Sensorless position estimation for variable-reluctance machines using artificial neural networks

This paper presents a new approach to the sensorless control of a variable-reluctance machine (VRM). The basic premise of the approach is that an artificial neural network (ANN) forms a very efficient mapping structure for the nonlinear VRM. Through measurement of the flux linkages and currents for the phases, the neural network is able to estimate the rotor position, thereby facilitating elimination of the rotor position sensor. The paper presents a discussion of the issues involved in designing, training and implementing the neural network. In order to demonstrate the feasibility of the concept, a 20 kW, 6/4, three-phase VRM is studied with training and evaluation data which are obtained from a simulation program. A neural network, based upon experimentally measured training and testing data for the same VRM, is also used to demonstrate the promise of this approach.