Precision control of piezoelectric actuator using support vector regression nonlinear model and neural networks

Abstract

Due to the inherent hysteresis nonlinearity, the piezoelectric actuator always causes positioning error in the opening system and instability in the closed system. In order to improve the positioning accuracy and response speed, a hybrid controller is proposed. The proposed hybrid controller consists of a feed-forward controller based on support vector machine for regression nonlinear model of piezoelectric actuator and a self-tuning PID feedback controller based on BP neural network. The feed-forward controller based on support vector regression nonlinear model is used to describe the nonlinearity of piezoelectric actuator and improve the response speed; the feedback controller consists of a gain tuning neural work and a variable gain PID controller, the neural network is trained to learn the optimal gains of the PID controller, and the PID controller is used to decrease steady-state error. In the end, the control precision of regular PID controller, PID feedback controller with support vector regression nonlinear model in feed-forward loop and hybrid controller is compared by experiments, the related experiment results show the proposed hybrid controller has good performance for precise tracking control of piezoelectric actuator.