Uncertainty analysis by Dempster-Shafer theory in probabilistic risk assessment

Tu Duong Le Duy (1) , Dominique Vasseur (2) , Mathieu Couplet (2) , Laurence Dieulle (1) , Christophe Bérenguer (1)

1 LM2S - Laboratoire Modélisation et Sûreté des Systèmes
2 EDF R&D MRI - Management des Risques Industriels

The traditional approach of parameter uncertainty analysis in Probabilistic Risk Assessment (PRA) often relies on the probability approach in which an assumed probability distribution, e.g. a log normal one, is used to represent parameter uncertainty. Such an approach has been recognized to be questionable and the Dempster-Shafer Theory (DST) has been recently proposed as an alternative to probability theory for modeling the uncertainty in a more appropriate way. However, since the mathematical structures of uncertainty representation within the DST are different from those of probabilistic approach, the output results in this framework are given in terms of interval values which may not be easy to use in decision making. In this paper, we propose to use the Transferable Belief Model interpretation of the DST from which the output results of uncertainty analysis can be applied to the current decision making process in risk assessment. The proposed approach is illustrated through a practical example from EDF Nuclear Power Plants PRA application

Domains: Performance [cs.PF]

Tu Duong Le Duy, Dominique Vasseur, Mathieu Couplet, Laurence Dieulle, Christophe Bérenguer. Uncertainty analysis by Dempster-Shafer theory in probabilistic risk assessment. European Safety and Reliability Conference - ESREL 2010, Sep 2010, Rhodes, Greece. ⟨hal-02586209⟩