Modeling inter-turn winding faults in switched reluctance machines based on neural network

This paper presents a new method for modeling inter-turn winding faults in switched reluctance machine (SRM) based on artificial neural network (ANN) , incorporating a simple analytical method to estimate the flux-linkage characteristics of SRM under winding faults. SRM has been proposed for use in applications requiring certain fault tolerance. It is important to distinguish and characterize the inter-turn winding faults in SRM for maintenance and diagnostic purposes. In order to build an accurate model of SRM with and without faults, the effective magnetic equivalent circuit method is used to calculate the nonlinear flux-linkage characteristics under various winding fault conditions. ANN is applied for its well-known interpolation capabilities for the highly nonlinear SRM, with phase current, rotor position and a fault condition parameter as inputs and flux-linkage as output. Then, the dynamic models for the SRM with inter-turn winding faults are constructed. The analysis of the results from the faulty machine under different control strategies is presented and verifies the good performance of the developed method.