Multi-objective optimization of PM AC machines using computationally efficient - FEA and differential evolution

A total of eleven independent stator and rotor variables are simultaneously employed for the optimization of a generic example IPM motor design. The multi-objective criterion maximizes efficiency, while minimizing torque ripple at the rated output condition. A Pareto-based differential evolution (DE) algorithm with 100 generations, each with a population of 100 individuals, is presented. Computationally efficient FEA (CE-FEA), which is based on a reduced number of magnetostatic solutions for a motor model in the abc reference frame, is employed. As a result, a total of 10,000 candidate motor designs, which are included in the comprehensive study, are evaluated in a record short time on a typical PC-based workstation. The paper includes an engineering trade-off discussion based on a typical-reference motor, two optimum designs in terms of average torque and torque ripple, and a best-compromise solution. For the case-study, an order of magnitude reduction of the rated-load torque ripple and open-circuit cogging torque has been achieved. This is while, at the same time, the specific torque output has been increased by as much thirty seven percent.