An inner product-based dynamic neural network hysteresis model for piezoceramic actuators

A new dynamic neural network hysteresis model for piezoceramic actuators is presented in this paper. In this model, the inner product of signals is introduced to construct an input vector in order to implement a transformation from a multi-valued mapping for the hysteresis to a single-valued mapping. The dynamic RBF neural network with output feedback is used not only to represent this single-valued mapping and but also to describe the dynamic behavior of the piezoceramic actuator. The results of both simulations and experiment show that, in comparison with the PI model, the proposed model is effective and is of good generalization capability under the condition of input signal varying with frequency.