Application Of RBF Neural Network In Sliding Mode Control System

Sliding mode variable structure control with fast response is invariant for system parameters and external disturbance. Its algorithm is simple easy to engineering realization and it made outstanding progress in solving complex nonlinear system synthesis problems in recent years.Neural network is a highly nonlinear continuous time dynamic system, and it has strong self-learning function for nonlinear systems and powerful mapping capability. Neural network in sliding mode variable structure control can realize adaptive sliding mode control.This paper does some research in the sliding mode variable structure control and RBF neural network combination by improving RBF neural network to produce superior neural network adaptive variable structure control.Firstly,RBF neural network based on center vector dynamic recursion method is proposed according to the characteristics of variable structure control and various RBF neural network on-line control algorithm. It uses dynamic recursion method to calculate the center of implicit unit nodes and also uses momentum factor method based on the classical gradient descent algorithm to adjust hidden unit nodes width and network weights. The simulation results prove the feasibility and efficiency of the algorithm.Secondly, this paper puts forward a method based on adaptive learning rate of RBF network equivalent sliding mode variable structure control by researching BP algorithm improvement measures and combining the application characteristic of RBF neural network for equivalent sliding mode variable structure control. Namely to network weights using adaptive learning rate, while for the iterative algorithm of implicit unit nodes center and width is using classical algorithm of gradient descent momentum factor method adjustment. The simulation results prove that the ability to weaken chattering is better than that of the conventional gradient descent method.Finally, this paper puts forward two methods based on GA_RBF equivalent control according to equivalent sliding mode variable structure on-line control data. They are sliding mode variable structure control based on GA_RBF networks compensating controller of radient descent algorithm and sliding mode variable structure control based on GA_ RBF networks compensating controller of K.Furuta control algorithm.These two kinds of control algorithm using the same optimum solution and optimum principle have different applications according to the collection of different data. The sliding mode variable structure control based on GA_RBF networks compensating controller of radient descent algorithm can further improve the system of position tracking precision. The sliding mode variable structure control based on GA_RBF networks compensating controller of K.Furuta control algorithm has the stability of K.Furuta control algorithm and it does not need to know the upper bound of the RBF neural network uncertain parts. Simulation results demonstrate sophistication of the above two kinds of controller.