Contribution of the mathematical modelling of knowledge to the evaluation of uncertainty margins of a LBLOCA transient (LOFT-L2-5)

This work is devoted to some recent developments in uncertainty analysis of the computer code responses used for accident management procedures in nuclear industry. The classical probabilistic approach to evaluate uncertainties is recalled. In this case, the statistical treatment of the code responses is based on the use of order statistics. It provides direct estimations of relevant statistical measures for safety studies. However, the lack of knowledge about uncertainty sources can deteriorate the decision-making. To respect the real state of knowledge, a second model, based on the Dempster-Shafer theory is introduced. It allows to mix the probabilistic approach with the possibility theory that is more appropriate when few information is available. An application of both methodologies to the uncertainty analysis of a LBLOCA transient (LOFT-L2-5) is given