The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/0264-4401.htm

Shuffled
Shuffled shepherd shepherd
optimization method: a new optimization
method
Meta-heuristic algorithm
Ali Kaveh and Ataollah Zaerreza 2357
School of Civil Engineering, Iran University of Science and Technology,
Tehran, Iran Received 24 October 2019
Revised 24 January 2020
2 February 2020
8 February 2020
Accepted 8 February 2020
Abstract
Purpose – This paper aims to present a new multi-community meta-heuristic optimization algorithm, which
is called shuffled shepherd optimization algorithm (SSOA). In this algorithm.
Design/methodology/approach – The agents are first separated into multi-communities and the
optimization process is then performed mimicking the behavior of a shepherd in nature operating on each
community.
Findings – A new multi-community meta-heuristic optimization algorithm called a shuffled shepherd
optimization algorithm is developed in this paper and applied to some attractive examples.
Originality/value – A new metaheuristic is presented and tested with some classic benchmark problems
and some attractive structures are optimized.
Keywords Shuffled shepherd optimization algorithm, Sheep, Shepherd, Meta-heuristic,
Optimization, Mathematical bench mark, Truss structures optimization
Paper type Research paper

1. Introduction
The term “optimization” refers to the study of problems in which one seeks to minimize or
maximize a function by systematically choosing the values of variables from/within a
permissible set (Kaveh, 2017). Optimization methods are divided into two groups of
gradient-based optimization algorithm and meta-heuristic optimization algorithm. Meta-
heuristic optimization algorithms are becoming more and more popular in engineering
applications because they

rely on rather simple concepts and are easy to implement;

do not require gradient information;

can bypass local optima; and

can be used in a wide range of problems covering different disciplines (Mirjalili and
Lewis, 2016).

Many meta-heuristic algorithms have one community and the process of the
optimization is carried out by performing on this community like Genetic Algorithm
(GA) (Holland, 1992), particle swarm optimization (PSO) (Kennedy and Eberhart, 1995)
and ant colony optimization (ACO) (Dorigo and Di Caro, 1999). Some meta-heuristic
algorithms divide agents into more than one community and perform optimization Engineering Computations
Vol. 37 No. 7, 2020
pp. 2357-2389
Compliance with ethical standards: Conflict of interest: No potential conflict of interest was reported © Emerald Publishing Limited
0264-4401
by the authors. DOI 10.1108/EC-10-2019-0481
EC operation on them, and merge them in each iteration. Examples of such algorithms are
37,7 Shuffled Complex Evolution (SCE) (Duan et al., 1993), Shuffled Frog-leaping Algorithm
(SFLA) (Eusuff et al., 2006) and Shuffled Differential Evolution (SDE) (Reddy and
Vaisakh, 2013).

2358

Figure 1.
Steps for dividing
sheep in herd (blue
balls show the
selected sheep)
Most of the meta-heuristic algorithms focus only on the better agents and the goal is to move Shuffled
the worse agents toward better positions by specified formulae that focus on good agents shepherd
while ignoring the worse agents. Additional attention to worse agents may improve the
overall performance of algorithms, resulting in better solution and/or using smaller number
optimization
of function evaluations. method

2359

Figure 2.
Flowchart of the
SSOA
EC Global
37,7 Function name Interval Function minimum

X [ [–10, 10]
2
1 1 1 1 0.352386
Aluffi-Pentiny f ð X Þ ¼ x41  x21 þ x1 þ x22
4 2 10 2

2360 2 3 4 7
Bohachevsky 1 X [ [–100, 100] f ð X Þ ¼ x21  2x22  cosð3p x1 Þ  cosð4p x2 Þ þ 0.0
10 10 10

2 3 3
Bohachevsky 2 X [ [–50, 50] f ð X Þ ¼ x21  2x22  cosð3p x1 Þcosð4p x2 Þ þ 0.0
10 10

Becker and Lago X [ [–10, 10]2 f(X) = (|x1| – 5)2 þ (|x2| – 5)2 0.0

2

–5 # x1 # 10 5:1 2 5 1
Branin
0 # x5 # 15 f ðX Þ ¼ x2  x þ x1 þ 10 1  cosðx1 Þ þ 10 0.397887
4p 2 1 p 8p

2 1
Camel X [ [5, 5] f ð X Þ ¼ 4x21  2:1x41 þ x61 þ x1 x2  4x22 þ 4x42 1.0316
3

2 1
Cb3 X [ [5, 5] f ð X Þ ¼ 2x21  1:05x51 þ x61 þ x1 x2 þ x22 0.0
6

n X
n
1 X n
Cosine mixture n = 4, X [ [1,1] f ðX Þ ¼ x2i  cosð5p xi Þ 0.4
i¼1
10 i¼1

3
Dejoung X [ [–5.12, 5.12] f ð X Þ ¼ x21 þ x22 þ x23 0.0

!
Exponential
n = 2, 4, 8, X
n
X [[–1,1]n f ð X Þ ¼ exp 0:5 x2i 1
i¼1

h 
Goldstein and
X [ [–2, 2]
2 f ð X Þ ¼ 1 þ ðx1 þ x2 þ 1Þ2 19  14x1 þ 3x21  14x2 3
price
ih
þ16x1 x2 þ 3x22 30 þ ð2x1  3x2 Þ2 ð18  32x1
i
12x21 þ 48x2  36x1 x2 þ 27x22
Table I.

X [ [–100, 100]
2 1 X 2 Y2
xi
Specification of the Griewank f ðX Þ ¼ 1 þ x2i  cos pffi 0.0
mathematical 200 i¼1 i¼1 i
optimization
problems (continued)
 Global
Shuffled
Function name Interval Function minimum shepherd
Hartman 3 X [ [0, 1]3 0 1 3.862782
optimization
X
4 X
3 method
f ðX Þ ¼  ci exp@ aij ðxj  pij Þ2A

i¼1 j¼1

2361
2 3 2 3
3 10 30 1
6 0:1 10 35 7 6 1:2 7
6
a¼4 7; c ¼ 6 7
3 10 30 5 4 3 5 and
0:1 10 35 3:2

2 3
0:3689 0:117 0:2673
6 0:4699 0:4387 0:747 7
p¼6
4 0:1091
7
0:8732 0:5547 5
0:03815 0:5743 0:8828

2 
X 
Rastrigin X [ [–1, 1]
2
f ðX Þ ¼ x2i  cosð18xi Þ 2.0
i¼1

n X
n1  2 Table I.
Rosenbrock X [ [–30, 30] , n = 2 f ðxÞ ¼ 100 xiþ1  x2i þ ðxi  1Þ2 0.0
i¼1

The purpose of this paper is to introduce a new multi-community and simple meta-heuristic
algorithm that is inspired by using the instinct of animals by a shepherd called Shuffled
Shepherd Optimization Algorithm. In this algorithm, in the first step the agents are
separated into multi-communities and the optimization progress is inspired by the behavior
of a shepherd in nature operating on each community. In the process of optimization,
attention is paid to both good and bad agents, leading to better performance of the
algorithm.
This paper is organized as follows: In Section 2, inspiration, mathematical model and the
steps for Shuffled Shepherd Optimization Algorithm are described. In Section 3, some
benchmark functions and 2 classic engineering problems are investigated using the SSOA.
In Section 4, four truss design problems and a large-scale double-layer gird design problem
are optimized using the SSOA, and finally conclusions are derived in Section 5.

2. Shuffled shepherd optimization algorithm

2.1 Inspiration
In nature, animals use instinct to find the best way for living. Human beings learn how to
use animals’ instinct for their purposes. Shepherds use animal instinct to find the best way
to go to pasture. Always in a herd, a shepherd puts one horse to find the best stiff and fast
way to pasture.
EC Horses have an instinct to find the best stiff and fast way. In nature, we can see the trail
37,7 of animals’ way. This fast and stiff ways are always followed by horses or other animals. In
addition to the use of this trail by shepherd, road engineers in the past have used this trail
for constructing new roads.
Shepherds use one or more horses in the herd to move their tools and find the way.
Shepherds try to guide sheep behind horses to pasture and bring them back because this
2362 trail is the best trail they can ever find. Shepherd behavior has been an inspiration in this
paper, and it is used for mathematical modeling of the optimization.

GEN–S–M GEN–S–M–LS CPA (Kaveh and CSS (Kaveh and Present work
Function (Tsoulos, 2008) (Tsoulos, 2008) Zolghadr, 2017) Talatahari, 2010b) (SSOA)

AP 1,277 1,253 560 804 295

Bf1 1,640 1,615 1,173 1,187 472
Bf2 1,676 1,636 1,376 742 451
BL 2,439 1,436 424 423 510
Branin 1,404 1,257 708 852 459
Camel 1,336 1,300 482 575 334
Cb3 1,163 1,118 548 436 266
CM 1,743 1,539 1,612 1,563 507
Dejoung 1,462 1,281 670 630 353
Exp2 817 807 435 132 123
Exp4 2,054 1,496 781 867 312
Exp8 2,054 1,496 1,105 1,426 673
Goldstein 1,408 1,325 805 682 430
and Price
Table II. Griewank 1,764 1,652 (0.99) 1,572 1,551 867
Performance Hartman3 1,332 1,274 1,128 860 326
Rastrigin 1,392 1,381 n/a 1,402 960
comparison for the Rosenbrock 1,675 1,462 n/a 1,452 1,810
mathematical Total 26.636 23.328(0.999) 12,869 15.584 9148
optimization
problems Note: N/A (not available)

Figure 3.
Schematic of the
welded beam
2.2 Mathematical model Shuffled
In this section, mathematical model of herd and shepherd are illustrated. shepherd
2.2.1 Herd. In nature, in one distract we can see a lot of herd, first we divide our sheep optimization
into “h” number of herd with “s” number of sheep in each herd, therefore we have n = h  s
method
sheep (agents). For dividing sheep (agents) in each herd, first we sort all the sheep in terms of
their objective function values in an ascending order (in maximization problems, and we sort
2363
No. !0 b0 b max Best Worst Mean Std.

1 0.5 1.5 2 1.960431 3.825033 2.667163 0.502

2 0.5 1.5 2.5 1.831520 3.005574 2.306228 0.308
3 0.5 1.5 3 1.724857 3.450027 2.073674 0.384
4 0.5 1.5 3.5 1.727421 2.220498 1.823992 0.124
5 0.5 1.5 4 1.724855 1.890547 1.771640 0.044
6 0.5 2 2.5 1.724852 1.994225 1.775230 0.083
7 0.5 2 3 1.724853 1.741982 1.726642 0.004
8 0.5 2 3.5 1.724856 1.753322 1.727385 0.005
9 0.5 2 4 1.724862 1.746769 1.728080 0.005
10 0.5 2.5 3 1.724854 1.726242 1.724955 2.68E-04
11 0.5 2.5 3.5 1.724872 1.726221 1.725263 3.85E-04
12 0.5 2.5 4 1.724870 1.732302 1.725973 0.002
13 0.5 3 3.5 1.724960 1.728071 1.725438 6.06E-04
14 0.5 3 4 1.725013 1.731427 1.726302 0.001
15 0.5 3.5 4 1.725214 1.744705 1.728456 0.004
16 1 1.5 2 1.724853 2.781531 2.178187 0.285
17 1 1.5 2.5 1.724852 2.570316 2.043011 0.240
18 1 1.5 3 1.724852 2.269081 1.872679 0.172
19 1 1.5 3.5 1.724853 1.907858 1.754319 0.043
20 1 1.5 4 1.724856 1.762913 1.729981 0.009
21 1 2 2.5 1.724852 1.729270 1.725003 8.06E-04
22 1 2 3 1.724853 1.725898 1.724937 2.55E-04
23 1 2 3.5 1.724861 1.725345 1.724928 9.98E-05
24 1 2 4 1.724888 1.727619 1.725423 7.03E-04
25 1 2.5 3 1.724856 1.725370 1.724949 1.01E-04
26 1 2.5 3.5 1.724891 1.727444 1.725278 5.49E-04
27 1 2.5 4 1.724890 1.728569 1.725809 8.20E-04
28 1 3 3.5 1.724954 1.727341 1.725683 6.79E-04
29 1 3 4 1.725115 1.730349 1.726995 0.001
30 1 3.5 4 1.725605 1.735186 1.728416 0.002
31 1.5 1.5 2 1.724852 2.307840 1.840313 0.162
32 1.5 1.5 2.5 1.724852 2.015799 1.750629 0.074
33 1.5 1.5 3 1.724853 1.865128 1.729554 0.026
34 1.5 1.5 3.5 1.724858 1.726011 1.724985 2.56E-04
35 1.5 1.5 4 1.724870 1.729607 1.725228 8.72E-04
36 1.5 2 2.5 1.724852 1.724871 1.724855 4.32E-06
37 1.5 2 3 1.724857 1.724999 1.724894 3.83E-05
38 1.5 2 3.5 1.724871 1.725779 1.725027 2.17E-04
39 1.5 2 4 1.724999 1.728106 1.725488 6.41E-04
40 1.5 2.5 3 1.724868 1.725358 1.724989 1.21E-04
Table III.
41 1.5 2.5 3.5 1.724885 1.726522 1.725429 3.76E-04
42 1.5 2.5 4 1.724991 1.730796 1.726354 0.001 Results of the
43 1.5 3 3.5 1.725099 1.730144 1.726135 0.001 sensitivity analysis
44 1.5 3 4 1.726169 1.732557 1.728059 0.002 for the welded beam
45 1.5 3.5 4 1.726031 1.734503 1.729984 0.003 problem
EC them in an descending order), then to segregate sheep in a herd, the first h number of sheep
37,7 are selected and put randomly in each herd as shown in Figure 1(a) (put one sheep in each
herd) then select the second h number of sheep and put them in a herd again as shown in
Figure 1(b). This process is continued until all sheep are segregated into a herd.

Method h(= x1 ) l(= x2 ) t(= x3 ) b(= x4 ) fcost

2364
GA1 (Deb, 1991) 0.248900 6.173000 8.178900 0.253300 2.433116
GA2 (Coello, 2000) 0.208800 3.420500 8.997500 0.210000 1.748310
HPSODEPSR (Ali and Tawhid, 2016) N/A N/A N/A N/A 2.380957
MGA (Coello Coello, 2000) 0.1829 4.0483 9.3666 0.2059 1.824551
hHHO-SCA (Kamboj et al., 2019) 0.190086 3.696496 9.386343 0.204157 1.779032
ESs (Mezura-Montes and Coello, 2008) 0.199742 3.61206 9.0375 0.206082 1.7373
RO (Kaveh and Khayatazad, 2012) 0.203687 3.528467 9.00423 0.20724 1.735344
CDE (Huang et al., 2007) 0.203137 3.542998 9.033498 0.206179 1.733462
IMSS (Rakhshani and Rahati, 2017) 0.2015 3.5620 9.0414 0.2057 1.7312
EBA (Yılmaz and Küçüksille, 2015) 0.2015 3.5620 9.0414 0.2057 1.7312
WOA (Mirjalili and Lewis, 2016) 0.205396 3.484293 9.037426 0.206276 1.730499
GA3 (Coello and Montes, 2002) 0.205986 3.471328 9.020224 0.206480 1.728226
CPSO (He and Wang, 2007) 0.202369 3.544214 9.04821 0.205723 1.728024
TEO (Kaveh and Dadras, 2017) 0.205681 3.472305 9.035133 0.205796 1.725284
IACO (Kaveh and Talatahari, 2010a) 0.205700 3.471131 9.036683 0.205731 1.724918
CSS (Kaveh and Talatahari, 2010b) 0.205820 3.46810 9.038024 0.20572 1.724866
CE-CBA(Xiao et al., 2019) 0.205726 3.47056 9.036630 0.20573 1.724858
Table IV. IAFOA (Wu et al., 2018) 0.205726 3.470562 9.036630 0.20573 1.724856
Optimization results IGMM (Varaee and Ghasemi, 2017) 0.205729 3.470496 9.306625 0.205730 1.724853
for the welded beam MCSS (Kaveh et al., 2013) 0.205729 3.470493 9.03662 0.20572 1.724853
design problem SSOA 0.2057296 3.4704888 9.0366236 0.2057297 1.7248524

Methods Best Mean Worst SD

GA1 (Deb, 1991) 2.433116 N/A N/A N/A

GA2 (Coello, 2000) 1.748309 1.771973 1.785835 0.011220
HPSODEPSR (Ali and Tawhid, 2016) 2.380957 2.380957 2.380957 1.82E-08
MGA (Coello Coello, 2000) 1.824551 1.919000 1.995000 5.37E-02
hHHO-SCA (Kamboj et al., 2019) 1.779032 N/A N/A N/A
ESs (Mezura-Montes and Coello, 2008) 1.737300 1.813290 1.994651 0.070500
RO (Kaveh and Khayatazad, 2012) 1.735344 1.9083 N/A 0.173744
CDE (Huang et al., 2007) 1.733461 1.768158 1.824105 0.022194
IMSS (Rakhshani and Rahati, 2017) 1.7312 N/A N/A N/A
EBA (Yılmaz and Küçüksille, 2015) 1.7312 N/A N/A N/A
WOA (Mirjalili and Lewis, 2016) 1.730499 1.7320 N/A 0.0226
GA3 (Coello and Montes, 2002) 1.728226 1.792654 1.993408 0.074713
CPSO (He and Wang, 2007) 1.728024 1.748831 1.782143 0.012926
TEO (Kaveh and Dadras, 2017) 1.725284 1.768040 1.931161 0.0581661
IACO (Kaveh and Talatahari, 2010a) 1.724918 1.729752 1.77596 0.009200
CSS (Kaveh and Talatahari, 2010b) 1.724866 1.739654 1.759479 0.008064
Table V.
CE-CBA (Xiao et al., 2019) 1.724858 1.724858 1.724858 3.5641E-15
Statistical results of IAFOA (Wu et al., 2018) 1.724856 1.724856 1.424856 8.991E-0.7
different methods for IGMM (Varaee and Ghasemi, 2017) 1.724853 1.732152 1.74769 7.14E-03
the welded beam MCSS (Kaveh et al., 2013) 1.724853 1.735438 1.753681 0.009527
problem SSOA 1.724852 1.724855 1.724871 4.32E-06
2.2.2 Shepherd. In each herd, shepherds try to lead the sheep toward the horse. In each herd Shuffled
the sheep are ordered according to their objective function values in an ascending order. The shepherd
sheep get selected from the first sheep to the last one. For finding the step size of the
movement of the sheep, the selected member is taken as a shepherd and it is denoted by i.
optimization
Obviously there are some better and worse sheep compared to the selected sheep (shepherd). method
The better sheep are called horses. Therefore, there are some horses and sheep for each
shepherd. Using the rule of nature, the vector for movement can be obtained. In nature, a
shepherd leads the sheep toward the horse. Therefore, one of the horses (denoted by d) and 2365
one of the sheep from the remaining sheep (shown with j) are selected randomly. The first
shepherd moves toward the selected sheep and for leading the sheep moves toward the
horse. Therefore, the movement vector is obtained as follows:

Stepsizei ¼ b  rand  ðXd  Xi Þ þ a  rand  ðXj  Xi Þ (1)

Where Xi, Xd, Xj are solution vectors of shepherd, selected horse and selected sheep in an m-
dimensional search space, respectively; rand is a random vector which signifies that each
component is in interval [0,1] and we have the number of components based on the number
of components of solution vectors; a parameter is equal to a0 in the start of the algorithm,
then decreases by the iteration number of the algorithm to zero and can be calculated by
equation (2); b parameter is equal to b 0 in the start of the algorithm then increases by the
iteration number of algorithm to b max; b can be calculated by equation (3) and sign “8”
denotes element-by-element multiplication.
First sheep selected from the herd does not have any other sheep in the herd better than
itself, so the first term of the step size is zero; and for the last sheep that is selected from the
herd which does not have any other sheep in the herd worse than itself, the second term of
the step size is zero. A decrease in a and increase in b , reduces gradually the exploration
and increases the exploitation of the algorithm.
a0
/ ¼ a0   iteration (2)
maxiteration

b max  b 0
b ¼ b0 þ  iteration (3)
maxiteration

After calculating the step size for all the sheep in a herd, the temple solution vector is
calculated for each sheep by the following equation:

Figure 4.
Schematic of the
pressure vessel
EC xtemple
i ¼ xold
i þ stepsizei (4)
37,7
If temple objective function is not worse than old objective function, then the position of the
sheep is updated, so we have xnew
i ¼ xtemple
i , otherwise xnew
i ¼ xold
i .

2366
No. !0 b0 b max Best Worst Mean Std.

1 0.5 1.5 2 5937.9221 9678.2863 6784.5424 937.052

2 0.5 1.5 2.5 6105.8884 7318.993 6642.038 392.250
3 0.5 1.5 3 5960.321 7407.2313 6589.6663 459.588
4 0.5 1.5 3.5 5957.5773 7167.1373 6272.9405 340.698
5 0.5 1.5 4 5887.6915 6737.8591 6074.1398 212.064
6 0.5 2 2.5 5885.5534 7110.4809 6058.9547 279.313
7 0.5 2 3 5885.4021 6538.2835 5942.1205 136.0747
8 0.5 2 3.5 5885.3516 6000.4418 5898.0198 25.609
9 0.5 2 4 5885.4596 5916.5175 5890.7515 8.903
10 0.5 2.5 3 5885.3417 5893.1437 5886.7253 1.869
11 0.5 2.5 3.5 5885.3491 5902.4701 5887.0622 3.170
12 0.5 2.5 4 5885.3328 8834.4409 5985.0822 538.170
13 0.5 3 3.5 5885.3318 8834.4091 5984.3788 538.288
14 0.5 3 4 5885.3374 8620.3227 6250.4736 9.453
15 0.5 3.5 4 5885.345 8834.4091 6264.8624 983.313
16 1 1.5 2 5987.6779 7305.9870 6403.9379 347.046
17 1 1.5 2.5 5885.8300 7283.3699 6304.6366 373.328
18 1 1.5 3 5885.6348 6940.9005 6113.3912 283.922
19 1 1.5 3.5 5885.3666 6299.4091 5940.4521 89.152
20 1 1.5 4 5885.5543 6291.8489 5920.6347 74.624
21 1 2 2.5 5885.3290 6043.9412 5894.7704 34.052
22 1 2 3 5885.3278 5890.2633 5885.7295 0.957
23 1 2 3.5 5885.3290 5896.7648 5886.4192 2.412
24 1 2 4 5885.3262 5890.5371 5885.7512 1.010
25 1 2.5 3 5885.3265 8893.5559 6076.9052 729.349
26 1 2.5 3.5 5885.3258 5885.9368 5885.4415 0.178
27 1 2.5 4 5885.3285 5889.6133 5885.5488 0.774
28 1 3 3.5 5885.3268 8627.6884 5976.8349 500.666
29 1 3 4 5885.3364 8893.5559 6359.7345 107.963
30 1 3.5 4 5885.3285 8893.5559 6359.5763 107.957
31 1.5 1.5 2 5885.3279 7282.0770 6114.4404 340.909
32 1.5 1.5 2.5 5885.3269 6730.1070 5966.3242 170.228
33 1.5 1.5 3 5885.3261 5986.5065 5899.7867 27.082
34 1.5 1.5 3.5 5885.3265 5900.9876 5886.7982 3.374
35 1.5 1.5 4 5885.3258 5929.4165 5887.9769 8.8108
36 1.5 2 2.5 5885.3258 6103.6344 5893.6254 39.922
37 1.5 2 3 5885.3258 5886.1050 5885.3677 0.146
38 1.5 2 3.5 5885.3258 8834.4091 5983.6813 538.417
39 1.5 2 4 5885.3258 8619.9955 5976.5543 499.259
40 1.5 2.5 3 5885.3258 8634.9585 6250.5333 947.002
Table VI.
41 1.5 2.5 3.5 5885.3259 8834.4091 6364.6986 109.087
Results of the 42 1.5 2.5 4 5885.3261 8893.5556 6266.4057 988.673
sensitivity analysis 43 1.5 3 3.5 5885.3260 11628.780 6624.1909 1456.87
for the pressure 44 1.5 3 4 5885.3276 11628.178 6740.3234 1704.06
vessel problem 45 1.5 3.5 4 5885.3281 8893.5559 7012.4183 1405.91
Methods Ts(= x1) Th(= x2) R(= x3) L(= x4) fcost
Shuffled
shepherd
hHHO-SCA (Kamboj et al., 2019) 0.945909 0.447138 46.8513 125.4684 6393.0928 optimization
GA (Deb, 1991) 0.812500 0.437500 42.097398 176.654050 6059.9463
CPSO (He and Wang, 2007) 0.812500 0.437500 42.091266 176.746500 6061.0777 method
ESs (Mezura-Montes and Coello, 2008) 0.812500 0.437500 42.098087 176.640518 6059.7456
CSS (Kaveh and Talatahari, 2010b) 0.812500 0.437500 42.103624 176.572656 6059.0888
CDE (Huang et al., 2007) 0.812500 0.437500 42.0984 176.6376 6059.7340 2367
WOA(Mirjalili and Lewis, 2016) 0.812500 0.437500 42.0982699 176.638998 6059.7410
IGMM (Varaee and Ghasemi, 2017) 0.812500 0.437500 42.098445 176.63659 6059.7143
HPSODEPSR (Ali and Tawhid, 2016) N/A N/A N/A N/A 6059.7143
HCS-LSAL (Long et al., 2014) 0.812500 0.437500 42.09844 176.6366 6059.7143
EBA (Yılmaz and Küçüksille, 2015) N/A N/A N/A N/A 6059.71
GA3 (Coello and Montes, 2002) 0.812500 0.437500 42.0974 176.6540 6059.9463
IACO (Kaveh and Talatahari, 2010a) 0.812500 0.437500 42.098353 176.637751 6059.7258
CE-CBA(Xiao et al., 2019) 0.812500 0.437500 42.09984456 176.6365958 6059.7143
MCSS (Kaveh et al., 2013) 0.812500 0.437500 42.107406 176.525589 6058.6233 Table VII.
LWOA (Zhou et al., 2018) 0.778858 0.385321 40.32609 200 5893.339 Optimization results
TEO (Kaveh and Dadras, 2017) 0.779151 0.385296 40.369858 199.301899 5887.5111 for the pressure
SSOA 0.778179 0.384660 40.320140 199.999927 5885.3258 vessel problem

Methods Best Mean Worst SD

hHHO-SCA (Kamboj et al., 2019) 6393.09 N/A N/A N/A

GA (Deb, 1991) 6059.95 6177.25 6469.32 130.9297
CPSO (He and Wang, 2007) 6061.08 6147.13 6363.8 86.4545
ESs (Mezura-Montes and Coello, 2008) 6059.75 6850 7332.88 426
CSS (Kaveh and Talatahari, 2010b) 6059.09 6067.91 6085.48 10.2564
CDE (Huang et al., 2007) 6059.73 6085.23 6371.05 43
IACO (Kaveh and Talatahari, 2010a) 6059.7258 6081.7812 6150.1289 67.2418
CE-CBA (Xiao et al., 2019) 6059.7143 6099.9218 6336.3404 104.25721
WOA (Mirjalili and Lewis, 2016) 6059.7410 6068.05 N/A 65.6516
IGMM (Varaee and Ghasemi, 2017) 6059.7143 6060.1598 6061.2868 0.5421
HPSODEPSR (Ali and Tawhid, 2016) 6059.7143 6059.7143 6059.7143 4.54E-12
HCS-LSAL (Long et al., 2014) 6059.7143 6087.3225 6317.4069 2.21E-02
Table VIII.
EBA (Yılmaz and Küçüksille, 2015) 6059.71 6173.67 6370.77 142.33
MCSS (Kaveh et al., 2013) 6058.623 6073.5931 6108.4579 24.6712 Statistical results of
LWOA (Zhou et al., 2018) 5893.339 6223.765 7070.343 418.7902 different methods for
TEO (Kaveh and Dadras, 2017) 5887.511 5942.565 6134.188 62.221 the pressure vessel
SSOA 5885.3258 5885.4415 5885.9368 0.178 problem

Problem !0 b0 b max No. of herds Size of herds Maximum iteration no.

Table IX.
72-bar spatial truss 0.5 2.3 2.6 4 5 549
200-bar planar truss 0.6 2.3 2.8 3 5 2,000 Parameters setting
272-bar transmission tower 0.5 2.0 2.3 4 5 700 for the SSOA for
582-bar tower truss 0.6 2.0 2.3 4 5 299 truss optimization
1016-bar double-layer grid 0.5 2.3 2.7 4 5 600 problems
EC 2.3 Steps of the optimization algorithm
37,7 The flowchart of the shuffled shepherd optimization algorithm (SSOA) is illustrated in
Figure 2 and the steps are as follows:
Step 1: Initialization
The SSOA parameters are set, and the initial position of the ith sheep is determined
randomly in an m-dimensional search space by the following equation:
2368 x0i ¼ xmin þ rand ð xmax  xmin Þ i ¼ 1; 2; . . . ; n (5)

where x0i is the initial solution vector of the ith sheep, xmax and xmin are the bound of
design variables, rand is a random vector with each component being in the interval
[0,1] and the number of components are equal to the number of variables; n is the
number of sheep.

Figure 5.
Schematic of the 72-
bar spatial truss

Case Node Fx kips Fy kips Fz kips

1 17 0.0 0.0 5.0

Table X. 18 0.0 0.0 5.0
Loading condition for 19 0.0 0.0 5.0
the 72-bar spatial 20 0.0 0.0 5.0
truss 2 17 5.0 5.0 5.0
 Optimal cross-sectional areas (in2)
BB-BC
Element GA (Erbatur ACO (Camp and PSO (Perez and (Camp, RO (Kaveh and CBO (Kaveh and Ilchi ECBO (Kaveh and Ilchi Present
group et al., 2000) Bichon, 2004) Behdinan, 2007) 2007) Khayatazad, 2012) Ghazaan, 2014) Ghazaan, 2014) work (SSOA)

A1–A4 1.755 1.948 1.7427 1.8577 1.8365 1.9170 1.8519 1.8823

A5–A12 0.505 0.508 0.5185 0.5059 0.5021 0.5031 0.5141 0.5126
A13–A16 0.105 0.101 0.1000 0.1000 0.1000 0.1000 0.1000 0.1003
A17–A18 0.155 0.102 0.1000 0.1000 0.1004 0.1001 0.1000 0.1001
A19–A22 1.155 1.303 1.3079 1.2476 1.2522 1.2721 1.2819 1.2442
A23–A30 0.585 0.511 0.5193 0.5269 0.5033 0.5050 0.5091 0.5161
A31–A34 0.100 0.101 0.1000 0.1000 0.1002 0.1000 0.1000 0.1000
A35–A36 0.100 0.100 0.1000 0.1012 0.1001 0.1000 0.1000 0.1001
A37–A40 0.460 0.561 0.5142 0.5209 0.5730 0.5184 0.5312 0.5312
A41–A48 0.530 0.492 0.5464 0.5172 0.5499 0.5362 0.5173 0.5166
A49–A52 0.120 0.100 0.1000 0.1004 0.1004 0.1000 0.1000 0.1000
A53–A54 0.165 0.107 0.1095 0.1005 0.1001 0.1000 0.1000 0.1006
A55–A58 0.155 0.156 0.1615 0.1565 0.1576 0.1569 0.1560 0.1562
A59–A66 0.535 0.550 0.5092 0.5507 0.5222 0.5374 0.5572 0.5486
A67–A70 0.480 0.390 0.4967 0.3922 0.4356 0.4062 0.4259 0.4027
A71–A72 0.520 0.592 0.5619 0.5922 0.5971 0.5741 0.5271 0.5726
Best weight
(lb) 385.76 380.24 381.91 379.85 380.458 379.75 379.77 379.6699
Average
weight (lb) N/A 383.16 N/A 382.08 382.533 380.03 380.39 379.9030
Std (lb) N/A 3.66 N/A 1.912 1.221 0.278 0.8099 0.1150
No. of
analyses N/A 18,500 N/A 19,621 19,084 16,000 18,000 11,000

for the 72-bar spatial

shepherd

Optimization results
2369
Shuffled

truss
Table XI.
method
optimization
EC Step 2: Evaluations
37,7 The value of the objective function for each sheep is evaluated
Step 3: Build herds
The sheep are divided into herds by using the process illustrated in Section 2.2.1
Step 4: Calculate the step size
The step size is calculated for each sheep as illustrated in Section 2.2.2 using equation (1).
2370 Step 5: Calculate the temple solution vector
The temple solution vector is calculated using equation (4) and the objective function is
evaluated and referred to as temple solution vector.
Step 6: Update the agent and merge
If the temple objective function is not worse than the old objective function then the
position of the sheep is updated and merged within the herds.
Step 7: Update the parameters
Update the values of ! and b using equations (2) and (3).
Step 8: Termination condition
Steps 3 to 7 are repeated until the specified maximum number of iterations is reached.

3. Validation of the SSOA

To verify the efficiency of the new algorithm, 17 mathematical benchmark problems and
two classic engineering design problems, four truss structures and double-layer grid are
optimized using the SSOA and compared to some other algorithms. Mathematical problems
are investigated in Section 3.1 and engineering problems investigated in Section 3.2. Trusses
and double-layer grid are optimized in Section 4.

3.1 Mathematical optimization problems

Mathematical problem chosen from Ref. (Tsoulos, 2008) are shown in Table I. Like any other
meta-heuristics algorithm, to have good performance and increase converge speed of algorithm
in a least computational cost, the number of agent and maximum iteration number should have

Figure 6.
Convergence histories
for the 72-bar planar
truss problem
balance to find the best computational cost for increasing converge speed of algorithm. Here the Shuffled
number of herd and size of each herd is set to 4 and the maximum number of the permitted shepherd
iterations is considered as 200. The value of a0 is considered as 0.5, b 0 is set to 2.4 and b max is
taken as 2.8. However, for Rosenbrock the values of b 0 and b max are set to 2 and 3.5,
optimization
respectively. The result of the optimization by some variants of GA (Tsoulos, 2008), CPA (Kaveh method
and Zolghadr, 2017), CSS (Kaveh and Talatahari, 2010b) and present work are compared in
2371

Figure 7.
Schematic of the 200-
bar planer truss
EC Table II. Each mathematical function is optimized 50 times independently using SSOA and the
37,7 average numbers of function evaluation are presented in Table II. The numbers in
the parentheses indicate the ratio of the successful runs in which the algorithm has located the
global minimum with predefined accuracy, which is taken as « = fmin – fmax = 104. The
absence of the parentheses means that the algorithm has been successful in all independent
runs.
2372 It can be seen from Table II that SSOA has generally performed better than variants of
GA, CPS, CSS and has better performance than CPS and variants of GA in each function, has
better performance than CSS in each function except BL and Rosenbrock

3.2 Engineering optimization problems

In this section, performance of the SSOA is examined by two classic engineering
problems and penalty approach is used for constraint handling, Maximum iteration
number is limited to 400 for the welded beam design and limited to 1,000 for the
pressure vessel design. Sensitivity analysis was performed to find the best combination
of the parameters of the SSOA. The values of a0 is taken as 0.5, 1 and 1.5, b 0 is varied
between 1.5 and 3.5 and b max is altered between 2 and 4 with increments of 0.5.
Therefore 45 possible combinations of parameters are created. For statistical
compression average and standard deviation are calculated for 30 independent runs of
the SSOA for each problem.

Element group Members in the group

1 1,2,3,4
2 5,8,11,14,17
3 19,20,21,22,23,24
4 18,25,56,63,94,101,132,139,170,177
5 26,29,32,35,38
6 6,7,9,10,12,13,15,16,27,28,30,31,33,34,36,37
7 39,40,41,42
8 43,46,49,52,55
9 57,58,59,60,61,62
10 64,67,70,73,76
11 44,45,47,48,50,51,53,54,65,66,68,69,71,72,74,75
12 77,78,79,80
13 81,84,87,90,93
14 95,96,97,98,99,100
15 102,105,108,111,114
16 82,83,85,86,88,89,91,92,103,104,106,107,109,110,112,113
17 115,116,117,118
18 119,122,125,128,131
19 133,134,135,136,137,138
20 140,143,146,149,152
21 120,121,123,124,126,127,129,130,141,142,144,145,147,148,150,151
22 153,154,155,156
23 157,160,163,166,169
24 171,172,173,174,175,176
25 178,181,184,187,190
Table XII. 26 158,159,161,162,164,165,167,168,179,180,182,183,185,186,188,189
Element grouping for 27 191,192,193,194
the 200-bar planer 28 195,197,198,200
truss 29 196,199
3.2.1 The welded beam design. The first classic engineering problem considers the Shuffled
design optimization of the welded beam shown in Figure 3. The aim of this problem is shepherd
to find the minimum constructing cost of the welded beam subjected to constraints on
shear stress (s), bending stress (r), buckling load (Pc), deflection (d) and
optimization
side constraints. The design variables are the thickness of the weld h(=x 1 ), method
length of attached part of the bar l(=x2 ), the height of the bar t(=x 3 ) and thickness of
the bar b(=x 4 ).
The mathematical formulation of the optimization problem is as follows: 2373
fcost ðxÞ ¼ 1:10471x21 x2 þ 0:04811x3 x4 ð14:0 þ x2 Þ

To be minimized and constraints are:

g1 ðxÞ ¼ t ðxÞ  t max # 0
g2 ðxÞ ¼ s ðxÞ  s max # 0
g3 ðxÞ ¼ x1  x4 # 0

g4 ðxÞ ¼ 1:10471x21 x2 þ 0:04811x3 x4 ð14:0 þ x2 Þ  5:0 # 0

g5 ðxÞ ¼ 0:125  x1 # 0
g6 ðxÞ ¼ d ðxÞ  d max # 0
g7 ðxÞ ¼ P  Pc # 0

Where:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 00 2
t ðxÞ ¼ ðt 0 Þ þ 2t 0 t 00 þ ðt Þ
2
2R
P MR
t 0 ¼ pffiffiffi ; t 00 ¼
2x1 x2 J
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2


x2 x22 x1 þ x3
M ¼P Lþ ;R¼ þ
2 4 2
( "
2 #)
pffiffiffi x22 x1 þ x3
J ¼2 2x1 x2 þ
12 2
6PL 4 PL3
s ð xÞ ¼ 2
d ð xÞ ¼ 3
x4 x3 Ex3 x4
rffiffiffiffiffiffiffiffiffi
4:013E
x23 x64 rffiffiffiffiffiffiffi !
36 x3 E
Pc ðxÞ ¼ 1
L2 2L 4G
P ¼ 6; 000 lb L ¼ 14in
E ¼ 30  106 psi G ¼ 12  106 psi
EC Variable boundaries are:
37,7 0:1 # x1 # 2:0
0:1 # x2 # 10
0:1 # x3 # 10
0:1 # x4 # 2:0
2374
Table III shows that in almost all combinations of parameters, the SSOA finds near optimal
solution but for finding the optimal solution, the parameter tuning plays an essential role. In
seven parameters’ combination, SSOA finds optimal solution but the worst solution is
different and this shows which parameters’ combination is appropriate for this problem. In
parameter combination number 36 (a0 = 1.5, b 0 = 2, b max = 2.5) the worst solution is

Optimal cross-sectional area (in2)

WEO
Element CMLPSA SAHS TLBO (Degertekin and (Kaveh and CSP (Kaveh Present work
group (Lamberti, 2008) (Degertekin, 2012) Hayalioglu, 2013) Bakhshpoori, 2016) et al., 2014b) (SSOA)

1 0.1468 0.1540 0.1460 0.1144 0.1480 0.100315

2 0.9400 0.9410 0.9410 0.9443 0.9460 0.945976
3 0.1000 0.1000 0.1000 0.1310 0.1010 0.101669
4 0.1000 0.1000 0.1010 0.1016 0.1010 0.101119
5 1.9400 1.9420 1.9410 2.0353 1.9461 1.948657
6 0.2962 0.3010 0.2960 0.3126 0.2979 0.290019
7 0.1000 0.1000 0.1000 0.1679 0.1010 0.101494
8 3.1042 3.1080 3.1210 3.1541 3.1072 3.113037
9 0.1000 0.1000 0.1000 0.1003 0.1010 0.100209
10 4.1042 4.1060 4.1730 4.1005 4.1062 4.118997
11 0.4034 0.4090 0.4010 0.4350 0.4049 0.407309
12 0.1912 0.1910 0.1810 0.1148 0.1944 0.100247
13 5.4284 5.4280 5.4230 5.3823 5.4299 5.406352
14 0.1000 0.1000 0.1000 0.1607 0.1010 0.109823
15 6.4284 6.4270 6.4220 6.4152 6.4299 6.406481
16 0.5734 0.5810 0.5710 0.5629 0.5755 0.470731
17 0.1327 0.1510 0.1560 0.4010 0.1349 0.430735
18 7.9717 7.9730 7.9580 7.9735 7.9747 7.968047
19 0.1000 0.1000 0.1000 0.1092 0.1010 0.119819
20 8.9717 8.9740 8.9580 9.0155 8.9747 8.974186
21 0.7049 0.7190 0.7200 0.8628 0.70648 0.888874
22 0.4196 0.4220 0.4780 0.2220 0.4225 0.226645
23 10.8636 10.8920 10.8970 11.0254 10.8685 11.14610
24 0.1000 0.1000 0.1000 0.1397 0.1010 0.221872
25 11.8606 11.8870 11.8970 12.0340 11.8684 12.14581
26 1.0339 1.0400 1.0800 1.0043 1.0340 1.096515
27 6.6818 6.6460 6.4620 6.5762 6.6859 5.72775
28 10.8113 10.8040 10.7990 10.7265 10.8111 10.35575
29 13.8404 13.8700 13.9220 13.9666 13.8465 14.18211
Best weight
(lb) 25445.63 25491.9 25488.15 25674.83 25476.9 25291.024
Average
Table XIII. weight (lb) N/A 25610.2 25533.14 26613.45 25547.6 25763.978
Optimization results Std (lb) N/A 141.85 27.44 702.80 135.09 270.444
No. of
for the 200-bar planer structural
truss Analyses 9650 19,670 28,059 19410 31700 30015
28,000 Shuffled
Best optimized design shepherd
Average optimized design
27,500 optimization
method
27,000

2375
26,500

26,000

25,500

25,000

24,500
Figure 8.
Convergence histories
24,000
200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 for the 200-bar planar
truss problem
Iteration

Figure 9.
Schematic of the 272-
bar transmission
tower
EC 1.724871, which is very close to the optimum value (1.724852) and this indicated that this
37,7 parameter combination is suitable for this problem. It can be seen form Table IV. SSOA
found minimum weight and constraints are g1(x) = –9.048E – 07, g2(x) = –6.979E – 04,
g3(x) = –1.779E – 08, g4(x) = –3.433, g5(x) = –0.801, g6(x) = –0.236 and g7(x) = –0.001, this
indicates that g1(x), g2(x) and g3(x) have controllers’ role. Table V shows that the SSOA has
minimum average. The average of SSOA is less than the best solution of other method
2376 except MCSS (Kaveh et al., 2013) and IGMM (Varaee and Ghasemi, 2017).
3.2.2 Pressure vessel design. The second engineering problem considers the optimization
of the pressure vessel shown in Figure 4. The aim of this problem is to find the minimum
constructing cost of the pressure vessel. The design variables are the thickness of the sell Ts
(=x1), thickness of the head Th (=x2), the inner radius R (=x3) and length of cylindrical
section of vessel L (=x4).

Nodes
Case Force direction 1 2 11 20 29 Other free nodes

1 Fx (kN) 20 20 20 20 20 5
Fy (kN) 20 20 20 20 20 5
Fz (kN) 40 40 40 40 40 0
2 Fx (kN) 0 20 20 20 20 5
Fy (kN) 0 20 20 20 20 5
Fz (kN) 0 40 40 40 40 0
3 Fx (kN) 20 0 20 20 20 5
Fy (kN) 20 0 20 20 20 5
Fz (kN) 40 0 40 40 40 0
4 Fx (kN) 20 20 20 0 20 5
Fy (kN) 20 20 20 0 20 5
Fz (kN) 40 40 40 0 40 0
5 Fx (kN) 20 0 0 0 0 5
Fy (kN) 20 0 0 0 0 5
Fz (kN) 40 0 0 0 0 0
6 Fx (kN) 0 20 0 0 0 5
Fy (kN) 0 20 0 0 0 5
Fz (kN) 0 40 0 0 0 0
7 Fx (kN) 0 0 0 20 0 5
Fy (kN) 0 0 0 20 0 5
Fz (kN) 0 0 0 40 0 0
8 Fx (kN) 0 0 20 20 20 5
Fy (kN) 0 0 20 20 20 5
Fz (kN) 0 0 40 40 40 0
9 Fx (kN) 0 20 20 0 20 5
Fy (kN) 0 20 20 0 20 5
Fz (kN) 0 40 40 0 40 0
10 Fx (kN) 0 0 20 0 20 5
Fy (kN) 0 0 20 0 20 5
Fz (kN) 0 0 40 0 40 0
11 Fx (kN) 0 0 0 20 20 5
Fy (kN) 0 0 0 20 20 5
Table XIV. Fz (kN) 0 0 0 40 40 0
Loading condition for 12 Fx (kN) 0 0 20 20 0 5
the 272-bar Fy (kN) 0 0 20 20 0 5
transmission tower Fz (kN) 0 0 40 40 0 0
The mathematical formulation of the optimization problem is as following Shuffled
shepherd
optimization
fcost ðxÞ ¼ 0:6224x1 x3 x4 þ 1:7781x2 x23 þ 3:1661x21 x4 þ 19:84x21 x3
method

To be minimized and the constraints are: 2377

g1 ðxÞ ¼ x1 þ 0:0193x3 # 0
g2 ðxÞ ¼ x2 þ 0:00954x3 # 0
4
g3 ðxÞ ¼ p x23 x4  p x33 þ 1; 296; 000 # 0
3
g4 ðxÞ ¼ x4  240 # 0

Group no. Cross section area (mm2)

1 1,000.551
2 1,240.013
3 2,491.871
4 1,017.829
5 9,618.809
6 1,000.000
7 12,063.816
8 1,001.777
9 1,000.188
10 1,000.457
11 10,217.022
12 1,000.064
13 1,000.015
14 1,000.005
15 9,320.549
16 1,000.028
17 1,000.307
18 1,002.518
19 8,389.809
20 1,000.814
21 1,000.004
22 1,003.288
23 7,982.259
24 1,000.445
25 1,000.591
26 1,000.053
27 7,504.298
28 1,000.076 Table XV.
Volume (cm3) 1,168,200.624 Optimum results for
Average volume (cm3) 1,168,668.715 the 272-bar
Std (cm3) 310.7557 transmission tower
EC Variable boundaries are
37,7 0: # x1 # 99
0 # x2 # 99
10 # x3 # 200
2378 10 # x1 # 200
Table VI shows that in almost all parameter combinations, the SSOA can find near optimal
solution and in seven combinations, SSOA can find optimal solution but in combination
number 26 (a0 = 1, b 0 = 2.5, b max = 3.5), the difference between the worst and the best
solution is equal to 0.611, indicating that this combination is the best parameter for this
problem, reducing a0 and increasing b 0 value compared to the welded beam problem
indicate that the pressure vessel problem needs fewer exploration and more exploitation
compared to the welded beam problem.
Table VII compares the results and Table VIII compares the statistic results of the
present work with other optimization algorithms. Table VII indicates that the SSOA finds
best results than other considered algorithm and constraints are g1(x) = –2.65E – 07, g2(x) =

Figure 10.
Compression of
allowable and
existing
displacements for the
272-bar transmission
tower
–5.87E – 06, g3(x) = –33.33 and g4(x) = –40.00, this indicates that g1(x) and g2(x) have Shuffled
controllered optimization progress more than other constraints. Table VIII show that the shepherd
SSOA has the least average and standard deviation in comparison to the other algorithms.
optimization
4. Numerical examples
method
In this section, five numerical examples are provided to examine the performance of the
SSOA on the weight minimization of two truss structures and double-layer grid and the 2379
volume minimization of two truss structure. The results are compared to other optimization
techniques. All numerical examples are run 30 times independently to provide statistically
meaningful results. Parameter settings of the SSOA and the number of iteration limits on
numeric examples are listed in Table IX.

4.1 A 72-bar spatial truss

For the 72-bar spatial truss structure shown in Figure 5, material density is 0.1 lb/in3
and the modulus of elasticity is 10,000 ksi. The members are subjected to the stress
limit of 6 25 ksi. The structure is subjected to two load cases listed in Table X. The
nodes are subjected to the displacement limits of 60.25 in, the element is divided into 16
groups shown in Table XI.
Comparison of the optimal design by this work with other considered optimization
algorithm is given in Table XI. It can be seen that SSOA found a structure of weight
(379.6699) that is the best solution found. It has the least average and standard deviation
than other solution obtained by other considered algorithms, and the SSOA converges the
best solution with less analyses number (11,000) among the others Figure 6 shows the
convergence histories of the best result and the mean performance of 30 independent runs
for the 72-bar truss.

4.2 A 200-bar planar truss

The second structural optimization problem solved is the optimal design of the 200-
bar planer truss as shown in Figure 7. All members are made of steel: the material
density is 0.283 lb/in2 and modulus of elasticity is 30,000 ksi. The allowable stress for
all members is 610 ksi. No displacement constraints are included in the optimization

1,200,000

1,195,000
Penalizwd Volume (cm 3)

1,190,000

1,185,000

1,180,000

1,175,000
Figure 11.
1,170,000 Convergence histories
of the optimization
for the 272-bar
transmisson tower
EC process. The structure is divided into 29 groups of elements as shown in Table XII.
37,7 The minimum cross-sectional area of all members is 0.1 in2 and the maximum cross-
sectional area is 20 in 2 . This truss is subjected to three independent loading
conditions: (1) 1.0 kips acting in positive x-direction at node 1,6,15,20,29,43,48,57,62
and 71; (2) 10.0 kips acting in the negative y-direction at node 1, 2, 3, 4, 5, 6, 8, 10, 12,
14, 15, 16, 17, 18, 19, 20, 22, 24, . . ., 71, 72, 73, 74 and 75; (3) loading conditions 1 and 2
2380 acting together.
Comparison of the optimal design by this work with CMLPSA (Lamberti, 2008), SAHS
(Degertekin, 2012), TLBO (Degertekin and Hayalioglu, 2013), HPSSO (Kaveh et al., 2014a),
WEO (Kaveh and Bakhshpoori, 2016) and CSP (Kaveh et al., 2014b) is provided in
Table XIII. It can be seen that SSO find the lightest weight (25,291.024 lb) after 2,000
iteration (30,015 analyses), with standard deviation being 270.444. Figure 8 shows the
convergence curves for the best result and the mean performance of 30 independent runs for
the 200-bar truss.

4.3 A 272-bar transmission tower

The 272-bar transmission tower as shown in Figure 9 presented by Kaveh and
Massoudi (Kaveh and Massoudi, 2014). All nodal coordinate and end nodes of the
member are presented in Ref. (Kaveh and Massoudi, 2014). Members are divided into 28
groups because of symmetry as shown in Figure 9. In this example we imposed 11 load

Figure 12.
Schematic of a 582-
bar tower truss. (a)
Three-dimensional
view, (b) side view, (c)
top view
cases for basis load case as shown in Table XIV. The displacement of nodes 1,2,11,20,29 Shuffled
in Z-direction is limited to 20 mm and in x- and y- directions is limited to 100 mm. the shepherd
modulus of elasticity is 2  108 kN/m2 and the maximum available stresses for all
member is 6275000 kN/m2.
optimization
Optimum volume found by the SSOA is presented in Table XV. SSOA finds optimum method
volume after 14,020 analyses. Maximum stresses ratio is 0.7639, which has happened in load
case 10 in element 169 and maximum displacement is 20 mm for node 11 in z-direction in
load case 3. Displacement for nodes 1,2,11,20,29 are shown in Figures 10. Figure 11 shows 2381
the convergence curves for the best result and the mean performance of 30 independent runs
for the 272-bar transmission tower.

Optimal cross-sectional areas (cm2)

Element CBO (Kaveh and LCA (Jalili et al., LCA-Tie-I (Jalili LCA-Tie-II (Jalili Present work
group Mahdavi, 2014) 2016) et al., 2016) et al., 2016) (SSOA)

1 20.5526 24.79452 20.51882 20.07259 20.1041

2 162.7709 136.33420 163.98951 158.77580 158.7277
3 24.8562 22.85682 21.81944 20.16866 23.6884
4 122.7462 128.99263 145.10098 133.27490 137.7912
5 21.6756 21.74672 20.42018 20.65500 20.0352
6 21.4751 23.17653 20.68205 20.18854 20.0467
7 110.8568 105.84231 106.27745 111.50684 109.4470
8 20.9355 20.56419 21.13842 20.69701 20.0159
9 23.1792 21.08227 20.28843 20.22639 20.0309
10 109.6085 103.21703 92.78366 99.29528 100.6825
11 21.2932 23.71312 20.58610 20.64242 20.0607
12 156.2254 158.23508 146.12202 146.59149 155.0364
13 159.3948 162.60176 152.83220 176.66602 162.7792
14 107.3678 121.80656 116.46250 123.13002 105.5719
15 171.915 154.71782 171.00815 171.51066 187.3000
16 31.5471 32.41569 30.84408 28.55170 26.5851
17 155.6601 118.86294 144.78013 135.42682 142.4558
18 21.4951 26.28786 24.46161 21.26313 24.3699
19 25.1163 23.85769 20.39790 20.55001 20.2663
20 94.0228 126.08487 102.74431 90.43625 105.5655
21 20.8041 30.86815 20.79407 20.50218 20.0115
22 21.223 21.01600 21.83477 20.42497 20.4135
23 53.5946 81.51333 55.03878 55.28304 58.4523
24 20.628 20.89293 23.30772 20.33347 20.0043
25 21.5057 28.42340 23.50969 20.90459 20.0142
26 26.2735 20.54591 27.99494 24.41335 26.7862
27 20.6069 30.38560 20.23193 22.26166 20.0566
28 21.5076 37.23686 22.54473 22.12556 20.4116
29 24.1394 24.63037 21.97166 22.21663 21.3106
30 20.2735 22.72973 20.70026 20.68582 20.0539
31 21.1888 32.65086 21.37245 21.59396 20.0056
32 29.6669 21.04152 20.88096 20.02645 20.2027
Volume
(m3) 16.152 16.80 16.06 15.98 15.9179 Table XVI.
Mean (m3) N/A 16.93 16.23 16.11 15.9888
Optimization results
Std (m3) N/A 0.19 0.18 0.18 0.0752
No. of for the 582-bar truss
analyses 20,000 19,000 18,400 18,000 6,000 tower
EC 4.4 A 582-Bar tower truss
37,7 The 582-bar tower truss is schematized in Figure 12, first time is optimized with continuous
sizing variables by Kaveh and Mahdavi (Kaveh and Mahdavi, 2014). The members are
categorized as 32 independent sized variables because of structural symmetry. A single load
case is considered consisting of 5.0 kN applied in both x- and y-directions and 30 kN
applied in the z-direction at all free nodes of the tower. The minimum cross-sectional area of
2382 all members is 20 cm2 and maximum cross-sectional area is 1,000 cm2. The modulus of
elasticity is 203,893.6 MPa, yield stress is 253.1 MPA. Displacement limitation of 8 cm was
imposed on all nodes in the all direction; the slenderness ratio is limited to 300 for tension
member and to 200 for compression member and maximum stress of truss elements are
imposed according to the provisions of AISC ASD (AISC, A., 1989) as follows:
Maximum stresses of tension member

sþ
i ¼ 0:6Fy

Where Fy is the specified minimum yield stress.

Figure 13.
Convergence histories
of the optimization
for the 582-bar tower
truss

Figure 14.
Schematic of the
1016-bar double-layer
grid
 Maximum stresses of compression member Shuffled
8" ! # ! shepherd
>
> 2 3
>
> l 5 3 l i l optimization
>
> 1  2 Fy = þ
i
 i
for l i < Cc
< 2Cc 3 8Cc 8Cc3 method
si
>
>
>
> 12p 2 E
>
> for l i  Cc 2383
: 23l 2
i

Where E is the modulus of elasticity; l i is the slender ratio (l i = kLi/rp

i); C c denotes the
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
slenderness ratio dividing the elastic and inelastic bulking regions (Cc ¼ 2p 2 E=Fy ); k is

Figure 15.
Top view of the 1016-
bar double-layer grid
and element groups
EC the member effective length factor taken as 1 for all members; Li is the member length; ri is
37,7 the radius of gyration.
Comparison of the optimal design by this work with CBO (Kaveh and Mahdavi,
2014) and LCA (Jalili et al., 2016) is provided in Table XVI. It can be seen that SSOA find
the lightest volume (15.9179) with the least number of analysis, average and standard
divination among the other. Maximum displacement is in the top node in x direction
2384 which is 7.999991 cm. Figure 13 shows the convergence histories of the best result and
the mean performance of 30 independent runs for the 582-bar tower truss

4.5 A 1,016-bar double-layer grid

Design optimization of a double-layer grid with configuration of a square on diagonal
grid as shown in Figure 14 is investigated in this section. A span is 40  40 m is and
height is equal to 3 m. all connections are assumed to be ball jointed. This grid has 1,016
members and 320 nodes and simple support conditions are used for the bottom layer at
the nodes demonstrated in Figure 15.a. Each top layer joint is subjected to a
concentrated vertical load of 30 kN. The elements are divided into 25 groups as shown
in Figure15, which are selected from a list of steel pipe sections in manual of steel
construction [(AISC), A. I. O. S. C., 2001] as shown in Table XVII. The modulus of
elasticity is 205 GPa, yield stress is 248.2 MPA and the material density is
7,833.413 kg/m3. Displacement limitation is 20/3 cm were imposed on all nodes in the
vertical direction and limitation on stress and stability of truss elements are imposed
according to the provisions of AISC 360-10 (AISC Committee, 2010) as follows.
Tension member constraint:

(
1t Fy Ag ; 1t ¼ 0:9
pu # pr ; pr ¼ min
1t Fu Ae ; 1t ¼ 0:75

where pu is the required strength; pr is the nominal axial strength; Ag is gross area of member; Ae
effective net area; Fy specified minimum yield stress and Fu is specified minimum tensile strength.

Figure 16.
Convergence histories
of the optimization
for the 1016-bar
double-layer grid
Compression member constraint: Shuffled
pu # 8pr ; pr ¼ 1c Fcr Ag ; 1c ¼ 0:9 shepherd
  sffiffiffiffiffi
>
> optimization
>
>
Fy KL E
< 0:658 e Fy ; r # 4:71 Fy
> method
F

Fcr ¼ sffiffiffiffiffi
>
> KL E
>
> 0:877 * Fe ; > 4:71
>
: 2385
r Fy
p 2E
Fe ¼  2
KL
r

No. Type Nominal diameter (in.) Area (cm2) Gyration radius (cm)
a
1 ST ½ 1.6129 0.662432
b
2 EST ½ 2.064512 0.635
3 ST ¾ 2.129028 0.846582
4 EST ¾ 2.774188 0.818896
5 ST 1 3.161284 1.066038
6 EST 1 4.129024 1.034542
7 ST 1¼ 4.322572 1.371346
8 ST 1½ 5.16128 1.582166
9 EST 1¼ 5.677408 1.331214
10 EST 1½ 6.903212 2.003806
11 ST 2 6.903212 1.53543
12 EST 2 9.548368 1.945132
13 ST 2½ 10.96772 2.41681
14 ST 3 14.387068 2.955798
15 EST 2½ 14.5161 2.346452
c
16 DEST 2 17.161256 1.782572
17 ST 3½ 17.290288 3.395726
18 EST 3 19.483832 2.882646
19 ST 4 20.451572 3.835908
20 EST 3½ 23.741888 3.318002
21 DEST 2½ 25.999948 2.143506
22 ST 5 27.74188 4.775454
23 EST 4 28.451556 3.749548
24 DEST 3 35.290252 2.65811
25 ST 6 35.999928 5.700014
26 EST 5 39.419276 4.675124
27 DEST 4 52.25796 3.490976
28 ST 8 54.19344 7.462012
29 EST 6 54.19344 5.577332
30 DEST 5 72.90308 4.379976
31 ST 10 76.77404 9.342628
32 EST 8 82.58048 7.309358
33 ST 12 94.19336 11.10361
34 DEST 6 100.64496 5.236464
35 EST 10 103.87076 9.216898
36 EST 12 123.87072 11.028934
37 DEST 8 137.41908 7.004812 Table XVII.
The steel pipe
Note: aST standard weight; bEST extra strong; and cDEST double-extra strong sections
EC where Fe is elastic bulking stress; Fcr is critical stress; E is the modulus of elasticity; L
37,7 laterally unbraced length of the member; r is radius of gyration and K is effective length
factor taken equal to 1.
Slenderness ratio constraints:

KL
# 200 ; for compression member
2386 r
KL
# 300 ; for tension member
r

Table XVIII presents the optimum designs obtained by CBO,ECBO, Kaveh and Ilchi
Ghazaan (2018) and present work .Table XVIII shows that SSOA has found a solution
after 12,020 numbers of analyses that is 0.82 per cent higher than the best solution
found but has the least average and standard deviation than other solutions obtained
by other considered algorithms. Figure16 shows the convergence histories of the best

Sections
CBO (Kaveh and Ilchi ECBO (Kaveh and Ilchi Present work
Element group Ghazaan, 2018) Ghazaan, 2018) (SSOA)

1 EST 5 EST 5 EST 5

2 DEST 3 EST 5 ST 5
3 ST 3 ½ ST 3 ST 4
4 ST 2 ½ ST 3 ½ EST 2 ½
5 ST 2 ½ ST 2 ½ ST 3 ½
6 ST 2 ST 2 EST 1 ½
7 ST 2 DEST 2 EST 1 ½
8 ST 2 ½ DEST 2 EST 1 ½
9 DEST 2 ½ EST 2 ST 4
10 DEST 2 ½ ST 6 DEST 2 ½
11 ST 1 ½ ST 2 ST 2 ½
12 DEST 5 EST 8 ST 10
13 EST 3 ½ EST 3 ½ EST 4
14 EST 3 ½ ST 5 ST 4
15 EST 4 ST 4 EST 4
16 ST 6 EST 5 ST 6
17 ST 5 ST 5 ST 5
18 EST 4 EST 5 EST 5
19 EST 5 EST 5 DEST 4
20 ST 8 ST 8 DEST 4
21 ST 6 ST 5 ST 6
22 ST 3 ST 3 ST 3 ½
23 EST 6 EST 2 ½ ST 3 ½
24 ST 3 ½ ST 5 ST 2 ½
Table XVIII. 25 EST 1 ½ ST 4 ST 3 ½
Results of Weight (kg) 74,849 67,839 68,398
Average optimized weight (kg) 79,422 73,042 72,084
optimization for the
Standard deviation on average
1016-bar double- weight (kg) 8,154 9,158 1,802
layer grid No. of structural analyses 9,760 15,760 12,020
result and the mean performance of 30 independent runs for the 1016-bar double-layer Shuffled
grid shepherd
optimization
5. Conclusions method
This study presents a new multi-community meta-heuristic algorithm called SSOA; this
algorithm is inspired by the behavior of a shepherd. For the analysis of exploration and
exploitation ability of the Algorithm 17, mathematical benchmark functions are analyzed 2387
and shown that the present algorithm works better than other considered algorithms.
To show the efficiency and robustness of the SSOA, two classic engineering
problems (i.e. design of welded beam and design of pressure vessel), four truss design
problems (i.e. design of 72-bar spatial truss, 200-bar planer truss, 272-bar transmission
tower and 582-bar tower truss) and double-layer grid design problem (i.e. design of
1016-bar double-layer grid) are solved by the SSOA. Results indicate that SSOA can
find optimal solution with less number of analyses for some problems as opposed to
other considered algorithms.

References
(AISC), A. I. O. S. C (2001), Manual for Steel Construction, Load and Resistance Factor Design, 3rd
Edition, American Institute of Steel Construction-AISC, Chicago.
AISC, A (1989), Manual of Steel Construction–Allowable Stress Design, American Institute of Steel
Construction (AISC), Chicago.
Ali, A.F. and Tawhid, M.A. (2016), “A hybrid PSO and DE algorithm for solving engineering
optimization problems”, Applied Mathematics and Information Sciences, Vol. 10 No. 2,
pp. 431-449.
AISC Committee (2010), Specification for Structural Steel Buildings (ANSI/AISC 360-10), American
Institute of Steel Construction, Chicago-IL.
Camp, C.V. (2007), “Design of space trusses using Big Bang–Big Crunch optimization”, Journal of
Structural Engineering, Vol. 133 No. 7, pp. 999-1008.
Camp, C.V. and Bichon, B.J. (2004), “Design of space trusses using ant colony optimization”, Journal of
Structural Engineering, Vol. 130 No. 5, pp. 741-751.
Coello Coello, C.A. (2000), “Constraint-handling using an evolutionary multiobjective optimization
technique”, Civil Engineering and Environmental Systems, Vol. 17 No. 4, pp. 319-346.
Coello, C.A.C. (2000), “Use of a self-adaptive penalty approach for engineering optimization problems”,
Computers in Industry, Vol. 41, pp. 113-127.
Coello, C.A.C. and Montes, E.M. (2002), “Constraint-handling in genetic algorithms through the use of
dominance-based tournament selection”, Advanced Engineering Informatics, Vol. 16,
pp. 193-203.
Deb, K. (1991), “Optimal design of a welded beam via genetic algorithms”, AIAA Journal, Vol. 29 No. 11,
pp. 2013-2015.
Degertekin, S.O. (2012), “Improved harmony search algorithms for sizing optimization of truss
structures”, Computers and Structures, Vols 92/93, pp. 229-241.
Degertekin, S.O. and Hayalioglu, M.S. (2013), “Sizing truss structures using teaching-learning-based
optimization”, Computers and Structures, Vol. 119, pp. 177-188.
Dorigo, M. and DI Caro, G. (1999), “Ant colony optimization: a new Meta-heuristic”, Proceedings of
the 1999 congress on evolutionary computation-CEC99 (Cat. No. 99TH8406), IEEE,
pp. 1470-1477.
EC Duan, Q., Gupta, V.K. and Sorooshian, S. (1993), “Shuffled complex evolution approach for effective and
efficient global minimization”, Journal of Optimization Theory and Applications, Vol. 76 No. 3,
37,7 pp. 501-521.
Erbatur, F., Hasançebi, O., Tütüncü, I. and KıLıÇ, H. (2000), “Optimal design of planar and
space structures with genetic algorithms”, Computers and Structures, Vol. 75,
pp. 209-224.
2388 Eusuff, M., Lansey, K. and Pasha, F. (2006), “Shuffled frog-leaping algorithm: a memetic Meta-heuristic
for discrete optimization”, Engineering Optimization, Vol. 38 No. 2, pp. 129-154.
He, Q. and Wang, L. (2007), “An effective co-evolutionary particle swarm optimization for constrained
engineering design problems”, Engineering Applications of Artificial Intelligence, Vol. 20 No. 1,
pp. 89-99.
Holland, J.H. (1992), “Genetic algorithms”, Scientific American, Vol. 267 No. 1, pp. 66-73.
Huang, F.Z., Wang, L. and He, Q. (2007), “An effective co-evolutionary differential evolution for
constrained optimization”, Applied Mathematics and Computation, Vol. 186 No. 1,
pp. 340-356.
Jalili, S., Kashan, A.H. and Hosseinzadeh, Y. (2016), “League championship algorithms for optimum
design of pin-jointed structures”, Journal of Computing in Civil Engineering, Vol. 31 No. 2,
pp. 04016048.
Kamboj, V.K. Nandi, A. Bhadoria, A. and Sehgal, S. (2019), “An intensify Harris Hawks
optimizer for numerical and engineering optimization problems”, Applied Soft
Computing, 106018.
Kaveh, A. (2017), Advances in Metaheuristic Algorithms for Optimal Design of Structures, 2nd edition,
Springer, Berlin.
Kaveh, A. and Bakhshpoori, T. (2016), “Water evaporation optimization: a novel physically inspired
optimization algorithm”, Computers and Structures, Vol. 167, pp. 69-85.
Kaveh, A. and Dadras, A. (2017), “A novel Meta-heuristic optimization algorithm: thermal exchange
optimization”, Advances in Engineering Software, Vol. 110, pp. 69-84.
Kaveh, A. and Ilchi Ghazaan, M. (2014), “Enhanced colliding bodies optimization for design
problems with continuous and discrete variables”, Advances in Engineering Software,
Vol. 77, pp. 66-75.
Kaveh, A. and Ilchi Ghazaan, M. (2018), Meta-Heuristic Algorithms for Optimal Design of Real-Size
Structures, Springer, Berlin.
Kaveh, A. and Khayatazad, M. (2012), “A new Meta-heuristic method: ray optimization”, Computers
and Structures, Vol. 112, pp. 283-294.
Kaveh, A. and Mahdavi, V. (2014), “Colliding bodies optimization method for optimum design of
truss structures with continuous variables”, Advances in Engineering Software, Vol. 70,
pp. 1-12.
Kaveh, A. and Massoudi, M. (2014), “Multi-objective optimization of structures using charged system
search”, Scientia Iranica, Vol. 21, pp. 1845.
Kaveh, A. and Talatahari, S. (2010a), “An improved ant colony optimization for constrained
engineering design problems”, Engineering Computations, Vol. 27 No. 1, pp. 155-182.
Kaveh, A. and Talatahari, S. (2010b), “A novel heuristic optimization method: charged system search”,
Acta Mechanica, Vol. 213 Nos 3/4, pp. 267-289.
Kaveh, A. and Zolghadr, A. (2017), “Cyclical parthenogenesis algorithm: a new Meta-heuristic
algorithm”,
Kaveh, A., Bakhshpoori, T. and Afshari, E. (2014a), “An efficient hybrid particle Swarm and
Swallow Swarm Optimization algorithm”, Computers and Structures, Vol. 143,
pp. 40-59.
Kaveh, A., Share, M.A.M. and Moslehi, M. (2013), “Magnetic charged system search: a new Meta- Shuffled
heuristic algorithm for optimization”, Acta Mechanica, Vol. 224 No. 1, pp. 85-107.
shepherd
Kaveh, A., Sheikholeslami, R., Talatahari, S. and Keshvari-Ilkhichi, M. (2014b), “Chaotic swarming of
particles: a new method for size optimization of truss structures”, Advances in Engineering optimization
Software, Vol. 67, pp. 136-147. method
Kennedy, J. and Eberhart, R. (1995), “Particle swarm optimization (PSO)”, Proc. IEEE International
Conference on Neural Networks, Perth, 1942-1948.
Lamberti, L. (2008), “An efficient simulated annealing algorithm for design optimization of truss
2389
structures”, Computers and Structures, Vol. 86, pp. 1936-1953.
Long, W., Liang, X., Huang, Y. and Chen, Y. (2014), “An effective hybrid cuckoo search algorithm for
constrained global optimization”, Neural Computing and Applications, Vol. 25 Nos 3/4,
pp. 911-926.
Mezura-Montes, E. and Coello, C.A.C. (2008), “An empirical study about the usefulness of evolution
strategies to solve constrained optimization problems”, International Journal of General Systems,
Vol. 37 No. 4, pp. 443-473.
Mirjalili, S. and Lewis, A. (2016), “The whale optimization algorithm”, Advances in Engineering
Software, Vol. 95, pp. 51-67.
Perez, R.L. and Behdinan, K. (2007), “Particle swarm approach for structural design optimization”,
Computers and Structures, Vol. 85, pp. 1579-1588.
Rakhshani, H. and Rahati, A. (2017), “Intelligent multiple search strategy cuckoo algorithm for
numerical and engineering optimization problems”, Arabian Journal for Science and
Engineering, Vol. 42 No. 2, pp. 567-593.
Reddy, A.S. and Vaisakh, K. (2013), “Shuffled differential evolution for large scale economic dispatch”,
Electric Power Systems Research, Vol. 96, pp. 237-245.
Tsoulos, I.G. (2008), “Modifications of real code genetic algorithm for global optimization”, Applied
Mathematics and Computation, Vol. 203 No. 2, pp. 598-607.
Varaee, H. and Ghasemi, M.R. (2017), “Engineering optimization based on ideal gas molecular
movement algorithm”, Engineering with Computers, Vol. 33 No. 1, pp. 71-93.
Wu, L., Liu, Q., Tian, X., Zhang, J. and Xiao, W. (2018), “A new improved fruit fly optimization
algorithm IAFOA and its application to solve engineering optimization problems”, Knowledge-
Based Systems, Vol. 144, pp. 153-173.
Xiao, W., Liu, Q., Zhang, L., Li, K. and Wu, L. (2019), “A novel chaotic bat algorithm based on
catfish effect for engineering optimization problems”, Engineering Computations, Vol. 36
No. 5.
YıLMAZ, S. and Küçüksille, E.U. (2015), “A new modification approach on bat algorithm for solving
optimization problems”, Applied Soft Computing, Vol. 28, pp. 259-275.
Zhou, Y., Ling, Y. and Luo, Q. (2018), “Lévy flight trajectory-based whale optimization
algorithm for engineering optimization”, Engineering Computations, Vol. 35 No. 7,
pp. 2406-2428.

Corresponding author
Ali Kaveh can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]