Solving advanced multi-objective robust designs by means of multiple objective evolutionary algorithms (MOEA): A reliability application

Abstract

This paper extends the approach proposed by the second author in [Rocco et al. Robust design using a hybrid-cellular-evolutionary and interval-arithmetic approach: a reliability application. In: Tarantola S, Saltelli A, editors. SAMO 2001: Methodological advances and useful applications of sensitivity analysis. Reliab Eng Syst Saf 2003;79(2):149-59 [special issue]] to obtain a robust system design. The approach based on the use of evolutionary algorithms and interval arithmetic finds the maximum-volume inner box (MIB) or the maximal ranges of variation for each variable that preserve pre-specified design/performance requirements. The original single-objective formulation considers the definition of a MIB around a specified centroid (case 1), or around an unspecified centroid (case 2). In this paper, both cases were successfully modified and solved as multiple-objective (MO) problems, showing the advantages of MO formulations in a design-selection decision framework. Special attention is devoted to the unspecified centre MO problem where the computational efficiency could be a critical issue. In that sense, a new procedure based on the "percentage representation" is proposed. This approach reduces drastically the computational burden, extending the possibilities of use of robust design.