Modeling a two-phase excitation switched reluctance motor with artificial neural network

This paper first introduces the necessity to adopt feed-forward (FF) artificial neural network (ANN) in approximation of magnetic characteristics for a two-phase excitation (TPE) switched reluctance motor (SRM) modeling. Then the magnetic characteristics of a TPESRM are trained by a learning algorithm named MARQUARDT algorithm. The first step of the training is the selection of net structure and learning algorithm. Then the preparations of the sample data are explained. Its main objective is to reduce the total number of samples effectively. Finally, the forward, inverse flux-linkage characteristics and the co-energy characteristics are successfully trained. The training results are acceptable for engineering applications.