A New Regressive RBF Neural Network Model for Rate-Dependent Hysteresis in Reluctance Actuators

This paper presents a novel hysteresis approximation method for the rate-dependent hysteresis effect in reluctance actuators. The method is described by a regressive radial basis function (RRBF) neural network model, and its parameters are trained by a global optimization algorithm named dynamic opposite learning teaching learning based optimization (DOLTLBO). The regressive structure is adopted for solving the multivalued problem between the input and the output of the hysteresis nonlinearity, and the DOLTLBO algorithm fully exploits the approximation capability of the RRBF model, thus guaranteeing the low complexity of the neural network. Experiments were performed on reluctance actuators, and results show that the new model has high performance in hysteresis approximations for reluctance actuators when compared to previous models such as the RPI Model and the Chua model.