A supervised learning method of neural networks in a non-linear and time depended control process

The task of controlling the performance of a pneumatic positioning system has always be and still remains, a challenge for researchers in the area of control. The unpredictable system behavior arises not only from the non linear nature of the system dynamics but also from the existence of energy loses after long term operations. Therefore a huge effort is spent throughout multiple control approaches in order to minimize the position error of a pneumatic piston. In this paper an attempt to achieve highly piston position accuracy is implemented based on a modified artificial neural network technique. The so-called "radial basis function" was applied, improved the response of a real pneumatic rig whilst all experimentation results are recorded in this paper.