Methodological developments for probabilistic risk analyses of socio-technical systems

International audienceNowadays, the risk analysis of critical systems cannot be focused only on a technical point of view. Indeed, several major accidents have changed this initial way of thinking. As a result, there exist numerous methods that allow to study risks by considering on the main system resources: the technical process, the operator constraining this process, and the organisation conditioning human actions. However, few works propose to jointly use these different methods to study risks in a global approach. In that way, this paper presents a methodology, which is under development between CRAN, EDF and INERIS, allowing an integration of these different methods to probabilistically estimate risks. This integration is based on unification and structuring knowledge concepts; and the quantitative aspect is achieved through the use of Bayesian Networks. An application of this methodology, on an industrial case, demonstrates its feasibility and concludes on model capacities, which are about the necessary consideration of the whole causes for a system weakness treatment, and the classification of these contributors considering their criticality for this system. This tool can thus be used to help decision makers to prioritise their actions