Improved micropositioning of 2 DOF stage by using the neural network compensation of plant nonlinearities

Abstract

This paper describes the system for micropositioning of a 2 DOF mechanism with piezoelectric actuators (PEAs) called a piezo actuated stage (PAS). The PAS is fabricated by a photo structuring process from photosensitive glass and PEAs are built-on to meet the request for its precise movement. The PAS is designed as a general 2 DOF stage. It can be used for different micropositioning or micro- assembling tasks according to the selected end-effector. The other components of the closed-loop control system for micropositioning of PAS are the high voltage drivers, the incremental position sensors and the control processing unit. Due to the nonlinear behaviour of the system for micropositioning, the precise position control of PAS with traditional PI controller is aggravated. Concerning the plant nonlinearities, the feedforward neural networks (NN) are used as a tool for their compensation. After the training procedure with the back-propagation (BPG) algorithm, the trained NN inverse model of plant nonlinearities is used as a feedforward part of the proposed controller. The experiment results have shown that the NN compensation improves the control performance of traditional PI controller. © 2010 Journal of Mechanical Engineering. All rights reserved.