Multiple fault detection and isolation

Philippe Weber (1) , Sylviane Gentil (1) , Patrique Ripoll (2) , Laurent Foulloy (2)

1 LAG - Laboratoire d'automatique de Grenoble
2 LAMII - Laboratoire d'Automatique et de MicroInformatique Industrielle

Model-based fault detection methods allow the generation of residuals as fault indicators. Isolation is generally based on the incidence matrix structure. Simultaneous faults generate a new fault signature, corresponding to the superposition of the fault effects. With classical decision methods, each fault combination results in a new residual configuration, leading to an extra column in an extended incidence matrix. This solution is extremely combinatorial. This work uses the structural properties of the incidence matrix corresponding to single faults to reason about fault combinations. This reasoning is implemented as a fuzzy inference system to take uncertainties into account. An implementation of this algorithm has been applied to an automotive engine. Copyright © 1999 IFAC

Mots clés: diagnosis; fault detection; fault isolation; fuzzy-logic; decision making; automotive engine.; automotive engine

Domaines: Automatique / Robotique

Philippe Weber, Sylviane Gentil, Patrique Ripoll, Laurent Foulloy. Multiple fault detection and isolation. 14th IFAC World Congress, 1999, Beijing, China. pp.223-228. ⟨hal-00092895⟩