UNCERTAINTY ANALYSIS IN CATHARE2 :DETERMINATION OF THE BASIC UNCERTAINTIES BY CIRCE

摘要

This paper presents Circe, a methodology for determining the uncertainties of the constitutive relationships in Cathare 2 code. It is a compulsory step for calculating the final uncertainty of a given response, or in other words for performing any uncertainty analysis.rnFirst, the general principles of Circe are explained. The problem is rather difficult because the constitutive relationships (also called correlations) cannot be directly measured in experiments used to establish them. The results of these experiments are consequently considered to obtain information about the correlations. More precisely, a statistical approach is used, combining the differences between code and experiment results. The uncertainties are given in the form of a covariance matrix of the parameters associated with the correlations.rnThen, the application of Circe to an experiment where several correlationsrnare influential is presented. Unlike other classical sensitivity methods, Circe can be applied even in such a complex case.rnFinally, an improvement of Circe is presented. The study of a correlation can be made more precise, if this correlation is more or less well fitted, depending on the considered field (field of void fractions, or pressures, for example).