Research And Application Of PID Self-turning Control Based On RBF Neural Network

Conventional PID control was widely used in the field of chemical engineering and electric control due to its simple method and the error of model possessed robustness and stability.And it was one of the earliest control method.Conventional PID control excessively depended on the mathematical model of the object,and its parameter was difficult to confirm.Consequently,it was difficult to adapt time-varying and nonlinear system.Neural network manifested many advantage,such as mighty adaptability and study ability of multiple objective.At the same time it could approach nonlinear continuous function with random precision.So researchers proposed a new method of nonlinear control system.This method could combine the neural network with self-adapting control.Which is the PID self-turning control based on RBF neural network.In this paper,two kinds of improvement methods for RBF neural network PID self-turning control were proposed.One method was that introducing the square of PID parameter momentum factor into gradient descent method.This method could increase the amendment amount at the same gradient direction,and it could ensure that the corrected results would proceed along convergent direction.When the weight value corrected too fast,this algorithm could reduced the amendment amount as much as possible.So the renovation of weight value would overcome the non-steady in nonlinear system due to faster study rate in a certain degree.The other method was to increase the amount of input parameters in order to restrain sharp increase of scale parameter(Kp)and integral parameter(Ki).At the same time diverging time of the system would be delayed because of PID parameter.So fast divergent of the system was deferred to some account.At last,the emulation experiment was processed based on the improvement of RBF neural network and PID Self-tuning control.The conclusion showed that the two kinds of improvement brought excellent results.This improved method was used for temperature control in the growth of artificial crystal,and describe some nonlinear question such as the temperature control in the growth of artificial crystal in detail.In order to realize temperature control,the square of PID parameter momentum factor was introduced into gradient descent method.The emulation result showed that the improved method could bring excellent results for the temperature control during crystal growth.