On the estimation of stochastic Wiener-Hammerstein type systems with non-smooth nonlinearity

Author

Chen, C.-H. ; Fassois, S.D.

Author_Institution

Dept. of Mech. Eng., Nat. Chinyi Inst. of Technol., Taiching, Taiwan

Volume

3

fYear

1997

fDate

4-6 Jun 1997

Firstpage

1494

Abstract

A time-domain method for the estimation of stochastic Wiener-Hammerstein-type systems with non-smooth nonlinearity is presented. The method includes two estimation modes: 1) initial parameter estimates are obtained through a sequence of stages consisting of linear regression and filtering operations; 2) the estimates are refined via a maximum likelihood procedure based upon gradient-free optimization. The proposed method overcomes the drawbacks of correlation/frequency domain schemes, offering statistically optimal estimates from a single data record of normal operating data. Its effectiveness is demonstrated via the estimation of a DC generator characterized by a combined dead-zone/saturation nonlinearity

Keywords

DC generators; estimation theory; filtering theory; maximum likelihood estimation; optimisation; stochastic systems; time-domain analysis; DC generator; Wiener-Hammerstein type systems; filtering; linear regression; maximum likelihood estimation; nonsmooth nonlinearity; optimization; parameter estimation; stochastic systems; time-domain method; Filtering; Frequency domain analysis; Frequency estimation; Linear regression; Maximum likelihood estimation; Nonlinear filters; Optimization methods; Parameter estimation; Stochastic systems; Time domain analysis;

fLanguage

English

Publisher

ieee

Conference_Titel

American Control Conference, 1997. Proceedings of the 1997

Conference_Location

Albuquerque, NM

ISSN

0743-1619

Print_ISBN

0-7803-3832-4

Type

conf

DOI

10.1109/ACC.1997.610751

Filename

610751