A model-based theory of meta-diagnosis: what is wrong in diagnostic systems?

Yannick Pencole (CNRS-LAAS, France)

LOGIC AND COMPUTATION SEMINAR ANU-NICTA Logic Seminar

DATE: 2011-06-07
TIME: 15:30:00 - 16:30:00
LOCATION: RSISE Seminar Room, ground floor, building 115, cnr. North and Daley Roads, ANU
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
Every model-based diagnostic approach relies on a representation of a real-world system and most common academic approaches assume that such a representation meets some "good requirements" (like completeness, correctness, coverage,...) so that the quality of the diagnosis is "guaranteed". In this talk, we just go the opposite way by stating first, that diagnostic systems are in charge of diagnosing real and complex applications based on models that do not meet these "requirements". In other words, a real diagnostic system may fail with some economical consequences. After recalling the classical model-based diagnosis of Reiter, we propose here to list the properties required by "good" diagnostic systems and to then describe a logical framework to diagnose the diagnostic system (what is called the meta-diagnosis problem). This problem will be illustrated in the context of maintenance in aeronautical systems. The presented work results from a collaboration between LAAS-CNRS and AIRBUS, Toulouse, France. This talk will basically summarize the two following papers:

"Defining and exploring properties in diagnostic systems" Nuno Belard, Yannick PencolA and Michel Combacau. DX-10.

"A theory of meta-diagnosis: reasoning about diagnostic systems" Nuno Belard, Yannick PencolA and Michel Combacau. IJCAI-11.

BIO:
See http://homepages.laas.fr/ypencole/