Intelligent Control Of Hydraulic Position Servo System

Hydraulic servo system has been widely used in various industrial fields such as missile launching,aerospace and robotics because of its strong carrying capacity,fast response speed.Due to the manufacturing reasons and time-varying working conditions,it is very difficult to obtain the accurate mathematical model of hydraulic servo system.In the practical applications,hydraulic servo systems are subjected to other uncertainties such as unknown control direction and inaccurate zero of the proportional valve.Traditional model-based control strategies are difficult to deal with these problems in order to obtain satisfactory control results.Therefore,how to design a controller with strong robustness and without depending on the system model is very important for extending the applications of this system.The hydraulic servo system produced by Festo company is used as the experimental benchmark in this paper.Considering the inaccurate zero point of the proportional valve,the unknown model,the following controllers are designed to achieve high precision tracking performance for time varying reference:(1)Aiming at the uncertainty of the hydraulic position servo system,a super-twisting sliding mode controller with adaptive gain is proposed with the known model information.The system stability has been proved.And the experimental results show us the effectiveness of this method.Compared with the traditional sliding mode control,the proposed method has smaller tracking error.(2)Because the accurate system model of hydraulic position servo system is difficult to obtain,a neural network integral sliding mode control is proposed.At first,the RBF neural network is introduced to approximate the system unknown function.Then the fractional calculus is introduced,and the neural network fractional order integral sliding mode control is proposed.Considering the unknown control direction of the hydraulic servo system,neural network based on the integral sliding mode control and the fractional order integral sliding mode control are desigened,respectively.The Nussbaum gain technology is used to deal with the unknown control direction,the RBF neural network is used to approximate the uncertain function.The stability of all the proposed controllers are proved.The experimental results show that the proposed controllers can track the given reference signals very well.The fractional order integral sliding mode control method has the smallest tracking error.(3)Considering the unknown model and unknown control direction of hydraulic servo system,a backstepping adaptive control based on neural network is designed,and the system stability is proved.In order to reduce the complexity of backstepping control design,the dynamic surface control is used,and then the adaptive neural network dynamic surface control is proposed.The stability of the controlled system also is guaranteed by theoretical proof.The experimental result shows the effectiveness of the designed controller.However,the backstepping adaptive neural network control has smaller tracking error.Comparing the proposed methods in this paper with other existing methods,we can obtain the following conclusion.Whether the unknown control direction is considered or the direction is known and positive,neural network based fractional order integral sliding mode control strategy always has the best control performance,therefore the fractional order controller for the hydraulic position servo system is a good direction to improve the system performance.On the other hand,the experimental comparison results of backstepping adaptive neural network control and adaptive neural network dynamic surface control are contrary to the theoretical conclusion.The reason of this phenomenon may be that the system order is not very high,and there is function approximation in each step in dynamic surface method,therefore,the error caused by approximation will be more serious than the adaptive backstepping neural network control,therefore the error is larger as compared to the adaptive backstepping neural network control.