[article] 
in Sciences, Technologies et développement > N° 06 [01/01/2010] . - p. 49

Titre :	A multiple-model based sensor fault detection and isolation approach for nonlinear systems
Type de document :	texte imprimé
Auteurs :	Yamina Menasria, Auteur ; Nasr Eddine Debbache, Auteur
Année de publication :	2010
Article en page(s) :	p. 49
Langues :	Anglais
Mots-clés :	Nonlinear systems  Robust multiple observer  LMI formulation  Eigen-value assignment  Residual generation
Résumé :	In this paper we discuss an approach applied to sensor fault detection and isolation. Our technique is based on modelling nonlinear system with multiple model. As starting point, with the nonlinear model we determine static and nonlinear evolution of state variables. With the nonlinearities, we calculate the activation functions and also local models by linearization technique. Each unknown input local observer is constructed by the use of each local model and the multiple observer is obtained by an interpolation with activation functions like we have done for the multiple model. The optimal observer is calculated by the mean of the LMI approach and the eigen-values assignment. For fault detection and isolation we produce signals named residuals witch are chosen the output errors and are built in a DOS scheme to facilitate sensor fault isolation.

[article] A multiple-model based sensor fault detection and isolation approach for nonlinear systems [texte imprimé] / Yamina Menasria, Auteur ; Nasr Eddine Debbache, Auteur . - 2010 . - p. 49.
Langues : Anglais
in Sciences, Technologies et développement > N° 06 [01/01/2010] . - p. 49

Mots-clés :

Nonlinear systems  Robust multiple observer  LMI formulation  Eigen-value assignment  Residual generation

Résumé :

In this paper we discuss an approach applied to sensor fault detection and isolation. Our technique is based on modelling nonlinear system with multiple model. As starting point, with the nonlinear model we determine static and nonlinear evolution of state variables. With the nonlinearities, we calculate the activation functions and also local models by linearization technique. Each unknown input local observer is constructed by the use of each local model and the multiple observer is obtained by an interpolation with activation functions like we have done for the multiple model. The optimal observer is calculated by the mean of the LMI approach and the eigen-values assignment. For fault detection and isolation we produce signals named residuals witch are chosen the output errors and are built in a DOS scheme to facilitate sensor fault isolation.