Open Collections

Abstract

Engineering optimization typically involves large number of nonlinear constraints, thus effective constraint handling techniques (CHTs) are sought for metaheuristic optimization algorithms. Modified repair-based CHT is proposed here for multi-objective evolutionary algorithm based on decomposition (MOEA/D). This CHT is: (a) adaptive to the share of infeasible solutions in a population, (b) free of problem-specific heuristic that user typically needs to provide for repair and (c) without control parameters. Infeasible solutions with superior decomposition function value are repaired using information contained in the neighborhoods of the current population. The approach is tested on four multi-objective problems: a common mathematical optimization benchmark problem, two truss optimization problems and a real-world structural design of a tanker ship. A few prominent CHTs and metaheuristic algorithms are used for comparison. With the proposed CHT, MOEA/D shows improved convergence speed and spread of the Pareto front, providing competitive results in comparison to the other algorithms.