Methods to Reduce the Computational Burden of Robust Optimization for Permanent Magnet Motors

Abstract

One of the main obstacles to applying the robust optimization for the permanent magnet motors is the high computational burden, which is mainly caused by the robustness evaluation considering the manufacturing uncertainties. In this article, efforts are made to speed up the process of robust optimization, from the aspects of reducing the computational cost, and avoiding the computing resources being wasted. Firstly, several widely used methods, aiming to identify the worst-case performance under uncertainties, are examined, and compared. And the efficient worst-uncertain-combination-analysis (WUCA) method is consequently adopted to significantly reduce the computational cost of robustness evaluation. From another aspect, the parameter rounding (PR) operation is inevitable due to the constraint of manufacturing accuracy, but its adverse effects on the optimization are often ignored. It is found that parts of the computing resources might be wasted during the optimization. A new PR during the calculation (PRDC) method is proposed, and incorporated into the process of robust optimization. Compared with the conventional one, similar results can be achieved with the computational time reduced by approximately $\text{11}\%$ in the PRDC modified robust optimization.