Compensation of hysteresis in piezoelectric actuators without dynamics modeling

Abstract Along with the fast development of ultra-precision motion systems, model and control of piezoelectric actuators draw significant research interest. Unfortunately, one of the main obstacles hindering their applications is the hysteresis, which is not ignorable in applications that require a high accuracy in motion control. In this paper, a new NARMAX model based on BP neural network is proposed for modeling the nonlinear hysteresis behavior in piezoelectric actuator. Because the proposed model is constructed in an online way, there is no need to conduct experiments for parameters identification. In order to demonstrate the precision and the rate-dependent property of the proposed model, experiments are performed under designed excitations with different amplitudes and frequencies. Furthermore, taking advantage of the proposed model, we design a nonlinear controller based on adaptive inverse control for the compensation of hysteresis in the piezoelectric actuator. A feature of the developed controller is that it allows piezoelectric actuator which exhibits complex nonlinear hysteresis behavior to be directly controlled without dynamics modeling. Thus the developed controller can be generally used for various piezoelectric actuators with different dynamics, leading to favorable convenience in industrial applications. Because of essentially open-loop characteristics of the proposed control system, the instable phenomenon aroused by feedback can be avoided. Experiments are conducted to validate the performance of the developed control system. It is shown that the developed controller not only compensates the hysteresis in the piezoelectric actuator effectively, but also exhibits rate-dependent property that allows piezoelectric actuator to track both periodic and non-periodic motions accurately.