Neural network-based position sensorless control for transverse flux linear SRM

Abstract

The purpose of this paper is to present a sensorless control method for transverse flux linear switched reluctance motor (TFLSRM) based on position estimation by employing a public back propagation neural network (BPNN). The system characterizes with that only one public BPNN is needed to transform the winding current and flux linkage of each phase into each segmental position signal, then final total position is obtained by combining each segmental position signal. The starting position is derived from the comparison of the calculated position values based on currents and flux linkages of all phases. A TFLSRM with three phases is used to verify the validity of the proposed method, the established position sensorless control system with a BPNN is simulated. The results illustrate the excellent performance of the BPNN-based position sensorless control system. © Springer-Verlag Berlin Heidelberg 2007.