This paper introduces a new method for multi-objective topology optimization problems. The present method is composed of two steps: global search based on genetic algorithm (GA), which is followed by local search using a gradient-based method. In the global search, the design region is subdivided into small elements, to which either of two states, On (material) or Off (non-material) is assigned. In the optimization process based on multi-objective GA, diverse device shapes are generated by changing these states independently to obtain Pareto solutions. In order to improve performance of the devices corresponding to the Pareto solutions, local search based on the level-set method is performed starting from those solutions.
  The present method is applied to multi-objective design problem of IPM (Interior permanent magnet motor). The goal of this design is to maximize the average torque and minimize torque ripple. It is shown from the numerical results that the quality of the Pareto solutions can be improved by the local search. Moreover, it is found that manufacture of the optimized rotor becomes easier after the local search because small flux barriers disappear and the surface of the remained flux barriers becomes smooth. It is possible to obtain diverse IPM motors which have different average torques and torque ripples by the present method. The present method, therefore, gives useful information for the design of devices and machines.