Uncertainty analysis in probabilistic risk assessment: Comparison of probabilistic and non probabilistic approaches

Dominique Vasseur (1) , Tu Duong Le Duy (2) , Anne Dutfoy (3) , Laurence Dieulle (2) , Christophe Bérenguer (4)

1 EDF R&D MRI - Management des Risques Industriels
2 LM2S - Laboratoire Modélisation et Sûreté des Systèmes
3 EDF R&D - EDF R&D
4 STMR - Sciences et Technologies pour la Maitrise des Risques

In order to better control the safety of its Nuclear Power Plants (NPP), EDF developed Probabilistic Safety Assessments (PSA). PSA indicators are used to make decisions relative to plants design or procedures modifications, or to maintenance program optimization for example. To get robust decisions, it is necessary to take account of uncertainties in decision-making process. Uncertainties in PSA model are mainly epistemic ones and can be roughly split into two categories: parameter and model uncertainties. The treatment of these two types of uncertainty can be done in a probabilistic framework by Monte Carlo simulations for parametric uncertainties and by sensitivity studies for model uncertainties, but it can also be done in a non probabilistic framework called Dempster-Shafer Theory. In this paper both approaches are used to assess the uncertainties associated to a case study related to a specific PSA application: the precursor events analysis. The results are compared in order to identify the advantages and the drawbacks of both approaches

Domains: Computer Science [cs]; Performance [cs.PF]

Dominique Vasseur, Tu Duong Le Duy, Anne Dutfoy, Laurence Dieulle, Christophe Bérenguer. Uncertainty analysis in probabilistic risk assessment: Comparison of probabilistic and non probabilistic approaches. Advances in Safety, Reliability and Risk Management - ESREL 2011, Sep 2011, Troyes, France. ⟨hal-02586089⟩