Pid Control Method Based On Improved Bp Neural Network Research

Neural network theory and application come to front in recent years. Because of its strong self-study ability, parallel processing ability, information integration ability and error tolerance ability, Neural network got abroad application in system detection and control. Neural network can approach any complex nonlinear system fully. All quantificational analyse and qualitative analyse are distributed and stored in the nerve cells of neural network evenly, so it can study and adapt dynamic characteristic of uncertain system.BP Neural network PID controller are researched in this paper. Firstly, BP arithmetic was improved, and a new method of adding multiple momentum was presented. This method can reduce the times of network training. Then, the structure and some parameters of neural network were lucubrated. At last, a kind of NNPID control method was proposed and emulated. The results indicate that BP Neural network PID controller has great tracking capability.Aiming at the simple servo system with Lugre friction model, the simulation displayed that High-plus PD controller is not good enough in friction compensation. A new friction-compensation method which was based on BPNN-PID controller was put forward with the principle graph. And the simulation results indicated that improved BPNN-PID controller can compensate the friction of servo system very well.