Study On RBF Neural Network Intelligent Control For The Electro-Hydraulic Servo System

Electro-hydraulic servo control system, with the characteristics of high precision, quick response, easy to adjust, can control large inertia and make high power output, therefore it have a wide range of applications in the field of industrial control. But the electro-hydraulic servo control system is essentially non-linear,it have the feature of multivariate,strong coupled and nonlinear. When using traditional PID control, the control feature of system is sensitive on model error. When the behavior of the system changes in more big range, the accuracy of the total system controlling is not high and is dissatisfactory with the need of system controlling. The high nonlinear adaptive feature of neural network controller is effective in solving this problem.In this paper, with the electro-hydraulic servo control system of flight simulator is as the research object, the design of the neural network controller used in electro-hydraulic control system and parameter adjustment method is conducted in-depth study by means of simulation experiments.With a view to improve the use of neural network controller at high time-varying strongly nonlinear systems to obtain a good overall control performance. This paper presents a dual-network structure of RBF neural network direct inverse control method. Initially, the mathematical model of the single-channel electro-hydraulic servo control system for flight simulator,and then obtained the input and output data based on the model system. Reuse the RBF neural network and this set of data to established the inverse model for the electro-hydraulic position servo control system. Tooking the inverse model from the identification obtained as the controller and embedded control and simulation. And then adjusting the parameter of the neural network controller based on the situation of response making use of gradient descent algorithm and least-squares algorithm. make full use of neural networks of arbitrary nonlinear function approximation capability. After using a neural network to identificate the inverse dynamic model of the system. then use the network as a controller directly on line control, according to the error to adjust the controller parameters; At last builting flight simulator experiments platform and through experimental verificate that the neural network direct inverse control method with a special learning algorithm is effective. Simulation and experimental results show that this control strategy can improve the system dynamic characteristics, reducing the steady-state error of the system and improve the adaptive capacity and anti-jamming, meet performance requirements.