Determination of induction motor parameters by using neural network based on FEM results

A neural network (NN) is introduced for making model-building between design variables and equivalent circuit parameters in designing three phase induction motors. Some equivalent circuit parameters are calculated by finite element method (FEM) at each slip conditions and various design variables. From these results, a self-organized distributed network (SODN) is trained, then combined with per-phase equivalent circuit method. The use of the proposed method is evaluated from the results that a rapid execution time compared with original FEMs in design procedures. This method is useful for nonlinear problems and model-building problems such as parameter determination.