate fin heat exchanger
C
with offset strip fins. Najafi et al [16] developed a multi objective Genetic Algorithm to obtain a set of
AC
design geometrical to achieve two conflicting objectives, namely total heat transfer rate and total annual
cost. Xie et al [4] utilized a GA to optimize a compact heat exchanger under pressure drop restriction. The
main objective of optimization was minimization of the total annual cost minimization and total volume.
Mishra et al [2] used Genetic Algorithm to optimize cross flow plate fin heat exchanger on the basis of
the second law of thermodynamics to minimize the number of entropy generation units for a specified
heat duty under given space restriction. In another study, Mishra et al [17] applied GA for optimal design
2
ACCEPTED MANUSCRIPT
plate fin heat exchangers. The authors had considered minimization of total annual cost and total thermo
economic cost as an objective function. Some researchers also have used the Particle Swarm Optimization
algorithm to optimize the heat exchangers. Rao and Patel [5] utilized PSO algorithm to optimize a cross
flow plate fin heat exchanger. The authors had considered minimization of numbers of entropy generation
PT
units, total volume and total annual cost as target functions. Peng and Xiang [18] presented optimized
PSO algorithm to find optimal design parameters of a plate fin heat exchanger with permitted pressure
RI
drops. Yousefi et al [19] employed an ICA to optimize a cross-flow plate fin heat exchanger with the aim
SC
of minimizing the entropy generation units (EGU). Their results also demonstrated the better performance
of ICA over traditional GA. Yousefi et al [20] used an ICA to optimize a PFHE considering minimization
U
of total annual cost and total weight under given constrained conditions. Comparing their result to the
GA, they demonstrated that ICA presents shorter computational time and better results for their case.
AN
Yousefi et al [21] used a GA Hybrid with PSO (GAHPSO) for design optimization of a PFHE. A total
number of seven design parameters are considered as an optimization variable and the constraints are
M
handled by penalty function method. Comparing the results using GA and the GAHPSO can converge to
D
optimum solution with higher accuracy. In another work, Yousefi et al [22] presented Improved Harmony
TE
Search Algorithm for design optimization of plate-fin heat exchangers. Their algorithm deals with a large
number of continuous and discrete variables. To handle the constraints in the optimization problem, a
EP
self-adaptive penalty function scheme is used. Their numerical results indicated that the present approach
can generate optimum solutions with higher accuracy when compared to GA, PSO and GA hybrids with
C
PSO.
AC
Pham et al [23] examined the first application of the Bees Algorithm for optimal design of mechanical
problems. In this study, two standard samples of mechanical design including the design of the welded
beams and the helical springs design were considered. The purposes of this work were comparing the
Bees Algorithm related to the other optimization algorithms. The results showed the high performance of
this algorithm. In the other study, Pham et al [24] presented the first application of the Bees Algorithm for
3
ACCEPTED MANUSCRIPT
optimizing multi objective problems. They design welded beams problem with using multi objective
optimization BA. The results of this work were excellent compared with other optimization algorithms.
In this study, bees optimization algorithm has been used to optimize plate fin heat exchanger and in
order to maximize the effectiveness and to minimize the number of entropy generation units as two
PT
separate objective functions. The main goals of this study are (i) optimization of effective parameters in
the plate-fin heat exchangers to maximize the effectiveness and to minimize the number of entropy
RI
generation units and (ii) demonstrating the effectiveness of BA for optimization of cross flow plate-fin
SC
heat exchanger.
2. Thermal modeling
U
AN
Fig. 1 and Fig. 2 show a view of a cross flow plate-fin heat exchanger and offset strip fins with
1) For minimizing heat loss to the environment, the Number of fin layers for the cold side is assumed to
D
3) Heat transfer coefficient and the area distribution are assumed constant and uniform.
EP
4) Plate’s thickness is assumed negligible, thus, thermal resistance and longitude heat transfer of walls
C
can be negligible.
AC
In this study, while output temperature of fluids is unspecified, the ε–NTU method has been used for
4
ACCEPTED MANUSCRIPT
According to the Bejan's methodology [7], rate of entropy generation can be explained in terms of
T a ,2 Pa ,2 T b ,2 Pb ,2
S& = m a c p ,a Ln − R a Ln + m b c p ,b Ln − R b Ln )1 (
T
a ,1 Pa ,1 T b ,1 Pb ,1
PT
In the above equation, Ta,2 , Tb,2 , Pa,2 and Pb,2 are the outlet temperature and pressure of the fluid a and
fluid b respectively, which are determined by considering effectiveness of heat exchanger as:
RI
c a (T a ,1 −T a ,2 ) cb (T b ,2 −T b ,1 )
ε= = )2 (
c min (T a ,1 −T b ,1 ) c min (T a ,1 −T b ,1 )
SC
Therefore:
(T a ,1 −T b ,1 )
c min
T a ,2 = T a ,1 − ε
U
)3 (
ca
T b ,2 = T b ,1 + ε
c min
cb
(T a ,1 −Tb ,1 )
AN (4)
Outlet pressure on both side of the fluid flow is similarly obtained as follow:
M
Pa ,2 = Pa ,1 − ( Pa ,1 − Pa ,2 ) = Pa ,1 − ∆Pa )5 (
D
Pb ,2 = Pb ,1 − ( Pb ,1 − Pb ,2 ) = Pb ,1 − ∆Pb )6 (
TE
For the cross flow heat exchanger with both fluids unmixed, effectiveness [25] is given as:
{
1
ε = 1 − exp NTU
0.22
exp ( −c r NTU 0.78 ) − 1 }
EP
)7 (
c r
And
C
c min
cr = (8)
AC
c max
By considering thermal resistance of the wall and fouling factor negligible, the number of transfer
1 c 1 1
= min = c min + )9 (
NTU AU ( hA )a ( hA )b
Generally, heat transfer coefficient is explained in terms of Colburn factor [2] which is given by:
5
ACCEPTED MANUSCRIPT
−2
h = jc pG Pr 3 )10(
m
G = )11(
A ff
PT
Where, Aff is free flow area and in the present work [2] is formulated as:
A ff ,a = ( H a − T a )(1 − n aT a ) Lb N a
RI
)12(
SC
Heat transfer areas of the present heat exchanger for the two sides [2] are calculated by:
U
Ab = La Lb N b 1 + {2nb ( H b − t b )}
Therefore, total heat transfer area of the heat exchanger is formulated as:
AN )15(
M
A = Aa + Ab )16(
Also, frictional pressure drops for both sides of the fluid [2] are defined as below:
4f a LaG a2
EP
∆p a = )18(
2 ρ a d h ,a
4f b LbGb2
∆pb =
C
)19(
2 ρb d h ,b
AC
There are many equations for evaluation of Colburn coefficient and fanning factor for available offset
strip fin. Equations (20) and (21) presented by Manglik and Bergles [26] have been used in this study.
0.1
j = 0.6522 ( Re ) (α ) (δ ) (γ ) × 1 + 5.3 × 10 −5 ( Re ) (α ) (δ ) (γ )
−0.5403 −0.1541 −0.0677 −1.055
0.1499 1.34 0.504 0.456
(20)
0.1
f = 9.6243 ( Re ) (α ) (δ ) (γ ) × 1 + 7.7 × 10 −8 ( Re ) (α ) (δ ) (γ )
−0.7422 −0.1856 −0.2659
0.3053 4.429 0.920 3.767 0.236
(21)
6
ACCEPTED MANUSCRIPT
In these equations,
s t t 1
α= , δ = , γ = , s = −t
H −t l s n
4s ( H − t ) l
PT
Dh =
2 ( sl + ( H − t ) l + t ( H − t ) ) + ts
)22(
RI
And Reynolds number is obtained by
GD h mD h
Re = = (23)
µ A ff µ
SC
These equations are valid for 120<Re<104 and for ratios 0.134<α<0.997, 0.012<δ<0.048 and
U
0.041<γ<0.121. The above equations correlate f and j factors have 20% accuracy in comparison with
experimental results in laminar, transition and turbulence flow regimes. Thus, there is no need to describe
AN
the flow regime for a special condition. Therefore; it is very useful in most practical applications.
Finally, the number of entropy generation units is determinated as follows [2, 5]:
M
s&
Ns = )24(
c max
D
TE
3. Optimization method
EP
A colony of honey bees can extend itself over long distances (more than 10 km) and in multiple
directions simultaneously to utilize a large number of food sources. A colony prospers by deploying its
foragers to good fields. In principle, flower patches with plentiful amounts of nectar or pollen that can be
collected with less effort should be visited by more bees, whereas patches with less nectar or pollen
7
ACCEPTED MANUSCRIPT
The foraging process begins in a colony by scout bees being sent to search for promising flower
patches. Scout bees move randomly from one patch to another. During the harvesting season, a colony
continues its exploration and keeping a percentage of the population as scout bees. When they return to
the hive, those scout bees that found a patch which is rated above a certain quality threshold (measured as
PT
a combination of some constituents, such as sugar content) deposit their nectar or pollen and go to the
RI
This mysterious dance is essential for colony communication, and contains three pieces of information
SC
regarding a flower patch: the direction in which it will be found, its distance from the hive and its quality
rating (or fitness). This information helps the colony to send its bees to flower patches precisely, without
U
using guides or maps. Each individual’s knowledge of the outside environment is gleaned solely from the
waggle dance. This dance enables the colony to evaluate the relative merit of different patches according
AN
to both the quality of the food they provide and the amount of energy needed to harvest it. After waggle
dancing on the dance floor, the dancer (i.e. the scout bee) goes back to the flower patch with follower
M
bees that were waiting inside the hive. More follower bees are sent to more promising patches. This
D
The Bees Algorithm (BA) was developed by a group of researchers at the Manufacturing Engineering
EP
Center, Cardiff University [27]. This algorithm emulated the behavior of honey bees in foraging for
pollen and nectar to find the optimal solution for both continuous and combinatorial problem. The
C
algorithm required six parameters, namely the number of scout bees (n), number of selected sites (m),
AC
number of top-ranking (elite) sites among the m selected sites (e), number of bees recruited for each non-
elite site (nsp), number of bees recruited for each elite site (nep), and neighborhood size (ngh).
Fig. 3 shows the pseudocode for the algorithm in its simplest form. The optimization process started the
optimization process started with n scout bees randomly spread across the solution space. Each scout bee
was associated with a possible solution to the problem. The fitnesses of the sites visited by the scout bees
8
ACCEPTED MANUSCRIPT
are evaluated in step 2. In step 4, bees that have the highest fitnesses are chosen as “selected bees” and
the best m sites visited by them are chosen for neighborhood search. Then, in steps 5 and 6, the algorithm
conducts searches in the neighborhood of the selected sites. The bees for the neighborhood search can be
chosen directly according to the fitnesses associated with the sites they are visiting. In the neighborhood
PT
search procedure, more forager bees were sent in the neighborhood of the elite (e) sites, and fewer bees
around the non-elite (m-e) sites. According to this strategy, the foraging effort was concentrated on the
RI
very best (i.e. elite) solutions. That is nep bees were sent to forage around the elite sites, while the area
SC
around the non-elite locations was exploited by nsp bees. Within the given neighborhood area (i.e. flower
patch size), some of the newly generated solutions were expected to be better than that found by the scout
U
bees. As shown in Fig. 4, only the best bee is chosen to advertise its source after which the center of the
neighborhood field is shifted to the position of the best bee (i.e. from A to B). However, in step 6, for
AN
each patch only the bee with the highest fitness will be selected to form the next bee population. In nature,
there is no such a restriction. This restriction is introduced here to reduce the number of points to be
M
explored. In step 7, in the global search procedure, the unselected scout bees (n-m) were used to explore
D
at random the solution space. This kind of search was to avoid bees being trapped at local optima. At the
TE
end of each cycle, a new list of scout bees was formed, comprising the fittest solutions from each
neighborhood (neighborhood search results), and the new randomly generated solutions (global search
EP
results). This list would be sorted in the next iteration and used for a new phase of optimization. The
combination of exploitative (neighborhood) and explorative (global) search would be able to capture the
C
best solution quickly and efficiently. These steps were repeated until the stopping criterion was met [28].
AC
In this present study, the optimization targets are two single-objective functions. The first is the
maximization of effectiveness and second objective is the minimization of the number of entropy
generation units.
9
ACCEPTED MANUSCRIPT
{
1
ε = 1 − exp NTU
c r
0.22
}
exp ( −c r NTU 0.78 ) − 1 (25)
and the second objective function, number of entropy generation units are calculated as follows.
PT
s&
Ns = (26)
c max
RI
Putting all the relevant values, the above equation can be simplified and expressed as,
ca c T b ,1 R ∆P c c T a ,1 R ∆P
Ns = [Ln {1 − ε min (1 − )} − a Ln {1 − a }] + b [Ln {1 − ε min ( − 1)} − b Ln {1 − b }] (27)
SC
c min ca T a ,1 C p ,a Pa ,1 c max cb T b ,1 C p ,b Pb ,1
U
The general optimization problem may be stated mathematically as [24]:
Where fi(x) are the l objective functions, X is the column vector of the k independent variables, and
TE
Cj(X) are p termed equality constraints, and those of form hr(X) are q inequality constraints. Here in the
present work, the objective functions are maximization of the effectiveness and minimization of the
EP
number of entropy generation units. Moreover, in order to consider the effect of constraints violation
C
during the optimization process, an arbitrary large value known as penalty function is also added to the
objective function. Therefore, the objective function for the present work is represented as [2, 5]:
AC
m
Minimize f ( x ) = f ( x ) + ∑ R1 ( g i ( x ) )
2
(30)
j =1
That R1 is the penalty parameter which has a large value (for example say 500). The term
∑ R ( g ( x ) ) takes into account the effect of constraints violation. For all values of penalty parameter
2
1 i
j =1
10
ACCEPTED MANUSCRIPT
( R1 ), the Bees Algorithm is not sensitive to penalty function. So, the right term of above equation is
replaced with a large constant number (static penalty function) similar to scheme 1 from reference [22].
A minimum of function –f is a maximum of function f. That is, for converting the minimization
problem to maximization problem, it is enough to multiple the minus in the objective function.
PT
3.4. Relation between the bee behavior and the algorithm applied for the optimization
RI
The correlation between the bee behavior and Bees Algorithm parameters and step by step description
of BA applied for the optimization of the plate-fin heat exchanger is presented in Fig. 6.
SC
4. Computational results
U
4.1. Case study A
AN
To investigate the application of proposed algorithm, a case study taken from the work of Kakac and
M
Liu is considered [29]. A plate-fin heat exchanger of gas to air with cross flow having heat duty 3300 of
kW is needed to be designed separately for the maximum effectiveness and minimum number of entropy
D
generation units. Maximum dimension of the heat exchanger is 2×3×2 m. Pressures drops of the hot and
TE
cold flow are limited to 7.5 kPa and 4.5 kPa, respectively. The offset strip fin surface is used on the gas
and air sides. The other specifications of operation conditions have been mentioned in Table 1.
EP
In this study the seven design variables such as the length of hot flow, length of cold flow, fin
frequency, number of hot side layers, fin strip length, fin height and fin thickness are selected as a
C
optimization variables. All variables are continuous except the number of hot side layers. The thermal
AC
conductivity of the wall is 15 kW/m.K, and the plate thickness is considered to be constant at 0.5 ×10-3 m
and is not to be optimized. The variation ranges of the variables have been mentioned in Table 2.
11
ACCEPTED MANUSCRIPT
To optimize the heat duty and permitted pressure drop, the main problem is the finding for variable
design that maximizes the heat effectiveness of heat exchanger. Fig. 7 shows the effectiveness
PT
convergence diagram as objective function without considering the heat duty constraint. A significant
decrease in the objective function has been observed in the beginning of the evaluation process after 9
RI
iterations. After approximately 60 iterations, the changes in the objective function become relatively low.
Finally, the maximum efficiency of the plate-fin heat exchanger is found after 150 iterations with the
SC
value of 0.830. Table 3 shows the preliminary design of the plate-fin heat exchanger and optimum
structure which is obtained from Bees Algorithm. The increment of 6.68 % has been observed in
U
efficiency of heat exchanger by optimization method in comparison with preliminary design. It has been
AN
shown that for maximizing effectiveness, fin frequency decreased from 500 to 388.9 m-1 while the fin
offset length is increased from 6×10-3 to 8.57 ×10-3 m. Fig. 8 shows the effectiveness convergence
M
diagram by considering heat duty constraint. Similarly, in this case, the maximum effectiveness of the
plate-fin heat exchanger is converged to the value of 0.7972 after 15 iterations. Table 4 shows the
D
preliminary design of the plate fin heat exchanger and optimized structure which is obtained from BA by
TE
considering the heat duty constraint. In this case, the increment of 2.46 % has been made in heat
exchanger effectiveness by optimization method in comparison with preliminary design. Besides, the fin
EP
frequency has decreased from 500 to 400.8 while the fin offset length has increased from 6×10-3 to
7.07×10-3 m.
C
AC
The efficiency of the Bees Algorithm in comparison with the GA [2] and PSO [5] is investigated. The
results are demonstrated in Table 5. The sample algorithm has been programmed by MATLAB software
and the result of this algorithm is the average of 10 times of the execution of the related computer code.
The mentioned results indicate that the present BA can converge to the optimum values with higher
12
ACCEPTED MANUSCRIPT
accuracy in comparison with references. The results from BA in the number of entropy generation unit's
minimization are 16.5 % and 0.27 % better than the results of GA and PSO, respectively. Fig. 9 shows the
convergence of number of entropy generation units as objective function for the considered example.
PT
4.2. Case study B
For comparison between the present algorithm and Imperialist Competitive Algorithm, a case study
RI
selected from references [19]. A gas to air single pass cross-flow heat exchanger has a heat duty equal to
SC
1069.8 kW. This exchanger must be designed for the minimum number of entropy generation units.
Maximum dimension of the exchanger is 1×1×1 m. Gas and air inlet temperatures of hot and cold flow
U
are considered 900 K and 200 K, respectively. Also, gas and air mass flow rates are 1.66 kg/s and 2.00
AN
kg/s, respectively. Pressure drops are set to be limited to 9.50 kPa and 8.00 kPa. The gas and air inlet
pressures are 160 kPa and 200 kPa absolute. The offset strip fin surface is used on the gas and air sides.
M
The plate thickness is set at 0.5×10-3 m and is not an optimization variable. Operating conditions and the
cost function constant values required for cost evaluation are listed in Table 6.
D
In this study, a total number of seven parameters, namely hot flow length, cold flow length, number of
TE
hot side layers, fin frequency, fin thickness, fin height and fin strip length are considered as optimization
variables. All variables except the number of hot side layers are continuous. Thickness of the plate is
EP
considered to be constant at 0.5×10-3 m and is not to be optimized. The variation ranges of the variables
As shown as in Table 8, In this case study, the cross-section of the flow has been increased due to a
decrease in the lance length of the fin and the number of layers and an increase in the fin height.
Therefore, the mass flux of flow has been decreased by increasing the cross-section of the flow. By
decreasing the hot stream flow length and increasing the cold stream flow length, the pressure drop at hot
13
ACCEPTED MANUSCRIPT
and cold side has decreased and increased, respectively. Combination of these factors causing the pressure
drop at hot and cold sides is decreased than the ICA and GA in value of 39 % and 31 %, and 82 % and 11
%, respectively.
PT
The efficiency of the proposed algorithm in comparison with the ICA and GA is evaluated. The results
RI
are illustrated in Table 9. The Bees Algorithm programmed in MATLAB and run on an Intel® Core™ i5
CPU. The result of this algorithm is the average of 10 times of the execution of the related computer code.
SC
The results indicated the higher accuracy of proposed algorithm to converge the optimum value in
comparison with references. The obtained results from BA in the number of entropy generation unit's
U
minimization are 2.5 % and 5.2 % better than the results of ICA and GA, respectively. Fig. 10 shows the
AN
convergence of number of entropy generation units as an objective function for this case study.
M
4.2.3. A comparison among the proposed algorithm, ICA, GA, IHS and GAHPSO
To prove the effectiveness of the proposed algorithm, a comparison between this algorithm with the
D
algorithms of ICA and GA is discussed in terms of computational time and accuracy convergence to the
TE
optimal value [20]. In addition to, the objective function of the number of EGU has been studied for
compare with GA and PSO, objective functions of the total pressure drop and the total heat transfer area
EP
has been studied for compare with GAHPSO and Improved Harmony Search Algorithm with static
penalty parameter (IHS scheme 1) [21, 22]. Results are shown in Table 9.
C
AC
Similar to ICA, GA, IHS and GAHPSO, this algorithm's population size and number of iterations are
set to 100 and 200, respectively. This algorithm is programmed in MATLAB and run on an Intel® Core™
i5. The mentioned CPU time is the average of 10 executions of the computer program. In case 1, the
results of the BA 2.5 % and 5.2 % better than the results of the ICA and GA, respectively. In case 2, the
results of the BA similar to GAHPSO and IHS but 25 % better than the results of GA. Also, in case 3, the
results of the BA 5.7 %, 5.4 % and 24.2 % better than the results of the GAHPSO, IHS and GA,
14
ACCEPTED MANUSCRIPT
respectively. In case 2 and 3, the computational time of BA is less than GA, IHS and GAHPSO. Also, in
all case studies the results obtained by BA have higher accuracy rather than the preliminary design.
Therefore; the BA is higher speed and accuracy in converging to the optimum value.
PT
5. Conclusions
This study showed the successful usage of the Bees Algorithm in the design of plate-fin compact heat
RI
exchanger. A generalized procedure has been developed to carry out the optimization to find the
SC
maximum of the effectiveness and minimum of the number of entropy generation units of the heat
exchangers based on the ε–NTU method and the Bees algorithm technique. This algorithm is used in most
U
of heat engineering problems that consist of several discrete and continuous variables and the most
AN
discontinuity. Based on the applications, the seven design parameters were mentioned as optimization
variables. The constraints have been used adding a penalty function to the objective function. To examine
M
capability and accuracy of this algorithm, two case studies were selected from the references. The
obtained values showed the better results of BA in comparison with preliminary design. Also, it has been
D
shown that the BA is more accurate in convergence to the optimized results in comparison with GA, PSO
TE
and ICA. It is concluded that the BA algorithm can provide a strong ability of auto-search and combined
optimization in the optimization design of heat exchangers compared to the traditional designs in which a
EP
trial-and error process may be involved. By application of the BA algorithm in the optimal design the heat
exchanger configurations or structures can be optimized according to different design objectives such as
C
minimum surface area or cost. The process of offered design for plate fin heat exchanger in this study
AC
using BA can be used for other types of heat exchangers such as fin and tube heat exchangers, shell and
tube heat exchangers and recuperates. Moreover, other types of fins such as plain, perforated, wavy and
louvered fins can be applied on both cold and hot sides of the heat exchanger rather than the offset strip
fins which are applied on the both side in the present work. Therefore, BA can be applied in design of
15
ACCEPTED MANUSCRIPT
different types of heat exchangers to search for the optimum designs based on the desired objectives and
Appendices
PT
In this section, we will review a brief description of the other algorithms that we studied:
RI
A. Particle Swarm Optimization
PSO is a heuristic optimization methodology which inspire from searching animals for foods. This
SC
method has some similarities with other heuristic optimization methods:
U
1- Population-based. Heuristic methods use and change some solutions. In this method, each solution of
AN
problem or each location in the search space is called particle.
2- Iteration based. The solutions are changed and updated in each iteration.
M
3- Stochastic initialization and search procedure. Generally, solutions of the heuristic methods
initialized randomly. Also, in each iteration, some random components are used to update the
D
solutions.
TE
In comparison with other heuristic methods, PSO has some specific features:
1- Each particle has a feature which is called velocity and is used to update solutions.
EP
2- The best location for each particle which is obtain until iteration k, Pbest and the best location until
iteration k, Gbest are saved and will be used to update the particle locations.
C
Assume that the search space is n-dimensional, then the particle can be display by a n-dimensional vector
AC
Xi = [xi1,xi2,.,xin]T, and velocity Vi = [vi1,vi2,.,vin]T, where i= 1,2,..,N and N is the size of population [30].
In PSO, the following expressions are used to update the particle's location:
V i k +1 = w k
×V i k + c1 × rand 1 × ( Pbest ,i − x ik )
(A.1)
+c 2 × rand 2 × (G best − x ik )
16
ACCEPTED MANUSCRIPT
x ik +1 = x ik + χ ×V i k +1 (A.2)
where, wk is the inertia coefficient, χ is the maximum and minimum allowed velocity, c1 and c2 express
the cognitive and social parameters and rand1 and rand2 are two uniform random functions in the range of
[0,1].
PT
The inertia coefficient is used to adjust the influence of the velocity at the previous iteration on the current
RI
velocity. Larger values of the w lead to better global searching and lower ones help to better local search.
Setting w to one at the start of the search and gradual decrease toward zero at the end of search is a good
SC
choice. Values of the other used parameters are as follows: χ= 0.7, c1 = c2 =2 [30], But to avoid local
optimum and toward global solution, this parameter can be determined dynamically [31]. As regards, c1
U
is associated to learning factor from self-experimental particle and c2 is learning factor from other
AN
particle. Thus, if at the initial optimization procedure c1 is big and c2 is small and at the end of
optimization procedure c1 is small and c2 is big, the algorithm moves to the absolute optimum solution
M
[32].
D
B. Genetic Algorithm
TE
problems. GA begins with a set of solutions called population. Each solution is represented by a set of bit
EP
string (chromosome). Solutions from one population are taken and used to form a new population. Then,
Solutions which are selected to form new solutions (offspring) are selected according to their fitness (The
C
higher fitness value, the higher chances to reproduce). In general, process of GA can be summarized in 6
AC
steps:
3. [New population] Create a new population by repeating following steps until the new population is
complete
17
ACCEPTED MANUSCRIPT
3.1. [Selection] Select two parent chromosomes from a population according to their fitness (the
3.2. [Crossover] With a crossover probability, crossover parents to form new offsprings (children). If
PT
3.3. [Mutation] With a mutation probability, mutate new offspring at each locus (position in
chromosome).
RI
3.4. [Accepting] Place new offspring in the new population
SC
4. [Replace] Use new generated population for a further run of the algorithm
5. [Test] If the end condition is satisfied, stop, and return the best solution in current population
U
6. [Loop] Go to step 2
It is inspired by the imperialistic competition. It starts with an initial population called colonies. The
D
colonies are then categorized into two groups namely; imperialists (best solutions) and colonies (rest of
TE
the solutions). The imperialists try to absorb more colonies to their empire. The colonies will change
according to the policies of imperialists. The colonies may take the place of their imperialist if they
EP
become stronger than it (propose a better solution). This algorithm has been successfully applied to PSS
design [34] and data clustering [35]. The flowchart of proposed algorithm is depicted in Fig. C. 1.
C
AC
18
ACCEPTED MANUSCRIPT
REFERENCES
[1] R. K. Shah, D. P. Sekulic, Fundamentals of heat Exchanger design, John Wiley and Sons, Inc., New York, 2003.
[2] M. Mishra, P. K. Das, S. Sarangi, Second law based optimization of cross flow plate-fin heat exchanger design
PT
[3] S. Sanaye, H. Hajabdollahi, Thermal-economic multi-objective optimization of plate fin heat exchanger using
RI
[4] G.N. Xie, B. Sunden, Q.W. Wang, Optimization of compact heat exchangers by a genetic algorithm, Appl.
SC
Therm. Eng. 28 (2008) 895-906.
[5] R.V. Rao, V.K. Patel, Thermodynamic optimization of cross flow plate-fin heat exchanger using a particle
U
[6] N. Sahiti, F. Krasniqi, Xh. Fejzullahu, J. Bunjaku, A. Muriqi, Entropy generation minimization of a double-pipe
AN
pin fin heat exchanger, Appl. Therm. Eng. 28 (2008) 2337-2344.
[7] A. Bejan, The Concept of Irreversibility in Heat Exchanger Design: Counterflow Heat Exchangers for Gas-to-
M
Gas Applications, J. Heat Trans. 99 (1977) 374.
[8] J.M. Reneaume, N. Niclout, MINLP optimization of plate fin heat exchangers, Chem. Biochem. Eng. Q. 17 (1)
D
(2003) 65–76.
TE
[9] J.M. Reneaume, N. Niclout, Plate fin heat exchanger design using simulated annealing. Elsevier, City, 2001.
[10] L. Wang, B. Sundén, Design methodology for multi stream plate-fin heat exchangers in heat exchanger
EP
[11] J.M. Reneaume, H. Pingaud, N. Niclout, Optimization of plate fin heat exchangers: a continuous formulation,
C
[12] K. Muralikrishna, U. Shenoy, Heat exchanger design targets for minimum area and cost, Chem. Eng. Res. Des.
AC
[13] R. Selbas, O. KizIlkan, M. Reppich, A new design approach for shell-and-tube heat exchangers using genetic
algorithms from economic point of view, Chem. Eng. Process. 45 (4) (2006) 268–275.
[14] M.C. Tayal, Y. Fu, U.M. Diwekar, Optimal design of heat exchangers: a genetic algorithm framework, Ind.
19
ACCEPTED MANUSCRIPT
[15] P. Ahmadi, H. Hajabdollahi, I. Dincer, Cost and Entropy Generation Minimization of a Cross-Flow Plate Fin
Heat Exchanger Using Multi-Objective Genetic Algorithm, J. Heat. Trans. 133 (2) (2011) 021801.
[16] H. Najafi , B.Najafi, P. Hoseinpoori, Energy and cost optimization of a plate and fin heat exchanger using
PT
[17] M. Mishra, P. K. Das, S. Sarangi, Optimum design of cross flow plate-fin heat exchangers through Genetic
RI
[18] H. Peng, L. Xiang, Optimal design approach for the plate-fin heat exchangers using neural networks cooperated
SC
[19] M. Yousefi, A.N. Darus, H. Mohammadi, Second law based optimization of a plate fin heat exchanger using
U
[20] M. Yousefi, A.N. Darus, H. Mohammadi, An imperialist competitive algorithm for optimal design of plate-fin
exchangers by an improved harmony search algorithm, Appl. Therm. Eng. 50 (2013) 877-885.
D
[23] D.T. Pham, A. Ghanbarzadeh, S. Otri, Optimal design of mechanical components using the bees algorithm,
TE
Proceedings of the Institution of Mechanical Engineers, J. Mech. Eng. Sci. 28 (2009) 101-114.
[24] D.T. Pham, A. Ghanbarzadeh, Multi-Objective Optimisation using the Bees Algorithm, In 3rd International
EP
Virtual Conference on Intelligent Production Machines and Systems (IPROMS 2007), Scotland, 242 (2007) 111-
116.
C
[25] F.P. Incropera, D.P. DeWitt, Fundamentals of heat and mass transfer, John Wiley and Sons Inc., 1988.
AC
[26] R. Manglik, A. Bergles, Heat transfer and pressure drop correlations for the rectangular offset strip fin compact
[27] D. Pham, A. Ghanbarzadeh, E. Koc, S. Otri, S. Rahim, and M. Zaidi, The bees Algorithm. A
novel tool for complex optimization problems, in proceedings of the 2nd International Virtual Conference
20
ACCEPTED MANUSCRIPT
[28] S. Otri, improving the bees algorithm for complex optimization, PhD thesis, Cardiff University, United
Kingdom, 2011.
[29] S. Kakac, H. Liu, Heat exchanger selection rating and thermal design, CRC Press LLC, Florida, 2002.
[30] K. E. Parasopoulos, M.N. Vrahatis, On the computation of all global minimizers through particle swarm
PT
optimization, IEEE Trans. Evol. Comput. 8(3) (2004).
[31] P. K. Tripathi, S. Bandyopadhyay, S.K. Pa, Multi-Objective particle swarm optimization with time variant
RI
inertia and acceleration coefficients, Inf. Sci. 177 (2007) 5033–49.
[32] A. Kashefi Kaviani, G.H. Riyahi, SH.M. Kouhsari, Optimal design of a reliable hydrogen-based stand-alone
SC
wind/PV generating system, considering component outages, Renew. Energ. 34 (2009) 2380-2390.
[33] E. Atashpaz-Gargari and C. Lucas, Imperialist competitive Algorithm: An Algorithm for optimization inspired
U
by imperialistic competition, in Evolutionary Computation, CEC 2007, IEEE Congress, (2007) 4661–4667.
[34] A. Jalilvand, S. Behzadpoor, and M. Hashemi, Imperialist competitive Algorithm-based design of PSS to
AN
improve the power system, Power Electronics, Drives and Energy Systems (PEDES), 2010 Joint International
imperialist competitive Algorithm and k-means for data clustering, Eng. Appl. Artif. Intel. 24 (2) (2011) 306– 317.
D
TE
C EP
AC
21
ACCEPTED MANUSCRIPT
PT
Fig. 3. Graphical Explanation of the Neighborhood
RI
Fig. 4. Pseudocode of the BA
SC
Fig. 6. Step by step description of BA applied for the optimization of the plate-fin heat exchanger
U
Fig. 7. Evolution process for maximum effectiveness without heat duty constrains
AN
Fig. 8. Evolution process for maximum effectiveness with heat duty constraints
22
ACCEPTED MANUSCRIPT
Table 1
The operating conditions of the PFHE (input data for the model) [29].
PT
460 300
RI
SC
Pressure (kPa) 100 900
U
Density (kg.m-3) 0.54 4.86
Prandtl number
3.2×10-5
0.69
AN 3.2×10-5
0.69
Table 2
PT
Cold flow length(m) 0.1 3
-3 -3
Fin height (m) 2×10 10×10
RI
-3 -3
Fin thickness (m) 0.1×10 0.2×10
SC
-3 -3
Fin offset length (m) 1×10 10×10
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Table 3
The result of preliminary design and Bees Algorithm for maximizing effectiveness without heat duty constraint.
PT
Cold flow length (m) 1.8 2.99
-3 -3
Fin height (m) 5.7×10 5.72×10
-3 -3
Fin thickness (m) 0.15×10 0.169×10
RI
Fin frequency (m-1) 500 388.9
SC
-3 -3
Fin offset length (m) 6×10 8.57×10
U
Cold flow pressure drop (kPa) 10 3.38
Table 4
The result of preliminary design and Bees Algorithm for maximizing effectiveness with heat duty constraint.
PT
Cold flow length (m) 1.8 2.626
-3 -3
Fin height (m) 5.7×10 7.337×10
RI
-3 -3
Fin thickness (m) 0.15×10 0.137×10
SC
-3 -3
Fin offset length (m) 6×10 7.07×10
U
Hot flow pressure drop (kPa) 15 7.4
1.79
3.8
1.92
M
Effectiveness 0.778 0.7972
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Table 5
Comparison results of the GA, PSO and BA for minimum number of entropy generation units.
PT
Cold stream flow length (m) 0.996 0.887 0.995
RI
Fin frequency 442.9 534.9 405.69
SC
-3 -3 -3
Lance length of the fin (m) 9.8×10 6.3×10 9.998×10
U
Pressure drop at cold side (kPa) 1.834 2.216 1.143
Table 6
The operating conditions of the PFHE (input data for the model) [19].
PT
Mass flow rate (kg.s-1) 1.66 2
900 200
RI
Input temperature (℃)
SC
Pressure (kPa) 160 200
U
Specific heat (J.kg-1K-1) 1122 1073
Table 7
PT
Cold flow length(m) 0.1 1
-3 -3
Fin height (m) 2×10 10×10
RI
-3 -3
Fin thickness (m) 0.1×10 0.2×10
SC
-3 -3
Fin offset length (m) 1×10 10×10
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
Table 8
Comparison results of the ICA, GA and BA for minimum number of entropy generation units.
PT
Cold stream flow length (m) 0.88 0.44 0.940
-3 -3 -3
Fin height (m) 5×10 7.2×10 8.33×10
RI
Fin thickness (m) 0.19×10-3 0.1×10-3 0.166×10-3
SC
-3 -3 -3
Lance length of the fin (m) 9.6×10 7.2×10 9.51×10
U
Pressure drop at cold side (kPa) 0.67 0.52 0.460
Convergence iterations 40 30 44
M
Effectiveness 0.821 0.821 0.826
Table 9
Comparison results of the ICA, GA, GAHPSO, IHS and BA for Case study B
PT
Parameters Preliminary design PSO [21] GA [21] GAHPSO [21] HIS [22] BA
Case 2 Total pressure drop (kPa) 1.82 0.063 0.070 0.056 0.056 0.056
RI
CPU time (s) - 3.49 4.35 3.25 3.05 2.45
SC
Case 3 Heat transfer area (m2) 142.75 123.5 132.5 112.7 112.4 106.6
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
PT
RI
Fig. 1. A view of a plate-fin heat exchanger
U SC
AN
M
D
TE
EP
C
AC
ACCEPTED MANUSCRIPT
PT
RI
Fig. 2. Detailed view of an offset strip fin
U SC
AN
M
D
TE
EP
C
AC
ACCEPTED MANUSCRIPT
PT
4. Select sites for neighborhood search.
5. Recruit bees for selected sites (more bees for best e sites) and evaluate the fitness.
RI
6. Select the fittest bee from each patch.
SC
8. End While.
U
Fig. 3. Pseudocode of the BA
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
PT
RI
Fig. 4. Graphical Explanation of the Neighborhood Search
U SC
AN
M
D
TE
EP
C
AC
ACCEPTED MANUSCRIPT
Neighborhood Search
PT
Fitness Evaluation Fitness Evaluation
Local search
Select Patch Fittest Select Patch Fittest
RI
Global search Select the m Fittest Bees
from all Patch
SC
New Population
Random (n-m)
U
Fitness Evaluation
No
Converged? AN
Yes
M
Solution
D
PT
d) Number of bees recruited for best e sites (nep),
e) Number of bees recruited for the other (m-e) selected sites (nsp),
f) Neighborhood size [ngh],
Range of decision variables : [La , Lb , H , N , t , l , n] = Variablek = [xkL , xkU]
RI
STEP 2:
SC
Pop(:, k)= xkL + (xkU - xkL) * rand (1, n),
END
STEP 3:
U
WHILE (stopping criterion not met)
STEP 4:
Neighborhood Pop (q, k) = (Pop(i, k) - ngh(k)) + ((Pop(i, k) + ngh(k)) - (Pop(i, k) - ngh(k))) * rand,
IF Neighborhood Pop (q, k) < xkL
Neighborhood Pop (q, k) = xkL,
ELSE IF Neighborhood Pop (q, k) > xkU
Neighborhood Pop (q, k) = xkU,
END
EP
END
END
Evaluate fitness of the each patch population + Penalty function*
For each patch only the bee with the highest fitness will be selected to form the next bee population.
END
C
STEP 6:
New Pop = m fittest Bees from all Patch + Random (n-m) remaining bees,
End While
Fig. 6. Step by step description of BA applied for the optimization of the plate-fin heat exchanger
ACCEPTED MANUSCRIPT
PT
RI
SC
Fig. 7. Evolution process for maximum effectiveness without heat duty constrains.
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
PT
RI
SC
Fig. 8. Evolution process for maximum effectiveness with heat duty constraints.
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
PT
RI
SC
Fig. 9. Convergence of the objective of number of entropy generation units
U
AN
M
D
TE
C EP
AC
ACCEPTED MANUSCRIPT
PT
RI
SC
Fig. 10. Convergence of the objective of number of entropy generation units
U
AN
M
D
TE
EP
C
AC
ACCEPTED MANUSCRIPT
Nomenclature
PT
Cr Cmin/Cmax S& rate of entropy generation (W/K)
RI
f fanning friction factor T temperature (K)
SC
g(x) Constraint
U
H height of fin (m) ρ density
Lf
Colburn factor
∆p
effectiveness
pressure drop
Na ,Nb number of fin layers for fluid a and b a,b fluid a and b
TE
PT
RI
U SC
AN
M
D
TE
EP