A neural inverse model for the dynamic hysteresis in the piezoelectric actuator using the hysteretic inverse operator

The hysteresis inherent in the piezoelectric actuator shows dynamic behavior dependent on the change-rate of the input. To compensate the effect of the dynamic hysteresis, an neural inverse model is proposed in this paper. Therein, to solve the problem that the neural networks can not approximate the multi-valued mapping of the inverse hysteresis, a hysteretic inverse operator is proposed firstly, which constructs the hysteretic memory and describes the rate-dependent property of the dynamic inverse hysteresis. Then, with the introduction of the hysteretic inverse operator and its input change-rate into the input space, a newly three-dimensional input space is constructed, on which, the output of the dynamic inverse hysteresis, i.e. the input voltage, is can be uniquely specified. The proposed model is of simple structure and available for the on-line tuning for the adaptation to the environmental changes. Finally, the experimental results are presented to show the effectiveness of the proposed approach.