A novel genetic algorithm developed on a reduced search space for optimal redundancy allocation in Multi-State Series-Parallel Systems

Abstract

We consider the redundancy allocation problem (RAP) for multi-state series-parallel systems (MSSPSs),with the objective of maximizing system availability under cost constraint. We show that, to obtain the global optimum,an evolutionary algorithm(EA)only needs to search within a subset of the feasible region.To exploit this, a repair algorithm is designed in a genetic algorithm (GA),to ensure it generates individuals only within the boundary subset,without increasing its computational complexity. We also add a novel mutation operator to improve local search efficiency, via generalizing the local search operators that appear in previous MSSPS RAP literature. In benchmark test, our modified GA out performs the comparative algorithms and finds a solution better than the published one.