Tracking Performance of a Positioning System with a Linear DC Servo Motor using Neural Network

One of fundamental problems in factory automation or office automation is how to obtain linear motion. Linear motors produce directly the linear motion force without a motion-transform mechanism. Linear d. c. motors (LDM) have excellent performance and controllability. The dynamics of small-sized LDMs is adversely affected by the dead-band due to the friction between brushes and commutators.This paper considers a positioning system with a linear d. c. motor using neural network. The positioning system consists of a neural servo controller and a fixed gain feedback controller. The neural servo controller requires the reference position signal, the reference speed signal, the reference accelaration signal and tracking error signal, the objective of which is to repress the influence of friction and to improve the tracking performance. The weights of the neural servo controller are adjusted by the back propagation algorithm so that the tracking error is minimized. The effectiveness of the proposed system is demonstrated by experiment.