Spectral techniques for nonlinear system analysis and identification

Abstract

This article reviews some recent and current research work with emphasis on new recommended spectral techniques that can analyze and identify the optimum linear and nonlinear system properties in a large class of single-input/single-output nonlinear models by using experimentally measured input/output random data. This is done by showing how to replace these nonlinear models with equivalent multiple-input/single-output linear models that are solvable by well-established practical procedures. The input random data can have probability density functions that are Gaussian or non-Gaussian with arbitrary spectral properties. Results in this article prove that serious errors can occur when conventional linear model analysis procedures are used to determine the physical properties of nonlinear systems. © 1993 John Wiley & Sons, Inc.