An Effective Integrated Metaheuristic Algorithm
For Solving Engineering Problems
Adis Alihodzic
University of Sarajevo, BiH
Department of Mathematics
ul. Zmaja od Bosne, 33-35, Sarajevo
Email: adis.alihodzic@pmf.unsa.ba
Sead Delalic
University of Sarajevo, BiH
Department of Mathematics
ul. Zmaja od Bosne, 33-35, Sarajevo
Email: delalic.sead@pmf.unsa.ba
Dzenan Gusic
University of Sarajevo, BiH
Department of Mathematics
ul. Zmaja od Bosne, 33-35, Sarajevo
Email: dzenang@pmf.unsa.ba
Abstract—To tackle a specific class of engineering problems,
in this paper, we propose an effectively integrated bat algorithm
with simulated annealing for solving constrained optimization
problems. Our proposed method (I-BASA) involves simulated an-
nealing, Gaussian distribution, and a new mutation operator into
the simple Bat algorithm to accelerate the search performance as
well as to additionally improve the diversification of the whole
space. The proposed method performs balancing between the
grave exploitation of the Bat algorithm and global exploration of
the Simulated annealing. The standard engineering benchmark
problems from the literature were considered in the competition
between our integrated method and the latest swarm intelligence
algorithms in the area of design optimization. The simulations
results show that I-BASA produces high-quality solutions as well
as a low number of function evaluations.
I. I NTRODUCTION
I
N THE last fifteen years, it was shown that most de-
sign nonlinear constrained optimization problems are an
essential class of issues in real-world applications, and almost
all are characterized as NP-hard problems. For such design
optimization problems, the finding of the best solution may
require centuries, even with a supercomputer. These highly
nonlinear and multimodal optimization problems are based
on the optimization of objective functions with complex con-
straints which usually involve thousands of or even millions
of elements, and they were written in the form of simple
bounds or more often as nonlinear inequalities. Nonlinearly
constrained optimization problems contain continuous and
discrete design variables, nonlinear objective functions, and
constraints, some of which may be active at the global optima.
Due to the complex nature of an objective function, as well as
the constraints that need to be met, it is challenging how to ef-
fectively and robustly explore overall search space. Therefore,
practically solving engineering problems are come down to
some efficient methods which are problem-specific [1]. Since
optimization methods can not escape falling in into some of
the local optima, metaheuristics as very modern and efficient
global techniques are considered to overcome these type of
problems [2]. Besides, those are capable of generating quality
solutions in a reasonable amount of time. The creating of
quality solutions is related to the establishment of the right bal-
ance between exploration and exploitation. [3]. Since a magic
formula does not exist, which works for all types of problems
[4], in this paper, several swarm intelligence algorithms [5]
have been adopted for solving nonlinear engineering problems.
Some of the most popular swarm intelligence optimization
techniques are artificial bee colony(ABC) [6][7][8][9], firefly
algorithm (FA) [10][1][11][12], cuckoo search (CS) [13][14],
bat algorithm (BA) [15][16][17][18][19], flower pollination
algorithm [20], and etc. In this article, we have combined the
bat algorithm as a representative of swarm intelligent multi-
agent algorithm with one agent simulated annealing method
to produce as much as possible suboptimal solutions.
The Bat meta-heuristic algorithm (BA) has proposed by
Xin-She Yang 2010 [15]. The primary mechanism of this
swarm intelligence technique propagates echolocation of bats
as agents. The agents seek for prey and avoid obstacles by
using echolocation. In the paper [19], it has been shown that
the BA very well performs local search, but at times it deviated
into some local optima, and it can not reach the optimal
solution while solving a hard problem. The original version of
bat algorithm, as well as the other metaheuristic algorithms,
were designed to address unconstrained problems. To tackle
the constrained problems, bat algorithm (BA) uses a penalty
approach as a constraint handling technique [16]. From the
experiments presented in [16], it can be seen that the BA is
almost always superior to other metaheuristics.
To promote the results obtained by the simple bat algorithm,
in this article, we propose an integrated I-BASA approach
to take on engineering problems. Unlike the original bat
algorithm which is not capable to found satisfying balance
between diversification and intensification, the proposed I-
BASA approach based on simulated annealing (SA) [21], a
new mutation operator, and Gaussian distribution achieves
a right balance and raises overall search performance. The
integrated I-BASA method was tested on the eight well-chosen
benchmark problems, and the simulation results report that our
approach almost always wins the state-of-the-art algorithms
regarding the convergence and accuracy. In this paper, we
have decided to exploit Deb’s rules as a constraint handling
process instead of a standard penalty method. Throughout
the simulation results, it can be seen that introduced rules
significantly improve the quality of the solutions.
The basic structure of the article looks like this. The basic
definitions related to constrained optimization are described in
Proceedings of the Federated Conference on
Computer Science and Information Systems pp. 207–214
DOI: 10.15439/2020F81
ISSN 2300-5963 ACSIS, Vol. 21
IEEE Catalog Number: CFP2085N-ART ©2020, PTI 207