Neural Computing and Applications

https://doi.org/10.1007/s00521-021-05963-2 (0123456789().,-volV)(0123456789().
,- volV)

S.I. : NEW TRENDS OF NEURAL COMPUTING FOR ADVANCED APPLICATIONS

Elite and dynamic opposite learning enhanced sine cosine algorithm

for application to plat-fin heat exchangers design problem
Lidong Zhang1 • Tianyu Hu1 • Zhile Yang2 • Dongsheng Yang3 • Jianhua Zhang4

Received: 28 October 2020 / Accepted: 25 March 2021

Ó The Author(s) 2021

Abstract
The heat exchanger has been widely used in the energy and chemical industry and plays an irreplaceable role in the
featured applications. The design of heat exchanger is a mixed integer complex optimization problem, where the efficient
design significantly improves the efficiency and reduces the cost. Many intelligent methods have been developed for heat
exchanger optimal design. In this paper, a novel variant of sine and cosine algorithm named EDOLSCA is proposed,
enhanced by dynamic opposite learning algorithm and the elite strategy. The proposed method is tested in CEC2014
benchmark and proved to be of significant advantages over the original algorithm. The new algorithm is then validated in
the plate-fin heat exchanger (PFHE) optimal design problem. The comparison results of the proposed algorithm and other
algorithms prove that EDOLSCA also has demonstrated superiority in heat exchanger optimal design.

Keywords Plate-fin heat exchanger  Design optimization  Sine cosine algorithm  Dynamic-opposite learning

1 Introduction type of heat exchanger, also known as brazed aluminum

heat exchanger in the thermal engineering. Compared with
In 2020s, energy consumption has been growing at an other types of heat exchangers, PFHE has compact struc-
exponential rate thanks to the higher energy efficiency and ture, relatively small size, relatively high thermal load and
strong demand. As early as 1973, Lincoln et al. [28] lightweight, due to which it has been widely used in
pointed out that energy conservation should be paid aerospace, railway, energy and chemical industries. In real-
attention to and the sustainable development of environ- world cases, the majority of the design for heat exchanger
ment and resources should be guaranteed through techno- is based on the experience of designers, without any unified
logical progress and social science [55]. With the design method and scheme [45]. Bio-optimization, which
significant improvement of computational capacity, an is of high compatible capacity in solving various opti-
increasing number of industrial equipment began to use mization problems, is promising to effectively solve PFHE
computer-aided design and optimization to improve the optimal design problem. To solve the design problem of
energy efficiency and reduce economic expenditure plate-fin heat exchanger, the geometric limitation of heat
[49, 51]. Plate-fin heat exchanger (PFHE) is a common exchanger, such as shape and size, should be determined
based on the actual engineering needs. Then the objective
function is derived based on the design objective. The fluid
& Zhile Yang
[email protected] data and constraints are brought into the operation, and the
appropriate algorithm is selected for them. Finally, the
1
Northeast Electric Power University, Jilin 132012, China optimal design data are obtained.
2
Shenzhen Institute of Advanced Technology, Chinese Bio-inspired optimization is based on a variety of
Academy of Sciences, Shenzhen 518055, China physical and biology phenomena in the nature. Several
3
Intelligent Electrical Science and Technology Research typical examples include genetic algorithm (GA) based on
Institute, Northeastern University, Shenyang 110819, China evolutionary biology [46], particle swarm optimization
4
State Key Laboratory of Alternate Electrical Power System (PSO) from simplified social models [25], simulated
with Renewable Energy Sources, North China Electric Power annealing algorithm from metallurgy for specific rate
University, Beijing 102206, China

123
 Neural Computing and Applications

cooling operations [42], ant colony optimization (ACO) and development. Banerjee and Nabi [4] used SCA to
derived from the behavior of ants in food searching and optimize the shuttle’s re-entry trajectory back to Earth to
paths discovering [13]. Moreover, a teaching learning- find the best trajectory during the return. Attia et al. [3]
based optimization (TLBO) algorithm has also been pro- used the modified sine cosine algorithm (MSCA) to solve
posed to simulate the teaching process of teachers and the problem of optimal power flow (OPF), which acceler-
students [39]. ated the speed of SCA search and avoided the local optimal
In the heat exchanger optimal design optimization, selection of the original algorithm as much as possible.
many scholars have utilized bio-inspired optimization Hekimoglu [19] used SCA to solve the parameters of
methods for solving the problems. The optimization of heat automatic voltage regulator and proportional integral
exchanger can be divided into two types: single objective derivative (PID). In the optimization problem of automatic
and multi-objective. Featured optimization targets of opti- voltage regulator system, SCA has strong robustness.
mal design include the amount of entropy produced, total Mahadad and Srairi [30] use SCA to optimize load margin
heat exchange, maintenance costs and production costs. stability to improve power system security. Wang et al.
Port diameter, port horizontal distance, port vertical dis- [44] employed the multi-objective sine cosine algorithm
tance, plate thickness, fin length and number of plates are (MOSCA) combined with wavelet neutral network (WNN)
often used as geometric design variables [36]. Some recent to accurately predict the wind behavior of wind power
proposed methods, such as Gaussian process regression and plants. Das et al. used SCA to optimize the hydrothermal
adaptive resampling, are adopted to minimize the cost [7]. scheduling problem in the economics of power system to
Some scholars have combined traditional and novel minimize the total cost of power generation and solve the
methods, such as computational fluid dynamics (CFD) short-term hydrothermal scheduling problem [11]. Chan-
combination with GA, to optimize heat exchangers with drasekaran et al. [8] solved the partial shading detection
Colburn factor j and the friction factor f as targets [2, 29]. (PSD) and MPPT (maximum power point tracking) prob-
Moreover, the optimal design problem is mostly solved by lems using SCA.
simple and improved optimization algorithm. Jonh and Since the proposal of canonical SCA, many scholars
Krishnakumar [22] used GA to optimize the perforated have proposed various algorithm improvement schemes or
plate matrix heat exchanger surfaces, where the optimal combinations with other algorithms to solve the complex
objectives were Colburn factor j and the friction factor problems. Elaziz et al. [1] combined the opposition-based
f. Rao and Saroj [38] used Jaya algorithm to optimize the learning (OBL) idea with SCA to improve the search
shell and tube heat exchanger (STHE) economically, taking accuracy of SCA. Nenavath and Jatoth [32] combined SCA
into account the consistency and maintainability caused by with differential evolution (DE) and demonstrated that the
scaling. In the same year, Rao and Saroj [37] improved the hybrid SCA-DE algorithm solves the problems effectively.
Jaya algorithm and proposed the elitist-Jaya algorithm to Sindhu et al. [40] combined elite strategy with SCA to
optimize the setup and operation costs of the STHE. make simple improvements to the algorithm and improve
Dhavle et al. [12] used cohort intelligence (CI) to optimize the search efficiency of SCA. Bureerat and Pholdee [6]
STHE design, and the total cost of the heat exchanger was proposed adaptive sine cosine algorithm integrated with
taken as the target, and the exterior scene of the tube and differential evolution (ASCA-DE) to solve the problem of
the spacing of baffles were taken as the design variables. structural damage detection. Gupta and Deep [17] proposed
Vasconcelos et al. [43] proposed a falcon optimization an improved SCA with crossover scheme and proved the
algorithm (FOA) based on the predation behavior of fal- validity using CEC2014 benchmark test. In addition, they
cons. The total cost of STHE was used as the objective also solved the multilevel threshold processing of image
function. Iver et al. [21] combined GA and CI and pro- segmentation using the new method and obtained com-
posed an adaptive range genetic algorithm (ARGA) to petitive result. Chegini et al. [9] combined the PSO and
solve the costs of the economic optimization problems and Levy flight methods and proposed the PSOSCALF with
extended them to solve other complicated core mechanical stronger jumping ability and higher searching ability.
engineering application problems. Gupta and Deep [16] added self-adaptive to SCA to
Sine cosine algorithm (SCA) is a new meta-heuristic enhance the exploitation ability. Pasandideh and Kha-
algorithm first proposed by Mirjalili in 2016 [31]. This lilpourazari [33] combined crow and SCA and proposed the
optimization method is based on sine and cosine functions sine cosine crow search algorithm to balance the explo-
in mathematics and has been immediately widely applied ration and exploitation. Li et al. [26] proposed the
in various optimization problems. Ekiz et al. [14] enhanced brain storm SCA (EBS-SCA) to solve the prob-
expounded the feasibility of SCA in solving constraint lem of premature convergence of SCA on complex opti-
optimization problems and concluded that the successful mization problem, which significantly enhanced the
feature of SCA lies in the smooth balance of exploration algorithm performance. Gupta and Deep [18] proposed

123
Neural Computing and Applications

SCA with global optimization based on their previous principle diagram of the plate-fin heat exchanger. The two
studies, referred to simulated quenching algorithm search fluids flow alternately in different fins, exchanging heat
mechanism for improved algorithm, and used it to train with maximum efficiency. The optimization work in this
multilayer perceptrons. Though a number of SCA variants paper is based on this featured heat exchanger. In this
have proposed, the optimization capacity of the algorithm section, the formula is derived first; then, the optimization
is to be further improved, in particular toward the con- constraints and objectives are determined.
vergence speed aspect.
This paper tries to use dynamic opposite learning (DOL) 2.1 Formula derivation for plate-fin heat
to significantly enable the diversity of solutions in the SCA, exchanger
where the elitist strategy is also used to accelerate the
convergence. The DOL strategy was first proposed by Xu Heat duty Q is defined first, then Q defines the overall
et al. [50], which significantly improved the comprehensive energy exchange level of the heat exchanger, and the for-
performance of TLBO . Combining DOL and elite policies mula is given by Yousefi et al. [53]:
with SCAs, a new EDOLSCA is proposed, taking the Q ¼Cmin ðTAin  TBin Þ ð1Þ
advantages of the adopted strategies to speed up SCA
search and improve precision for specific problems. The Cmin ¼m  Cp ð2Þ
ideas in this paper are evident in the CEC2014 benchmark
where C is the heat capacity rate, TAin ; TBin refer to fluid A
test and PFHE design issues.
(usually hot fluid) and fluid B (usually cold fluid), in refers
The major contribution of this article is summarized as
to the fluid flowing into the heat exchanger, and min refers
follows:
to the less numeric of the two fluids, max is the larger of the
Firstly, the objective function of PFHE optimization
two fluids.
design is derived by using the method of number of transfer
The effectiveness  is given by Incropera et al. [20],
units (NTU), and two design objectives and fluid parame-   
0:78
ters are given. ð 1
Þ NTU 0:22 eðCr NTU Þ 1 ð3Þ
¼1e C r

Secondly, elite strategy and DOL are adopted to

strengthen SCA, generating the EDOLSCA. The where e is the Eulers number,
improvement makes the convergence speed of SCA greatly
Cr ¼ Cmin =Cmax ð4Þ
increase.
Finally, EDOLSCA was tested and compared using the Heat transfer coefficient method is a common method of
CEC2014 benchmark and then adopted for solving an thermodynamic solution, which is usually used to solve
engineering example of PFHE optimal design. problems without knowing outlet temperature. The heat
The structure of this paper is as follows: transfer coefficient is expressed as:
Section 2 demonstrates the thermodynamic formula of
1
PFHE, establishes the optimization objective and evalua- ¼
NTU
tion function, as well as the constraint conditions. Sec-   ð5Þ
tions 3 and 4 elaborate the preliminaries of the elite and AffA AffB
Cmin þ
DOL strategies and illustrate the principle and specific jA CpA PrA0:667 mh AA jB CpB PrB0:667 mB AB
algorithm scheme of EDOLSCA. Section 5 provides a where j is Colburn factor, A; AHT is heat transfer area (m2 ),
comprehensive test of EDOLSCA, including tests based on
Cp is specific heat of fluid (W kg1 K1 ), Pr is Prandtl
the CEC2014 benchmark and PFHE design, with compre-
number, and Ns is number of entropy generation units.
hensive analysis of the results. Section 6 summarizes the
The free flow area of two fluids is defined as:
whole paper and outlooks the future directions.
AffA ¼ðHt  tA Þð1  nA tA ÞLB NA ð6Þ
AffB ¼ðHB  tB Þð1  nB tB ÞLA NB ð7Þ
2 Previous work on plate-fin heat exchanger
optimization In the formula, H is height of the fin, H is height of the fin, t
is fin thickness, n is fin frequency, L denotes heat
A standard two-flow straight finned plate-fin heat exchan- exchanger length, N is number of fin layers, lis dynamic
ger is shown in Fig.1. The plate-fin part of the plate-fin heat viscosity (N  s 1  m 2 ), and q is density (kg  m 3 ).
exchanger is shown in Fig.(a), in which the design vari- Fluid B has one more layer than fluid B:
ables mentioned below are marked. The fins are generally
welded from thin aluminum plates. The picture shows a
relatively simple form of fin. Figure(b) is the working

123
 Neural Computing and Applications

(a) (b)
Fig. 1 Plate-fin heat exchanger and rectangular offset strip fin

NB ¼ NA þ 1 ð8Þ The pressure drop due to friction between the fluid in the
channel and the fin can be calculated as:
The heat transfer area of the fluid comes from Yousefi [52]:  2
mA
2fA LA Aff
AA ¼LA LB NA ð1 þ ð2nA ðHA  tA ÞÞÞ ð9Þ DPA ¼
A ð18Þ
qA Dh;A
AB ¼LA LB NA ð1 þ ð2nB ðHB  tB ÞÞÞ ð10Þ  2
mB
2fB LB Aff
j is the Colburn factor and f is fanning friction factor [23].
DPB ¼
B ð19Þ
For laminar flow ðRe  1500Þ, qB Dh;B
 0:15  0:14
0:5 l s where DP is pressure drop (N m2 )
j ¼0:53ðReÞ t ð11Þ
Dh H
 0:41  0:02 2.2 Objective function for plate-fin heat
l s exchanger
f ¼8:12ðReÞ0:74 t ð12Þ
Dh H
For turbulent flowðRe [ 1500Þ, For most heat exchangers, the flow of fluid in the channel is
 0:24  accompanied by friction, which leads to the process of
l s 0:02
entropy increase. The design of the heat exchanger deter-
j ¼0:21ðReÞ0:4 t ð13Þ
Dh H mines the amount of entropy increase. Therefore, entropy
 0:65  0:17 increase is considered as an effective evaluation index in
l s
f ¼1:12ðReÞ0:36 t ð14Þ the design of heat exchanger. Under the same fluid con-
Dh H dition, the smaller the number of entropy increases, the
where Re is Reynolds number, s is fin spacing. better the performance of the heat exchanger.
The fin spacing of the PFHE: In this article, one of the target functions is defined as
entropy generation units [5]:
s ¼ ð1=n  tÞ ð15Þ " #   
ðTBin  TAin Þ2 RcteA DPA
where Ns ¼ð1  Þ þ
TBin TAin CpA PAin
m  Dh    ð20Þ
Re ¼ ð16Þ RcteB DPB
Aff l þ
CpB PBin
For the fin shape in this paper, the hydraulic diameter Dh of
the heat exchanger can be expressed as: where Rcte is specific gas constant (J kg1 K1 )
In addition to entropy generation, the effective heat
4sðH  tÞl transfer area of heat exchanger is also one of the important
Dh ¼ ð17Þ
2ðsl þ ðH  tÞl þ tðH  tÞÞ þ ts indexes to evaluate heat exchanger. The heat transfer area
where Dh is hydraulic diameter and l is lance length of the is also closely related to the Reynolds number and effi-
fin. ciency of the heat exchanger. Therefore, the second
objective function in this paper is the heat transfer area of
the heat exchanger

123
Neural Computing and Applications

The total heat transfer area is: 3 Algorithm preliminaries

AHT ¼ AA þ AB ð21Þ
Heuristic optimization algorithms are inspired by biology
as well as natural sciences, such as physics and mathe-
2.3 Application example of plate-fin heat matics. Different ideas of particle updating determine the
exchanger different characteristics of each algorithm. For different
algorithm ideas, the parameters that need to be adjusted are
For the two PFHE design examples prepared in this paper, also different. Some algorithms do not need to adjust
the difference lies in the difference of fluid parameters and parameters, while others may need to adjust 4–6 parame-
size constraints. The specific constraints and fluid data are ters to ensure the best results.
shown in Table 1.
In order to meet the heat duty requirement of the heat 3.1 Sine cosine algorithm
exchanger, effective constraints are carried out in the cal-
culation process of the algorithm, and the penalty function For SCA, the main idea of the algorithm is the sine and
is used to achieve this requirement. The formula is as cosine functions in mathematics. Four random numbers
follows: determine the motion direction and motion distance of
gð X Þ ) nð X Þ  Q ¼ 0 ð22Þ particles, and there are no parameters to be adjusted in the
algorithm. Therefore, SCA has relatively strong adapt-
In the design case, the overall size of the heat exchanger, ability, which provides the possibility for wide application.
the density of fins, the length, width and height of fins, the The core of the algorithm is as follows:
number of fluid layers of the two fluids and the thickness of (  
tþ1
Xit þ r1  sinðr2 Þ  r3 Pit  Xit ; r4 \0:5
fins are all the constraints of the design. The volume of the Xi ¼  
heat exchanger is limited to 1m*1m, and heat duty is X t þ r1  cosðr2 Þ  r3 Pi  X t ; r4  0:5
i t ii
160kW of Case 1 and 1050kW of Case 2, respectively. In ð23Þ
the calculations, fluids are considered ideal gases. a
r1 ¼ a  t ð24Þ
T

Table 1 Design constraints and

Parameters Project 1 Project 2
fluid data
Lower bound Upper bound Lower bound Upper bound

Heat exchanger length , m, FluidA 0.1 1 0.1 1

Heat exchanger length , m, FluidB 0.1 1 0.1 1
Height of the fin, m 0.002 0.01 0.002 0.01
Fin frequency 100 1000 100 1000
Fin thickness, m 0.0001 0.0002 0.0001 0.0002
Lance length of the fin, m 0.001 0.01 0.001 0.01
Number of fin layers 1 10 1 200

Parameters Project 1 Project 2

Fluid A Fluid B Fluid A Fluid B

Mass flow rate of fluid, kg  s1 0.8962 0.8296 1.66 2

Temperature, K 513 277 1173.15 473.15
Pressure, Pa 100000 100000 160000 200000
Specific heat of fluid, J  kg1  K 1 1017.7 1011.8 1122 1073
3 0.8196 0.9385 0.6296 0.9638
Density, kg  m
Dynamic viscosity ,N  s  m2 2.410E-05 2.182E-05 4.010E-05 3.360E-05
Prandtl number 0.6878 0.6954 0.731 0.694

123
 Neural Computing and Applications

learning (QOBL) [34] and quasi-reflection-based learning

(QRBL) [15] are evolved.
OBL was first proposed to accelerate the convergence of
DE. A strategy for finding the opposite number is
proposed:
XO ¼ a þ b  X ð25Þ

where X O is the opposite X, X 2 ½a; b.

Fig. 2 Rules for the direction of particle motion in SCA In practical application, if D becomes the dimension of
the problem, Eq.25 can be rewritten as:
where X is location of position, t is the t  th iteration , P is XjO ¼ aj þ bj  Xj ; j ¼ 1 : D ð26Þ
final solution, and T is maximum number of iterations.
Since the main updating method in the formula involves In practical application, OBL, QOBL and QRBL are all
sine and cosine function, it is named sine and cosine likely to fall into the local optimal position in the search
algorithm. The rules of the algorithm are shown more space. Xu et al. [50] proposed the dynamic opposite
intuitively in Figs.2, 3. Figure 2 shows how the parameters learning (DOL) strategy to solve this problem:
in the equation affect the search space for the solution, X DO ¼ X þ w  rand  ðrand  X O  XÞ ð27Þ
which is a two-dimensional representation. Figures 2 and 3
together show the influence of sines and cosines on the next where ’rand’ is a random number in [0, 1] , w is the weight
position in the calculation. The search space can be that determines the combination of DOL and the original
achieved by changing the range of values of sines and algorithm, that is, the weight of balance, and the
cosines. And the random position in space or out of space is ’rand  X O ’ makes the search space become asymmetric,
also realized by equation parameters.In SCA, r1 determines so as to solve the problem of falling into local optimal, and
the direction of movement, r2 determines the distance of make the search become dynamic and constantly changing.
movement, ranging in [0,2p], r3 determines the direction In the same way, DOL is applied to the real world and
whether to move inward or outward, bringing randomness dimension D is added, which can be rewritten as:
to the position, ranging in [-2,2], r4 determines whether to XjDO ¼ Xj þ w  rand  ðrand  XjO  Xj Þ; j ¼ 1 : D ð28Þ
move by sine or cosine manner, ranging in [0,1].
Applying the above DOL strategy to the initial population
3.2 Dynamic-opposite learning and iteration makes the algorithm search better and faster.

In some featured cases, optimization algorithm can be

improved effectively by modifying certain steps, such as 4 Elite and DOL-based SCA
convergence speed and search accuracy. Oppositional-
based learning (OBL) [35] is a popular modification In this section, we describe in details how EDOLSCA
method in recent years, from which quasi-opposite-based works. EDOLSCA combines elite and DOL policies to
bring greater speed and precision to SCA.

Fig. 3 Relationship between particle motion and sine and cosine functions in SCA

123
Neural Computing and Applications

4.1 DOL population initialization

After the population is initialized, the first step will be

calculated:
oPDO
ij ¼ oPij þ r1i  ðr2i  ðaj þ bj  oPij Þ  oPij Þ ð29Þ

where oP is a random population generated within the

upper and lower bound by using random numbers, and i
and j are the jth variables of the ith individual. a and b are
the boundaries of the variables, and r1 and r2 are random
numbers. This step can improve the adaptability of the
improved initial population through initial processing of
the population. After modifying the population, the particle
should be checked to see whether it exceeds the set
boundary.

4.2 DOL generation jumping

After the DOL processes the initial population, the algo-

rithm enters the SCA processing phase and is updated
according to the SCA rules. In order to cooperate with
different algorithms, jumping rate (Jr) is introduced into
DOL as the basis for whether the algorithm enters into
DOL step. Jr is a number in the range of [0, 1] , and the
specific value should be determined by experiments. In the
algorithm step, the program randomly generates a random
number in the range of [0, 1] . If the random number is less
than the set Jr, then the algorithm will enter the DOL step:
PDO
ij ¼ Pij þ w  r3i  ðr4i  ðaj þ bj  Pij Þ  Pij Þ ð30Þ

The boundary between A and B will be dynamically

updated:
aj ¼ minðPij Þ
ð31Þ
bj ¼ maxðPij Þ
As with the other steps, the updated population needs to
check the boundaries.

4.3 EDOLSCA steps

In addition to the DOL strategy, the introduction of elites is

intended to address the exploration speed. The concept of
5 Experimental results and discussion
elite is to select a part of the best individuals before each
iteration, skipping the intermediate steps and keeping until
5.1 Benchmark test
the end of each iteration to form a new population, so as to
ensure the overall superiority of the population.
In this test, the benchmark proposed by CEC2014 [27] was
At this point, the specific process of EDOLSCA has
selected as the test function. The 14 alternative benchmarks
been introduced, and the flowchart of the algorithm is
are shown in Table 2, the first four benchmarks are uni-
shown in Fig. 4, and steps of EDOLSCA are shown in
modal functions, and the remaining twelve are multi-modal
Algorithm.1.
functions. These benchmarks are effective to test the
exploration and development capabilities of algorithms. In

123
 Neural Computing and Applications

C1 ¼ C2 ¼ 2:05, K ¼ 0:729. For the parameters of

EDOLSCA, w range is 1–15; Jr range is 0.1–1. The algo-
rithms run 10 times for independent testing, and the results
are shown in Tables 3, 4. The average with the smallest
index is chosen, that is w ¼ 10 and Jr ¼ 1. The population
size in the test was set to 50, and the number of iterations
was 20,000. Results of the mean and standard deviation of
each algorithm at each benchmark are recorded to compare
the algorithms. Average index is considered as one of the
criteria of comprehensive evaluation algorithm, which is
calculated from the ranking of each algorithm. The calcu-
lated results are shown in Table 5, and the convergence
diagram is shown in Fig.5.
First of all, it can be seen from the calculation results in
the table that, for F1-F4, three of the four baselines
EDOLSCA reach the optimal value, and in F5-F14, most
EDOLSCA also get the minimum value of all algorithms.
’Ave index’ shows that EDOLSCA has the best compre-
hensive result among all comparison algorithms, and only
the improved SCA in one benchmark performs less well
than the original algorithm. It is observed in the figure that
EDOLSCA converges much faster than SCA in most
benchmarks, which proves that the improvement is suc-
cessful. In addition to using the average index method to
detect EDOLSCA, the Wilcoxon rank-sum test [47]
method is also introduced to analyze the results of the
algorithm. The significance is set to 0.05, two-tailed. The
test results are shown in Table 6.In the result table, ’"’
denotes the EDOLSCA is better than other counterpart,
’ ’ means other algorithms (the algorithm pointed by the
arrow) is better, and ’=’ means there is no significant dif-
ference between the two algorithms after the test. The
results show that the convergence accuracy has a small
improvement compared with SCA, but the overall
improvement ratio is not high. It shows that EDOLSCA has
an absolute advantage in convergence speed, but it can still
be improved in terms of accuracy.

5.2 Application to PFHE design problems

In this section, the optimization object is extended to the

design problem of PFHE. In fact, the design optimization
problem of PFHE is different from the repeatability opti-
Fig. 4 The flowchart of EDOLSCA mization. The design optimization problem only needs an
optimal design, and the stability of the repeatability opti-
order to make the comparison of algorithms more intuitive, mization is more important [54]. Therefore, the stability of
this paper selects four standard and classical algorithms the algorithm is not considered in the PFHE problem,
with improvements, namely, PSO [25], cfPSO [10], where only the accuracy of the algorithm is considered. For
cfwPSO and ETLBO [24]. Algorithm parameters are as the heat transfer area objective function, both cases 1 and 2
follows: for PSO, C1 ¼ C2 ¼ 2; for cfPSO and cfwPSO, are applicable. In order to optimize the environment

123
Neural Computing and Applications

Table 2 Unimodal and

Function number Function name
multimodal benchmark
F1 Rotated High Conditioned Elliptic Function
F2 Rotated Bent Cigar Function
F3 Rotated Discus Function
F4 Shifted and Rotated Rosenbrock’s Function
F5 Shifted and Rotated Weierstrass Function
F6 Shifted and Rotated Griewank’s Function
F7 Shifted Rastrigin’s Function
F8 Shifted and Rotated Rastrigin’s Function
F9 Shifted Schwefel’s Function
F10 Shifted and Rotated Schwefel’s Function
F11 Shifted and Rotated Katsuura Function
F12 Shifted and Rotated HappyCat Function
F13 Shifted and Rotated HGBat Function
F14 Shifted and Rotated Expanded Griewank’s plus Rosenbrock’s Function

Table 3 The sensitivity analysis

Weight Mean Ave index
of w
F1 F3 F4 F7

w=1 5.27E ? 07 2.03E ? 04 1.66E ? 02 5.07E ? 00 13.75

w=2 3.36E ? 07 1.76E ? 04 1.41E ? 02 2.75E ? 00 11
w=3 1.69E ? 07 1.79E ? 04 1.42E ? 02 2.24E ? 00 8.75
w=4 2.02E ? 07 1.96E ? 04 1.17E ? 02 2.13E ? 00 10
w=5 1.94E ? 07 1.93E ? 04 8.24E ? 01 1.33E ? 00 8.25
w=6 1.62E ? 07 1.69E ? 04 9.63E ? 01 1.54E ? 00 6.75
w=7 2.52E ? 07 2.02E ? 04 4.62E ? 01 1.43E ? 00 8.5
w=8 2.17E ? 07 1.55E 1 04 5.37E ? 01 1.08E ? 00 6.5
w=9 1.88E ? 07 1.80E ? 04 5.08E ? 01 1.01E ? 00 5.75
w = 10 1.46E ? 07 1.88E ? 04 4.64E ? 01 9.22E-01 4
w = 11 1.72E ? 07 2.04E ? 04 5.59E ? 01 9.09E-01 6.75
w = 12 2.74E ? 07 2.36E ? 04 5.45E ? 01 9.48E-01 9.5
w = 13 2.10E ? 07 2.48E ? 04 4.41E ? 01 8.97E-01 6.5
w = 14 2.70E ? 07 2.05E ? 04 3.49E 1 01 9.71E-01 7.5
w = 15 1.44E 1 07 2.73E ? 04 4.61E ? 01 1.05E ? 00 6.5
Bold values indicate the best results compared to the other counterparts

Table 4 The sensitivity analysis

Jumping rate Mean Ave index
of Jr
F1 F3 F4 F7

Jr = 0.1 2.39E ? 07 2.66E ? 04 4.96E ? 01 1.88E ? 00 9.25

Jr = 0.2 3.07E ? 07 2.40E ? 04 4.58E ? 01 1.11E ? 00 8
Jr = 0.3 1.66E ? 07 2.27E ? 04 4.60E ? 01 9.15E-01 7
Jr = 0.4 1.10E ? 07 1.90E ? 04 5.05E ? 01 6.57E-01 6.5
Jr = 0.5 1.68E ? 07 1.43E ? 04 4.79E ? 01 5.26E-01 6.25
Jr = 0.6 1.35E ? 07 1.28E ? 04 3.41E 1 01 3.43E-01 2.75
Jr = 0.7 1.58E ? 07 1.41E ? 04 5.62E ? 01 5.94E-01 6.5
Jr = 0.8 9.57E ? 06 1.30E ? 04 4.42E ? 01 6.08E-01 3.75
Jr = 0.9 1.27E ? 07 9.60E 1 03 4.61E ? 01 4.47E-01 3.5
Jr = 1 6.42E 1 06 1.13E ? 04 4.42E ? 01 3.92E-01 1.5
Bold value indicates the best results compared to the other counterparts

123
 Neural Computing and Applications

realistically, the number of iterations is set to 500,000, and Fig. 5 The convergence trends of all algorithms on all unimodal and c
the population is still 50. For the objective function of simple multimodal functions
entropy increase, only test case 1, the number of iterations
is set to 4,000, and the population is 50. In order to com-
pare the fairness of the study, we add two varieties of It can be seen from the result image that SCA and
algorithms in this section: elite sine cosine algorithm EDOLSCA have the same advantage in the AHT objective
(ESCA), dynamic opposite learning sine cosine algorithm function. The EDOLSCA continued to converge under a
(DOLSCA) and two common algorithms: differential high number of iterations, and the improved algorithm has
evolution (DE) [41] and neighborhood field for cooperative higher accuracy than the original algorithm. On the Ns
optimization (NFO) [48]. The results are shown in Figs.6, objective function, EDOLSCA converges faster than SCA,
7. and the accuracy difference is larger. At the same time, in

Table 5 The mean and standard value of unimodal/multi-modal test functions

Algorithms F1 F2 F3 F4
Mean Std Mean Std Mean Std Mean Std

PSO 6.32E ? 07 3.26E ? 07 8.52E ? 07 6.17E ? 07 1.17E 1 03 7.06E ? 02 4.09E ? 02 8.59E ? 01

cfPSO 2.74E ? 08 9.05E ? 07 1.19E ? 10 5.61E ? 09 2.31E ? 04 6.30E ? 03 1.48E ? 03 6.14E ? 02
cfwPSO 2.29E ? 08 8.71E ? 07 1.51E ? 10 5.50E ? 09 3.24E ? 04 8.70E ? 03 2.00E ? 03 8.59E ? 02
ETLBO 3.84E ? 08 1.29E ? 08 3.54E ? 10 8.50E ? 09 6.58E ? 04 1.38E ? 04 3.93E ? 03 8.02E ? 02
SCA 2.82E ? 07 1.09E ? 07 5.04E ? 08 5.22E ? 08 4.73E ? 04 1.28E ? 04 1.19E ? 02 5.88E ? 01
EDOLSCA 2.63E 1 07 2.08E 1 07 7.38E 1 07 1.10E 1 08 5.11E ? 04 2.51E 1 04 1.03E 1 02 5.05E 1 01
Algorithms F5 F6 F7 F8

PSO 2.66E ? 01 2.08E ? 00 2.55E ? 00 1.41E ? 00 9.31E ? 01 1.88E ? 01 1.16E ? 02 2.35E ? 01

cfPSO 2.66E ? 01 2.81E ? 00 1.56E ? 02 6.13E ? 01 1.29E ? 02 2.72E ? 01 1.45E ? 02 2.84E ? 01
cfwPSO 2.92E ? 01 2.30E ? 00 1.67E ? 02 5.62E ? 01 1.38E ? 02 2.64E ? 01 1.55E ? 02 3.18E ? 01
ETLBO 3.41E ? 01 2.43E ? 00 3.62E ? 02 1.04E ? 02 2.57E ? 02 1.73E ? 01 2.62E ? 02 3.26E ? 01
SCA 2.12E 1 01 2.77E ? 00 6.53E ? 00 9.34E ? 00 1.02E ? 02 1.90E ? 01 1.17E ? 02 1.41E ? 01
EDOLSCA 3.27E ? 01 5.83E 1 00 1.58E 1 00 1.86E 1 00 5.57E 1 01 1.43E 1 01 7.80E 1 01 1.49E 1 01

Algorithms F9 F10 F11 F12

PSO 2.09E ? 03 5.04E ? 02 3.68E 1 03 6.70E ? 02 5.09E-01 1.59E-01 5.83E-01 1.26E-01

cfPSO 3.04E ? 03 6.28E ? 02 4.04E ? 03 6.52E ? 02 5.10E-01 1.04E-01 3.32E ? 00 4.02E-01
cfwPSO 2.01E 1 03 5.01E ? 02 4.82E ? 03 1.19E ? 03 1.04E 1 00 5.68E-01 3.65E ? 00 3.98E-01
ETLBO 6.51E ? 03 4.25E ? 02 6.96E ? 03 6.15E ? 02 3.08E ? 00 5.26E-01 5.32E ? 00 6.99E-01
SCA 3.33E ? 03 5.65E ? 02 4.19E ? 03 2.93E ? 02 1.41E ? 00 2.04E-01 5.37E-01 8.89E-02
EDOLSCA 3.86E ? 03 1.45E 1 03 4.60E ? 03 1.67E 1 03 2.21E ? 00 3.80E-01 3.15E-01 8.13E-02

Algorithms F13 F14 Best num Ave index

PSO 3.23E-01 1.90E-01 4.78E ? 01 1.89E ? 01 2 2.36

cfPSO 5.94E ? 01 2.01E ? 01 8.63E ? 02 1.01E ? 03 0 3.86
cfwPSO 6.82E ? 01 1.23E ? 01 4.84E ? 02 6.49E ? 02 2 4.07
ETLBO 1.30E ? 02 2.77E ? 01 4.22E ? 04 3.32E ? 04 0 5.86
SCA 1.12E ? 00 2.75E ? 00 2.65E ? 01 8.50E ? 00 1 2.64
EDOLSCA 2.74E-01 4.60E-02 1.85E 1 01 1.04E 1 01 9 2.21

Bold value indicates the best results compared to the other counterparts

123
Neural Computing and Applications

123
 Neural Computing and Applications

-0.2
Table 6 Wilcoxon rank-sum test on unimodal and multi-modal

log10(Number of Heat Transfer Area)

PSO
functions cfPSO
cfwPSO
DE
Algorithms EDOLSCA NFO
-0.4 ETLBO
SCA
F1 F2 F3 F4 F5 F6 F7 F8 ESCA
DOLSCA
EDOLSCA
PSO " = " " " "
-0.6
cfPSO " " = " " " "
cfwPSO " " = " = " " "
ETLBO " " = " " " " "
-0.8
SCA = " = = " " " 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25

FES

Algorithms EDOLSCA Same Better Fig. 7 The solution result of the Ns objective, cases 1

F9 F10 F11 F12 F13 F14

6 Conclusion
PSO = " " " 2 8
cfPSO = = " " " 3 9 This paper proposed a new EDOLSCA adopted elite and
cfwPSO = " " " 3 9 DOL strategies to strengthen the canonical SCA, which
ETLBO " " " " " " 1 13 greatly improves the performance of the original algorithm
SCA = = = " = = 8 5 in terms of convergence accuracy and convergence speed.
The proposed method was compared with the state-of-the-
art algorithms counterparts in the CEC2014 benchmark test
proved to perform well. The algorithm is further adopted in
the three cases, the accuracy of EDOLSCA is the highest, solving the optimal design problem of PFHE, where the
which proves its potentials in engineering design opti- dynamic and random search space of the DOL strategy has
mization problems. The results of benchmark and engi- achieved excellent results. The proposed algorithm again
neering tests show that EDOLSCA has excellent dynamic demonstrated superiority in solving the complex optimal
search capabilities and has unique advantages in speed and design problem. Future work lies on the wide application of
accuracy, as well as demonstrating excellent comprehen- the proposed EDOLSCA in various industrial and science
sive performance. optimization scenarios. The well robustness of the algo-
rithm is promising to provide a useful computational tool
for engineering optimization.

0.4 1

PSO PSO
0.9
log10(Number of Heat Transfer Area)

cfPSO
log10(Number of Heat Transfer Area)

0.35 cfPSO
cfwPSO cfwPSO
0.8
DE DE
0.3
NFO 0.7 NFO
ETLBO ETLBO
0.25 SCA 0.6 SCA
ESCA ESCA
0.2 DOLSCA 0.5 DOLSCA
EDOLSCA EDOLSCA
0.4
0.15
0.3
0.1
0.2
0.05 0.1

0 0
0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25
FES FES
(a) (b)
Fig. 6 The solution result of the AHT objective, cases 1 and 2

123
Neural Computing and Applications

Acknowledgements This research work is supported by the National 12. Dhavle SV, Kulkarni AJ, Shastri A, Kale IR (2018) Design and
Key Research and Development Project under Grant economic optimization of shell-and-tube heat exchanger using
2018YFB1700500 and National Science Foundation of China under cohort intelligence algorithm. Neural Comput Appl
Grants 52077213 and 62003332. 30(1):111–125
13. Dorigo M, Birattari M, Stutzle T (2006) Ant colony optimization.
IEEE Comput Intell Mag 1(4):28–39
Declarations 14. Ekiz S, Erdoğmuş P, Özgür B (2017) Solving constrained opti-
mization problems with sine-cosine algorithm. Periodicals of
Conflict of interest The authors declared that they have no conflicts of Engineering and Natural Sciences 5(3)
interest to this work. We declare that we do not have any commercial 15. Ergezer M, Simon D, Du D (2009) Oppositional biogeography-
or associative interest that represents a conflict of interest in con- based optimization. In: 2009 IEEE international conference on
nection with the work submitted. systems, man and cybernetics, pp. 1009–1014. IEEE
16. Gupta S, Deep K (2019) A hybrid self-adaptive sine cosine
Open Access This article is licensed under a Creative Commons algorithm with opposition based learning. Exp Syst Appl
Attribution 4.0 International License, which permits use, sharing, 119:210–230
adaptation, distribution and reproduction in any medium or format, as 17. Gupta S, Deep K (2019) Improved sine cosine algorithm with
long as you give appropriate credit to the original author(s) and the crossover scheme for global optimization. Knowledge-Based Syst
source, provide a link to the Creative Commons licence, and indicate 165:374–406
if changes were made. The images or other third party material in this 18. Gupta S, Deep K (2020) A novel hybrid sine cosine algorithm for
article are included in the article’s Creative Commons licence, unless global optimization and its application to train multilayer per-
indicated otherwise in a credit line to the material. If material is not ceptrons. Appl Intell 50(4):993–1026
included in the article’s Creative Commons licence and your intended 19. Hekimoğlu B (2019) Sine-cosine algorithm-based optimization
use is not permitted by statutory regulation or exceeds the permitted for automatic voltage regulator system. Trans Ins Measure Con-
use, you will need to obtain permission directly from the copyright trol 41(6):1761–1771
holder. To view a copy of this licence, visit http://creativecommons. 20. Incropera FP, Lavine AS, Bergman TL, DeWitt DP (2007)
org/licenses/by/4.0/. Fundamentals of heat and mass transfer. Wiley
21. Iyer VH, Mahesh S, Malpani R, Sapre M, Kulkarni AJ (2019)
Adaptive range genetic algorithm: a hybrid optimization
References approach and its application in the design and economic opti-
mization of shell-and-tube heat exchanger.Eng Appl Artif Intell
1. Abd Elaziz M, Oliva D, Xiong S (2017) An improved opposition- 85:444–461
based sine cosine algorithm for global optimization. Exp Syst 22. John AK, Krishnakumar K (2017) Performing multiobjective
Appl 90:484–500 optimization on perforated plate matrix heat exchanger surfaces
2. Abeykoon C (2020) Compact heat exchangers-design and opti- using genetic algorithm. Int J Simul Multidiscip Design Optimiz
mization with cfd. Int J Heat Mass Trans 146:118766 8:A3
3. Attia AF, El Sehiemy RA, Hasanien HM (2018) Optimal power 23. Joshi HM, Webb RL (1987) Heat transfer and friction in the
flow solution in power systems using a novel sine-cosine algo- offset stripfin heat exchanger. Int J Heat Mass Trans 30(1):69–84
rithm. Int J Electric Power Energy Syst 99:331–343 24. Kadambur R, Kotecha P (2015) Multi-level production planning
4. Banerjee A, Nabi M (2017) Re-entry trajectory optimization for in a petrochemical industry using elitist teaching-learning-based-
space shuttle using sine-cosine algorithm. In: 2017 8th interna- optimization. Exp Syst Appl 42(1):628–641
tional conference on recent advances in space technologies 25. Kennedy J, Eberhart R (1995) Particle swarm optimization. In:
(RAST), pp. 73–77. IEEE Proceedings of ICNN’95-International Conference on Neural
5. Bejan A (2013) Entropy generation minimization: the method of Networks, vol. 4, pp. 1942–1948. IEEE
thermodynamic optimization of finite-size systems and finite-time 26. Li C, Luo Z, Song Z, Yang F, Fan J, Liu PX (2019) An enhanced
processes. CRC Press brain storm sine cosine algorithm for global optimization prob-
6. Bureerat S, Pholdee N (2017) Adaptive sine cosine algorithm lems. IEEE Access 7:28211–28229
integrated with differential evolution for structural damage 27. Liang JJ, Qu BY, Suganthan PN (2013) Problem definitions and
detection. In: International Conference on Computational Science evaluation criteria for the cec 2014 special session and compe-
and Its Applications, pp. 71–86. Springer tition on single objective real-parameter numerical optimization.
7. Campet R, Roy PT, Cuenot B, Riber É, Jouhaud JC (2020) Computational Intelligence Laboratory, Zhengzhou University,
Design optimization of an heat exchanger using gaussian process. Zhengzhou China and Technical Report, Nanyang Technological
Int J Heat Mass Trans 150:119264 University, Singapore 635
8. Chandrasekaran K, Sankar S, Banumalar K (2020) Partial shad- 28. Lincoln GA (1973) Energy conservation. Science
ing detection for pv arrays in a maximum power tracking system 180(4082):155–162
using the sine-cosine algorithm. Energy Sustain Develop 29. Liu C, Bu W, Xu D (2017) Multi-objective shape optimization of
55:105–121 a plate-fin heat exchanger using cfd and multi-objective genetic
9. Chegini SN, Bagheri A, Najafi F (2018) Psoscalf: A new hybrid algorithm. Int J Heat Mass Trans 111:65–82
pso based on sine cosine algorithm and levy flight for solving 30. Mahdad B, Srairi K (2018) A new interactive sine cosine algo-
optimization problems. Appl Soft Comput 73:697–726 rithm for loading margin stability improvement under contin-
10. Clerc M, Kennedy J (2002) The particle swarm-explosion, sta- gency. Electric Eng 100(2):913–933
bility, and convergence in a multidimensional complex space. 31. Mirjalili S (2016) Sca: a sine cosine algorithm for solving opti-
IEEE Trans Evol Comput 6(1):58–73 mization problems. Knowledge-based Syst 96:120–133
11. Das S, Bhattacharya A, Chakraborty AK (2018) Solution of 32. Nenavath H, Jatoth RK (2018) Hybridizing sine cosine algorithm
short-term hydrothermal scheduling using sine cosine algorithm. with differential evolution for global optimization and object
Soft Comput 22(19):6409–6427 tracking. Appl Soft Comput 62:1019–1043

123
 Neural Computing and Applications

33. Pasandideh SHR, Khalilpourazari S (2018) Sine cosine crow 45. Wang Z, Li Y (2016) Layer pattern thermal design and opti-
search algorithm: a powerful hybrid meta heuristic for global mization for multistream plate-fin heat exchangers–a review.
optimization. arXiv preprint arXiv:1801.08485 Renew Sustain Energy Rev 53:500–514
34. Rahnamayan S, Tizhoosh HR, Salama MM (2007) Quasi-oppo- 46. Whitley D (1994) A genetic algorithm tutorial. Statist Comput
sitional differential evolution. In: 2007 IEEE congress on evo- 4(2):65–85
lutionary computation, pp. 2229–2236. IEEE 47. Wilcoxon F(1992) Individual comparisons by ranking methods.
35. Rahnamayan S, Tizhoosh HR, Salama MM (2008) Opposition- In: Breakthroughs in statistics, pp. 196–202. Springer
based differential evolution. IEEE Trans Evol Comput 48. Wu Z, Chow TW (2013) Neighborhood field for cooperative
12(1):64–79 optimization. Soft Comput 17(5):819–834
36. Raja BD, Jhala R, Patel V (2018) Thermal-hydraulic optimization 49. Wu Z, Li Q, Wu W, Zhao M (2019) Crowdsourcing model for
of plate heat exchanger: A multi-objective approach. Interna- energy efficiency retrofit and mixed-integer equilibrium analysis.
tional J Therm Sci 124:522–535 IEEE Trans Ind Inform 16(7):4512–4524
37. Rao RV, Saroj A (2017) Constrained economic optimization of 50. Xu Y, Yang Z, Li X, Kang H, Yang X (2020) Dynamic opposite
shell-and-tube heat exchangers using elitist-jaya algorithm. learning enhanced teaching-learning-based optimization.
Energy 128:785–800 Knowledge-Based Syst 188:104966
38. Rao RV, Saroj A (2017) Economic optimization of shell-and-tube 51. Yang Z, Mourshed M, Liu K, Xu X, Feng S (2020) A novel
heat exchanger using jaya algorithm with maintenance consid- competitive swarm optimized rbf neural network model for short-
eration. Appl Therm Eng 116:473–487 term solar power generation forecasting. Neurocomputing
39. Rao RV, Savsani VJ, Vakharia D (2012) Teaching-learning- 397:415–421
based optimization: an optimization method for continuous non- 52. Yousefi M, Darus A, Mohammadi H (2012) An imperialist
linear large scale problems. Information sciences 183(1):1–15 competitive algorithm for optimal design of plate-fin heat
40. Sindhu R, Ngadiran R, Yacob YM, Zahri NAH, Hariharan M exchangers. Int J Heat Mass Transf 55(11–12):3178–3185
(2017) Sine-cosine algorithm for feature selection with elitism 53. Yousefi M, Enayatifar R, Darus AN, Abdullah AH (2013) Opti-
strategy and new updating mechanism. Neural Comput Appl mization of plate-fin heat exchangers by an improved harmony
28(10):2947–2958 search algorithm. Appl Therm Eng 50(1):877–885
41. Storn R, Price K (1997) Differential evolution-a simple and 54. Yuen SY, Lou Y, Zhang X (2019) Selecting evolutionary algo-
efficient heuristic for global optimization over continuous spaces. rithms for black box design optimization problems. Soft Comput
J Glob Optimiz 11(4):341–359 23(15):6511–6531
42. Van Laarhoven PJ, Aarts EH (1987) Simulated annealing. In: 55. Zhang L, Li Y, Zhang H, Xu X, Yang Z, Xu W (2021) A review
Simulated annealing: Theory Appl, pp. 7–15. Springer of the potential of district heating system in northern china. Appl
43. de Vasconcelos Segundo EH, Mariani VC, dos Santos Coelho L Therm Eng p. 116605
(2019) Design of heat exchangers using falcon optimization
algorithm. Appl Therm Eng 156:119–144 Publisher’s Note Springer Nature remains neutral with regard to
44. Wang J, Yang W, Du P, Niu T (2018) A novel hybrid forecasting jurisdictional claims in published maps and institutional affiliations.
system of wind speed based on a newly developed multi-objec-
tive sine cosine algorithm. Energy Convers Manag 163:134–150

123