Bayesian networks application to the dependability of multi-state systems
Résumé
The application of Bayesian Networks (BN) in dependability is a relatively recent
development. Their popularity grew in the area of reliability analysis of systems, since the
years 90s. A large number of scientific publications in this area shows the interest in the
applications of BN in the field of dependability and risk analysis. The most important
publications demonstrate equivalence with probabilistic methods conventionally used in
dependability. We have now a number of survey papers that gives a good view of the ability
of BN application to dependability. Unfortunately, this modeling formalism is not fully
accepted in the industry. The questions resting by today's engineers are focuses on the validity
of BN models and the resulting estimates. Indeed, the modeling formalism by BN is not based
on a specific semantic in dependability but offers a general formalism for modeling problems
under uncertainty. This plenary talk explains the principles of knowledge structuration to
ensure a valid BN model and illustrate the flexibility and efficiency of representation by BN
in various frameworks such as Fault Diagnosis and Fault Tolerant Control.