SoftwareX 19 (2022) 101192

Contents lists available at ScienceDirect

SoftwareX
journal homepage: www.elsevier.com/locate/softx

Original software publication

CIOA: Circle-Inspired Optimization Algorithm, an algorithm for

engineering optimization
∗
Otávio Augusto Peter de Souza a , , Letícia Fleck Fadel Miguel b
a
Postgraduate Program in Civil Engineering (PPGEC), Federal University of Rio Grande do Sul (UFRGS), Av. Osvaldo Aranha, 99, 90035-190, Porto
Alegre/Rio Grande do Sul, Brazil
b
Department of Mechanical Engineering (DEMEC), Postgraduate Program in Mechanical Engineering (PROMEC), Postgraduate Program in Civil
Engineering (PPGEC), Federal University of Rio Grande do Sul (UFRGS), Av. Sarmento Leite, 425, 90050-170, Porto Alegre/Rio Grande do Sul, Brazil

article info a b s t r a c t

Article history: This paper presents a new, robust and very efficient metaheuristic optimization algorithm, called
Received 20 May 2022 Circle Inspired Optimization Algorithm (CIOA), for solving constrained and unconstrained engineering
Received in revised form 15 July 2022 optimization problems. The inspiration for the proposed algorithm consists of well-known formulations
Accepted 4 August 2022
of the trigonometric circle. CIOA is compared with five other very famous algorithms in ten benchmark
Keywords: function optimization problems, five real-world engineering constrained optimization problems, and
Metaheuristic algorithms also four structural optimization problems for plane and spatial trusses subjected to multiple and
Real-world problems different types of constraints. The results obtained demonstrate that the proposed algorithm is more
Structural optimization efficient than other famous algorithms, contributing to the accurate and fast solution of complex
Circle-Inspired Optimization Algorithm optimization problems.
© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

Code metadata

Current code version v01

Permanent link to code/repository used for this code version https://github.com/oapsouza/CIOA-Circle-Inspired-Optimization-Algorithm-
Matlab-version
Permanent link to reproducible capsule -
Legal code license BSD-3-Clause
Code versioning system used -
Software code languages, tools and services used MATLAB
Compilation requirements, operating environments and dependencies MATLAB
If available, link to developer documentation/manual -
Support email for questions [email protected] or [email protected]

1. Motivation and significance Numerous metaheuristic algorithms have been developed in

recent decades. Among the main ones are Simulated Annealing
The classic optimization algorithms are generally determin- (SA) [2], Particle Swarm Optimization (PSO) [3], Differential Evo-
istic and gradient-based. These algorithms are very efficient in lution (DE) [4], Harmony Search (HS) [5], Artificial Bee Colony
the optimization of unimodal functions and without disconti- (ABC) [6,7], Ant Colony Optimization (ACO) [8], Firefly Algorithm
nuities. However, most engineering optimization problems are (FA) [9,10], Gravitational Search Algorithm (GSA) [11], Cuckoo
non-linear, involve complex multimodal functions and have nu- Search (CS) [12], Bat Algorithm (BA) [13], Flower Pollination Algo-
merous restrictions that need to be addressed simultaneously. rithm (FPA) [14], Krill Herd (KH) [15], Backtracking Search Opti-
Metaheuristic algorithms are efficient in solving these problems mization Algorithm (BSA) [16], Grey Wolf Optimizer (GWO) [17],
because they are stochastic algorithms in which the optimization Search Group Algorithm (SGA) [18], Whale Optimization Algo-
process occurs through an exchange between diversification and rithm (WOA) [19], Spotted Hyena Optimizer (SHO) [20], Emperor
intensification [1]. Penguin Optimizer (EPO) [21], Butterfly Optimization Algorithm
(BOA) [22], Sooty Tern Optimization Algorithm (STOA) [23], Seag-
∗ Corresponding author. ull Optimization Algorithm (SOA) [24], Multi-Operator Differ-
E-mail addresses: [email protected] (Otávio Augusto Peter de Souza), ential Evolution (EnMODE) [25], Self-Adaptive Spherical Search
[email protected] (Letícia Fleck Fadel Miguel). (SASS) [26], Tunicate Swarm Algorithm (TSA) [27], Multi Leader

https://doi.org/10.1016/j.softx.2022.101192
2352-7110/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Optimizer (MLO) [28], Darts Game Optimizer (DGO) [29], Spring

Search Algorithm (SSA) [30] and Rat Swarm Optimizer (RSO) [31].
Binary versions and hybrid versions of existing algorithms have
also been released recently, including: BOSA [32], BEPO [33],
ESA [34], HGOANM [35], HTSSA [36], CLFD [37] and EOBL-GOA
[38].
The use of metaheuristic algorithms is recurrent in many areas
of knowledge, such as in research related to safety, health, AI,
Multicore Systems, among others [39–44]. In Engineering, several
authors have used metaheuristics to solve complex optimization
problems that involve numerous constraints that need to be con-
sidered simultaneously [45–61]. Since no metaheuristic algorithm Fig. 1. Changing radius and updating the center of the circle: (a) the search
can be considered the best for all possible optimization problems, agent improves its classification by reducing the radius size, (b) the search agent
worsens its classification by increasing the radius size.
the development of algorithms that accurately and quickly solve
specific problems in a given area becomes relevant [62].
Thus, this paper presents the proposal and formulation of a
new, robust and very efficient metaheuristic optimization algo- The CIOA is initialized generating a vector ⃗r in which the value
rithm called the Circle-Inspired Optimization Algorithm (CIOA) to of each element rj is calculated by Eq. (1).
optimize engineering problems. The CIOA is formulated through cr .j2
some well-known properties of the trigonometric circle. The algo- rj = , 1 ≤ j ≤ Nag (1)
rithm’s search agents describe arc trajectories and are governed Nag
by two specific parameters: a user-defined angle and a radius in which Nag is the number of search agents and cr is a constant
of the circumference that varies with each iteration depending determined using Eq. (2). In this way, the elements of the radius
on the evaluation of each search agent. The CIOA’s robustness vector are ordered in ascending order, in which the first element
and efficiency are evaluated in ten benchmark functions well corresponds to r1 = cr /Nag and the last element is given by
known in the literature where the CIOA’s performance is com- rNag = cr .Nag .
pared with five other famous algorithms. In addition, the CIOA is √
applied to five real-world optimization problems provided for the U b − Lb
cr = (2)
’CEC2020 One Goal Restricted Optimization Competition in the Nag
Real World’ [63] and also four Truss optimization problems sub- in which Ub and Lb are the upper and lower bounds of each
ject to multiple constraints. Thus, the main contribution of this variable, respectively.
paper is to propose and validate a new and effective optimization After each search agent assigns random values to the design
tool for engineering problems. variables in the first solution, an evaluation of the objective
function is carried out and then each search agent is classified in
2. Software description a ranking according to the quality of the solution that it obtained.

This section presents a description of the developed algorithm: 2.2.2. CIOA main loop
CIOA. Initially, an overview of the code’s architecture is presented, Throughout the iterations, each search agent will have its
then details of its implementation and, finally, aspects of the coordinates in the next iteration defined by the classification
algorithm’s functionality. performed in the previous iteration. The best-classified agents,
i.e., those who obtained the best solutions will make shorter
2.1. Software architecture movements (using smaller radii of the vector ⃗r ) while the agents
that occupy the worst classifications will make longer move-
The CIOA is an optimization algorithm programmed in MAT- ments. A search agent classified as the jth best solution in an
LAB, composed of two files: Main_Program_CIOA.m and Objec- iteration k will have its new coordinates in the iteration k + 1
tive_Function_CIOA.m calculated using Eqs. (3) and (4).
Main_Program_CIOA.m is the main file, in which the algorithm
formulation is programmed. It is in this file that the user must x2i (k + 1) = x2i (k) − rand1 .rj . sin (k.θ) + rand2 .rj . sin((k + 1) .θ )
adjust the algorithm parameters and inform details about the op- (3)
timization problem. The Objective_Function_CIOA.m is a function x2i−1 (k + 1) = x2i−1 (k) − rand3 .rj . cos (k.θ )
file in which the user must implement the objective function he
wants to optimize. In this file, inequality and equality constraints
+ rand4 .rj . cos((k + 1) .θ ) (4)
(if any) and the penalty adopted will also be programmed. in which 2i and 2i−1 refer to even and odd numbers, respectively.
Therefore Eq. (3) updates the ⃗ x = [x2 ; x4 ; x6 ; . . .] coordinates
2.2. Software implementation while Eq. (4) updates the ⃗ x = [x1 ; x3 ; x5 ; . . .] coordinates; rand
variables are random numbers with a uniform distribution be-
In this section, a brief description of the formulation and tween zero and 1; the angle θ is a parameter provided by the
implementation of the CIOA is given. user; the variable rj corresponds to the jth element of the vector
⃗r , i.e., the jth smallest radius.
2.2.1. Initialization of the CIOA The process of changing radii and updating the center of
In the CIOA, each search agent moves along arcs governed the circle is outlined below: In a hypothetical case in which all
by two main parameters: An angle θ defined by the user and random variables have the same value, it is assumed that in an
a radius r calculated by the algorithm, whose value depends on iteration a search agent departs from point 1 to point 2 in a
the evaluation of the objective function: the better the evaluation movement governed by angle θ and radius r1,2 with center at
of the objective function performed by a given search agent, the O1,2 . In the next iteration, the agent leaves point 2 and moves
lower the value of r for this agent in the next iteration. to point 3 maintaining the angle θ , the movement is governed
2
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Fig. 2. Behavior of a search agent in extreme situations: (a) Case 1, (b) Case 2, (c) Case 3. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)

by a radius r2,3 and has a center at O2,3 . Two possible distinct in which ⃗rnew is the new radii vector after the update.
cases of this process are illustrated in Fig. 1: The first in (a), in
which the agent improves its classification by reducing the radius 2.2.3. Local search phase
size, i.e., r2,3 < r1,2 ; the second illustrated in (b), in which the
The main loop of the CIOA formulated in the previous section
search agent worsens its classification by increasing the radius
promotes global and local searches simultaneously. This is be-
size, i.e., r2,3 > r1,2 . To facilitate understanding, the action of
cause the agents that produce the best solutions describe small
random variables was not considered in Fig. 1.
Over consecutive iterations, each agent is likely to alternate movements corresponding to a local search while the agents with
their performance at random in the ranking of the best positions. the worst solutions describe large movements corresponding to
But three hypothetical cases of behavior can occur: a global search. However, it is important that the algorithm
dedicates some iterations only to local search, in which all agents
(a) In the first case, a search agent always worsens its classifica- are restricted to the most promising areas of the search space.
tion, increasing the radius that governs its movement along with
A parameter GlobIt that represents the proportion of iterations
the iterations;
before the exclusively local search is inserted in the algorithm
(b) In the second case, a search agent always improves its clas- and its value can vary freely in the interval (0, 1]. However, the
sification, reducing the radius that governs its movement along use of 0.75 ≤ GlobIt ≤ 0.95 is recommended.
with the iterations; The exclusively local search will start at iteration k when the
(c) In the third case, a search agent always presents the same ratio between k and the total number of iterations is greater than
classification among the group of agents, keeping their radius GlobIt . At this moment, all search agents are restarted assuming
constant. the coordinates that produced the best solution so far. In addition,
Fig. 2 illustrates the behavior of a search agent in each of a change is made to the upper and lower bounds of each variable
the hypothetical extreme situations applied to a two-dimensional so that the new limits are given by Ub1i and Lb1i , calculated using
problem: in (a) the behavior presented in Case 1 is described, Eqs. (6) and (7).
in (b) the behavior exposed in Case 2 and in (c) the behavior
Ub − Lb
described in Case 3. In each frame of Fig. 2, the agent starts Ub1i = xibest + (6)
its journey in position 1. The blue line represents the agent’s 10000
trajectory without including randomization while the yellow line Ub − Lb
Lb1i = xibest − (7)
represents one of the possible trajectories when randomization 10000
is considered (variables rand in Eqs. (3) and (4)). In this way, the in which xibest is the variable in dimension i that produced the
agent ends its journey in position 2 when randomization is not
best solution so far.
considered and in position 2’ when considering random variables.
After initializing, this local search step is governed by the same
Two examples are presented for each extreme case, in which, in
equations as the main loop described in Section 2.2.2, replacing
each case, it is observed that without randomization the behavior
of a given agent is the same in both examples, whereas when Ub with Ub1i and Lb with Lb1i . In this way, the search space for
considering random variables, the behavior tends to change each agents is restricted, forcing the values for the design variables
execution. to be close to the values that generated the best solution in the
In the hypothesis that in a given iteration a variable xi has a main loop. For the specific cases in which Lb1i < Lb or Ub1i > Ub ,
value greater than Ub or less than Lb this variable will receive the whenever the value of a design variable is in the ranges Lb1i <
value corresponding to the variable xi of the search agent who xi < Lb or Ub1i > xi > Ub , its value will be automatically updated
obtained the best solution in the most recent iteration. to, respectively, xi = Lb or xi = Ub .
Whenever after k iterations the search agents complete a
complete lap, that is, whenever k.θ exceeds a multiple of 360◦ 2.2.4. User information and pseudocode
an update of the vector ⃗r is calculated using Eq. (5), to accelerate
In this section, the most important information that should
the convergence of the algorithm.
be taken into account by CIOA users is added, to take the best
⃗rnew = ⃗r .0.99 (5) advantage of the proposed algorithm. The suggested values for
3
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

the CIOA parameters were obtained through sensitivity analysis real-world problems used in CEC 2020 and also in engineer-
of the algorithm. ing problems related to structural optimization well known in
the literature. Finally, statistical and convergence analyses are
(a) Define 0.75 < GlobIt < 0.95. To promote a good balance
performed.
between accuracy and computational time GlobIt = 0.85 is rec-
In all analyses, the performance of the CIOA is compared to
ommended. Larger values can reduce computational time while
that of five famous algorithms: PSO [3], HS [5], FA [9,10], SGA [18]
smaller values tend to increase accuracy.
and WOA [19]. The algorithms are implemented according to
(b) Although θ can take on any value between 0◦ and 360◦ , tests their original papers in Matlab using the following computing
carried out on several problems show that, for better perfor- platform: Windows 10, 8th generation Core i5 processor and 8 GB
mance, θ should not be a 360◦ divider. Relatively low θ values, of memory. In all simulations, the parameters used for the CIOA
such as θ = 13◦ , θ = 17◦ or θ = 19◦ , produced good results take into account the instructions presented in Section 2.2.4that
in different types of problems, with different numbers of agents were obtained through sensitivity analyses. Therefore, it was
and iterations considered. High values of θ produce excellent adopted for the CIOA θ = 17◦ and GlobIt = 0.85.
results only in specific cases, presenting inferior performance in
structural optimization problems, for example. 3.1. Experimental analysis through benchmark functions
(c) The algorithm performs better when the number of itera-
Ten functions are used to validate the CIOA, which are pre-
tions is 2 to 5 times greater than the number of search agents.
sented in Table 1. These functions are part of a well-known
Preferably use more than 100 search agents.
set of benchmark test functions used to validate metaheuristic
The CIOA pseudocode is presented in Algorithm 1.
algorithms [17,19–24,27–32,34]. These functions are divided into
three groups. The functions of the first group, f1 to f3 , are uni-
2.3. Software functionalities
modal and evaluate the exploitation capacity of the algorithm.
The functions of the second and third groups determine the ex-
In the Main_Program_CIOA.m file, the user will have to define
ploration capability of the algorithm. The second group contains
the θ angle, in degrees, and the GlobIt parameter (ThetaAngle
multimodal functions (f4 and f5 ), while the third group is formed
and GlobIt in the code, as shown in Fig. 3(a). In addition, the
by multimodal functions with fixed dimensions (f6 to f10 ).
following must be informed: The number of variables in the For each problem, 50 independent runs were performed. In
problem (Nv ar ); the number of search agents and iterations to be each run, 200,000 objective function evaluations were executed,
considered (Nag and Nit ) and the upper and lower limits for the consisting of 250 research agents and 800 iterations. The results
design variables (Ub and Lb ). In the Objective_Function_CIOA.m obtained, shown in Table 2, are the best value for the global
file, the user will describe the objective function (obj) and the optimum, the mean value for the global optimum, the standard
number of equality and inequality constraints (nug and nuh). If deviation between runs and the average computational time for
there are no restrictions, the number informed will be zero. Then, each run (in seconds).
the restrictions must be implemented, as well as the weight value As can be seen in Table 2, for the optimal value of the objective
of the penalty to be considered, as shown in Fig. 3(b). For use, the function, the CIOA presented the best result in one multimodal
code must be executed in the Main_Program_CIOA.m. function (f5 ) and in all five multimodal functions with fixed
dimensions (f6 to f10 ). For the results in terms of mean value, CIOA
3. Illustrative examples performed best on a unimodal function (f2 ), a multimodal func-
tion (f5 ) and all five multimodal functions with fixed dimensions
In this section, the Circle-Inspired Optimization Algorithm (f6 to f10 ). Regarding computational time, the CIOA presented the
(CIOA) is validated through an experimental analysis involv- second best performance in all functions. More detailed statistical
ing the optimization of ten benchmark functions known from analyses are presented in Section 3.3.
the literature. Then, the CIOA is applied in the solution of five
4
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Fig. 3. CIOA functionalities.

Table 1
Benchmark test functions: D is the number of design variables and fmin is the optimal value.
Group Function D Range fmin
D D
∑ ∏
f1 (x) = | xi | + |xi | 10 [−10, 10] 0
Unimodal i=1 i=1
D−1
∑ )2
[100 xi+1 − x2i + (xi − 1)2 ]
(
f2 (x) = 10 [−30, 30] 0
i=1
D
∑
f3 (x) = (|xi + 0.5|)2 10 [−100, 100] 0
i=1
D
∑[
x2i − 10cos (2π xi ) + 10
]
f4 (x) = 10 [−5.12, 5.12] 0
Multimodal
i=1
D D ( )
1 ∑ ∏ xi
f5 (x) = x2i − cos √ +1 10 [−600, 600] 0
4000 i
i=1 i=1
11 ⌉2
∑⌈ x1 (b2i b i x2 )
+
f6 (x) = ai − 4 [−5, 5] 0.00030
b2i + bi x3 x4 +
i=1 ⎛ ⎞
Multimodal with fixed dimensions 4
∑ 6
∑ )2
f7 (x) = −
(
ci exp ⎝− aij xj − pij ⎠ 6 [0, 1] −3.32
i=1 j=1
5
∑ ]−1
f8 (x) = − (X − ai ) (X − ai )T + ci
[
4 [0, 10] −10.1532
i=1
7
∑[ ]−1
f9 (x) = − (X − ai ) (X − ai )T + ci 4 [0, 10] −10.4028
i=1
10
∑[ ]−1
f10 (x) = − (X − ai ) (X − ai )T + ci 4 [0, 10] −10.536
i=1

5
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Table 2
Results for the benchmark functions.
F CIOA WOA SGA FA HS PSO
f1 best 2.74E−05 0.00E+00 1.10E−03 6.86E−02 1.73E−04 0.00E+00
mean 1.77E−04 0.00E+00 1.39E−03 8.31E−02 2.26E−04 0.00E+00
std 1.32E−04 0.00E+00 1.86E−04 1.04E−02 6.12E−05 0.00E+00
time 3.19E+00 5.63E+00 2.00E+00 7.09E+00 1.10E+01 9.60E+00

f2 best 5.48E−03 1.87E−03 1.11E+00 3.59E+00 7.87E−02 3.05E−01

mean 2.81E+00 4.65E+00 3.13E+00 8.50E+00 5.91E+00 2.86E+00
std 1.74E+00 9.74E+00 6.46E−01 3.11E+00 6.33E+00 1.77E+00
time 3.32E+00 5.86E+00 2.28E+00 7.46E+00 1.11E+01 1.01E+01

f3 best 6.67E−10 0.00E+00 1.97E−05 4.63E−02 4.93E−09 0.00E+00

mean 4.01E−09 0.00E+00 2.90E−05 6.39E−02 7.94E−09 0.00E+00
std 2.64E−09 0.00E+00 8.42E−06 1.05E−02 3.59E−09 0.00E+00
time 3.07E+00 5.59E+00 2.00E+00 7.12E+00 1.09E+01 9.40E+00

f4 best 1.61E−06 0.00E+00 1.15E−05 2.91E+00 9.09E−07 3.98E+00

mean 3.45E−01 6.53E+00 1.69E+00 9.00E+00 1.42E−06 1.06E+01
std 4.68E−01 7.15E+00 1.03E+00 3.64E+00 3.23E−07 4.33E+00
time 3.27E+00 5.76E+00 2.17E+00 7.27E+00 1.11E+01 9.47E+00

f5 best 1.41E−07 7.58E−02 9.13E−05 1.82E−01 4.18E−02 4.18E−02

mean 5.39E−07 2.00E−01 1.59E−03 3.58E−01 7.52E−02 8.89E−02
std 7.89E−07 1.12E−01 6.56E−03 7.03E−02 2.22E−02 5.02E−02
time 3.42E+00 5.73E+00 2.33E+00 7.52E+00 1.09E+01 9.52E+00

f6 best 3.07E−04 3.07E−04 3.32E−04 3.08E−04 3.07E−04 3.07E−04

mean 3.16E−04 6.55E−04 6.00E−04 3.40E−04 4.62E−04 4.21E−04
std 5.95E−06 4.49E−04 1.19E−04 7.78E−05 3.53E−04 3.12E−04
time 3.59E+00 4.53E+00 3.27E+00 2.01E+01 9.74E+00 1.24E+01

f7 best −3.3220 −3.3220 −3.3220 −3.3220 −3.3220 −3.3220

mean −3.3220 −3.2315 −3.2055 −3.2744 −3.2792 −3.2578
std 2.91E−09 5.67E−02 1.68E−02 5.88E−02 5.76E−02 5.99E−02
time 4.75E+00 5.43E+00 3.44E+00 2.02E+01 1.04E+01 1.12E+01

f8 best −10.1532 −10.1532 −10.1532 −10.1529 −10.1532 −10.1532

mean −10.1532 −9.8473 −10.1532 −10.1518 −4.1063 −6.2816
std 5.59E−09 1.22E+00 1.92E−07 5.81E−04 2.76E+00 3.21E+00
time 4.67E+00 5.54E+00 4.39E+00 2.16E+01 1.07E+01 1.34E+01

f9 best −10.4029 −10.4029 −10.4029 −10.4028 −10.4029 −10.4029

mean −10.4029 −9.6588 −10.4029 −10.4017 −5.8754 −8.9268
std 5.87E−09 1.86E+00 1.73E−07 6.61E−04 3.63E+00 2.72E+00
time 5.07E+00 5.83E+00 4.64E+00 2.12E+01 1.12E+01 1.41E+01

f10 best −10.5364 −10.5364 −10.5364 −10.5360 −10.5364 −10.5364

mean −10.5364 −10.1038 −10.5364 −10.5352 −6.9520 −8.5055
std 5.46E−09 1.48E+00 2.40E−07 5.19E−04 3.79E+00 3.33E+00
time 5.84E+00 6.65E+00 5.40E+00 2.20E+01 1.17E+01 1.45E+01

Table 3
Real-World and Structural Engineering Optimization Problems: D is the number of variables, ng the number of inequality constraints, and nh the number of equality
constraints.
Problem Name or Description D ng nh
R1 Reactor Network Design (RND) [63] 6 1 4
R2 Two-reactor Problem [63] 7 4 4
R3 Tension/compression spring design [63] 3 3 0
R4 Pressure vessel design [63] 4 4 0
R5 Planetary gear train design optimization problem [63] 9 10 1
S1 Size optimization of a 25-bar space truss with stress and displacement constraints [46,49,53,54] 8 124 0
S2 Shape and size optimization of an 18-bar plane truss with stress and buckling constraints [47,49,53] 12 54 0
S3 Shape and size optimization of a 52-bar space truss with multiple natural frequency constraints [48,50–53] 13 2 0
S4 Size optimization of a realistic transmission tower of a 163-bar with stress, displacement, buckling and 11 810 0
fundamental natural frequency constraints [53]

6
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Table 4
Results obtained for the Real-World Problems and Structural Engineering Problems.
Prob. CIOA WOA SGA FA HS PSO
R1 BV 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
best −0.375348 −0.335389 −0.012813 −0.270118 −0.369001 −0.367176
MV 2.48E−07 1.23E−04 0.00E+00 9.84E−05 0.00E+00 1.18E−06
mean −0.337139 −0.153401 −0.004673 −0.248812 −0.334576 −0.311166
std 1.66E−02 1.68E−01 3.15E−03 9.65E−02 2.58E−02 5.65E−02
time 1.84 2.49 1.48 9.73 5.28 5.90

R2 BV 0.00E+00 0.00E+00 0.00E+00 7.66E−04 1.70E−08 1.25E−01

best 100.0198 100.2392 100.1649 99.2610 103.5389 136.8535
MV 2.00E−02 7.07E−02 7.50E−02 1.20E−01 1.10E−01 1.27E−01
mean 118.3750 137.9467 132.6793 145.6808 151.0005 151.6865
std 8.89E+00 2.62E+01 2.77E+01 1.79E+01 2.38E+01 1.49E+01
time 2.05 2.79 1.45 9.59 4.74 5.05

R3 BV 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00

best 0.0126654 0.0126679 0.0127192 0.0126914 0.0126673 0.0126688
MV 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
mean 0.0126966 0.0128988 0.0127197 0.0127583 0.0155228 0.0133129
std 2.56E−05 3.02E−04 4.17E−07 7.24E−05 2.09E−03 5.15E−04
time 1.46 1.63 1.26 9.72 3.62 4.96

R4 BV 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00

best 6059.713 6060.718 6059.975 6060.814 6059.715 6090.526
MV 0.00E+00 2.35E−07 9.65E−09 0.00E+00 0.00E+00 0.00E+00
mean 6118.333 6217.558 6150.204 6248.245 6990.840 6380.185
std 1.02E+02 3.50E+02 1.25E+02 1.95E+02 4.15E+02 2.55E+02
time 1.69 1.92 1.35 10.57 4.22 5.31

R5 BV 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00

best 0.523250 0.525730 0.526281 0.526281 0.530000 0.529296
MV 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
mean 0.529220 0.534866 0.530465 0.530287 0.560294 0.535737
std 2.58E−03 5.26E−03 3.14E−03 4.02E−03 2.93E−02 5.97E−03
time 3.46 3.33 1.70 7.61 6.30 5.66

S1 best 247.294 247.409 247.275 247.440 247.953 247.387

mean 247.315 247.632 247.281 248.465 248.713 247.748
std 2.45E−02 1.90E−01 6.89E−03 1.21E+00 6.64E−01 3.97E−01
time 82.82 87.88 79.83 108.38 98.74 103.59

S2 best 2075.366 2083.760 2062.459 2118.526 2109.862 2156.886

mean 2146.280 2138.049 2092.427 2269.436 2233.645 2244.520
std 38.273 53.986 33.868 100.388 106.119 64.468
time 240.72 243.62 218.70 246.71 257.88 263.29

S3 best 195.429 198.459 192.776 199.535 200.105 202.131

mean 196.857 201.455 199.507 202.976 204.241 207.010
std 1.574 2.208 3.257 3.918 3.321 5.503
time 584.91 603.38 563.97 635.08 630.97 722.98

S4 best 17485.043 17530.890 17495.760 17506.285 17509.516 17531.429

mean 17521.203 17625.222 17580.862 17657.104 17576.885 17610.460
std 25.413 68.463 86.463 212.133 57.747 56.191
time 2379.09 2238.69 2230.02 2473.72 2441.72 2413.71

3.2. Application in real-world problems and also in structural engi- constraints to be violated during the optimization process. In
neering problems these cases, the algorithm comparison is as follows:
(a) For each run, a constraint violation rate and the corresponding
In this section, the CIOA is applied to solve real-life optimiza- objective function value linked to this rate are calculated. Eq. (8)
presents the calculation of the violation rate according to the
tion problems. The problems addressed are presented in Table 3
inequality gi and equality hj constraints.
and are divided into two groups: Real-world optimization prob-
∑ng ∑nh
max (gi (x) , 0) + max ⏐hj (x)⏐ − 0.0001, 0
(⏐ ⏐ )
lems provided for the ’CEC2020 One Goal Restricted Optimization i=1 j=1
Competition in the Real World’ (Problems R1 to R5 ) and Structural Viol =
ng + nh
optimization problems for trusses subject to multiple constraints
(8)
(Problems S1 to S4 ).
Details on the implementation of real-world problems (R1
(b) The best result of the objective function is obtained in the run
to R5 ) can be seen in [63]. Some of these problems also have that generated the lowest violation rate (BV ). If all runs generate
equality constraints in addition to inequality constraints. In real- equal violation rates, the best result is obtained by the lowest
world problems, due to their high complexity, it is common for objective function value.
7
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Fig. 4. Problem S4 : Realistic Tower (dimensions in meters).

(c) The mean result for the objective function value is linked to rates, the best algorithm is the one with the lowest value for the
the mean of the violations of each execution (MV ), according objective function.
to Eq. (9), in which Violk is the violation of each execution and Three of the four truss optimization problems (S1 to S3 ) are
nruns is the number of executions. well-known in the literature and have already been studied
∑nruns by several authors [46–54]. The S4 optimization problem ad-
k=1 Violk
MV = (9) dresses the design of a real structure of an 82m-high transmission
nruns line tower which was damaged during a typhoon in Japan in
(d) One algorithm is better than another whenever its con- 1991 [64]. Thus, the proposed optimization problem consists of
straint violation rates are lower. If two algorithms have the same minimizing the mass of the structure considering the wind effects
8
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Table 5
Results of the Wilcoxon Signed-Rank Test for the benchmark functions.
F CIOA × WOA CIOA × SGA CIOA × FA CIOA × HS CIOA × PSO
f1 T+ 0 1275 1275 918 0
T− 1275 0 0 357 1275
winner WOA CIOA CIOA CIOA PSO

f2 T+ 763 756 1273 903 650

T− 512 519 1 372 625
winner CIOA CIOA CIOA CIOA CIOA

f3 T+ 0 1275 1275 1225.5 0

T− 1275 0 0 49.5 1275
winner WOA CIOA CIOA CIOA PSO

f4 T+ 1044 1182 1275 0 1275

T− 231 93 0 1275 0
winner CIOA CIOA CIOA HS CIOA

f5 T+ 1275 1275 1275 1275 1275

T− 0 0 0 0 0
winner CIOA CIOA CIOA CIOA CIOA

f6 T+ 779 1275 490 405 329

T− 496 0 785 820 946
winner CIOA CIOA FA CIOA PSO

f7 T+ 1184 1274 1275 1275 999

T− 91 1 0 0 276
winner CIOA CIOA CIOA CIOA CIOA

f8 T+ 147 1275 1275 1255 1085

T− 1128 0 0 20 190
winner WOA CIOA CIOA CIOA CIOA

f9 T+ 329 1275 1275 1024.5 534

T− 946 0 0 151.5 741
winner WOA CIOA CIOA CIOA PSO

f10 T+ 194 1275 1275 908 609

T− 1081 0 0 317 666
winner WOA CIOA CIOA CIOA PSO

as constraints. Details about the implementation can be seen by several simulations do not fully reveal which of the algorithms
in [53]. In Fig. 4, the realistic structure of the tower is shown. is the most effective. An advanced form of analysis can be done by
For problems R1 to R5 , 25 independent runs were performed: pairwise comparison using the Wilcoxon Signed-Rank Test [16].
In each round, 100,000 objective function evaluations were con- The detailed procedure for running the Wilcoxon Signed-Rank
sidered, formed through 400 iterations and 250 search agents. Test can be found in [65]. Thus, the Wilcoxon Signed-Rank Test
For problems S1 to S4 , 10 independent runs were considered: In was used to compare the CIOA with each of the other algorithms
problems S1 and S4 , 200,000 objective function evaluations were in benchmark functions analyzed in Section 3.1. The results are
considered in each algorithm. In problems S2 and S3 , with shape presented in Table 5, in which the performance of the CIOA is
optimization, 400,000 and 600,000 objective function evaluations superior to that of the algorithm with which it was compared
were considered, respectively. Results are presented in Table 4. whenever T + > T − .
Table 4 shows the excellent performance of the CIOA for real- As can be seen in Table 5, the CIOA performed better than the
life optimization problems. In Engineering problems R1 to R5 , other algorithms in 38 of the 50 comparisons, proving the CIOA’s
CIOA always presented the best optimal result and did not violate
competitiveness against rival algorithms.
the constraints in the best run (i.e., BV = 0). In addition, CIOA
Fig. 5 shows boxplots referring to the results of some problems
has the best mean in four of the five problems and the best
solved in this paper. Different types of problems were selected:
standard deviation in three. In the structural optimization prob-
Functions f1 and f2 (unimodal), functions f4 and f5 (multimodal),
lems of trusses, the CIOA presented the best mean and the best
functions f7 and f9 (multimodal with fixed dimensions), prob-
standard deviation in 2 problems (S3 and S4 ), in addition, CIOA
lems R3 and R5 (real-world problems for CEC 2020) and problem
presented the best optimal design in the problem S4 , proving to
be very competitive for solving problems with a high number of S3 (Shape and size optimization of a 52-bar space truss). The
constraints. presented results demonstrate the robustness of the CIOA in
the analyzed optimization problems, when compared with other
3.3. Statistical and convergence analysis algorithms.
Fig. 6 shows the CIOA convergence curves for different types
In this section, statistical and convergence analyses of the CIOA of optimization problems: unimodal functions (f1 and f3 ), mul-
algorithm are presented for the different types of optimization timodal functions (f4 and f5 ), multimodal functions with fixed
problems analyzed in this paper. dimensions (f6 and f8 ), real-world problems of CEC 2020 (R1
In many cases of comparison of algorithms, simple statistical and R4 ) and structural optimization problems of trusses (S1 and
tests such as mean and standard deviation of the results obtained S4 ). It is noticed that for the different types of problems the
9
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Fig. 5. Boxplots for some of the different optimization problems.

CIOA presents different convergence behaviors. In general, it is analyzed and in 6 of the 9 proposed application problems. Statis-
observed that the CIOA presents rapid convergence in all ana- tical analysis through Wilcoxon Signed-Rank Test and Boxplots
lyzed problems. The behavior in Problem R1 is highlighted, which demonstrated the superiority and robustness of the CIOA in most
presents equality constraints that are violated at some point in
of the optimization problems in which it was tested. In addition,
the optimization process, so in later iterations there may be an
the CIOA has the advantages of fast convergence and the reduced
increase in the objective function as long as there is a reduction
in the violation rates, calculated by the CIOA. number of parameters that must be defined by the user: only the
θ angle and the GlobIt parameter. The limitations or disadvantages
4. Impact and conclusions of the CIOA are the high number of objective function evaluations
and the number of search agents that must be used, a recurring
The new metaheuristic optimization algorithm presented in fact in several metaheuristic algorithms. However, the CIOA’s
this paper, called Circle-Inspired Optimization Algorithm (CIOA),
low computational operating time allows it to perform numerous
proved to be a powerful tool for solving complex optimization
operations quickly. Although the examples discussed in this paper
problems such as benchmark function optimization, real-world
optimization problems and structural truss optimization. In direct are mainly directed to engineering optimization problems, using
comparisons with other algorithms, the CIOA presented the best the CIOA to solve optimization problems in other areas of study
result for global optimization in 7 of the 10 benchmark functions can be promising.
10
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

Fig. 6. CIOA convergence curves.

Declaration of competing interest Appendix A. Supplementary data

The authors declare that they have no known competing finan- The supplementary related to this article consist of the files
cial interests or personal relationships that could have appeared ‘Main_Program_CIOA.m’ and ‘Objective_Function_CIOA.m’ which
to influence the work reported in this paper. contain the MATLAB code of the CIOA algorithm.
Supplementary material related to this article can be found
online at https://doi.org/10.1016/j.softx.2022.101192.
Data availability
References
Data will be made available on request.
[1] Yang XS. Nature-inspired metaheuristic algorithms. 2nd ed.. Frome:
Luniver Press; 2010.
Acknowledgments [2] S Kirkpatrick, Gelatt CD, Vecchi MP. Optimization by simulated annealing.
Science 1983;220:671–80. http://dx.doi.org/10.1126/science.220.4598.671.
[3] Kennedy J, Eberhart R. Particle swarm optimization. In: Proceedings of
The authors acknowledge the financial support of CAPES, Brazil the international conference on neural networks. 1995, p. 1942–8. http:
and CNPq, Brazil. //dx.doi.org/10.1109/ICNN.1995.488968.

11
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

[4] Storn R, Price K. Differential evolution – A simple and efficient heuristic for [29] Dehghani M, Montarezi Z, Givi H, Guerrero JM, Dhiman G. Darts game
global optimization over continuous spaces. J Global Optim 1997;11:341–9. optimizer: A new optimization technique based on darts game. Int J Intell
http://dx.doi.org/10.1023/A:1008202821328. Eng Syst 2020;13:286–94. http://dx.doi.org/10.22266/ijies2020.1031.26.
[5] Geem ZW, Kim JH, Loganathan GV. A new heuristic optimization al- [30] Dehghani M, Montazeri Z, Dhiman G, Malik OP, Morales-Menendez R,
gorithm: Harmony search. Simulation 2001;76:60–8. http://dx.doi.org/10. Ramirez-Mendoza RA, et al. A spring search algorithm applied to engi-
1177/003754970107600201. neering optimization problems. Appl Sci 2020;10:6173. http://dx.doi.org/
[6] Karaboga D, Basturk B. A powerful and efficient algorithm for numerical 10.3390/app10186173.
function optimization: artificial bee colony (ABC) algorithm. J Global Optim [31] Dhiman G, Garg M, Nagar A, Kumar V, Dehghani M. A novel algorithm
2007;39:459–71. http://dx.doi.org/10.1007/s10898-007-9149-x. for global optimization: Rat swarm optimizer. J Ambient Intell Humaniz
[7] Karaboga D, Basturk B. Artificial bee colony (ABC) optimization algorithm Comput 2021;12:8457–82. http://dx.doi.org/10.1007/s12652-020-02580-0.
for solving constrained optimization problems. In: Proceedings of the [32] Dehghani M, Montazeri Z, Malik OP, Dhiman G, Kumar V. BOSA: Binary ori-
International Fuzzy Systems Association World Congress. 2007, p. 789–98. entation search algorithm. Int J Innov Technol Explor Eng 2019;9:5306–10.
http://dx.doi.org/10.1007/978-3-54072950-1_77. http://dx.doi.org/10.35940/ijitee.A4215.119119.
[8] Socha K, Dorigo M. Ant colony optimization for continuous domains. [33] Dhiman G, Oliva D, Kaur A, Singh KK, Vimal S, Sharma A, et al. BEPO:
European J Oper Res 2008;185:1155–73. http://dx.doi.org/10.1016/j.ejor. A novel binary emperor penguin optimizer for automatic feature se-
2006.06.046. lection. Knowl-Based Syst 2021;211:106560. http://dx.doi.org/10.1016/j.
[9] Yang XS. Firefly algorithms for multimodal optimization. In: Proceedings knosys.2020.106560.
of the International Symposium on Stochastic Algorithms. 2009, p. 169–78. [34] Dhiman G. ESA: a hybrid bio-inspired metaheuristic optimization approach
http://dx.doi.org/10.1007/978-3-642-04944-6_14. for engineering problems. Eng Comput 2021;37:323–53. http://dx.doi.org/
[10] Yang XS. Firefly algorithm, stochastic test functions and design optimiza- 10.1007/s00366-019-00826-w.
tion. Int J Bio-Inspired Comput 2010;2:78–84. http://dx.doi.org/10.1504/ [35] Yildiz BS, Pholdee N, Bureerat S, Yildiz AR, Sait SM. Robust design of a robot
IJBIC.2010.032124. gripper mechanism using new hybrid grasshopper optimization algorithm.
[11] Rashedi E, Nezamabadi-Pour H, Saryazdi S. GSA: A gravitational search Expert Syst 2021;38. http://dx.doi.org/10.1111/exsy.12666.
algorithm. Inform Sci 2009;179:2232–48. http://dx.doi.org/10.1016/j.ins. [36] Yildiz AR, Erdas MU. A new hybrid Taguchi-salp swarm optimization
2009.03.004. algorithm for the robust design of real-world engineering problems. Mater.
[12] Yang XS, Deb S. Cuckoo search via Lévy flights. In: Proceedings of the Testings 2021;63:157–62. http://dx.doi.org/10.1515/mt-2020-0022.
World Congress on Nature and Biologically Inspired Computing. 2009, p. [37] Yildiz BS, Kumar S, Pholdee N, Bureerat S, Sait SM, Yildiz AR. A new chaotic
210–4. http://dx.doi.org/10.1109/NABIC.2009.5393690. Lévy flight distribution optimization algorithm for solving constrained
[13] Yang XS. A new metaheuristic bat-inspired algorithm. In: Proceedings of engineering problems. Expert Syst 2022;e12992. http://dx.doi.org/10.1111/
the Workshop on Nature Inspired Cooperative Strategies for Optimization. exsy.12992.
2010, p. 65–74. http://dx.doi.org/10.1007/978-3-642-12538-6_6. [38] Yildiz BS, Pholdee N, Bureerat S, Yildiz AR, Sait SM. Enhanced grasshopper
[14] Yang XS. Flower pollination algorithm for global optimization. In: Proceed- optimization algorithm using elite opposition-based learning for solving
ings of the Unconventional Computation and Natural Computation. 2012, real-world engineering problems. Eng Comput 2021. http://dx.doi.org/10.
p. 240–9. 1007/s00366-021-01368-w.
[15] Gandomi AH, Alavi AH. Krill herd: A new bio-inspired optimization [39] Gupta VK, Shukla SK, Anupriya, Rawat RS. Crime tracking system and
algorithm. Commun Nonlinear Sci Numer Simul 2012;17:4831–45. http: People’s safety in India using machine learning approaches. Int J Modern
//dx.doi.org/10.1016/j.cnsns.2012.05.010. Res 2022;2:1–7.
[16] Civicioglu P. Backtracking search optimization algorithm for numerical [40] Vaishnav PK, Sharma S, Sharma P. Analytical review analysis for screening
optimization problems. Appl Math Comput 2013;219:8121–44. http://dx. COVID-19 disease. Int J Modern Res 2021;1:22–9.
doi.org/10.1016/j.amc.2013.02.017. [41] Sharma T, Nair R, Gomathi S. Breast cancer image classification using
[17] Mirjalili S, Mirjalili SB, Lewis A. Grey wolf optimizer. Adv Eng Softw transfer learning and convolutional neural network. Int J Modern Res
2014;69:46–61. http://dx.doi.org/10.1016/j.advengsoft.2013.12.007. 2022;2:8–16.
[18] Gonçalves MS, Lopez RH, Fadel Miguel LF. Search group algorithm: A [42] Chatterjee I. Patenting machine-learning: Review and discussions. Int J
new metaheuristic method for the optimization of truss structures. Com- Modern Res 2021;1:15–21.
put Struct 2015;153:165–84. http://dx.doi.org/10.1016/j.compstruc.2015. [43] Kumar R, Dhiman G. A comparative study of fuzzy optimization through
03.003. fuzzy number. Int J Modern Res 2021;2:1–14.
[19] Mirjalili S, Lewis A. The whale optimization algorithm. Adv Eng Softw [44] Shukla SK, Gupta VK, Joshi K, Gupta A, Singh MK. Self-aware execution
2016;95:51–67. http://dx.doi.org/10.1016/j.advengsoft.2016.01.008. environment model (SAE2) for the performance improvement of multicore
[20] Dhiman G, Kumar V. Spotted hyena optimizer: A novel bio-inspired systems. Int J Modern Res 2022;2:17–27.
based metaheuristic technique for engineering applications. Adv Eng Softw [45] Ghasemi MR, Hinton E, Wood RD. Optimization of trusses using ge-
2017;114:47–70. http://dx.doi.org/10.1016/j.advengsoft.2017.05.014. netic algorithms for discrete and continuous variables. Eng Comput
[21] Dhiman G, Kumar V. Emperor penguin optimizer: A bio-inspired algorithm 1999;16:272–301. http://dx.doi.org/10.1108/02644409910266403.
for engineering problems. Knowl-Based Syst 2018;159:20–50. http://dx.doi. [46] Lee KS, Geem ZW. A new structural optimization method based on the
org/10.1016/j.knosys.2018.06.001. Harmony Search Algorithm. Comput Struct 2004;82:781–98. http://dx.doi.
[22] Arora S, Singh S. Butterfly optimization algorithm: a novel approach for org/10.1016/j.compstruc.2004.01.002.
global optimization. Soft Comput 2019;23:715–34. http://dx.doi.org/10. [47] Lee KS, Geem ZW. A new meta-heuristic algorithm for continuous en-
1007/s00500-018-3102-4. gineering optimization: harmony search theory and practice. Comput
[23] Dhiman G, Kaur A. STOA: A bio-inspired based optimization algorithm Methods Appl Mech Engrg 2005;194:3902–33. http://dx.doi.org/10.1016/
for industrial engineering problems. Eng Appl Artif Intell 2019;82:148–74. j.cma.2004.09.007.
http://dx.doi.org/10.1016/j.engappai.2019.03.021. [48] Lingyun W, Mei Z, Guangming W, Guang M. Truss optimization on shape
[24] Dhiman G, Kumar V. Seagull optimization algorithm: Theory and its and sizing with frequency constraints based on genetic algorithm. Comput
applications for large-scale industrial engineering problems. Knowl-Based Mech 2005;35:361–8. http://dx.doi.org/10.1007/s00466-004-0623-8.
Syst 2019;165:169–96. http://dx.doi.org/10.1016/j.knosys.2018.11.024. [49] Lamberti L. An efficient simulated annealing algorithm for design opti-
[25] Sallam KM, Elsayed SM, Chakrabortty RK, Ryan MJ. Multi-operator mization of truss structures. Comput Struct 2008;86:1936–53. http://dx.
differential evolution algorithm for solving real-world constrained opti- doi.org/10.1016/j.compstruc.2008.02.004.
mization problems. In: Proceedings of the IEEE congress on evolution- [50] Gomes HM. Truss optimization with dynamic constraints using a particle
ary computation. 2020, p. 1–8. http://dx.doi.org/10.1109/CEC48606.2020. swarm algorithm. Expert Syst Appl 2011;38:957–68. http://dx.doi.org/10.
9185722. 1016/j.eswa.2010.07.086.
[26] Kumar A, Das S, Zelinda I. A self-adaptive spherical algorithm for real- [51] Gomes HM. A firefly metaheuristic structural size and shape optimisation
world constrained optimization problems. In: Proceedings of the genetic with natural frequency constraints. Int J Metaheuristics 2012;2:38–55.
and evolutionary computation conference companion. 2020, p. 13–4. http: http://dx.doi.org/10.1504/IJMHEUR.2012.048215.
//dx.doi.org/10.1145/3377929.3398186. [52] Miguel LFF, Fadel Miguel LF. Shape and size optimization of truss
[27] Kaur S, Awashti LK, Sangal AL, Dhiman G. Tunicate swarm algorithm: A structures considering dynamic constraints through modern metaheuristic
new bio-inspired based metaheuristic paradigm for global optimization. algorithms. Expert Syst Appl 2012;39:9458–67. http://dx.doi.org/10.1016/
Eng Appl Artif Intell 2020;90:103541. http://dx.doi.org/10.1016/j.engappai. j.eswa.2012.02.113.
2020.103541. [53] Miguel LFF, Fadel Miguel LF. Assessment of modern metaheuristic algo-
[28] Dehghani M, Montazeri Z, Dehghani A, Ramirez-Mendoza RA, Samet H, rithms – HS ABC and FA – in shape and size optimization of structures
Guerrero JM, et al. MLO: Multi leader optimizer. Int J Intell Eng Syst with different types of constraints. Int J Metaheuristics 2013;2:256–93.
2020;13:364–73. http://dx.doi.org/10.22266/ijies2020.1231.32. http://dx.doi.org/10.1504/IJMHEUR.2013.056404.

12
Otávio Augusto Peter de Souza and Letícia Fleck Fadel Miguel SoftwareX 19 (2022) 101192

[54] Bekdas G, Nigdeli SM, Yang XS. Sizing optimization of truss structures [60] Yildiz BS, Patel V, Pholdee N, Sait MS, Bureerat S, Yildiz AR. Conceptual
using flower pollination algorithm. Appl Soft Comput 2015;37:322–31. comparison of the ecogeography-based algorithm equilibrium algorithm,
http://dx.doi.org/10.1016/j.asoc.2015.08.037. marine predators algorithm and slime mold algorithm for optimal product
[55] Kaveh A, Zolghadr A. Meta-heuristic methods for optimization of design. Mater Testing 2021;63:336–40. http://dx.doi.org/10.1515/mt-2020-
truss structures with vibration frequency constraints. Acta Mech 0049.
2018;229:3971–92. http://dx.doi.org/10.1007/s00707-018-2234-z. [61] Gupta S, Abderazek H, Yildiz BS, Yildiz AR, Mirjalili S, Sait SM. Comparison
[56] Souza OAP, Miguel LFF. Comparison of the performance of different of metaheuristic optimization algorithms for solving constrained mechan-
metaheuristic optimization algorithms. In: Proceedings of the XLI Ibero- ical design optimization problems. Expert Syst Appl 2021;183:115351.
Latin American congress on computational methods in engineering. 2020, http://dx.doi.org/10.1016/j.eswa.2021.115351.
ISSN: 2675-6269. [62] Ho YC, Pepyne DL. Simple explanation of the no-free-lunch theorem and
[57] Yildiz ABS, Pholdee N, Bureerat S, Yildiz AR, Sait SM. Sine-cosine optimiza- its implications. J Optim Theory Appl 2002;115:549–70. http://dx.doi.org/
tion algorithm for the conceptual design of automobile components. Mater 10.1023/A:1021251113462.
Testing 2020;62:744–8. http://dx.doi.org/10.3139/120.111541. [63] Kumar A, Wo G, Ali MZ, Mallipeddi R. A test-suite of non-convex con-
[58] Panagant N, Pholdee N, Bureerat S, Kaen K, Yildiz AR, Sait SM. Seag- strained optimization problems from the real-world and some baseline
ull optimization algorithm for solving real-world design optimization results. Swarm Evol Comput 2020;56:1–27. http://dx.doi.org/10.1016/j.
problems. Mater Testing 2020;62:640–4. http://dx.doi.org/10.3139/120. swevo.2020.100693.
111529. [64] Murotsu Y, Okada H, Shao S. Reliability-based design of transmission line
[59] Yildiz BS, Pholdee N, Bureerat S, Erdas MU, Yildiz AR, Sait SM. Compari- structures under extreme wind loads. Struct Safety Reliab 1994;3:1675–81.
sion of the political optimization algorithm, the archimedes optimization [65] Derrac J, García S, Molina D, Herrera F. A practical tutorial on the use of
algorithm and the levy flight algorithm for design optimization in nonparametric statistical tests as a methodology for comparing evolution-
industry. Mater Testing 2021;63:356–60. http://dx.doi.org/10.1515/mt- ary and swarm intelligence algorithms. Swarm Evol Comput 2011;1:3–18.
2020-0053. http://dx.doi.org/10.1016/j.swevo.2011.02.002.

13