The Feedforward Friction Compensation of Linear Motor Using Genetic Learning Algorithm

Abstract

This paper proposes a feedforward friction compensator based on LuGre friction model. The various parameters in both the friction model and the system plant model would be coarsely estimated by the various experiments, and then the genetic algorithm (GA) finely optimizes the key parameters by a single identification experiment. When compared with the conventional black box learning algorithm, this model-based compensator uses only five parameters to model the nonlinear friction phenomenon and the corresponding convergent rate of parameters is fast in the learning process. Finally, the friction compensated performance of proposed algorithm is evaluated and compared with the traditional uncompensated system. The simulated and experimented results show that the velocity tracking error is drastically improved by the feedforward friction compensator in a linear motor motion system.