Abstract: In order to identify the rate-dependent hysteresis in piezoelectric actuator, a neural dynamic model based on expanded input space method is built. In this method, a novel hysteretic operator is proposed to describe the change tendency and the dynamic property of rate-dependent hysteresis. Thus, an expanded input space is constructed with the introduction of such hysteretic operator. The multi-valued mapping of the rate-dependent hysteresis can be transformed into a one-to-one mapping in this expanded input space. Furthermore, the rate-dependent property of hysteresis can be extracted by the hysteretic operator. Then neural networks can be used to approximate the one-to-one mapping and the dynamic model is constructed for the rate-dependent hysteresis. The proposed neural model can describe the rate-dependent hysteresis and has a simple architecture which is totally different from the modeling framework of the hysteresis model based on the weighted superposition of operators. Moreover, it is convenient to adjust the parameters on-line to adapt to the variation of the operating environment. Finally, this method is applied to dynamic modeling of the hysteresis in piezoelectric actuator. The experimental results prove the validity of this modeling technique.

Key words: Dynamic hysteretic operator, Expanded input space method, Modeling, Rate-dependent hysteresis