Inverse Model Approximation Using Iterative Method And Neural Networks With Practical Application For Unstable Nonlinear System Control

In this paper a method for controlling and stabilizing an unstable nonlinear system using a NARX neural network is presented. It is difficult to design a common feedback controller or even perform system identification on unstable systems, more even so if these systems are nonlinear. To compensate for nonlinearity a feedforward controller is required. In this paper we present a method of estimating inverse model of the system for the feedforward controller using a NARX artificial neural network in an iterative approach which takes less time than methods commonly used and performs as good. This method is verified and tested on an educational model of magnetic levitation of steel ball. Both static and dynamic forms of the inverse model are presented and evaluated with positive results.