Modeling hysteresis for magnetostrictive actuators

In this paper, a new method to construct neural-network-based hysteresis model is proposed. At first, a new backlash operator is presented so as to construct a basic model. And then, based on the basic model, a two-dimension input space is established so that the one-to-multiple mapping of hysteresis can be transformed into a one-to-one mapping which can be identified by using traditional method of neural networks. Finally, two experiments were implemented to validate the proposed neural hysteresis model. The experimental results demonstrate that the approach is effective.