A Neural Hysteresis Model for Smart-Materials-Based Actuators

In this paper, a constraint factor (CF) is presented. The CF and an odd m-order polynomial form a new hysteretic operator (HO) together. And then, an expanded input space is constructed based on the proposed HO. In the expanded input and output spaces, the one-to-multiple mapping of hysteresis is transformed into a one-to-one mapping so that a neural network can be used to develop a neural hysteresis model. The model parameters are computed by using the least square method. Finally, the neural hysteresis model is employed to approximate a real data from a magnetostrictive actuator in an experiment. The experimental results demonstrate the proposed approach is effective.