Neural Network Composite Control of a Hydraulic Servo Motor System by Means of Smoothed Selection of Velocity and Position Compensators

This paper proposes a new Neural Network composite control to achieve good control performance in terms of both velocity and position in a hydraulic servo motor system with inherent non-linearity. This control system is equipped with two state feedback Neural Network compensators that are independently designed from a velocity and position compensator. The input signal sent to the hydraulic servo system is generated by selecting the output of two compensators in accordance with the driving condition of the hydraulic motor. This report firstly describes conducting a simulation and experiment with the condition that the angular velocity and position were constants. It secondly confirmed that Neural Network composite control is necessary in order to attain better velocity and position control performance. In addition, selection of the compensator output was studied, revealing that composite control was most effectively performed through smoothed selecting of the signal using a sigmoid function. Furthermore, experimental verification indicates that the Neural Network composite control has exceptional control accuracy and robustness.