A rotor position estimator for switched reluctance motors using CMAC

Abstract

This paper presents an approach to rotor position estimation in switched reluctance motors (SRMs) by using a cerebellum model articulation controller (CMAC). Previous research has shown that an artificial neural network (ANN) forms an efficient mapping structure through measurement of the flux linkages and currents for the phases. A CMAC is investigated in this paper in order to overcome the high computational power requirement problem that is encountered in a feedforward ANN based rotor position estimator. The CMAC structure does not contain neurons with activation functions, and all mathematical operations are performed without multiplication. These simplicities increase the throughput in real time implementation performed with conventional embedded controllers. However, the distributed memory structure of a CMAC requires more space. The issues involved in designing, training and implementing a CMAC are presented. In order to demonstrate the feasibility of the concept, a 20 kW, 6/4, three phase SRM is studied with training and evaluation data, which are obtained from a simulation program. A CMAC that is based on experimentally measured training and testing data for the same SRM is also used to demonstrate the promise of this approach.