Structures 43 (2022) 752–766

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

Design of buckling restrained steel braces using application programming

interface between simulation and discrete optimization
Osman Tunca
Karamanoglu Mehmetbey University, Department of Civil Engineering, 70200 Karaman, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: In this study, buckling restrained braces (BRBs) are optimized through application programming interface be­
Buckling restrained steel braces tween simulation and discrete optimization. It is aimed to maximize the energy dissipation capacity of BRBs
Finite element analysis considering the American Institute of Steel Construction (ANSI/AISC 341). Unlike other studies, BRBs modeled in
Application programming interface
the finite element packaged software are directly linked to optimization algorithms. So, the geometric and
Metaheuristic algorithms
Black widow optimization algorithm
material nonlinearities are also considered. Harmony Search Algorithm (HSA), which simulates the improvisa­
Harmony search algorithm tion musician performances in finding pleasing harmony, and Black Widow Optimization Algorithm (BWOA),
which imitates inimitable paring attitude of black widow spiders, are taken as the optimizer tools of this study.
They encoded in Microsoft Visual Basic programming language. Initially, the algorithmic performances of the
HSA and BWOA are compared and evaluated on two benchmark structural engineering design problems. Af­
terward, two different shaped BRBs are modeled in a finite element analysis (FEA) based software, namely
ANSYS Workbench. Then, the obtained simulations are integrated with the HS and BWO algorithms throughout
the application programming interface without identification of complex objective function and design con­
straints in formulations. These are directly calculated by ANSYS Workbench over very simple formulas. So, the
attained optimum designs of BRBs are investigated with a new approach. Furthermore, in order to see the su­
preme algorithmic performances of the HSA and BWOA, all benchmark and BRB design problems are solved so-
called well-established conventional standard Genetic Algorithm (sGA). Eventually, the proposed novel design
methodology gives opportunity and eases to the designers since it provides convenience in solving complicated
problems having nonlinear objective function and design constraints that are tiresome to encode.

1. Introduction longitudinal deformations by showing similar behavior under tension

and compression [6,7]. A BRB element generally consists of three main
Buckling Restrained Brace (BRB) is a structural system used to ensure parts. These are core plate, steel tube surrounding the core plate, and
stability both in the new structures and in the strengthening of the old filling material as seen in Fig. 1. Here, the main load carrier element is
structures. The BRB has a similar load carrying capacity in tension and the core plate. It carries the axial force on the brace. The outer steel tube
compression under the effect of lateral forces. So, they display a and filling material create lateral support for the core plate. Thus, the
balanced hysteretic behavior compared to conventional steel braces. buckling of the brace is restrained. A frictionless surface is required
They also constitute effective elements of a structural system to reduce between the core element and the filling material in order for the system
plastic deformations in structures in duration of an earthquake [1-4]. to work correctly. So, a thin layer material is used or spacing out a gap to
The technology and material requirements of the construction of a BRB prevent shear force transfer and allow lateral deformation under pres­
are quite simple compared to the other passive energy absorption sys­ sure [3].
tems which are used to increase the seismic respond demands of the The design studies on BRBs consist of two main concepts as the
structures [5]. The preference of the BRBs due to their economy and ease studies on modeling together with the structure and the studies just as a
to be implemented in a construction compared to other damping de­ building element. Although it is a more accurate approach to evaluate
vices. Additionally, it increases the lateral stiffness of the structure like BRBs with the structure, it is quite compelling and costly to experiment
classical braces. But it does not buckle under compression, unlike the with it in the laboratory. Likewise, modeling the finite element analysis
traditional brace members. This is because it can perform large of the 3D buckling restrained braced frames (BRBFs) in Finite Element

E-mail address: [email protected].

https://doi.org/10.1016/j.istruc.2022.07.010
Received 23 February 2022; Received in revised form 2 July 2022; Accepted 5 July 2022
Available online 12 July 2022
2352-0124/© 2022 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
O. Tunca Structures 43 (2022) 752–766

directly computed by the finite element analysis (FEA) based software,

namely ANSYS Workbench [26]. In the FEA, the problem is divided to
finite number of sub-problems. Then, each sub-problem is solved, and
the obtained solutions are combined. The experimental data obtained in
many BRB studies are verified with FEA [27-29]. The main goal of the
BRB design is aimed to maximize the energy dissipation capacity. The
energy dissipation capacity is directly related to the area under the
hysteretic curve. Calculating the hysteretic curve of a BRB is quite
difficult via executing of homemade programming. Because such a
calculation includes the material nonlinearity, geometric nonlinearity,
and friction. Instead, these complex operations can straightly be out­
sourced by a FEA based ANSYS Workbench software. Thus, total energy
dissipation capacity of BRBs can be easily calculated with high accuracy.
However, the solution time may be costly, depending on the type of
handling design problem. Today, FEA packaged softwares are developed
Fig. 1. The general structure of BRBs. considerably with the developments in computer technology. Their
experience and knowledge are enriched day by day. Most of these offer
Analysis (FEA) packaged softwares throw a lot of computational loads an Application Programming Interface (API) to their users [30-32].
on the computers. To get rid of this computational load, researchers Thus, a user can connect its algorithm with FEA based softwares [26,30].
assume about many structural details such as connection points, stress There have been numerous studies conducted on new generations of
distribution and nonlinearity [8,9]. Instead, more detailed studies can BRBs [33-35]. Most of these are quite successful. But their developments
be carried out on an element basis, considering the recommended still continued. Thus, the innovative aspect of this current study can be
loading protocols by the regulations [10-12]. When the literature is specified as instead of developing a new type of BRB, the existing one is
examined, it is seen that three main methods, experimental, finite considered to be re-designed as a sizing design optimization problem.
element and optimization, are many times offered in the design studies Hence, the metaheuristic algorithms are needed to be used for this
of the BRBs. Among these, optimization is a numerical method aiming to complex design optimization purpose. So, in this study, initially the
find the most suitable result [13]. It consists of the objective function, design performances of the Harmony Search Algorithm (HSA) [36] and
design variables, and design constraints [14]. Optimization algorithms Black Widow Optimization Algorithm (BWOA) [37] are compared on
are effective tools to find solutions to various types of engineering design two type of structural benchmark design problems (Cantilever Beam
problems [15-18]. Among these, metaheuristics are developed by Design, Reinforced Concrete Beam Design Problem). Additionally, in
inspiring natural phenomena [17]. There is a lot of algorithms in this order to see the algorithmic performances of the abovementioned HS
field such as the firefly algorithm [19], rain optimization algorithm and BWO algorithms, the optimum designs obtained through them for
[20], water strider algorithm [21], tunicate swarm algorithm [22], etc. Cantilever Beam Design and Reinforced Concrete Beam Design bench­
Abedini et al. optimize on planar BRBFs via salp swarm algorithm (SSA) mark problems are compared and evaluated with those of obtained from
and enhanced colliding bodies optimization (ECBO) algorithm [5]. so-called standard Genetic Algorithm (sGA) [38] which is known as very
Rezazadeh et al. also minimize structural weight and seismic energy of a popular and powerful conventional nature-inspired metaheuristic
similar design problem [23]. The obtained results with these algorithms technique. Microsoft Visual Basic for Application v7.0 (MS VBA v7) [39]
may not be decided as global optimum due to their probabilistic char­ is utilized as the programming platform. After that, two different types
acteristics based on randomness [24]. They do not need any complex BRBs are modeled on the FEA based numerical simulation software, so-
derivative operation on the objective function [25]. This is the main called ANSYS Workbench [40]. Then, two different finite element
reason why metaheuristic methods are chosen to find solutions to models are executed with two metaheuristic optimization algorithms by
complex engineering design problems such as structural design opti­ using IronPython [41] script files. Hence, the sizing design optimization
mization problems as to be handled in this study. Because within the of the BRBs is acquired via two different algorithms and their algo­
scope of this study, the objective function and design constraints are rithmic design performances are compared and evaluated in detail.

Fig. 2. The design variables of BRBs.

753
O. Tunca Structures 43 (2022) 752–766

Fig. 3. The loading protocol.

The combining of the optimization procedure and the finite element respectively. Here, H1, H2, H3 represent horizontal distances, V1, V2, V3
method, which are often used separately in scientific studies, distin­ show vertical distances, ti and to indicate the thickness of the inner plate
guishes this study from other BRB design studies. The FEA based pack­ and the thickness of the outer box-section, respectively. All of them
aged softwares are updating itself day by day. Also, researchers are represent the design variables of the design optimization problem. The
developing more advanced metaheuristic techniques almost every design constraints should be taken into account for the results obtained
passing day. Besides, the computer technology is developing, and more to be reasonable. For this purpose, a penalty method implemented
powerful computers are produced as well. Namely, when this study is objective function as given in Eq. (1) is utilized to take into account the
evaluated together with all these developments, it becomes clear that slightly infeasible design and/or designs in terms of the total violated
the integration of the optimization procedure and FEA method through constraints [42,43]. So, the fitness of the solution is evaluated on the
application programming interface (API) functions is very important basis of the objective function and a penalty function is introduced for
and useful for optimum designs of structures and/or structural members. avoiding infeasible solutions.
The general structure of this study is as follows; in the first section, ( )2
the general concept, principal aim, and innovative aspect of the study
Ng
∑
ϕi = Obj(x) 1 + P Max(0; g) (1)
are described, in from the second to the fourth sections, the methodol­ n=1
ogy used in the study are examined, the optimum design examples and
their detailed explanations are comprised in the fifth section as well as Here, ϕi is the fitness value, Obj(x) is the objective function, gi is
the obtained results are given detail in this section. Finally, the discus­ penalized design constraints, Ng is the number of penalties, and P is the
sion of the results and concluding remarks of the study are given in the penalty parameter taken as 1.0. In the maximization problems, the ϕi
sixth and seventh sections, respectively. can be considered as 1/ϕi.
The BRB models are designed according to AISC 341-10 Seismic
2. Optimum discrete design optimization of buckling restrained Provisions [44]. Table C-K3.1 of AISC 341-10 is utilized as a loading
braces protocol of the BRB as given in Fig. 3.
In Fig. 3, Δby stands for the deformation of the BRB at the yield point,
In the main part of the study, it is aimed to obtain the optimum and Δbm represents the brace deformation corresponding to the design
design of two types Buckling Restrained Braces (BRBs). The main story drift of the building. Here, that the brace deformation corre­
objective of the optimum design of BRB is to maximize the energy sponding to the design story drift is four times the yield deformation is
dissipation capacity of these structural elements. This directly related to considered. Additionally, the cycles should continue up to 200Δby in­
hysteretic curve of the BRB. The mechanical properties of the materials elastic deformation without interruption [44]. The interruption can
used in the fabrication of BRB and the geometric nonlinearity affect the cause convergence error during the FEA process. Additionally, some
hysteretic curve. So, to accurately obtain the expression of such a design problems may occur, such as the core plate not fitting into the outer
problem is quite hard. However, the energy dissipation capacity and any tube. So, the Eq. (2) is encoded.
needed quantity of the BRB can be easily obtained from Finite Element {
Analysis (FEA). So, in this part of the current study, the objective Obji = The value computed by ANSYS ←if FEA is converged
Obji =
Obji = 1 J (joule) ←otherwise
function and design constraints are directly computed by ANSYS
Workbench via application programming interface. Therefore, neither (2)
the objective function nor the design constraints are needed to be The Eq. (2) is activated if any convergence error is received. It means
identified in intricate formulations in the optimization coding process of that if FEA analysis does not converge to any feasible value, the value of
handled design problem. In this way, the designers are supplied with not dissipated energy of BRB is assigned as 1.0 J (joule). Thus, the generated
to deal with tiresome identification and tedious solution of these main unpractical solution vector is eliminated in the solution of maximization
inputs owning nonlinear content during the encoding phase. problem, and the iterative process continue utilizing with only the re­
Here, two different types of BRB are considered as shown in Fig. 2. sults from converged solution vectors.
In Fig. 2, the quarter part of the diagonally shaped core plated The structural behavior restrictions should be followed in the BRB
buckling restrained brace (BRBd), parabolic shaped core plated buckling design. In AISC 341-10 [44], the ratio of the maximum compression
restrained brace (BRBp), and the cross-section of both are given force to the maximum tension force shall not be greater than 1.3 for each

754
O. Tunca Structures 43 (2022) 752–766

BWO algorithm enables to escape from local optimums through direct­

ing a lot of search agents towards global optimal. Namely, it can be said
that cannibalism phase of the BWO algorithm provides a promising
exploitation ability with better converge rate.
In this study, the performances of both algorithms are compared for
the first time on the structural design problem of this study that is
Cantilever Beam Design (CBD) problem. Additionally, the optimum
design of buckling restrained braces (BRBs) is also newly investigated
through application programming interface (API) mastery between FEA
based ANSYS Workbench and Microsoft Visual Basic programming
language utilizing IronPython functions as mentioned and described in
below.

3.1. Harmony search algorithm

Fig. 4. Relationship between music improvement and engineering optimiza­
tion [36]. Harmony Search Algorithm (HSA) is developed by Geem et al. in
2001 [36]. It is inspired by organizations of the musical orchestra during
loading cycles in conformity with Section K3 item 8 of AISC 341-10 the development of the algorithm. Here, each musical instrument is
specifications as stated in Eq. (3). Here, it is worthy to mention that if liked a design variable. Additionally, the harmony that an orchestra will
this design constraint is violated (obtained as greater than 1.3) for achieve with different experiences is associated with finding optimum
random assigned values to design variables among the limit bounds by results of the optimization algorithm. This relevance is given in Fig. 4.
algorithms, this selection is canceled and control is renewed by selecting HSA has five main steps.
new design variable values. Step 1: Initialize HSA parameters: Before starting the iteration phase,
some parameters specific to HSA should be defined. So, Harmony
Pcomi
gi = − 1.3 ≤ 0, i = 1, 2, 3, ⋯, 14 (3) memory size (HMS), harmony memory consideration rate (HMCR),
Pteni
pitch adjustment ratio (PAR), and cycling criteria are defined to the
In Eq. (3), Pcom and Pten represent the maximum force in compression optimization algorithm.
and tension which are occurred in each cycle of the loading. This means Step 2: Initialize harmony memory matrices: After the defining of the
that if each sequential compression and tension push–pull is called a algorithm parameters, harmony memory matrix (HMM) is generated by
loading cycle, the loading protocol consists of total 14 loading cycles. considering the number of design variables and HMS as shown in Eq. (4).
Thus, the g1 to g14 represent the design constraints calculated for each ⎡ ⎤
x1,1 x1,2 ... ... x1,n− 1 x1,n
loading cycle. Hence, the design of BRB bases on maximum dissipated ⎢ x2,1 ⎥
x2,2 ... ... x2,n− 1 x2,n
energy capacity can be ensured. To accomplish this, the optimally ⎢
⎢ ...
⎥
⎥
... ... ... ... ...
designed BRB is expected to exhibit similar behavior in compression and HMM = ⎢ ⎢ ⎥
⎥ (4)
⎢ ... ... ... ... ... ... ⎥
tension. Eq. (3) is considered to ensure that the optimally designed BRB ⎣ xHMS− 1,1 xHMS− 1,2 ... ... xHMS− 1,n− 1 xHMS− 1,n ⎦
complies with practice code provisions of AISC 341–10. xHMS,1 xHMS,2 ... ... xHMS,n− 1 xHMS,n

3. Optimization algorithms Here, each element of HMM is generated randomly (Eq. (5)).
xi,j = xmin + (xmax − xmin ) × Rnd (5)
As known well, optimization is a process to find the most useful
result and/or results for a design problem. This process includes certain In Eq. (5), xi,j represents any element of HMM, Rnd is a random
design constraints to be satisfied to attain particular objective(s). number between 1 and 0, xmax and xmin are respectively the maximum
Thence, this kind of design problems are named as discrete design and the minimum value of the design variables. The values of the
programming problems. The traditional gradient-based programming objective function are calculated for each row of HMM. So, the fitness
algorithms are not able to solve such problems as they are not effective values of these are unearthed. Then, all of the solution vectors are sorted
for complex design problems. Consequently, to attain design optimiza­ by considering these.
tion solutions of structural problems, it is required probabilistic based Step 3: Improvise New Harmony: This is the main stage of the HSA.
metaheuristic techniques instead of deterministic ones. This why the Firstly, a new random number (Rnd) is generated. If the randomly
metaheuristic methods are not required any direct relation between generated number is less than HMCR, the randomly selected value from
objective function(s) and design constraints. Since the probabilistic- the HMM is considered (Eq. (3)). Otherwise, a new value is calculated by
based metaheuristic techniques mimics the natural events (i.e., laws of using Eq. (6).
social culture and biology) these techniques are named as nature- (
inspired metaheuristic algorithms [45,46]. There are numerous meta­ xi =
xi ∈ {x1i , x2i , x3i , ..., xHMS
i }, if Rnd < HMCR
(6)
heuristic optimization methods. Because of their simplicity for imple­ xi israndomlygenerated otherwise.
menting and requiring relatively less algorithmic parameters, the
Secondly, Rnd is generated again. On the condition that PAR is bigger
Harmony Search Algorithm (HSA), the Black Widow Optimization Al­
than Rnd the pitch adjustment process is applied. If it is less than Rnd, no
gorithm (BWOA) and the standard Genetic Algorithm (sGA) are utilized
action is taken on the design variable (Eq. (7)).
as optimizer tools in this study. While the robustness of Harmony Search
(
Algorithm (HSA) is provided on design problems in many fields, the xi = xi ± Rnd × bw , if Rnd < PAR
xi = (7)
BWOA can be specified as one of the novel metaheuristic algorithms. In xi = xi otherwise.
nature, the black widows have cannibalistic behaviors. Mimicking this
natural phenomenon gives supremacy to BWO algorithm as a principal Here, the bw is the maximum amplitude value of the pitch adjustment
advantage in exploitation phase if it is compared to so-called conven­ process.
tional algorithms such as Genetic Algorithm (GA) and/or evolutionary Step 4: Update Harmony Memory: After the operating previous stage, a
algorithm (EA). Therefore, it is expected that BWOA show prepotent new solution vector is obtained. If obtained fitness value from the new
algorithmic performance to design search space effectively. Besides, the solution is better than the worst one of the current values, HMM is

755
O. Tunca Structures 43 (2022) 752–766

3.2. Black widow optimization algorithm

As a novel metaheuristic optimization algorithm, Black Widow

Optimization Algorithm (BWOA) is developed by inspiring the behavior
of black widows [37]. In BWOA, a group of black widow spiders is
considered as solution vectors. Just like the life cycle of spiders, BWOA is
designed to survive the most appropriate solutions. This consists of five
main steps as follows.
Step 1: Initialize Population: First of all, some parameters specific to
the algorithm are determined. Examples of these are the number of
population (Npop), the number of variables (Nvar), the maximum number
of iteration (Nite), rate of procreating (PP), rate of mutation (PM), and
rate of cannibalism (CR). After that, Pop matrix is generated randomly
(Eq. (8)).
⎡ ⎤
x1,1 x1,2 ... ... x1,Nvar − 1 x1,Nvar
⎢ x2,1 x2,2 ... ... x2,Nvar − 1 x2,Nvar ⎥
Fig. 5. The flowchart of the standard HSA. ⎢
⎢ ...
⎥
⎥
... ... ... ... ...
pop = ⎢⎢ ⎥
⎥ (8)
⎢ ... ... ... ... ... ... ⎥
⎣ xNpop − 1,1 xNpop − 1,2 ... ... xNpop − 1,Nvar − 1 xNpop − 1,Nvar ⎦
xNpop ,1 xNpop ,2 ... ... xNpop ,Nvar − 1 xNpop ,Nvar

Each row of the pop represents the solution vector as shown in Eq.
(9).
Widow = [ x1 x2 ... ... xNvar − 1 xNvar ] (9)
After the generation of the pop, the fitness value of each widow is
calculated, and pop is reorganized according to obtained fitness values.
Step 2: Pro-creating: The parental matrix (pop1) determined by the
population number by considering pro-creating rate is generated from
the best solutions of the (pop). To produce a new generation, parents are
selected randomly from the (pop1). Then, the following equation is
executed.
(
y1 = αx1 + (1 − α)x2
(10)
y2 = (1 − α)x1 + αx2

In Eq. (10), α is a random number that is selected between 0 and 1. x1

and x2 are the parents and, y1 and y2 are the children. In here, the same
parent cannot be selected more than once. So, the parental selection
process is repeated Nvar/2 times. Then, children and moms are used to
generating a new matrix which is sorted by considering their fitness
Fig. 6. The flowchart of the BWOA. values.
Step 3: Cannibalism: There are three types of cannibalism in BWOA.
These are the cannibalism between females and males in parents, the
sibling cannibalism and the mother cannibalism. Totally which the
number of survivors is determined by considering cannibalism rating.
All surviving solution vectors are saved in a pop2.
Step 4: Mutation: The number of mutant children is determined in this
stage. Then, randomly selected black widows, which represent the so­
lution vector, are transferred to pop3. After that, randomly selected two
solutions in each row of the pop3 are exchanged. Thus, mutant children
are obtained and, the fitness values of these are calculated. After the all
above stages, Updated Pop is determined by using Eq. (11).
Updated pop = pop2 + pop3 (11)
Step 5: Convergence: In this stage, a cyclic criterion is checked. If a
cyclic criterion is not satisfied, iteration continues between Step 2 to
Step 4 stages of BWOA. The flowchart of the BWOA is given in Fig. 6.
Fig. 7. The general scheme of API concept.
4. Integration of encoded metaheuristic algorithms with ANSYS
updated. In this way, HMM is improved. Workbench via application programming interface
Step 5: Check cycling criteria: As long as cycling criteria is not satisfied
the Step 3 and Step 4 phases of the HSA continue iteratively. The opti­ As mentioned before, the design variables are the inputs of the
mum solution vector is obtained at last. The flowchart of the HSA is optimization problem and the objective function and design constraints
given in Fig. 5. are the outputs. The algorithms operate only on input and output values,
not on the functions themselves. This paves the way for the creation of
objective functions with third-party software. Some packaged softwares

756
O. Tunca Structures 43 (2022) 752–766

• A new BRB design modeled in ANSYS Workbench analyzed for

attaining structural responds.
• By the medium of IronPython API script file, the structural responds
are retrieved back to HS and BWO algorithms for acquiring the
objective function and design constraints that are the output data.
• The obtained output data is utilized to calculate the penalized design
constraints and the fitness value.
• Then, the algorithm resumes its specific computational steps.

5. Problem definitions and results

In this study, two benchmark design problems (cantilever beam

design and reinforced concrete beam design) and two different buckling
restrained braces (BRBd and BRBp) are utilized to compare and evaluate
the performances of the HS and BWO algorithms. Also, all design ex­
amples are optimally designed with standard Genetic Algorithm, which
is one of the well-known conventional and powerful optimization al­
gorithms, in order to fairly compare and evaluate the algorithmic per­
formances of HS and BWO algorithms. To check encoded HS and BWO
optimization algorithms, initially two different benchmark design
problems are used to accuracy observation. So, the Cantilever Beam
Design (CBD) and Reinforced Concrete Beam Design (RCBD) problems
related to civil engineering field, are firstly considered for BWOA. After
the encoded algorithms are verified, each optimization algorithm is
connected to Finite Element Models (FEM) of two different Buckling
Restrained Braces (BRBs). The running of the FEMs of BRBs is a time
cost. Therefore, algorithm parameters of HSA and BWOA are chosen
considering minimizing the computation time as much as possible. At
this point, the control parameters of HSA and BWOA that are respec­
tively the HMS and Npop, which determine the number of rows of the
main algorithm matrices, come to the fore. The harmony memory size
(HMS) is taken as 10 as recommended in previous studies [36,49-52].
Accordingly, Duan et al. suggested that the HMS can be taken as the
number of design variables [53]. As BWOA is a relatively novel opti­
mization algorithm, it is used in limited numbers in literature. The
number of population (Npop) can be 20 in the studies of Fu et al. [54] and
Fig. 8. The flowchart of ANSYS-MS VBA integrated programming Sadeghi et al. [55], 15 in the study of Pena-Delgado [56]. In this current
methodology. study, since mainly the performance comparison of the HSA and BWOA
is considered, the HMS and Npop are taken as 10. In all of the design
examples, the same algorithm parameters are used for HSA and BWOA
Table 1 as given in Table 1.
The parameters of algorithms.
Algorithms Parameters
5.1. Benchmark design problem
HSA [57] HMS = 10, HMCR = 0.9, par = 0.3, Nite = 500
BWOA [37] Npop = 10, PP = 0.6, CR = 0.44, PM = 0.4, Nite = 500
To make sure parameters of optimization algorithms, two different
type engineering benchmark design problems are computed utilizing
used in engineering field allow users connection possibility to their in­ HSA and BWOA. Also, the optimum results obtained for those algo­
terfaces via specific functions. These are named application program­ rithms are compared with the optimal designs attained via sGA.
ming interface (API) functions. For example, by aid of API functions, a
very trendy structural analysis program, SAP2000 [47], works with 5.1.1. Cantilever beam design problem
many popular programming languages [26], and the FEA-based ANSYS As a first benchmark design problem, the cantilever beam design
can be run with a python-based API script file [48] as seen in Fig. 7. (CBD) consists of five types of box-section beams, which was originally
In this study, ANSYS Workbench v18.1 [40] is operated with the help developed by Chickermane and Gea [58], is initially solved by using
of the IronPythons script files [41] using MS VBA v7 [39]. The flowchart HSA, BWOA and sGA to show the accuracy of the proposed design
of ANSYS-MS VBA integrated programming methodology is depicted in optimization methodology. The static structure of CBD problem is
Fig. 8. shown in Fig. 9. The cantilever beam is fixed supported in one end, and a
The API concept conducted between metaheuristic HS and BWO al­ vertical point load is applied to another end.
gorithms and FEA-bases ANSYS Workbench is described below. The objective of the benchmark problem is to obtain a minimum
weighted design. In this problem, the height and width of each beam
• The metaheuristic HS and BWO algorithms, which are encoded in parts are considered equal, and thicknesses of these are constant (t = 2/
Microsoft Visual Basic v7 programming language, generate solution 3) [59]. The weight of this is briefly computed by using Eq. (12).
vector. Weight(x) = 0.0624(x1 + x2 + x3 + x4 + x5 ) (12)
• The solution vector is considered as input data. And, it is transferred
to the FEA-based ANSYS Workbench software via IronPython API Here, x1 to x5 represent the length of each part of the beam. There is an
script file. upper limit on the vertical displacement of the free end of the cantilever

757
O. Tunca Structures 43 (2022) 752–766

Fig. 9. Cantilever beam design problem.

Table 2
The final obtained optimal design results of CBD problem.
Algorithm x1 (m) x2 (m) x3 (m) x4 (m) x5 (m) min. weight (kg) g (constraints)

HSA 6.026 5.321 4.477 3.477 2.174 1.34004 − 5.00348E-05

BWOA 6.026 5.321 4.477 3.477 2.174 1.34004 − 5.00348E-05
sGA 6.14 5.2 4.528 3.414 2.219 1.34166 − 1.23025E-03
hHHO-SCA [60] 5.937725 4.85041 4.622404 3.45347 2.089114 1.30412236 0.087632349
ALO [61] 6.0181 5.3114 4.4884 3.4975 2.1583 1.33995 2.37867E-06

1.349 BWOA
HSA
1.347 sGA
1.345

1.343
Fitness

1.341

1.339

1.337

1.335
0 100 200 300 400 500
Number of Iteration
Fig. 10. The design history graphs of CBD problem.

beam [58]. The minimum weighted beam should satisfy the displace­ Optimizer (ALO) [61]. But, as seen in Table 2, in both algorithms there
ment constraint. are slight constraint violations of 8.763 × 10− 2 and 2.378 × 10− 6,
respectively. But the HS and BWO algorithms conducted in this study do
61 37 19 7 1
g= + + + + − 1≤0 (13) not violate the constraint in all. The convergence performances of HSA,
x31 x32 x33 x34 x35
BWOA and sGA are displayed over design history graphs as shown in
In Eq. (13), vertical displacement of the free end of the cantilever Fig. 10. It is clear from this figure that the BWOA reaches the optimum
beam g is the design constraint of this optimization problem. In addition, solution at the 41st iteration out of 500, while HSA attains at the 256th
all of the design variables range between 0.01 and 100 with increments iteration. The worst convergence performance is displayed by sGA since
of 0.001. it reaches the optimal solution at 301st iteration our of 500. This means
Here, to find minimum weighted cantilever beam is aimed. For this that the BWOA overperform for this benchmark design example in terms
purpose, the HSA and BWOA are conducted on this CBD design problem. of computational effort.
In addition, the performances of these algorithms are compared in
finding minimum weight. The final obtained optimal design results are 5.1.2. Reinforced concrete beam design problem
tabulated in Table 2. A 30 ft (9.144 m) simple reinforced concrete beam design (RCBD)
While the HS and BWO algorithms attain same minimum weighted was firstly considered as benchmark design problem by Amir and
design of 1.34004 kg, the sGA yields slightly heavier design weight as Hasegawa [62]. There is distributed live load of 2 klbf (8.896 kN) and
1.34166 kg. The optimal design attained via both HSA and BWOA looks dead load of 1 klbf (4.448 kN) (with the self-weight of the beam) on the
heavier than the previously reported optimum ones reached via Hybrid beam. The compressive strength of the concrete is 5 ksi (34.474 MPa)
Harris Hawks-Sine Cosine Algorithm (hHHO-SCA) [60] and Ant Lion and, the yield strength of the steel is 50 ksi (344.738 MPa). The unit

758
O. Tunca Structures 43 (2022) 752–766

Fig. 11. Concrete beam design problem.

Table 3
The final obtained optimal design results of RCBD problem.
Algorithm As (in2 ) As (cm2 ) h(in) h(cm) b(in) b(cm) min. cost ($) g1 (constraints) g2 (constraints)

HSA 6.32 40.774 34 86.36 8.5 21.59 359.208 − 0.224 0

BWOA 6.32 40.774 34 86.36 8.5 21.59 359.208 − 0.224 0
sGA 7.11 45.871 32 81.28 8 20.32 362.634 − 0.917 0

420

410

400
BWOA
390
Fitness

HSA
sGA
380

370

360

350
0 100 200 300 400 500
Number of Iteration
Fig. 12. The design history graphs of RCBD problem.

costs of the materials are $0.02 in2/feet ($0.423 cm2/m) for concrete The moment of beam under the bending, dead load, and live load are
and $1 in2/feet ($21.167 cm2/m) for steel. The minimum cost is aimed Mu , Ma , and Ml , respectively. So, Ma =1.350 kip-in (152.529 kN-m)
in this benchmark design problem. The minimum cost of the RCBD is andMl = 2700 kip-in (305.059 kN-m). The depth to wide ratio is
simplified as Eq. (14). limited to four. The constraints of the concrete beam design problem can
There are three design variables as seen Fig. 11. These consists of the be simplified as follows.
area of reinforcements (As ), the wind of the reinforced concrete beam
A2s
(b), and the height of the beam (h). The cross-section area of the lon­ g1 = − As h + 7.375 + 180 ≤ 0 (16)
b
gitudinal reinforcement is selected from standard bar dimensions [63].
Cost(As , b, h) = 29.4 As + 0.6bh (14) g2 =
h
− 4≤0 (17)
b
The effective depth is calculated as 0.8 h. ACI 318-77 (Building Code
Requirements for Structural Concrete, American Concrete Institute) Here, to find minimum coasted reinforced concrete beam is aimed. For
[64] is considered as structural requirements. The required strength as this purpose, the HSA, BWOA, and sGA are conducted on this RCBD
follows, problem. Moreover, the performances of these algorithms are compared
( ) in finding minimum cost. The final obtained optimal design results are
As σy
Mu = 0.9 As σy (0.8h) 1 − 0.59 ≥ 1.4Ma + 1.7Ml (15) tabulated in Table 3.
0.8bhσ c The HS and BWO algorithms attain same minimum coasted design of

759
O. Tunca Structures 43 (2022) 752–766

Fig. 13. The general views of BRB models.

Table 4
The geometric bounds of BRBd and BRBp.
Name of variable V1 V2 H1 H2 H3 V3 to ti

Min. Value (m) 0.050 0.200 0.025 0.025 0.050 0.100 0.005 0.010
Max. Value (m) 0.150 0.250 0.075 0.075 0.100 0.150 0.015 0.015
Increment (m) 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001

5.2. Design of BRBs

Table 5
The optimum solutions for BRBd.
After making sure that the parameters of optimization algorithms to
Algorithms Design variables (m) Energy dissipation (Joule) be used in the study, the BRBs are modeled. ANSYS Workbench v18.1
HSA V1 V2 H1 H2 201,169 which is Finite Element Analysis (FEA) packaged software is used for
0.133 0.231 0.071 0.029 this. The geometric model consists of three parts such as core plate, outer
H3 V3 to ti tube, and concrete as a filling material. The geometric model of the BRB
0.085 0.115 0.005 0.013
BWOA V1 V2 H1 H2 201,169
is shown in Fig. 13(a). While one end of the core palate is fixed, the load
0.133 0.231 0.071 0.029 is applied from the other end (Fig. 13(c)).
H3 V3 to ti
0.085 0.115 0.005 0.013 5.2.1. Description of numerical simulation
sGA V1 V2 H1 H2 192,891
There is various type of model in the rich material library of ANSYS
0.086 0.206 0.073 0.028
H3 V3 to ti Workbench v18.1. Three different materials are used in default. Struc­
0.064 0.116 0.005 0.013 tural Steel NL, which has a bi-linear material model, is assigned to the
core plate, Structural Steel is assigned for the outer tube, and Concrete is
g1 = − 0.29, g2 = − 0.30, g3 = − 0.27, g4 = − 0.28, g5 = − 0.25, g6 = − 0.27, g7 =
− 0.25,
used for filling material. The young modulus’s (E) of concrete and both
g8 = − 0.27, g9 = − 0.26, g10 = − 0.27, g11 = − 0.28, g12 = − 0.29, g13 = − 0.29, steels are 30 GPa and 200 GPa, respectively. The yield strength (Fy) of
g14 = − 0.28. Structural Steel NL is 250 MPa and the tangent modulus is 1450 MPa.
In the numerical simulations conducted by using FEA based ANSYS,
359.208 $, and sGA finds it as 362.634 $. All algorithms conducted in the meshing process has high significance. The mesh size and the using
this study do not violate the constraint in all. The convergence perfor­ element type for meshing are two fundamental factors. The numbers of
mances of HSA, BWOA and sGA are displayed over design history graphs nodes and the numbers of meshing elements are taken as small as
as shown in Fig. 12. It is clear from this figure that the BWOA reaches the possible since these directly affect the computation time of FEA based
optimum solution at the 34th iteration out of 500, while HSA attains at numerical simulation analysis. Yet, on condition the meshing is too
the 120th iteration. The sGA reaches the optimum design at 148th coarse, the obtained results may not be accurate. There should be a
iteration out of 500. This means that the BWOA overperform for this balance between computation time and accuracy of the obtained results.
benchmark design example in terms of computational effort. Thus, the shell type meshing elements present higher accuracy with
lower numbers of nodes. So, steel parts of the BRB models are considered
as shell elements. Thus, Quad4, Quad8, and Hex20 element types are
used for the core plate, outer tube, and concrete part of the BRB models,

760
O. Tunca Structures 43 (2022) 752–766

250

200
Fitness Value (kJ)
150

BWOA
100 HSA
sGA

50

0
0 100 200 300 400 500
Number of Iteration
(a) Computational iteration history

250

200
Fitness Value (kJ)

150

100
BWOA
HSA
50 sGA

0
0 5 10 15 20 25
Number of Days
(b) Computational time history
Fig. 14. The design history graphs of BRBd.

respectively. And, the mesh element size is taken 3 cm. The mesh dis­ are taken too small, convergence errors may occur due to the loss of
tribution on the BRB and the core plate are shown in Fig. 13 (b) and (d), stability. All of them are given in Table 4. There are 4.22 × 1012 different
respectively. The Newton-Raphson method and Large Deflection are acti­ solution vector combinations. Among these, the optimum solution vec­
vated in the analysis setting. Thus, both geometric and material tor is selected by the HS, BWO and sGA optimization algorithms.
nonlinearity is taken in the account in FEA based numerical simulation In this study, the Application Programming Interface (API) concept is
process. The Bounded Contact is assigned to the connection between the used to design two types of BRB. In each evaluation, the nonlinear finite
outer steel tube and concrete. The interaction between the core plate element models created in ANSYS Workbench v18.1 packaged software
and the concrete is defined by using Frictional Contact. So, the frictional are executed. All the analyses are implemented on Intel(R) Core (TM) i3-
coefficient is taken as 0.6. There is a gap between the core plate and 2120 CPU @ 3.30 GHz with 8 GB of RAM computer. The time con­
concrete. The 5 mm gap is defined by using contact settings. There are sumption of the FEA changes with different values of design variables.
geometric bounds (i.e., the core plate should enough fit into the outer Each finite element analysis takes about 54 min. It takes about 3 months
tube) besides mechanical limits (i.e., maximum compression force to the to get all the final feasible solutions. So, the algorithm performances are
maximum tension force should not exceed 1.3). additionally considered in the time cost. To better explain the attained
solutions, only the first 20 days are given in the design history graphs.
5.2.2. Optimum design procedure
The optimum design problem of a BRB on element basis is discussed 5.2.3. Results of diagonally shaped BRB (BRBd)
for the first time in this study. While considering these values, it is aimed The BRBd design problem is solved by HSA, BWOA and sGA. While
to stay within reasonable limits. In addition, if some geometric values the calculated maximum energy dissipation capacity is 201.169 kJ in

761
O. Tunca Structures 43 (2022) 752–766

Fig. 15. The detailed results of BRBd.

respectively, the sGA reach the optimal design in about 21 days. The
Table 6 computational time difference between the HSA and BWOA is 30.05%.
The optimum solutions of BRBp. The total deformation distributions of the optimum designed BRBd
Algorithms Design variables (m) Energy dissipation (Joule) are given in Fig. 15(a). The accumulation of the total deformation is seen
HSA V1 V2 H1 H2 204,962
at end of the plastic zone. The hysteretic behavior of the optimum
0.136 0.216 0.075 0.029 designed BRBd is indicated in Fig. 15(c) for standard cycles and in
H3 V3 to ti Fig. 15(d) for additional cycles. Besides, the energy dissipation of opti­
0.065 0.117 0.005 0.013 mally designed BRBd during simulation is given in Fig. 15(b). It can be
BWOA V1 V2 H1 H2 205,023
deducted from Fig. 13 that the optimally designed BRBd performs
0.136 0.225 0.075 0.029
H3 V3 to ti reasonably close to symmetry.
0.066 0.117 0.005 0.013
sGA V1 V2 H1 H2 194,335 5.2.4. Results of parabolic shaped BRB (BRBp)
0.122 0.216 0.072 0.028 The BRBp is taken into account as another design problem of this
H3 V3 to ti
0.087 0.116 0.005 0.013
study. The objective function and design constraints of the optimization
problem are treated same as the BRBd design. The only difference is the
g1 = − 0.29, g2 = − 0.27, g3 = − 0.26, g4 = − 0.24, g5 = − 0.24, g6 = − 0.24, g7 = geometry of the core plate of the BRB as given in Fig. 2. The calculated
− 0.24,
energy dissipation capacity from the FEA of optimally designed BRBp
g8 = − 0.23, g9 = − 0.23, g10 = − 0.24, g11 = − 0.26, g12 = − 0.26, g13 = − 0.31,
are obtained via HS and BWO algorithms as 204.962 kJ and 205.023 kJ,
g14 = − 0.45.
respectively. So, it is indicated that the final energy dissipation of BRBp
designed via BWOA is 0.03% better than those of HSA. Moreover, this
HSA and BWOA, the sGA generates worser energy dissipation capacity
design example is solved by using sGA and it yields worser energy
with 192.891 kJ. The obtained optimum designs are given in Table 5.
dissipation than HSa and BWOA. The optimal BRBp designs obtained via
From this table, it is seen that all design constraints are satisfied for each
all algorithms are given in Table 6. Moreover, same table indicates that
algorithm.
all design constraints are fully satisfied.
As obvious from Table 5, the energy dissipations acquired by HSA
The design history graphs of optimal designs attained through HSA,
and BWOA are same and better than those accomplished by sGA. The
BWOA, and sGA for BRBp are illustrated in Fig. 16. In Fig. 16(a), the
design histories of all algorithms (HSA, BWOA, and sGA) to accomplish
algorithmic performances of all algorithms can easily be compared. In
optimal BRBd designs are given in Fig. 14. Even if HSA and BWOA
this context, the BWOA reaches the optimal maximum energy dissipa­
generate same optimal structural energy dissipation capacity at the end,
tion capacity at 30th iteration out of 500, while the HSA reaches at
as seen in Fig. 14(a), the HSA yields the maximum energy dissipation
222nd iteration. The sGA accomplishes the worst maximum energy
capacity at the 192nd iteration out of 500 while the BWOA reaches this
dissipation capacity as it reaches 246th iteration out of 500. Besides,
optimum value at the 25th iteration. The worst convergence is exhibited
while the convergence rate of BWOA is smoother than both HSA and
by sGA as it attains the maximum energy dissipation at the 234th iter­
sGA. It is interesting to say that the HSA relatively exploits design search
ation out of 500. In Fig. 14(b), it is apparent that while the HSA and
space elaborately than BWOA and sGA. Additionally, the computational
BWOA reach the optimum design solutions in about 7.2 and 9.4 days,
time cost design history graphs of HSA and BWOA is presented in Fig. 16

762
O. Tunca Structures 43 (2022) 752–766

250

200
Fitness Value (kJ)
150

BWOA
100
HSA
sGA
50

0
0 100 200 300 400 500
Number of Iteration
(a) Computational iteration history

250

200
Fitness Value (kJ)

150

BWOA
100 HSA
sGA

50

0
0 5 10 15 20 25
Number of Days
(b) Computational time history
Fig. 16. The design history graphs of BRBp.

(b). From this graph, it is seen that the HSA displays better performance and BWO algorithms through conducting proposed API methodology.
with 8.3 days, while the BWOA consumes 11.3 days in finding optimum The BRBs are modelled as diagonal bordered (BRBd) and parabolic
design solutions of BRBp. It is seen that sGA needs much more time to bordered (BRBp). It is fundamentally purposed that these design prob­
obtain the optimum result compared to other algorithms. lems are to maximize the energy dissipation capacities. The ANSI/AISC
Also, the total deformation of the optimally designed BRBp is given 341 practical design specifications are used to define geometric and
in Fig. 17(a). It is obviously seen that the global buckling of the core mechanic structural restrictions. The FEA-based BRB models include
plate is limited. And, the hysteretic behavior of the BRBp can be material and geometric nonlinearity and complex frictional contacts.
observed from Fig. 17(c) and (d). The energy dissipation capacity These are generated using FEA-based numerical simulation software, so-
(Fig. 17(b)) is confirmed by hysteretic curves. So, it can be concluded called ANSYS Workbench v18.1, to provide simultaneous integration
that optimally designed BRBp performs reasonably close to symmetry throughout API functions with the HS and BWO algorithms encoded in
between compression and tension. MS VBA v7.0. To prove the accuracy of proposed design optimization
methodology, two benchmark structural design problems (Cantilever
6. Discussion of results Beam Design (CBD) and Reinforced Concrete Beam Design (RCBD)) are
initially solved. These benchmark design problems aim to find the
As the fundamental aim of this study, two different core shaped minimum weighted optimum CBD structure and the minimum cost of
Buckling Restrained Brace (BRB) elements are optimally designed via HS the RCBD structure. Throughout CBD and RCBD problems, it is seen that

763
O. Tunca Structures 43 (2022) 752–766

Fig. 17. The detailed results of BRBp.

the encoded HS and BWO algorithms are accurately executed. Then, the • In BRBp design, while the BWOA reaches the maximum energy
sizing design optimization of two different complex BRBs, which are dissipation capacity at 30th iterations, the HSA reaches optimal en­
very difficult to define in a homemade programming, are identified ergy dissipation capacity at 222nd iterations out of 500. However,
straightly over simple objective function and design constraints for­ the HSA displays 36.15% better computational time effort than
mulas with the aid of the API functions throughout the integration of BWOA in finding final optimum design solution.
ANSYS Workbench with HS and BWO algorithms encoded in MS VBA v7. • It is obviously seen that the global buckling of the core plates of two
Since the finite element analyses take up to about one hour, row number optimum designed BRBs (BRBd and BRBp) is limited. The energy
of the solution matrix is decreased as much as possible. Also, in order dissipation histories are confirmed by hysteretic curves of BRBs.
show the supremacy of mainly considered algorithmic performance • The hysteretic behavior and the energy dissipation histories of the
capacities of HS and BWO algorithms, a conventional more established BRBs indicate that obtained optimal designs are nearly symmetric
standard Genetic Algorithm (sGA) is executed for all abovementioned and possess reasonable structural characteristics. In design optimi­
design examples. So, the supreme algorithmic performances of HSA and zation of BRBp, the BWOA attain 0.03% better design than those of
BWOA are displayed by means of the computational iterations and HSA.
computational time costs. • From all obtained optimal design solutions, it can be concluded that
the BRBp has 1.92% more energy dissipation capacity than BRBd.
7. Concluding remarks Moreover, the total deformation distribution is observed over the
optimally designed BRBd.
In this study, the buckling restrained braces (BRBs) are optimized
throughout a design methodology in which the IronPython script file- Eventually, the BRBs stand out from other passive energy devices as
based application programming interface (API) functions that supply they are economical and easy to apply to the structures. Furthermore,
integration between FEA-based numerical simulation and discrete the BRBs increase the lateral stiffness of the equipped structures. The
design optimization Harmony Search (HS) and Black Widow Optimi­ acceptance design criteria of ANSI/AISC 341 practical design specifi­
zation (BWO) algorithms. For this purpose, the energy dissipation ca­ cations are repeatable and stable behavior with increasing stiffness of
pacity of BRBs is maximized in direction of structural restrictions taken the BRBs. Here, the maximum amount of energy dissipation throughout
from ANSI/AISC 341 practical design specifications. In summary, the BRBs can be accepted as an indication that they also provide economical
fundamental concluding remarks of this study are itemized as structures since reducing structural design weight. But still, the effects of
followings: BRBs on the cost and rigidity of the structural frames should be partic­
ularly handled. Thus, as a future study, it is planned to consider the
• In CBD problem, the both algorithms are ended with the same min­ potential effects of some additional design variables on handled opti­
imum design weight of 1.34 kg. But, the BWOA exhibits relatively mum BRB design problem in detail. For instance, by considering various
better algorithmic performance in finding optimum design solution types of practical joint details, BRB equipped frame structures subjected
according to iteration-based evaluation. to earthquake loads can be modeled via FEA packaged software based
• In BRBd design, the both algorithms yield the same optimal design numerical simulation. To do this, more comprehensive elemental and
solutions of 201.169 kJ. While the HSA converges to the optimum structural BRB design problems satisfying all practical structural re­
design at the 192nd iteration, the BWOA reaches at the 25th iteration quirements will be solved by considering novel various design variables.
out of 500. Even the BWOA performs rapid convergence to the op­
timum design, it requires 30.05% extra time effort than HSA.

764
O. Tunca Structures 43 (2022) 752–766

Declaration of Competing Interest dynamic constraints. Structures 2021;34:3533–47. https://doi.org/10.1016/j.

istruc.2021.09.101.
[27] Hu B, Min Y, Wang C, Xu Q, Keleta Y. Design, analysis and application of the
The author declares that he has no known competing financial in­ double-stage yield buckling restrained brace. J Build Eng 2022;48:103980. https://
terests or personal relationships that could have appeared to influence doi.org/10.1016/j.jobe.2021.103980.
the work reported in this paper. [28] Yun Z, Cao Y, Takagi J, Zhong G, He Z. Experimental and numerical investigation
of a novel all-steel assembled core-perforated buckling-restrained brace. J Constr
Steel Res 2022;193:107288. https://doi.org/10.1016/j.jcsr.2022.107288.
References [29] Bai J, Chen H, Ma G, Duan L. Development of a four-tube-assembled buckling-
restrained brace for convenient post-earthquake damage examination and
[1] Watanabe A, Hitomi Y, Saeki E, Wada A, Fujimoto M. Properties of brace encased replacement. J Build Eng 2022;50:104209. https://doi.org/10.1016/j.
in buckling-restraining concrete and steel tube. Tokyo-Kyoto JAPAN 1988. jobe.2022.104209.
[2] Xie Q. State of the art of buckling-restrained braces in Asia. J Constr Steel Res [30] Carbas S, Artar M. Optimum design of cold-formed steel frames via five novel
2005;61:727–48. https://doi.org/10.1016/j.jcsr.2004.11.005. nature-inspired metaheuristic algorithms under consideration of seismic loading.
[3] Black CJ, Makris N, Aiken ID. Component testing, seismic evaluation and Structures 2021;33:4011–30. https://doi.org/10.1016/j.istruc.2021.06.096.
characterization of Buckling-Restrained Braces. J Struct Eng 2004;130:880–94. [31] Minafò G, Camarda G. An open-source GA framework for optimizing the seismic
https://doi.org/10.1061/(ASCE)0733-9445(2004)130:6(880). upgrading design of RC frames through BRBs. Eng Struct 2022;251:113508.
[4] Brown P, Aiken I, Jafarzadeh S. Seismic retrofit of the Wallace F. Bennett federal https://doi.org/10.1016/j.engstruct.2021.113508.
building. Modern Steel Construction 2001. [32] Minafò G, Camarda G. Genetic optimization for the design of seismic retrofitting of
[5] Abedini H, Hoseini Vaez SR, Zarrineghbal A. Optimum design of buckling- plane RC frames with buckling restrained braces (BRBS), Athens, Greece: 2021, p.
restrained braced frames. Structures 2020;25:99–112. https://doi.org/10.1016/j. 1333–44. 10.7712/120121.8564.19069.
istruc.2020.03.004. [33] Atlayan O, Charney FA. Hybrid buckling-restrained braced frames. J Constr Steel
[6] Farhat F, Nakamura S, Takahashi K. Application of genetic algorithm to Res 2014;96:95–105. https://doi.org/10.1016/j.jcsr.2014.01.001.
optimization of buckling restrained braces for seismic upgrading of existing [34] Wang F, Shi Q-X, Wang P. Seismic behaviour of reinforced concrete frame
structures. Comput Struct 2009;87:110–9. https://doi.org/10.1016/j. structures with all steel assembled Q195 low yield buckling restrained braces.
compstruc.2008.08.002. Structures 2021;30:756–73. https://doi.org/10.1016/j.istruc.2021.01.051.
[7] Güneyisi EM. Seismic reliability of steel moment resisting framed buildings [35] Liao X, Zhao C, Liu Z. Parametric finite element analysis of Buckling-Restrained
retrofitted with buckling restrained braces: seismic reliability of buildings with Braces. J Phys: Conf Ser 2021;1732:012143. https://doi.org/10.1088/1742-6596/
buckling restrained braces. Earthquake Engng Struct Dyn 2012;41(5):853–74. 1732/1/012143.
[8] Wu S, He H, Chen Y, Chen J. Calculation method of time-variant additional [36] Geem ZW, Kim JH, Loganathan GV. A new heuristic optimization algorithm:
damping ratio and bi-objective design of RC frame with BRBs. Structures 2021;33: harmony search. Simulation 2001;76:60–8. https://doi.org/10.1177/
2546–58. https://doi.org/10.1016/j.istruc.2021.06.025. 003754970107600201.
[9] Guneyisi EM, Tunca O, Azez I. Nonlinear dynamic response of reinforced concrete [37] Hayyolalam V, Pourhaji Kazem AA. Black Widow Optimization Algorithm: a novel
building retrofitted with buckling restrained braces. Earthquakes Struct 2015;8: meta-heuristic approach for solving engineering optimization problems. Eng Appl
1349–62. https://doi.org/10.12989/EAS.2015.8.6.1349. Artif Intell 2020;87:103249. https://doi.org/10.1016/j.engappai.2019.103249.
[10] Zhang Y, Ren X, Zhang XY, Huang TT, Sun L, Xie YM. A novel buckling-restrained [38] Goldberg DE. Genetic algorithms in search, optimization, and machine learning.
brace with auxetic perforated core: experimental and numerical studies. Eng Struct Reading, Mass: Addison-Wesley Pub. Co; 1989.
2021;249:113223. https://doi.org/10.1016/j.engstruct.2021.113223. [39] Microsoft Visual Basic Programming Language. Microsoft Corporation One
[11] Yuan Y, Qing Y, Wang C-L, Lin X, Zhu M. Development and experimental Microsoft Way Redmond, WA 98052-6399 USA; 2016.; 2016.
validation of a partially buckling-restrained brace with dual-plate cores. J Constr [40] Ansys Inc. ANSYS Mechanical Theory Reference: Release 18.1, Canonsburg PA,
Steel Res 2021;187:106992. https://doi.org/10.1016/j.jcsr.2021.106992. USA; 2016.; n.d.
[12] Al-Sadoon ZA, Saboor Karzad A, Sagheer A, AlHamaydeh M. Replaceable fuse [41] IronPython Team. Python Software Foundation, Microsoft Corporation; 2001-
buckling-restrained brace (BRB): experimental cyclic qualification testing and 2020.; n.d.
NLFEA modeling. Structures 2022;39:997–1015. https://doi.org/10.1016/j. [42] Çarbaş S. Optimum structural design of spatial steel frames via biogeography-based
istruc.2022.03.081. optimization. Neural Comput Appl 2017;28:1525–39. https://doi.org/10.1007/
[13] Aydogdu I. Cost optimization of reinforced concrete cantilever retaining walls s00521-015-2167-6.
under seismic loading using a biogeography-based optimization algorithm with [43] Tunca O, Çarbaş S. Biogeography-based optimization algorithm for designing of
Levy flights. Eng Optim 2017;49:381–400. https://doi.org/10.1080/ planar steel frames. Int J Intell Syst Appl Eng 2016;4:53–7. https://doi.org/10.
0305215X.2016.1191837. 18201/ijisae.266128.
[14] Erdal F. A firefly algorithm for optimum design of new-generation beams. Eng [44] ANSI/AISC 341-10. Seismic Provisions for Structural Steel Buildings, American
Optim 2017;49:915–31. https://doi.org/10.1080/0305215X.2016.1218003. Institute of Steel Construction, Chicago, USA; 2010.; n.d.
[15] Chen H, Xu Y, Wang M, Zhao X. A balanced whale optimization algorithm for [45] Saka MP. Optimum design of steel skeleton structures. In: Geem ZW, editor. Music-
constrained engineering design problems. Appl Math Model 2019;71:45–59. Inspired Harmony Search Algorithm, Springer Berlin Heidelberg; n.d.
https://doi.org/10.1016/j.apm.2019.02.004. [46] Carbas S, Toktas A, Ustun D, editors. Nature-Inspired Metaheuristic Algorithms for
[16] Talatahari S, Azizi M. Optimum design of building structures using Tribe-Interior Engineering Optimization Applications. Singapore: Springer Singapore; 2021.
Search Algorithm. Structures 2020;28:1616–33. https://doi.org/10.1016/j. 10.1007/978-981-33-6773-9.
istruc.2020.09.075. [47] SAP2000. Integrated finite element analysis and design of structures; 2008.; n.d.
[17] Kalemci EN, İkizler SB, Dede T, Angın Z. Design of reinforced concrete cantilever [48] Tunca O. Optimum design of beams with varying cross-section by using application
retaining wall using Grey wolf optimization algorithm. Structures 2020;23:245–53. interface. In: Carbas S, Toktas A, Ustun D, editors. Nature-Inspired Metaheuristic
https://doi.org/10.1016/j.istruc.2019.09.013. Algorithms for Engineering Optimization Applications. Singapore: Springer; 2021.
[18] Moghaddam H, Afzalinia F, Hajirasouliha I. Optimal distribution of friction p. 225–49. https://doi.org/10.1007/978-981-33-6773-9_11.
dampers to improve the seismic performance of steel moment resisting frames. [49] Kim Y-H, Yoon Y, Geem ZW. A comparison study of harmony search and genetic
Structures 2022;37:624–44. https://doi.org/10.1016/j.istruc.2022.01.007. algorithm for the max-cut problem. Swarm Evol Comput 2019;44:130–5. https://
[19] Yang X-S. Firefly algorithm, stochastic test functions and design optimisation. Int J doi.org/10.1016/j.swevo.2018.01.004.
Bio-Inspired Comput 2010;2:78–84. https://doi.org/10.1504/IJBIC.2010.032124. [50] Singh S, Sharma RM. HSCA: a novel harmony search based efficient clustering in
[20] Moazzeni AR, Khamehchi E. Rain optimization algorithm (ROA): a new heterogeneous WSNs. Telecommun Syst 2018;67:651–67. https://doi.org/
metaheuristic method for drilling optimization solutions. J Petrol Sci Eng 2020; 10.1007/s11235-017-0365-5.
195:107512. https://doi.org/10.1016/j.petrol.2020.107512. [51] Theodossiou N, Kougias I. Harmony search algorithm. Hydrol Hydraulics Water
[21] Kaveh A, Dadras EA. Water strider algorithm: a new metaheuristic and Resour Manage 2012;56. https://doi.org/10.2495/978-1-84564-664-6/07.
applications. Structures 2020;25:520–41. https://doi.org/10.1016/j. [52] Weyland D. A critical analysis of the harmony search algorithm—How not to solve
istruc.2020.03.033. sudoku. Oper Res Perspect 2015;2:97–105. https://doi.org/10.1016/j.
[22] Kaur S, Awasthi LK, Sangal AL, Dhiman G. Tunicate Swarm Algorithm: a new bio- orp.2015.04.001.
inspired based metaheuristic paradigm for global optimization. Eng Appl Artif [53] Duan Q, Liao TW, Yi HZ. A comparative study of different local search application
Intell 2020;90:103541. https://doi.org/10.1016/j.engappai.2020.103541. strategies in hybrid metaheuristics. Appl Soft Comput 2013;13:1464–77. https://
[23] Rezazadeh F, Mirghaderi R, Hosseini A, Talatahari S. Optimum energy-based doi.org/10.1016/j.asoc.2012.05.016.
design of BRB frames using nonlinear response history analysis. Struct Multidisc [54] Fu Y, Hou Y, Chen Z, Pu X, Gao K, Sadollah A. Modelling and scheduling
Optim 2018;57:1005–19. https://doi.org/10.1007/s00158-017-1791-4. integration of distributed production and distribution problems via black widow
[24] Taiyari F, Kharghani M, Hajihassani M. Optimal design of pile wall retaining optimization. Swarm Evol Comput 2022;68:101015. https://doi.org/10.1016/j.
system during deep excavation using swarm intelligence technique. Structures swevo.2021.101015.
2020;28:1991–9. https://doi.org/10.1016/j.istruc.2020.10.044. [55] Sadeghi B, Shafaghatian N, Alayi R, El Haj AM, Zishan F, Hosseinzadeh H.
[25] Saka MP. Shape and topology optimization design of skeletal structures using Optimization of synchronized frequency and voltage control for a distributed
metaheuristic algorithms: a review. Comp Tech Rev 2014;9:31–68. https://doi. generation system using the Black Widow Optimization algorithm. Clean Energy
org/10.4203/ctr.9.2. 2022;6:869–82. https://doi.org/10.1093/ce/zka.
[26] Artar M, Carbas S. Discrete sizing design of steel truss bridges through teaching- [56] Peña-Delgado AF, Peraza-Vázquez H, Almazán-Covarrubias JH, Torres Cruz N,
learning-based and biogeography-based optimization algorithms involving García-Vite PM, Morales-Cepeda AB, et al. A novel bio-inspired algorithm applied

765
O. Tunca Structures 43 (2022) 752–766

to selective harmonic elimination in a three-phase eleven-level inverter. Math [60] Kamboj VK, Nandi A, Bhadoria A, Sehgal S. An intensify Harris Hawks optimizer
Probl Eng 2020;2020:e8856040. https://doi.org/10.1155/2020/8856040. for numerical and engineering optimization problems. Appl Soft Comput 2020;89:
[57] Kattan A, Abdullah R. A dynamic self-adaptive harmony search algorithm for 106018.
continuous optimization problems. Appl Math Comput 2013;219:8542–67. [61] Mirjalili S. The ant lion optimizer. Adv Eng Softw 2015;83:80–98. https://doi.org/
https://doi.org/10.1016/j.amc.2013.02.074. 10.1016/j.advengsoft.2015.01.010.
[58] Chickermane H, Gea HC. Structural optimization using a new local approximation [62] Amir HM, Hasegawa T. Nonlinear mixed-discrete structural optimization. J Struct
method. Int J Numer Meth Eng 1996;39:829–46. https://doi.org/10.1002/(SICI) Eng 1989;115:626–46. https://doi.org/10.1061/(ASCE)0733-9445(1989)115:3
1097-0207(19960315)39:5<829::AID-NME884>3.0.CO;2-U. (626).
[59] Bayzidi H, Talatahari S, Saraee M, Lamarche C-P, Precup R-E. Social network [63] Hosseinaei S, Ghasemi MR, Etedali S. Optimal design of passive and active control
search for solving engineering optimization problems. Comput Intell Neurosci systems in seismic-excited structures using a new modified TLBO. Period Polytech
2021;2021:1–32. Civil Eng 2020. https://doi.org/10.3311/PPci.16507.
[64] ACI 318-77. Building Code Requirements for Structural Concrete, American
Concrete Institute, 1981; n.d.

766