A Novel Space Reduction Technique for Design Optimization of Permanent Magnet Synchronous Motors

Due to the non-linearity and large dimensions of permanent magnet motor optimization, the use of metaheuristic methods such as GA, PSO, etc. would not be the most appropriate method especially if the fitness assessment is done by a time-consuming solver such as finite element analysis (FEA). The FEA, which is widely used by most researchers, requires a lot of time and space and leads to huge computational costs. On the other hand, the accuracy of approximate analytical models is not sufficient for high-dimensional optimization tasks. To overcome these problems, a new space reduction optimization method is developed and presented in this paper. The proposed method gradually shrinks the search space and approaches an interesting subspace so that the wide variable space becomes smaller. As a result, FEA modeling accuracy is achieved as well as computational cost reductions. To validate the method, the design optimization is performed on a 2004 Toyota Prius Interior Permanent Magnet (IPM) motor. The results are compared with other optimization algorithms in terms of accuracy and number of performance evaluations. The comparison results show the superiority of the proposed algorithm, which can be a desirable alternative to industrial optimization tasks that necessarily require the least number of function evaluations.