Stochastic Topology Optimization Based on Level-Set Method

Abstract

The level-set method has been shown suitable for topology optimization because it can find optimal shapes with smooth boundaries and non-porous material regions. In the level-set method, however, optimization is conducted based on the gradient method. For this reason, the solutions obtained by the level-set method tend to fall into local optima. In this work, the simulated annealing is introduced into the level-set method to solve this problem. In order to test the present method, the shape of the flux barrier in the rotor of an interior permanent magnetic motor (IPM-motor) is optimized. It is shown that the resultant properties are better than those obtained by the deterministic level-set method.