Application-Oriented Robust Design Optimization Method for Batch Production of Permanent-Magnet Motors

From the perspective of industrial production, the design and optimization of electrical machines are application oriented, including maximizing production quality and minimizing production cost in terms of different manufacturing conditions. To achieve these goals, this study presents an efficient application-oriented robust design optimization method for permanent-magnet (PM) motors. The method consists of two main contributions. The first one is the development of an overall optimization strategy, including qualitative and quantitative analyses to provide possible options for an application. Multiphysics analysis, uncertainty analysis, production cost, and optimization models need to be investigated. The second one proposes a multilevel optimization method for the high-dimensional robust design problem of each option. To illustrate the advantages of the proposed method, PM motors with soft magnetic composite cores are investigated for domestic applications. The design optimization is conducted in terms of three motor options and three batch production volumes for both conventional deterministic and robust approaches, and it consists of 18 high-dimensional multiphysics optimization problems in total. Main optimization results are presented and discussed. Experimental and simulation results are presented to validate the effectiveness of the proposed models and methods.

I. Introduction

Electrical machines are the heart of many modern appliances, as well as industry equipment and systems. In a global market and in the context of sustainability, they must fulfill various requirements physically and technologically. To satisfy these requirements, optimization is of great significance for electrical machine design, and many optimization methods have been developed. Optimization methods mainly include the following: 1) direct optimization of analysis models, such as analytical model, magnetic equivalent circuit model, and finite-element model (FEM), and 2) indirect optimization of approximation models (surrogate models of FEM), by using different kinds of optimization algorithms including intelligent algorithms. Some popular intelligent algorithms are the genetic algorithm (GA), the differential evolution algorithm (DEA), and the particle swarm optimization (PSO) algorithm. Surrogate models mainly include the parametric models, such as response surface model (RSM) and radial basis function (RBF) model; semiparametric model, e.g., Kriging model; and nonparametric models, such as artificial neural network and support vector machine models [1]–[8] .