Inverse model based neural controllers with adaptive integral part

This paper deals with intelligent controller design using artificial neural networks (ANN) in the role of feedback controllers. Neural controllers are built up and trained as inverse neural process models. Their performance and robustness are, gradually, improved and augmented by introducing, first, an adaptive simple integrator and, then, a controller with fuzzy integrator part. The proposed ANN control system performance is demonstrated using examples of both: control of a perturbed linear system of second order and a non-linear continuous biochemical process with simulated uncertainties. MATLAB programme package environment has been used to build up and train the ANN feedback controllers.