Neural Network Modeling and Disturbance Observer Based Control of a Pneumatic System

The nonlinearity and plant dynamics make it very difficult to perform accurate control for pneumatic systems. This paper presents the results on the modeling and control of a pneumatic system. With Levenberg-Marquardt method, a neural network model, from which a third-order ARX model is derived, has been developed for the pneumatic system. Based on the built ARX model, a direct digital controller is designed with Ragazzini method. To reject external noise, a disturbance observer (DOB) is constructed to improve the control accuracy. The experimental results of the position control for the pneumatic system are given to demonstrate the performance of the DOB-based controller. The accuracy and response speed are satisfactory for both steady-state and dynamic tracking