Improved hybrid Strength Pareto Evolutionary Algorithms for multi-objective optimization

Abstract

Purpose

The purpose of this paper is to design an improved multi-objective algorithm with better spread and convergence than some current algorithms. The proposed application is for engineering design problems.

Design/methodology/approach

This study proposes two novel approaches which focus on faster convergence to the Pareto front (PF) while adopting the advantages of Strength Pareto Evolutionary Algorithm-2 (SPEA2) for better spread. In first method, decision variables corresponding to the optima of individual objective functions (Utopia Point) are strategically used to guide the search toward PF. In second method, boundary points of the PF are calculated and their decision variables are seeded to the initial population.

Findings

The proposed methods are tested with a wide range of constrained and unconstrained multi-objective test functions using standard performance metrics. Performance evaluation demonstrates the superiority of proposed algorithms over well-known existing algorithms (such as NSGA-II and SPEA2) and recent ones such as NSLS and E-NSGA-II in most of the benchmark functions. It is also tested on an engineering design problem and compared with a currently used algorithm.

Practical implications

The algorithms are intended to be used for practical engineering design problems which have many variables and conflicting objectives. A complex example of Welded Beam has been shown at the end of the paper.

Social implications

The algorithm would be useful for many design problems and social/industrial problems with conflicting objectives.

Originality/value

This paper presents two novel hybrid algorithms involving SPEA2 based on: local search; and Utopia point directed search principles. This concept has not been investigated before.

Citation

Shankar, K. and Baviskar, A.S. (2018), "Improved hybrid Strength Pareto Evolutionary Algorithms for multi-objective optimization", International Journal of Intelligent Computing and Cybernetics, Vol. 11 No. 1, pp. 20-46. https://doi.org/10.1108/IJICC-12-2016-0063