Reduction of Computational Efforts in Finite Element-Based Permanent Magnet Traction Motor Optimization

This paper presents a method for reducing computational time in constrained single objective optimization problems related to permanent magnet motors modeled using computationally intensive finite element method. The method is based on differential evolution algorithm. The principal approach is to interrupt the evaluation of inequality constraints after encountering the constraint which is violated and continue their evaluation only if better solutions are obtained in comparison with the candidate vector from the previous generation both in terms of inequality constraints and the cost function. This approach avoids unnecessary time consuming finite element calculations required for solving the inequality constraints and saves the overall optimization time without affecting the convergence rate. The main features of this approach and how it complements the existing method for handling inequality constraints in Differential Evolution algorithm is demonstrated on design optimization of an interior permanent magnet motor for the low-floor tram TMK 2200. The motor geometry is optimized for maximum torque density.