Séminaire MOCOSY : Challenges of Model Based Diagnosis in Automotive Systems
Intervenant:    
Mattias Nyberg
SCANIA et Linkoping University
Date : 24 avril 2009
Heure : 11h00
Lieu :
 Salle Europe
LAAS, 7 Av. de Colonel Roche
Résumé :

 Automotive systems is a most challenging area of diagnosis. Complexity, strict regulations for on-board diagnosis, importance of uptime, economic pressure and competition are the key driving factors. A model based approach seems to be a promising solution to these challenges. The talk will argue for the different needs of model based diagnosis in automotive systems. A framework is laid out and key issues are identified.

Two main usages of model based diagnosis are: to design residuals in diagnostic tests, and to perform fault isolation. For residual generation, non-linear systems must be handled and the method of sequential derivation of signal estimates together with redundant model equations has shown to be a very promising approach. Some new results on this sequential residual generation approach are presented. For fault isolation, consistency based approaches or Baysian networks should be the method of choice. These two approaches to fault isolation are compared. Practical experiences from working with model based diagnosis in industry are shared, and some results from application studies are presented.