Identification of Fuzzy Relational Models for Fault Detection

Patrick Amann (1, 2) , Jean-Marc Perronne (1) , Gérard Gissinger (1) , Paul Martin Frank (2)

1 IRIMAS - Institut de Recherche en Informatique Mathématiques Automatique Signal
2 Fachgebiet Mess- und Regelungstechnik

This paper presents the concept of fuzzy relational models for use in a fuzzy output estimator. A suitable field of application is in fault diagnosis, where output observation rather than state observation is needed for the generation of fault reflecting residual signals. Due to their non-linear structure, fuzzy relational models can be used appropriately for building models of non-linear dynamic systems. In this paper, the identification of fuzzy models for residual generation is discussed. Emphasis is placed upon the model-building procedure including the identification of the model structure and of the parameters. As an application example, a real technical system is considered. The case study presents the detection of oversteering of a passenger car. The results of the application to residual generation are discussed.

Mots clés: Fuzzy relational model; Fuzzy output observer; Model identi"cation; Fault detection; Residual generation

Domaines: Génie logiciel [cs.SE]

Patrick Amann, Jean-Marc Perronne, Gérard Gissinger, Paul Martin Frank. Identification of Fuzzy Relational Models for Fault Detection. Control Engineering Practice, 2001, 9, pp.555-562. ⟨10.1016/S0967-0661(01)00016-8⟩. ⟨hal-00954560⟩