Author : Osama. A. Montasser and B. A. El-Sayed

The main target of this work is to design an appropriate position controller for a pneumatic system using artificial neural networks. The rod position of a double acting pneumatic cylinder, controlled by proportional linear valves, was chosen as the present control system. A mathematical dynamic model for the pneumatic system was derived. The model shows, as it is expected, that the pneumatic system is of highly nonlinear features. This is due to cylinder-piston mechanical friction, the compressibility feature of air, and nonlinear characteristics of the flow through a valve orifice of variable area. The model shows, as well, that pneumatic system is of time varying characteristics. A Proposed Neural Network Controller, PNNC, is designed and implemented. The PNNC is a rule-based controller, where both the slope and amplitude of the activation function of each neuron is adapted to enhance the control system performance. A considerable improvement of the system response for different input conditions is achieved by applying the PNNC on the present control system. The robustness and effectiveness of the proposed controller were verified through computer simulations using MATLAB package and SIMULINK toolbox. A comparison with the Conventional Neural Network Controller, CNNC and the typical PID controller, assured that the present PNNC is robust and more efficient in terms of both the system stability and speed of response.

Keywords: Accurate Position Control, Adaptive Learning Algorithm, Neural Network Controllers, Pneumatic Actuators, Sigmoid Activation Function.

Article published in International Journal of Current Engineering and Technology, Vol.4, No.5 (Oct-2014)