Research For Non-linear Friction Compensation Model In Motion Control Based On Neural Network

Motion control has been becoming a focal research at present, and research for non-linear friction compensation model in motion control system has been a prosperous task which achieved more attention by control researcher. There are many non-linear character influence such as stable friction, moving friction, viscosity friction and saturation of perform framework during velocity and position tracking control in the motion control system, for example, robot, machine tool, wireless antenna and chronometer telescope.In this paper, author make some simulation research for DC servo motor with non-linear friction by SIMULINK software in MATLAB. Firstly, I choose PID controller. There are "flat peak" phenomena in position tracking and "dead zone" in velocity tracking. It is unacceptable for the motion control system which need high precision tracking and motion stability. Then I adopt direct moment forward compensation method, add an effect which has same value but contrary sign with the disturbing in the forward loop by doping the disturbing value. This way can conquer non-linear friction influence to great extend and eliminate "flat peak" phenomena in position tracking and "dead zone" phenomena in velocity tracking. Last I try to apply BP neural network to system model ing and control in order to eliminate influence of non-linear friction . As a result, this way can make system acquire better capability target.