Neuro-control approach of switched reluctance motor drives

The purpose of the paper is to present several studies on neural networks used for the modelling of a switched reluctance motor (SRM) with variable structure control. A positioning system using a four-phase SRM is presented, in which the position error is processed by a sliding-mode controller. The control unit represents the subject of a neural network-based model. The proposed network system has a feedforward type architecture, structured on three layers of processing units. The networks are trained using the BKP algorithm. Once the network system is trained, it is integrated as a part of the positioning system. The training and testing sets of examples are obtained by numerical simulation of the positioning system using the Matlab environment.