Evolutionary Optimization based Solution approaches for Many Objective Reliability-Redundancy Allocation Problem

Recently, a number of evolutionary optimization approaches were proposed to solve the many objective problems. The reliability redundancy allocation problem (RRAP), which usually has four different objectives, namely, maximization of system reliability, minimizations of cost, weight and volume, are however solved mainly as a multi-objective problem considering only two or three objectives. Therefore, this paper reports a novel study of the RRAP as a many objective optimization problem. Here, we formulate the many objective RRAP (MaORRAP) with various structures such as series-parallel systems, overspeed gas turbine system, and large-scale system. For the formulated MaORRAP, we then provide the details of a novel solution procedure based on the non-dominated sorting genetic algorithm-III (NSGA-III), a well-discussed many objective evolutionary optimization algorithm. We also solve MaORRAP using three other popular evolutionary approaches, viz. non-dominated sorting genetic algorithm-II (NSGA-II), multi-objective evolutionary algorithm based on decomposition (MOEA/D) and strength Pareto evolutionary archive 2 (SPEA2) algorithm. Accordingly, we present all the results in a competitive fashion to have a thorough comparative assessment of the performances of the considered approaches and show that, in most of the cases, NSGA-III based solutions are superior to others.

Automated summary

This paper presents a novel study of the reliability redundancy allocation problem (RRAP) as a many objective optimization problem, where a new procedure based on NSGA-III is proposed for solving MaORRAP. The performances of NSGA-III and other popular evolutionary algorithms in solving MaORRAP are compared and it is shown that NSGA-III based solutions are superior in most cases.