A neural network approach to robust model-based diagnosis of faults in a three-tank system

Author

Marcu, Teodor ; Mirea, Letitia ; Klósz, Attila

Author_Institution

Dept. of Autom. Control & Ind. Inf., Tech. Univl of Iasi, Romania

fYear

1996

Firstpage

111

Lastpage

116

Abstract

The problem of robust model-based diagnosis of faults is addressed with application to a three-tank system. The present approach is based on artificial neural networks used as predictors of dynamic nonlinear models for residual generation, and as pattern classifiers for residual evaluation. A diagnosing subsystem is implemented in real-time using the Simulink/Matlab environment

Keywords

fault diagnosis; multilayer perceptrons; nonlinear dynamical systems; parameter estimation; pattern classification; process control; Matlab; Simulink; dynamic nonlinear models; identification; model-based fault diagnosis; multilayer perceptron; neural network; nonlinear systems; parameter estimation; pattern classification; real-time systems; residual generation; three-tank system; Artificial neural networks; Autoregressive processes; Fault diagnosis; Intelligent networks; Neural networks; Nonlinear dynamical systems; Pattern recognition; Predictive models; Robustness; Signal generators;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer-Aided Control System Design, 1996., Proceedings of the 1996 IEEE International Symposium on

Conference_Location

Dearborn, MI

Print_ISBN

0-7803-3032-3

Type

conf

DOI

10.1109/CACSD.1996.555240

Filename

555240