Multiobjective design optimization of real-life devices in electrical engineering: A cost-effective evolutionary approach

Abstract

When tackling the multicriteria optimization of a device in electrical engineering, the exhaustive sampling of Pareto optimal front implies the use of complex and time-consuming algorithms that are unpractical from the industrial viewpoint. In several cases, however, the accurate identification of a few nondominated solutions is often sufficient for the design purposes. An evolutionary methodology of lowest order, dealing with a small number of individuals, is proposed to obtain a cost-effective approximation of non-dominated solutions. In particular, the algorithm assigning the fitness enables the designer to pursue either shape or performance diversity of the device. The optimal shape design of a shielded reactor, based on the optimization of both cost and performance of the device, is presented as a real-life case study.