A Type of Radial Basis Function Technique for Control and Time Series Prediction of Positioning Systems

It is well known that pneumatic positioning systems are still irreplaceable in many application fields like industrial automation. The maintenance cost, the low level pollution and the high speed of operation, force the use of such systems. The pneumatic piston position control has always been a challenge and engineers have applied many control methods in order to achieve position accuracy. Apart from air compressibility, the most important issue to be solved is the highly nonlinear phenomena inside the cylinder body that become unpredictable over time and long term operations of the system. Multiple friction forces, energy losses and sealing deformations are always present in this type of actuating process. In this research work, an intelligent control approach is implemented for the task, in an attempt to overcome the classical control methods inefficiency. A subcategory method of artificial neural networks is adopted for investigation, which is described in details. All experimentation results, system performance behaviour discussion and possible further improvements, form the rest of this paper body. Key-Words: Artificial Neural Networks, Nonlinear Control, Time Varying Positioning System