Jury

• Jean-François PETIN, Professor at ENSEM Nancy, President
  
• Christophe BÉRENGUER, Professor at ENSE3 Grenoble, Reviewer
• Antoine RAUZY, CR CNRS HdR, Ecole Polytechnique, Reviewer
• Marc BOUISSOU, Senior Research at EDF R&D (Equipe MRI), Examinator
• Gilles DELEUZE, Senior Research at EDF R&D (Equipe MRI), Examinator
• Jean-Jacques LESAGE, Professor  at ENS Cachan, Supervisor
• Jean-Marc ROUSSEL, Associate Professor at ENS Cachan, Co-supervisor

Keywords: System safety, Qualitative studies, Repairable systems, Qualitative
analyses, Minimal Cut Sequences, BDMP

Abstract

Preventive risk assessment of a complex system relies on a dynamic models which describe the link between the system failure and the scenarios of failure and repair events from its components. The qualitative analysis of a binary dynamic and repairable system is aiming at compute and analyse the scenarios that lead to the system failure. Since such systems describe a large set of scenarios, only the most representative ones, called Minimal Cut Sequences (MCS), are of interest for the safety engineer. The lack of a formal definition for the MCS promoted the emergence of multiple definitions either specific to a given model (and thus not generic) or informal. This work proposes:

• a formal framework and definition for the MCS while staying independent of the reliability model used,
• a method to compute them using property extracted from their formal definition,
• an extension of the formal framework for multi-states components in order to perform the qualitative analyses of Boolean logic Driven Markov Processes (BDMP)models.

Under the hypothesis that the scenarios implicitly described by any reliability model can always be represented by a finite automaton, this work defines the coherency for dynamic and repairable systems as the way to give a minimal representation of all scenarios that are leading to the system failure.