Friction Compensation Based On Bp Neural Network Algorithm

Friction caused low speed creep and tracking errors which presents in servo systems has become an impediment to improve their performance. So it is important of compensating the friction to eliminate the effect of friction. The present situation of friction modeling and friction compensation at home and abroad are first summarized. The shortcomings of these studies are also presented and future trends in this field are also described, and the friction compensating using neural network is put forward. Then the theories and development of neural network is introduced. After analyzing the advantages of neural network identification compared to traditional identification, the BP neural network is used to identify the system. The network structure, algorithm principle and network design are deeply studied. Aiming at the shortcoming of studying slowly and easily stepping into the local minimum problem, some improved algorithms are put forward. The simulation results that adopts the improved algorithm(LM algorithm) is gained. The results prove that it is effective. The traditional PID is introduced and some adjusting methods are also brought out. Aiming at the complicated structure and serious nonlinear of controlled object, the neural network combied the neural PID is put forward, at the same time, the design of neural PID is gained in detail. Finally, using the strongly simulation function of MATLAB, the experiment of nonlinear time-variable system is done and proved that it is effective.The components, working principle, feature of PWM power amplifier are introduced briefly in the dissertation. The working characteristics of...