Identification of nonlinear systems using NARMAX model

Abstract

Most systems encountered in the real world are nonlinear in nature, and since linear models cannot capture the rich dynamic behavior of limit cycles, bifurcations, etc. associated with nonlinear systems, it is imperative to have identification techniques which are specific for nonlinear systems. The problem considered in this work is the modeling of nonlinear discrete systems based on the set of input–output data. This is often the only approach to modeling, as in most cases only external (i.e. input–output) data are available. This paper also discusses the practical aspects of identification of nonlinear systems. The NARMAX (Nonlinear AutoRegressive Moving Average with eXogenous input) model provides a unified representation for a wide class of nonlinear systems and has obvious advantages over functional series representations such as Volterra and Wiener series. This model is proven to provide a better parameter estimation and prediction accuracy than the linear model.