Modeling Preisach Hysteresis Nonlinearities of Neural Networks

An element hysteresis model (EHM) is constructed through the technique of continuous transformation, and produces a one-to-one relation between input and output space of hysteresis nonlinearity. The output of EHM is used as one of inputs of neural network (NN), and therefore, an EHM-based neural network model of hysteresis is built, and which can predict any kind of hysteresis nonlinearities which adapt a kind of input signal. For two sets of data from different material, respective test is carried out. These results indicate the proposed approach is simple and effective