Arabian Journal for Science and Engineering

https://doi.org/10.1007/s13369-019-04138-5

RESEARCH ARTICLE - CIVIL ENGINEERING

Size and Shape Optimization of Truss Structures with Natural

Frequency Constraints Using Modified Simulated Annealing Algorithm
Carlos Millan‑Paramo1,2 · João Elias Abdalla Filho1

Received: 18 May 2019 / Accepted: 11 September 2019

© King Fahd University of Petroleum & Minerals 2019

Abstract
In this article, a recently developed metaheuristic named modified simulated annealing algorithm (MSAA) is proposed for
size and shape optimization of truss structures with frequency constraints. These kinds of problems represent nonlinear
and non-convex search spaces with several local optima and are well-known as challenging optimization problem. MSAA
is a newly improved version of the simulated annealing (SA) algorithm with three modifications: preliminary exploration,
search step and a new probability of acceptance. Six benchmark truss optimization problems with frequency constraints are
explored for the validity of the present algorithm. Numerical results indicate that MSAA is more reliable, stable and efficient
than those found by other state-of-the-art metaheuristics optimization methods.

Keywords Truss structures · Size and shape optimization · Frequency constraints · Modified simulated annealing algorithm
(MSAA)

1 Introduction time consuming to be applied to these optimization prob-

lems. Also, a good starting point is vital for these methods
The natural frequencies and mode shapes of a structure are to obtain successful outcome. Under such circumstances,
important dynamical parameters that must be controlled the metaheuristic algorithms can serve as appropriate alter-
to maintain the desired structural behavior [1, 2]. These natives due to the ability to search global minima in high
parameters mainly depend on the configurations of mass modal and multidimensional spaces.
(or weight) and stiffness of the structure. In this way, the Size optimization is a fundamental type of truss optimi-
structural optimization problem with frequency constraints zation where the final goal is to obtain the best bar sections,
arises as an engineering problem. Optimizing the weight of whereas shape optimization works to search the best nodal
structures with frequency constraints can be considered as positions of predefined nodes of the truss structure. The first
a difficult problem to solve because the reduction in weight to address this problem was Bellagamba [4] and since then
generates conflict with the frequency constraints [3]. In addi- several researchers have been introducing different optimiza-
tion, the frequency restrictions are highly nonlinear, non- tion algorithms; however, this research area has not been fully
convex and implied with respect to the variables of design investigated so far. Table 1 lists the most important works
[1]. Therefore, deterministic approaches are difficult and that involve optimization methods (mathematical program-
ming and metaheuristic algorithms) to solve this problem.
Although several metaheuristics have been introduced
* Carlos Millan-Paramo in the last decade to solve this problem, most of them are
[email protected]
population-based and undergo many steps along with several
João Elias Abdalla Filho parameters that make them difficult to understand and code.
[email protected]
Also, there are some procedures in recent metaheuristics
1
Postgraduate Program in Civil Engineering (PPGEC), which make them similar. Because of this, the research-
Universidade Tecnológica Federal do Paraná, St. Dep. Heitor ers usually are confused to select a metaheuristic and still
Alencar Furtado, 5000, Curitiba, Paraná CEP: 81280-340, use the old algorithms instead of the recent ones. On the
Brazil
other hand, according to the No Free Lunch theorem [27] in
2
Department of Civil Engineering, Universidad de Sucre, the field of optimization, there is no algorithm to solve all
Sincelejo, Colombia

13
Vol.:(0123456789)
 Arabian Journal for Science and Engineering

Table 1 The list of main works in size and shape optimization for truss structures with natural frequency
Author Method

Lin et al. [5] Bi-factor algorithm based on the Kuhn–Tucker criteria

Grandhi and Venkayya [6] Optimality criterion (OC)
Wang et al. [7]
Sedaghati et al. [8, 9] Sequential quadratic programming (SQP)
Wei et al. [10] Niche hybrid parallel genetic algorithm (NHPGA)
Wei et al. [11] Parallel genetic algorithm
Gomes [12] Particle swarm optimization (PSO)
Miguel and Fadel Miguel [13] Harmony search (HS)
Firefly algorithm (FA)
Kaveh and Zolghadr [14] Hybridization of the charged system search and the Big Bang–Big Crunch algorithms (CSS-BBC)
Kaveh and Zolghadr [15] Charged system search (CSS)
Kaveh and Zolghadr [16] Democratic particle swarm optimization (DPSO)
Khatibinia and Naseralevi [17] Orthogonal multi-gravitational search algorithm (OMGSA)
Kaveh and Ilchi Ghazaan [18] Hybridization of the particle swarm optimization with an aging leader and challengers (ALC-PSO
and HALC-PSO)
Farshchin et al. [19] School-based optimization (SBO)
Kaveh and Zolghadr [20] Cyclical parthenogenesis algorithm (CPA)
Kaveh and Ilchi Ghazaan [21] Vibrating particles system (VPS)
Kaveh and Zolghadr [22] Tug of war optimization (TWO)
Ho-Huu et al. [23] Differential evolution (ReDe)
Tejani et al. [24, 25] Symbiotic organisms search (SOS)
Lieu et al. [26] Adaptive hybrid evolutionary firefly algorithm (AHEFA)

optimization problems. This indicates that a new adapted 2 Truss Optimization Problem
algorithm has potential to solve a group of problems (e.g., with Frequency Constraints
structures design) better than the current algorithms. Con-
trary to previous works, this paper aims to use a simple sin- The goal of the structural optimization problem is to mini-
gle-solution algorithm named modified simulated annealing mize the weight of the truss by achieving the optimum nodal
algorithm (MSAA). Millán Páramo et al. [28] recently intro- coordinates and optimum element cross-sectional areas
duced for solving global optimization problems and applied while satisfying some constraints on the natural frequen-
in structural optimization problems with success [29–31]. cies. All six example problems solved in this work have been
MSAA is a newly improved version of the simulated anneal- solved previously by other authors and are thus considered
ing (SA) algorithm with three modifications. Firstly, a pre- as benchmark problems. In all of those problems, lumped
liminary exploration is realized to choose the starting point masses are added as external masses that are not an intrinsic
of search. Secondly, the transition from the starting point to part of the structure to be optimized (for instance, a trans-
the new point is done by a search step. Thirdly, the range of mission tower would have the effects of cables and fixtures
probability of accepting a worse solution is reduced. These represented as lumped masses). Therefore, such masses are
modifications allow a good balance between exploration and not an integral part of the weight of the structure and are not
exploitation throughout the optimization process [28]. The included in the formulation. The mathematical formulation
validity of MSAA is confirmed by testing for size and shape for this problem can be expressed as follows:
optimization problems of truss structures with frequency � �
constraints. Optimal results attained by MSAA are compared Find, X = {A, NC}, where A = A1 , A2 , … , An
� �
with other metaheuristics in the literature. and NC = NC1 , NC2 , … , NCm
The remainder of this article is structured as follows. Sec- ∑
n
tion 2 describes the mathematical formulation of truss opti- Minimize W(X) = 𝜌i Ai Li
i=1
mization with frequency constraints. The MSAA is briefly ⎧ f − f min ≥ 0 (1)
presented in Sect. 3. Section 4 presents six most widely ⎪ q q
investigated benchmark numerical examples to illustrate the ⎪ f − f max ≤ 0
Subject to ⎨ r minr
efficiency of the MSAA. Finally, in Sect. 5, our conclusions ⎪ Ai ≤ Ai ≤ Aimax
are presented. ⎪ NCmin ≤ NCj ≤ NCmax
⎩ j j

13
Arabian Journal for Science and Engineering

Fig. 1 Flowchart of MSAA Begin Preliminary Exploration Eq. (2)

The MSAA parameters as set

Generate initial solution x chosen by Eq. (2)

Generate new solution y by Eq. (3)

Yes y is better than x? No

1
P = 1+e ∆f ⁄T Eq. (4)
Search Step Eq. (3)
Generate r in [0, 1) randomly
X2

Yes r<P

x=y No

Stop condition of inner loop is met? No

Ri+1
Ri
Yes
Decrease the temperature T X1

No
Stop condition of inner loop is met?

Yes
Output the solution x

End

Table 2 Design conditions of the truss problems

Size optimization Size and shape optimization
10-bar planar truss 200-bar planar 72-bar space truss 120-bar dome truss 37-bar planar truss 52-bar dome truss
truss

Young’s modulus 6.98 × 1010 2.1 × 1011 6.98 × 1010 2.1 × 1011 2.1 × 1011 2.1 × 1011
E (N/m2)
Material density ρ 2770 7860 2770 7971.81 7800 7800
(kg/m3)
Size variables 0.645 ≤ A ≤ 50 0.1 ≤ A≤30 0.645 ≤ A≤30 1 ≤ A≤129.3 1 ≤ A ≤10 1 ≤ A ≤10
(cm2)
Shape variables – – – – 0.1 ≤ y ≤3 All the free
(m) nodes can
displace ± 2 m
in symmetric
manner
Frequency con- f1 ≥ 7 f1 ≥ 5 f1 = 4 f1 ≥ 9 f1 ≥ 20 f1 ≤ 15.9155
straints (Hz) f2 ≥ 15 f2 ≥ 10 f3 ≥ 6 f2 ≥ 11 f2 ≥ 40 f2 ≥ 28.6479
f3 ≥ 20 f3 ≥ 15 f3 ≥ 60

13
 Arabian Journal for Science and Engineering

frequencies of the structure, respectively; the superscripts,

“max” and “min” denote the maximum and minimum allow-
able limits, respectively.

3 Modified Simulated Annealing Algorithm

The MSAA is a single-solution algorithm based on the

behavior of atomic arrangements in liquid or solid materials
during the annealing process. MSAA has three main char-
acteristics that differentiate it from the simulated annealing
proposed by Kirkpatrick [32]:

Fig. 2 A 10-bar planar truss [26]

1. The starting point is not randomly generated but is
selected by a preliminary exploration where the algo-
where 𝐖 is the weight of the structure; 𝐧 is the total number rithm performs a scan in the search space and is given
of members of the structure; 𝛒𝐢, 𝐀𝐢 and 𝐋𝐢 stand for the mate- by the following equation:
rial density, the cross-sectional area and the length of the ith ( )
XP×N = IP×N XL + randP×N XU − XL (2)
member, respectively; 𝐍𝐂𝐣 is a nodal coordinates (xj, yj, zj)
of node jth of the truss; 𝐟𝐪 and 𝐟𝐫 are the qth and rth natural

Table 3 Comparison of optimal Variables (cm2) DPSO SBO VPS ReDe ISOS AHEFA MSAA
designs of the 10-bar planar
truss obtained by different [16] [19] [21] [23] [25] [26]
algorithms
1 A1 35.944 35.5994 35.1471 35.1565 35.2654 35.1714 32.9710
2 A2 15.53 14.9956 14.6668 14.7605 14.6803 14.7203 15.5925
3 A3 35.285 35.4806 35.6889 35.1187 34.4273 35.1074 32.8514
4 A4 15.385 14.7646 15.0929 14.7275 14.9605 14.6986 15.5942
5 A5 0.648 0.6450 0.645 0.6450 0.6450 0.6451 0.6454
6 A6 4.583 4.6305 4.6221 4.5558 4.5927 4.5593 4.6552
7 A7 23.61 24.3272 23.5552 23.7199 23.3417 23.7330 26.1179
8 A8 23.599 23.8528 24.468 23.6304 23.8236 23.6795 26.1350
9 A9 13.135 12.6797 12.7198 12.3827 12.8497 12.3987 11.9983
10 A10 12.357 12.6375 12.6845 12.4580 12.5321 12.4231 11.9339
Best weight (kg) 532.39 532.05 530.77 524.45 524.73 524.45 532.04
Average weight (kg) 537.80 533.45 535.64 524.76 530.03 525.16 532.06
SD (kg) 4.02 2.20 2.55 1.11 3.48 1.92 0.01
NI 6000 10,000 30,000 8300 4000 5860 7130

Table 4 Natural frequencies Frequency DPSO SBO VPS ReDe ISOS AHEFA MSAA
of the optimum designs for the number (Hz)
10-bar planar truss [16] [19] [21] [23] [25] [26]

1 7.0000 7.0000 7.0000 7.0000 7.0001 7.0000 7.0000

2 16.1870 16.1660 16.1599 16.1924 16.1703 16.1920 15.8458
3 20.0000 20.0000 20.0000 20.0000 20.0024 20.0000 20.0000
4 20.0210 20.0004 20.0001 20.0002 – 20.0000 20.0001
5 28.4700 28.5834 28.6008 28.5517 – 28.5551 27.9633
6 29.2430 29.0034 29.0628 – – 28.9588 28.0334
7 48.7690 48.5051 48.4904 – – 48.5777 46.8421
8 51.3890 51.0259 51.0476 – – 51.0712 48.8795

13
Arabian Journal for Science and Engineering

where P is the number of points that are desired in the

search space; N is the number of dimensions of the prob-
lem; 𝐈𝐏𝐱𝐍 is the identity matrix of size P × N; 𝐗𝐋 is the
lower limit of the problem; 𝐗𝐔 is the upper limit of the
problem and 𝐫𝐚𝐧𝐝𝐏𝐱𝐍 is the matrix of random numbers
(pure randomness) between 0 and 1 of size P × N. To
start the optimization process with MSAA, all points
generated with (2) are evaluated in the objective func-
tion of the problem and the smallest value (in the case of
searching the minimum value of the function) is chosen
as the starting point of the search.
2. The new solution for comparison is not randomly gener-
Fig. 3 Convergence history of the 10-bar planar truss ated. From the starting point determined in the prelimi-
nary exploration step, a search step is generated in order
to determine the neighboring state. This step depends
on a radius (R) of action that gradually decreases as the
temperature of the system decreases. The transition from
starting point to the new point (search step) is performed
by the addition of random numbers that are between
[− R, R]. This enables the algorithm to execute a global
exploration at high temperatures and a local exploration
at low temperatures, providing a balance between the
exploration and exploitation of the algorithm. The radius
is updated as follows:
Ri+1 = Ri ⋅ 𝛼 (3)
where 𝐑𝐢 is the initial radius cycle and 𝛂 is the radius
reduction coefficient.
3. If the cost function of the new solution is higher than the
best value, the acceptance of the new solution depends
on the following equation:
1
P= (4)
1 + e(Δf ∕T)
where P is the probability of accepting the new solu-
tion; Δ𝐟 is the difference between the quality of the new
solution and the quality of the current solution; T is the
temperature of the system; and e is the Euler number.
This probability is in a range between 0 and 0.5, allow-
ing the algorithm to have a lower range of acceptance
of worse solutions.

The three modifications improve the initial exploration,

allow a balance between initial and final exploration and
control the convergence in the final search stage. For more
details, see [28]. The flowchart of the MSAA is illustrated
in Fig. 1.

Fig. 4 A 200-bar planar truss [26]

13
 Arabian Journal for Science and Engineering

Table 5 Comparison of optimal Variables (cm2) CSS-BBBC HALC-PSO SBO SOS ISOS AHEFA MSAA
designs of the 200-bar planar
truss obtained by different [14] [18] [19] [24] [25] [26]
algorithms
1 A1 0.2934 0.3072 0.3040 0.4781 0.3072 0.2993 0.3034
2 A2 0.5561 0.4545 0.4478 0.4481 0.5075 0.4508 0.5177
3 A3 0.2952 0.1000 0.1000 0.1049 0.1001 0.1001 0.1000
4 A4 0.1970 0.1000 0.1000 0.1045 0.1000 0.1000 0.1000
5 A5 0.8340 0.5080 0.5075 0.4875 0.5893 0.5123 0.5699
6 A6 0.6455 0.8276 0.8219 0.9353 0.8328 0.8205 0.8187
7 A7 0.1770 0.1023 0.1003 0.1200 0.1431 0.1011 0.1000
8 A8 1.4796 1.4357 1.4240 1.3236 1.3600 1.4156 1.4361
9 A9 0.4497 0.1007 0.1001 0.1015 0.1039 0.1000 0.1000
10 A10 1.4556 1.5528 1.5929 1.4827 1.5114 1.5742 1.4599
11 A11 1.2238 1.1529 1.1597 1.1384 1.3568 1.1597 1.1381
12 A12 0.2739 0.1522 0.1275 0.1020 0.1024 0.1338 0.1205
13 A13 1.9174 2.9564 2.9765 2.9943 2.9024 2.9672 2.9032
14 A14 0.1170 0.1003 0.1001 0.1562 0.1000 0.1000 0.1006
15 A15 3.5535 3.2242 3.2456 3.4330 3.4120 3.2722 3.7168
16 A16 1.3360 1.5839 1.5818 1.6816 1.4819 1.5762 1.5246
17 A17 0.6289 0.2818 0.2566 0.1026 0.2587 0.2562 0.2056
18 A18 4.8335 5.0696 5.1118 5.0739 4.8291 5.0956 5.1494
19 A19 0.6062 0.1033 0.1001 0.1068 0.1499 0.1001 0.1021
20 A20 5.4393 5.4657 5.4337 6.0176 5.5090 5.4546 5.3291
21 A21 1.8435 2.0975 2.1016 2.0340 2.2221 2.0933 1.9882
22 A22 0.8955 0.6598 0.6794 0.6595 0.6113 0.6737 0.6782
23 A23 8.1759 7.6585 7.6581 6.9003 7.3398 7.6498 7.9359
24 A24 0.3209 0.1444 0.1006 0.2020 0.1559 0.1178 0.3222
25 A25 10.9800 8.0520 7.9468 6.8356 8.6301 8.0682 8.9235
26 A26 2.9489 2.7889 2.7835 2.6644 2.8245 2.8025 2.5618
27 A27 10.5243 10.4770 10.5277 12.1430 10.8563 10.5040 10.4026
28 A28 20.4271 21.3257 21.3027 22.2484 20.9142 21.2935 21.3538
29 A29 19.0983 10.5111 10.6207 8.9378 10.5305 10.7410 10.6476
Best weight (kg) 2298.61 2156.73 2156.51 2180.32 2169.46 2160.74 2157.28
Average weight (kg) – 2157.14 2156.79 2303.30 2244.64 2161.04 2161.74
SD (kg) – 0.24 0.21 83.59 43.48 0.18 2.96
NI – 13,000 23,000 10,000 10,000 11,300 6200

Table 6 Natural frequencies Frequency CSS-BBBC HALC-PSO SBO SOS ISOS AHEFA MSAA
of the optimum designs for the number (Hz)
200-bar planar truss [14] [18] [19] [24] [25] [26]

1 5.010 5.000 5.000 5.0001 5.0000 5.0000 5.0000

2 12.911 12.254 12.2141 13.4306 12.4477 12.1821 12.3405
3 15.416 15.044 15.0192 15.2645 15.2332 15.0160 15.0001
4 17.033 16.718 16.6870 17.0225 – 16.6837 16.7190
5 21.426 21.461 21.4109 21.8468 – 21.3547 21.1624
6 21.613 21.524 21.4570 – – 21.4168 21.3616

4 Truss problems and Discussions (10-bar planar truss, 200-bar planar truss, 72-bar space
truss and 120-bar dome truss) and the others concern-
In this section, six most commonly used numerical exam- ing size and shape optimization (37-bar planar truss and
ples including four ones regarding size optimization 52-bar dome truss) are explored to evaluate feasibility and

13
Arabian Journal for Science and Engineering

temperature (Tinitial) and final temperature (Tfinal) are set

as 200, 1 and 1 × 10−3, respectively. Sensitivity analyses of
the MSAA on these parameters are investigated in [28, 30,
31]. According to [31], the maximum number of perturba-
tions (npmax) at the same temperature can be chosen in the
range of 100–300. In this study, MSAA is used by consid-
ering npmax as 230 for 10-bar planar truss, 72-bar space
truss and 52-bar dome truss; 200 for 200-bar planar truss;
and 100 for 120-bar dome truss and 37-bar planar truss.
These numbers of perturbations have been obtained in this
work after several attempts to find equilibrium between
accuracy and computational cost for each of the problems.
Fig. 5 Convergence history of the 200-bar planar truss If 300 perturbations were used in every problem, good
results would be guaranteed, however, always at a high
computational cost. The iterative process is terminated
when the algorithm reaches the final temperature. It is also
important to note that all presented MSAA designs are fea-
sible. The algorithm is coded in Matlab, and the two-node
linear bar element is utilized for FE analysis. Statistical
results are presented in terms of the best weight, average
weight, standard deviation (SD) and the corresponding
iterations number (NI). The results and discussions of the
test problems are explained in the following sections:

4.1 10-bar planar truss

The first problem is a planar truss consisting of 10 bars

(Fig. 2) considered to be a benchmark problem in the field
of structural design with multiple frequency constraints. A
lumped mass of 454 kg is added to each of the free nodes
(nodes 1–4). Table 3 provides a comparison between opti-
mal results obtained by the present algorithms with dif-
ferent approaches. It can be seen that the optimal weight
achieved by the MSAA (532.04 kg) is lighter than DPSO
(532.39 kg) and SBO (532.05 kg). Additionally, the MSAA
requires 7130 NI which is less than the SBO (10,000 NI),
VPS (30,000 NI) and ReDe (8300 NI). Although the con-
vergence speed of DPSO, ISOS and AHEFA are faster than
that of the MSAA (6000 NI for DPSO, 4000 NI for ISOS
and 5860 for AHEFA), the MSAA is more stable than these
methods with the smallest SD (0.01 kg for MSAA, 4.02 kg
for DPSO, 3.48 kg for ISOS and 1.92 for AHEFA). It is
important to mention that with respect to the SD, MSAA
ranks first among the considered metaheuristics. Table 4 rep-
resents the first eight natural frequencies optimally obtained
Fig. 6 A 72-bar space truss [26] by the present work, and it is clear that none of the frequency
constraints are violated. Figure 3 shows the convergence his-
tory of the best design for this problem.
validity of MSAA. The main input data of the problems
are given in Table 2. Hundred independent optimization 4.2 200-bar planar truss
runs are executed for each test problem due to the sto-
chastic nature of metaheuristic algorithms. For all exam- The 200-bar planar truss is shown in Fig. 4. A lumped mass
ples, the population size (preliminary exploration), initial is added to the upper nodes of the structure. The bars are

13
 Arabian Journal for Science and Engineering

Table 7 Comparison of optimal designs of the 72-bar space truss obtained by different algorithms
Variables (cm2) CSS-BBBC DPSO SBO VPS ReDe ISOS AHEFA MSAA
[14] [16] [19] [21] [23] [25] [26]

1 A1–A4 2.854 3.5498 3.4917 3.5017 3.5327 3.3563 3.5612 3.4927

2 A5–A12 8.301 7.8356 7.9414 7.9340 7.8303 7.8726 7.8736 7.8573
3 A13–A16 0.645 0.6450 0.6450 0.6450 0.6453 0.6450 0.6450 0.6450
4 A17–A18 0.645 0.6450 0.6450 0.6450 0.6459 0.6450 0.6451 0.6474
5 A19–A22 8.202 8.1183 8.1154 8.0215 8.0029 8.5798 7.9710 7.8897
6 A23–A30 7.043 8.1338 8.0533 7.9826 7.9135 7.6566 7.8928 8.0057
7 A31–A34 0.645 0.6450 0.6450 0.6450 0.6451 0.7417 0.6450 0.6450
8 A35–A36 0.645 0.6450 0.6450 0.6450 0.6451 0.6450 0.6451 0.6454
9 A37–A40 16.328 12.6231 12.8569 12.8175 12.7626 13.0864 12.5404 12.6034
10 A41–A48 8.299 8.0971 8.0425 8.1129 7.9657 8.0764 7.9639 7.9616
11 A49–A52 0.645 0.6450 0.6451 0.6450 0.6452 0.6450 0.6459 0.6451
12 A53–A54 0.645 0.6450 0.6450 0.6450 0.6450 0.6937 0.6462 0.6450
13 A55–A58 15.048 17.3908 17.2136 17.3362 16.9041 16.2517 17.1323 17.1604
14 A59–A66 8.268 8.0634 8.0804 8.1010 8.0434 8.1703 8.0216 8.0368
15 A67–A70 0.645 0.6450 0.6450 0.6450 0.6451 0.6450 0.6450 0.6450
16 A71–A72 0.645 0.6450 0.6450 0.6450 0.6473 0.6450 0.6451 0.6450
Best weight (kg) 327.51 327.65 327.55 327.65 324.25 325.01 324.24 324.97
Average weight (kg) – 327.76 327.68 327.67 324.32 329.47 324.41 325.13
SD (kg) – 0.06 0.07 0.02 0.05 2.66 0.24 0.18
NI – 20,000 15,000 30,000 10,840 4000 8860 7130

Table 8 Natural frequencies Frequency CSS-BBBC DPSO SBO VPS ReDe ISOS AHEFA MSAA
of the optimum designs for the number (Hz)
72-bar space truss [14] [16] [19] [21] [23] [25] [26]

1 4.0000 4.0000 4.0000 4.0000 4.0000 4.0000 4.0000 4.0000

2 4.0000 4.0000 4.0000 4.0002 4.0000 – 4.0000 4.0000
3 6.0040 6.0000 6.0000 6.0000 6.0001 6.0008 6.0000 6.0000
4 6.2491 6.2460 6.2460 6.2428 6.2762 – 6.2740 6.2664
5 8.9726 9.0710 9.0719 9.0698 9.1073 – 9.1137 9.1035

grouped into 29 by seeing symmetry as reported in the

previous studies. Table 5 lists a comparison between opti-
mal results obtained by the present algorithms with dif-
ferent approaches. From the table, it can be seen that the
best weight obtained by MSAA is better than those given
by CSS-BBBC (2298.61 kg), SOS (2180.32 kg), ISOS
(2169.46 kg) and AHEFA (2160.74 kg) and slightly higher
than HALC-PSO (2156.73 kg) and SBO (2156.51 kg). Note
that although the best weight provided by HALC-PSO and
SBO is smaller than that of the MSAA, the convergence
speed of the MSAA is faster than these methods (13,000
NI for HALC-PSO and 23000 NI for SBO). Regarding the
Fig. 7 Convergence history of the 72-bar space truss convergence speed, the MSAA (6200 NI) is the first among
the considered algorithms. From the statistical point of view,
MSAA is more stable than SOS and ISOS with the small-
est SD (2.86 kg for MSAA, 83.59 kg for SOS and 43.48 kg
for ISOS). Frequency values obtained by MSAA satisfy all

13
Arabian Journal for Science and Engineering

Fig. 8 A 120-bar dome truss

[26]

allowable constraints as in Table 6. Figure 5 shows the con- weight of 324.97 kg which is better than the results reported
vergence history of the best design for this problem. by CSS-BBBC (327.51 kg), DPSO (327.65 kg), SBO
(327.55 kg), VPS (327.65 kg) and ISOS (325.01 kg) and
4.3 72-bar space truss 0.72 kg heavier than those reported by ReDe and AHEFA.
In terms of convergence speed, it is observed that the NI
Figure 6 shows the geometry of the 72-bar space truss. This used by the DPSO, SBO, VPS, ReDe and AHEFA algo-
truss was investigated by many scholars as a large-scale siz- rithms is significantly slower compared to MSAA (20,000
ing problem [25]. The bars are categorized into 16 groups NI for DPSO, 15,000 NI for SOB, 30,000 for VPS, 10,840
by considering geometrical symmetry. A lumped mass of NI for ReDe, 8860 NI for AHEFA and 7130 for MSAA).
2770 kg is attached to each of nodes 1–4. The optimal results Finally, from statistical point of view, the MSAA provides
obtained with different algorithms are given in Table 7 for a lower SD and a lower average weight value than other
comparison. The optimum design achieved by MSAA has a metaheuristics. Table 8 lists the first five natural frequencies

13
 Arabian Journal for Science and Engineering

Table 9 Comparison of optimal designs of the 120-bar dome truss obtained by different algorithms
Variables (cm2) CSS-BBBC DPSO HALC-PSO VPS ReDe ISOS AHEFA MSAA
[14] [16] [18] [21] [23] [25] [26]

1 A1 17.478 19.607 19.8905 19.6836 19.5131 19.6662 19.5094 20.0425

2 A2 49.076 41.290 40.4045 40.9581 40.3914 39.8539 40.3867 39.4775
3 A3 12.365 11.136 11.2057 11.3325 10.6066 10.6127 10.6033 13.6425
4 A4 21.979 21.025 21.3768 21.5387 21.1415 21.2901 21.1168 20.4928
5 A5 11.190 10.060 9.8669 9.8867 9.8057 9.7911 9.8221 9.0488
6 A6 12.590 12.758 12.7200 12.7116 11.7781 11.7899 11.7735 15.2658
7 A7 13.585 15.414 15.2236 14.9330 14.8163 14.7437 14.8405 12.9846
Best weight (kg) 9046.34 8890.48 8889.96 8888.74 8707.32 8710.06 8707.26 8707.39
Average weight (kg) – 8895.99 8900.39 8896.04 8707.52 8728.56 8707.56 8709.96
SD (kg) – 4.26 6.38 6.65 0.15 14.23 0.25 3.43
NI – 6000 17,000 30,000 5080 4000 3560 3100

Table 10 Natural frequencies Frequency CSS-BBBC DPSO HALC-PSO VPS ReDe ISOS AHEFA MSAA
of the optimum designs for the number (Hz) [14]
120-bar dome truss [16] [18] [21] [23] [25] [26]

1 9.000 9.0001 9.000 9.0000 9.0000 9.0001 9.0000 9.0000

2 11.007 11.0007 11.000 11.0000 11.0000 10.9998 11.0000 11.0000
3 11.018 11.0053 11.000 11.0000 11.0000 – 11.0000 11.0000
4 11.026 11.0129 11.010 11.0096 11.0000 – 11.0000 11.0000
5 11.048 11.0471 11.050 11.0491 11.0667 – 11.0669 11.0419

which indicate the feasibility of the obtained designs. The

convergence history of the best design for this problem is
shown in Fig. 7.

4.4 120-bar dome truss

The fourth example is the 120-bar dome truss shown in

Fig. 8. The bars are categorized into seven groups by assum-
ing symmetry about the z-axis. A lumped mass is attached
as 3000 kg at node 1, 500 kg at nodes 2–13 and 100 kg at
the rest of the free nodes. Table 9 presents a comparison of
optimal results obtained by MSAA and other metaheuristics.
Fig. 9 Convergence history of the 120-bar dome truss It can be seen that MSAA gives weight benefit of 338.95 kg,

Fig. 10 A 37-bar planar truss [26]

13
Arabian Journal for Science and Engineering

Table 11 Comparison of Variables (cm2) PSO DPSO HALC-PSO VPS ReDe ISOS AHEFA MSAA
optimal designs of the 37-bar
planar truss obtained by [12] [16] [18] [21] [23] [25] [26]
different algorithms
1 y3, y19 0.9637 0.9482 0.9750 0.9042 0.9533 0.9257 0.9589 0.9413
2 y5, y17 1.3978 1.3439 1.3577 1.2850 1.3414 1.3188 1.3450 1.3393
3 y7, y15 1.5929 1.5043 1.5520 1.5017 1.5319 1.4274 1.5355 1.5434
4 y9, y13 1.8812 1.6350 1.6920 1.6509 1.6528 1.5806 1.6668 1.6744
5 y11 2.0856 1.7182 1.7688 1.7277 1.7280 1.6548 1.7397 1.7571
6 A1, A27 2.6797 2.6208 2.9652 3.1306 2.9608 2.6549 2.8210 2.9344
7 A2, A26 1.1568 1.0397 1.0114 1.0023 1.0052 1.0383 1.0019 1.0256
8 A3, A24 2.3476 1.0464 1.0090 1.0001 1.0014 1.0000 1.0001 1.0095
9 A4, A25 1.7182 2.7163 2.4601 2.5883 2.5994 3.0083 2.5308 2.5838
10 A5, A23 1.2751 1.0252 1.2300 1.1119 1.1949 1.0024 1.2210 1.1569
11 A6, A21 1.4819 1.5081 1.2064 1.2599 1.2165 1.4499 1.2429 1.2548
12 A7, A22 4.6850 2.3750 2.4245 2.6743 2.4303 3.1724 2.4718 2.5104
13 A8, A20 1.1246 1.4498 1.4618 1.3961 1.3644 1.2661 1.4018 1.4626
14 A9, A18 2.1214 1.4499 1.4328 1.5036 1.5548 1.4659 1.5061 1.5245
15 A10, A27 3.8600 2.5327 2.5000 2.4441 2.5247 2.9013 2.5604 2.4586
16 A11, A15 2.9817 1.2358 1.2319 1.2977 1.1946 1.1537 1.2146 1.1888
17 A12, A15 1.2021 1.3528 1.3669 1.3619 1.3163 1.3465 1.3605 1.3765
18 A13, A16 1.2563 2.9144 2.2801 2.3500 2.4465 2.6850 2.3992 2.2341
19 A14 3.3276 1.0085 1.0011 1.0000 1.0003 1.0000 1.0000 1.0007
Best weight (kg) 377.20 360.40 359.93 359.94 359.81 360.74 359.81 359.91
Average weight (kg) 381.2 362.21 360.23 360.23 359.99 363.40 359.92 359.98
SD (kg) 4.26 1.68 0.24 0.22 0.15 1.57 0.09 0.10
NI 12,500 6000 10,000 30,000 13,740 4000 8640 3100

Table 12 Natural frequencies Frequency PSO DPSO HALC-PSO VPS ReDe ISOS AHEFA MSAA
of the optimum designs for the number (Hz) [12]
37-bar planar truss [16] [18] [21] [23] [25] [26]

1 20.0001 20.0194 20.0216 20.0002 20.0005 20.0119 20.0000 20.0051

2 40.0003 40.0113 40.0098 40.0005 40.0004 40.0964 40.0001 40.0047
3 60.0001 60.0082 60.0017 60.0000 60.0022 60.0066 60.0002 60.0078

183.09 kg, 182.57 kg, 181.35 kg and 2.67 kg as compared

to those obtained by CSS-BBBC, DPSO, HALC-PSO, VPS
and ISOS, respectively. On the other hand, the MSAA design
is 0.07 kg and 0.13 kg slightly heavy when compared to
the ReDe and AHEFA algorithms, respectively. Regarding
convergence speed, the MSAA (3100 NI) ranks first among
the metaheuristics considered. Finally, it can be observed
from the results that the MSAA obtains a better result in
average weight and SD than other metaheuristics, only being
surpassed by the ReDe and the AHEFA algorithms. Table 10
represents the first five optimal natural frequencies obtained
by the present work and it is clear that none of the frequency
Fig. 11 Convergence history of the 37-bar planar truss constraints are violated. Figure 9 shows the convergence his-
tory of the best design for this problem.

13
 Arabian Journal for Science and Engineering

Fig. 12 A 52-bar dome truss

[26]

4.5 37-bar planar truss Table 11 presents a comparison of optimal results

obtained by MSAA and other metaheuristics. The results
The 37-bar truss, simply supported bridge, is presented in indicate that MSAA gives the best weight without violation
Fig. 10. This problem considers simultaneous size and shape of constraints as 359.91 kg. The weight benefit for MSAA
optimization. A lumped mass of 10 kg is attached to each is 17.29, 0.49, 0.02, 0.03 and 0.83 kg as compared to PSO,
of the free nodes of the lower chord as shown in Fig. 10. DPSO, HALC-PSO, VPS and ISOS, respectively. These
The lower chord bars are assumed to have a fixed rectan- results show that MSAA reports better weight as compared
gular section of 0.4 cm2, whereas the remaining elements to other results reported in the literature (except for ReDe
are clustered into 14 groups with respect to their structural and AHEFA with 359.81 kg). Moreover, regarding the
symmetry about the middle vertical plane. Nodes on the convergence speed, MSAA ranks first among considered
upper chords (shape variables) can move vertically in view metaheuristics. MSAA only requires 3100 NI to obtain the
of the structural symmetry about the middle plane, whereas optimal solution while the PSO, DPSO, HALC-PSO, VPS,
the lower chord nodes are fixed. Therefore, this problem has ReDe, ISOS and AHEFA require 12,500, 6000, 10,000,
14 size variables and 5 shape variables. 30,000, 13,740, 4000 and 8640 NI, respectively. The table

13
Arabian Journal for Science and Engineering

Table 13 Comparison of Variables (cm2) PSO CSS-BBBC HALC-PSO ReDe ISOS AHEFA MSAA
optimal designs of the 52-bar
dome truss obtained by different [12] [14] [18] [23] [25] [26]
algorithms
1 ZA 5.5344 5.3310 5.9362 6.0188 6.1631 5.9953 5.9649
2 XB 2.0885 2.1340 2.2416 2.2976 2.4224 2.3062 2.3239
3 ZB 3.9283 3.7190 3.7309 3.7417 3.8086 3.7308 3.7003
4 XF 4.0255 3.9350 3.9630 3.9996 4.1080 4.0000 3.9636
5 ZF 2.4575 2.5000 2.5000 2.5001 2.5018 2.5000 2.5001
6 A1–A4 0.3696 1.0000 1.0001 1.0000 1.0074 1.0000 1.0001
7 A5–A8 4.1912 1.3056 1.1654 1.0852 1.0003 1.0832 1.1797
8 A9–A16 1.5123 1.4230 1.2323 1.1968 1.1982 1.2014 1.210
9 A17–A20 1.5620 1.3851 1.4323 1.4503 1.2787 1.4527 1.4799
10 A21–A28 1.9154 1.4226 1.3901 1.4216 1.4421 1.4212 1.3978
11 A29–A36 1.1315 1.0000 1.0001 1.0001 1.0000 1.0000 1.0229
12 A37–A44 1.8233 1.5562 1.6024 1.5614 1.4886 1.5570 1.6747
13 A45–A52 1.0904 1.4485 1.4131 1.3878 1.4990 1.3904 1.3033
Best weight (kg) 228.38 197.31 194.85 193.20 194.75 193.20 194.81
Average weight (kg) 234.30 – 196.85 195.43 207.55 198.73 195.32
SD (kg) 5.22 – 2.38 3.86 8.74 4.41 2.12
NI 11,270 – 7500 16,200 4000 12,120 7130

Table 14 Natural frequencies Frequency PSO CSS-BBBC HALC-PSO ReDe ISOS AHEFA MSAA
of the optimum designs for the number (Hz)
52-bar dome truss [12] [14] [18] [23] [25] [26]

1 12.751 12.987 11.4339 11.6107 12.5459 11.6629 11.8993

2 28.649 28.648 28.6480 28.6482 28.6518 28.6480 28.6479
3 28.649 28.679 28.6480 28.6499 – 28.6481 28.6479
4 28.803 28.713 28.6482 28.6500 – 28.6482 28.6612
5 29.230 30.262 28.6848 28.6510 – 28.6611 28.6612

structural frequencies (Hz) for various methods. None of

the frequency constraints are violated. Figure 11 shows the
convergence history of the best design for this problem.

4.6 52-bar dome truss

Figure 12 indicates a 52-bar dome truss structure. This struc-

ture is considered for simultaneous shape and size optimiza-
tion. A lumped mass of 50 kg is added to all the free nodes
of the structure. The bars are linked into eight groups by
considering symmetry about the z-axis, whereas the free
nodes can shift ± 2 m in each direction of the vertical plane
Fig. 13 Convergence history of the 52-bar dome truss to keep the dome symmetric.
Table 13 compares the results of MSAA with other opti-
mization methods. It can be seen that the design yielded
also indicates that MSAA is more stable than PSO, DPSO, by the MSAA (194.81 kg) is relatively better than those by
HALC-PSO, VPS, ReDe and ISOS with the smallest SD almost all the other algorithms, but slightly worse than those
(0.10 kg for MSAA, 4.26 kg for PSO, 1.68 kg for DPSO, by the ReDe (193.20 kg) and AHEFA (193.20 kg) methods.
0.24 kg for HALC-PSO, 0.22 kg for VPS, 0.15 kg for In terms of convergence speed, the MSAA only requires
ReDe and 1.57 kg for ISOS). Table 12 shows the optimized 7130 NI which is less than the PSO (11,270 NI), HALC-
PSO (7500 NI), ReDe (16,200 NI) and AHEFA (12,120

13
 Arabian Journal for Science and Engineering

NI). Although the convergence speed of ISOS (4000 NI) 5. Lin, J.H.; Che, W.Y.; Yu, Y.S.: Structural optimization on geomet-
is faster than that of the MSAA, the MSAA is more sta- rical configuration and element sizing with statical and dynami-
cal constraints. Comput. Struct. 15, 507–515 (1982). https://doi.
ble than the ISOS with the smallest standard deviation org/10.1016/0045-7949(82)90002-5
(2.12 kg for MSAA and 8.74 kg for ISOS). Regarding SD, 6. Grandhi, R.V.; Venkayya, V.B.: Structural optimization with
the MSAA ranks first among the metaheuristics considered. frequency constraints. AIAA J. 26, 858–866 (1988). https://doi.
Table 14 represents the first five natural frequencies opti- org/10.2514/3.9979
7. Wang, D.; Zhang, W.H.; Jiang, J.S.: Truss optimization on shape
mally obtained by the present work and it is clear that none and sizing with frequency constraints. AIAA J. 42, 622–630
of the frequency constraints are violated. Figure 13 shows (2004). https://doi.org/10.2514/1.1711
the convergence history of the best design for this problem. 8. Sedaghati, R.; Suleman, A.; Tabarrok, B.: Structural optimiza-
tion with frequency constraints using the finite element force
method. AIAA J. 40, 382–388 (2002)
9. Sedaghati, R.: Benchmark case studies in structural design opti-
5 Conclusions mization using the force method. Int. J. Solids Struct. 42, 5848–
5871 (2005). https://doi.org/10.1016/j.ijsolstr.2005.03.030
The modified simulated annealing algorithm (MSAA), a 10. Lingyun, W.; Mei, Z.; Guangming, W.; Guang, M.: Truss opti-
mization on shape and sizing with frequency constraints based
metaheuristic optimization method developed recently, was on genetic algorithm. Comput. Mech. 35, 361–368 (2005). https
applied for the first time ever to size and shape optimiza- ://doi.org/10.1007/s00466-004-0623-8
tion of truss structures with frequency constraints. The 11. Wei, L.; Tang, T.; Xie, X.; Shen, W.: Truss optimization on
numerical results demonstrate that, in most of the cases, shape and sizing with frequency constraints based on paral-
lel genetic algorithm. Struct. Multidiscip. Optim. 43, 665–682
the MSAA attains the global optimum or a nearly global (2011). https://doi.org/10.1007/s00158-010-0600-0
optimum design with high accuracy and reliability with less 12. Gomes, H.M.: Truss optimization with dynamic constraints using
iterations than some other methods in the literature. The a particle swarm algorithm. Expert Syst. Appl. 38, 957–968
presented data on average weight and standard deviation of (2011). https://doi.org/10.1016/j.eswa.2010.07.086
13. Miguel, L.F.F.; Fadel Miguel, L.F.: Shape and size optimization
optimized weight obtained from one hundred independent of truss structures considering dynamic constraints through mod-
runs prove the robustness of MSAA. This is because the ern metaheuristic algorithms. Expert Syst. Appl. 39, 9458–9467
preliminary exploration helps the algorithm to explore the (2012). https://doi.org/10.1016/j.eswa.2012.02.113
search space, while the search step guides the algorithm 14. Kaveh, A.; Zolghadr, A.: Truss optimization with natural fre-
quency constraints using a hybridized CSS-BBBC algorithm
to escape from the local optimum (see Figs. 3, 5, 7, 9, 11 with trap recognition capability. Comput. Struct. 102–103, 14–27
and 13). The MSAA is very general and could be applied to (2012). https://doi.org/10.1016/j.compstruc.2012.03.016
more complex engineering problems such as laminated com- 15. Kaveh, A.; Zolghadr, A.: Shape and size optimization of truss
posite, functionally graded structures and reliability-based structures with frequency constraints using enhanced charged
system search algorithm. Asian J. Civ. Eng. 12, 487–509 (2011)
design optimization problems [33–36] where the computa- 16. Kaveh, A.; Zolghadr, A.: Democratic PSO for truss layout and
tional cost is always a concern. These and other applications size optimization with frequency constraints. Comput. Struct. 130,
remain to be performed. Only after the MSAA method has 10–21 (2014). https://doi.org/10.1016/j.compstruc.2013.09.002
been applied to a broad range of problems, it will be pos- 17. Khatibinia, M.; Naseralavi, S.: Truss optimization on shape and
sizing with frequency constraints based on orthogonal multi-
sible to establish that it is a better method to solve a given gravitational search algorithm. J. Sound Vib. 333, 6349–6369
class of problems. (2014). https://doi.org/10.1016/j.jsv.2014.07.027
18. Kaveh, A.; Ilchi Ghazaan, M.: Hybridized optimization algo-
Acknowledgements The authors are thankful to Universidade Tec- rithms for design of trusses with multiple natural frequency
nológica Federal do Paraná and for the scholarship granted to the first constraints. Adv. Eng. Softw. 79, 137–147 (2015). https://doi.
author by CAPES. org/10.1016/j.advengsoft.2014.10.001
19. Farshchin, M.; Camp, C.V.; Maniat, M.: Optimal design of truss
structures for size and shape with frequency constraints using
a collaborative optimization strategy. Expert Syst. Appl. 66,
References 203–218 (2016). https://doi.org/10.1016/j.eswa.2016.09.012
20. Kaveh, A.; Zolghadr, A.: Cyclical parthenogenesis algorithm
1. Grandhi, R.V.: Structural optimization with frequency con- for layout optimization of truss structures with frequency
straints—a review. AIAA J. 31, 2296–2330 (1993). https://doi. constraints. Eng. Optim. 49, 1317–1334 (2017). https ://doi.
org/10.2514/3.9979 org/10.1080/0305215X.2016.1245730
2. Zuo, W.; Xu, T.; Zhang, H.; Xu, T.: Fast structural optimization 21. Kaveh, A.; Ilchi Ghazaan, M.: Vibrating particles system algo-
with frequency constraints by genetic algorithm using adaptive rithm for truss optimization with multiple natural frequency
eigenvalue reanalysis methods. Struct. Multidiscip. Optim. 43, constraints. Acta Mech. 228, 307–322 (2017). https ://doi.
799–810 (2011). https://doi.org/10.1007/s00158-010-0610-y org/10.1007/s00707-016-1725-z
3. Talbi, E.G.: Metaheuristics: From Design to Implementation. 22. Kaveh, A.; Zolghadr, A.: Truss shape and size optimization with
Wiley, New York (2009) frequency constraints using tug of war optimization. Asian J.
4. Bellagamba, L.; Yang, T.Y.: Minimum-mass truss structures with Civ. Eng. 18, 311–313 (2017)
constraints on fundamental natural frequency. AIAA J. 19, 1452– 23. Ho-Huu, V.; Nguyen-Thoi, T.; Truong-Khac, T.; et al.: An
1458 (1981). https://doi.org/10.2514/3.7875 improved differential evolution based on roulette wheel

13
Arabian Journal for Science and Engineering

selection for shape and size optimization of truss structures 31. Millan-Paramo, C.; Filho, J.: Modified simulated annealing
with frequency constraints. Neural Comput. Appl. 29, 167–185 algorithm for optimal design of steel structures. Rev int métodos
(2018). https://doi.org/10.1007/s00521-016-2426-1 numér cálc diseño ing 35, 1–12 (2019). https://doi.org/10.23967
24. Tejani, G.G.; Savsani, V.J.; Patel, V.K.: Adaptive symbiotic /j.rimni.2019.03.003
organisms search (SOS) algorithm for structural design opti- 32. Kirkpatrick, S.; Gelatt, C.D.; Vecchi, M.P.: Optimization by
mization. J. Comput. Des. Eng. 3, 226–249 (2016). https://doi. simulated annealing. Science 220, 671–680 (1983). https://doi.
org/10.1016/j.jcde.2016.02.003 org/10.1126/science.220.4598.671
25. Tejani, G.G.; Savsani, V.J.; Patel, V.K.; Mirjalili, S.: Truss opti- 33. Lieu, Q.X.; Lee, J.: Modeling and optimization of functionally
mization with natural frequency bounds using improved symbi- graded plates under thermo-mechanical load using isogeometric
otic organisms search. Knowl. Based Syst. 143, 162–178 (2018). analysis and adaptive hybrid evolutionary firefly algorithm. Com-
https://doi.org/10.1016/j.knosys.2017.12.012 pos. Struct. 179, 89–106 (2017). https://doi.org/10.1016/j.comps
26. Lieu, Q.X.; Do, D.T.T.; Lee, J.: An adaptive hybrid evolutionary truct.2017.07.016
firefly algorithm for shape and size optimization of truss struc- 34. Lieu, Q.X.; Lee, J.; Lee, D.; et al.: Shape and size optimiza-
tures with frequency constraints. Comput. Struct. 195, 99–112 tion of functionally graded sandwich plates using isogeometric
(2018). https://doi.org/10.1016/j.compstruc.2017.06.016 analysis and adaptive hybrid evolutionary firefly algorithm. Thin-
27. Wolpert, D.H.; Macready, W.G.: No free lunch theorems for Walled Struct. 124, 588–604 (2018). https://doi.org/10.1016/j.
optimization. IEEE Trans. Evol. Comput. 1, 67–82 (1997). https tws.2017.11.054
://doi.org/10.1109/4235.585893 35. Lieu, Q.X.; Lee, J.: An isogeometric multimesh design approach
28. Millán Páramo, C.; Begambre Carrillo, O.; Millán Romero, E.: for size and shape optimization of multidirectional functionally
Proposal and validation of a modified Simulated annealing algo- graded plates. Comput. Methods Appl. Mech. Eng. 343, 407–437
rithm for solving optimization problems. Rev int métodos numér (2019). https://doi.org/10.1016/j.cma.2018.08.017
cálc diseño ing 30, 264–270 (2014). https://doi.org/10.1016/j. 36. Lieu, Q.X.; Lee, J.: A reliability-based optimization approach for
rimni.2013.10.003 material and thickness composition of multidirectional function-
29. Millán Páramo, C.; Begambre Carrillo, O.: Solving topology opti- ally graded plates. Compos. Part B Eng. 164, 599–611 (2019).
mization problems using the modified simulated annealing algo- https://doi.org/10.1016/j.compositesb.2019.01.089
rithm. Rev int métodos numér cálc diseño ing 32, 65–69 (2016).
https://doi.org/10.1016/j.rimni.2014.11.005
30. Millan-Paramo, C.: Modified simulated annealing algorithm for
discrete sizing optimization of truss structure. Jordan J. Civ. Eng.
12, 683–697 (2018)

13