On the use of system identification for accurate parametric modeling of nonlinear systems using noisy measurements

This paper gives a survey of identification methods for static, strongly nonlinear systems. A robust estimation algorithm is proposed for the estimation of static, nonlinear systems which can be described as a nonlinear function corrected with a rational form. The errors-in-variables-based algorithm solves the starting-value problem using an iterative, weighted least-squares procedure, constructs the rational form such that the set of normal equations becomes best conditioned, and uses a maximum-likelihood estimation step to increase the efficiency of the estimates. Properties discussed are illustrated by the measured static, strongly nonlinear dc characteristic of a forward-biased p-i-n diode.