Identification method based on NARMAX polynomials

Abstract

Advanced process control techniques use at some point a model of the process that is controlled. In real industrial processes, usually there are present nonlinearities, the time changing parameters of the equipment, noise and uncertainties. These processes are sometimes modeled by NARMAX models. The current paper approaches system modeling with NARMAX polynomials of a distillation process with the aid of the genetic algorithms technique. The high flexibility of NARMAX models is combined with the properties of genetic algorithms to search for solutions in highly dimensional and multimodal spaces in order to obtain a good prediction of the process output for a limited time interval and also to minimize process response time.