Authors: Mohamed El Hadi Lebbal, Ghaleb Hoblos, Houcine Chafouk, Dimitri Lefebvre
Addresses: IRSEEM, ESIGELEC, avenue Galilee, St-Etienne du Rouvray 76801, France. ' IRSEEM, ESIGELEC, avenue Galilee, St-Etienne du Rouvray 76801, France. ' IRSEEM, ESIGELEC, avenue Galilee, St-Etienne du Rouvray 76801, France. ' Groupe de Recherche en Electrotechnique et Automatique du Havre (GREAH) Universite du Havre, 25 rue P. Lebon BP 1123, Le Havre 76063, France
Abstract: This article proposes new modelling and diagnosis methods in order to detect and isolate actuators, sensors and functional faults, and able to appear on hybrid non-linear systems. These processes are with discrete inputs and continuous outputs. Our modelling method uses a multi-model approach permitting to uncouple continuous and discrete behaviours. The discrete inputs variables are used to define operating modes and commutation conditions sets. A variable change is proposed in order to reduce the discrete states number and rewrite non-linearities such as x1/2 and sign(x) in polynomial form. Then, residual generation is obtained using polynomial approach based on Grobner bases. To solve fault detection task of studied hybrid system, multi-model approach is explored. Once the fault is detected, signatures table is used for faults isolation. In order to improve the isolation and the discernability of faults, a technique based on fault magnitude estimation combined with the knowledge of discrete inputs has been proposed. Finally, this work is illustrated on a two tanks system.
Keywords: fault detection; fault isolation; fault identification; Grobner basis; hybrid dynamic systems; HDS; structured residual generation; nonlinear systems; modelling; fault diagnosis.
International Journal of Intelligent Systems Technologies and Applications, 2009 Vol.7 No.2, pp.137 - 156
Available online: 26 May 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article