Fast 3-D Optimization of Magnetic Cores for Loss and Volume Reduction

This paper presents an effective design method for inductors which is based on the multi-objective optimization accelerated by the artificial neural network (ANN). In the learning phase prior to the optimization phase, ANN is trained for 1000 input-output data sets obtained from the finite-element analysis for randomly generated dimensional parameters. The magnetic hysteresis of the ferrite core is modeled by the play model to evaluate the hysteresis losses. The multi-objective optimization problems are solved by the genetic algorithm in which the magnetic loss is effectively computed by the trained ANN to reduce the core volume as well as magnetic loss. The Pareto solutions for an EI-shaped ferrite core are obtained for different inductances. It is shown that the proposed method works much faster than the conventional optimization, and the magnetic loss and the inductance of the optimized inductor agree well with the experimental results.