Impact of experimental correlation on transposition method carried out with critical integral experiments

In order to estimate the bias on the effective multiplication factor (k_eff) of a criticality application case, and the associated uncertainty due to Nuclear Data (ND), a method which takes advantage of the integral experiments information and is based on ND sensitivity/uncertainty analyses and adjustment, has been implemented in a tool called RIB Such methodology has been worldwide studied for years in various fields. The RIB tool, developed at CEA, is related to the experimental validation database of the French criticality-safety package, CRISTAL V2.0, containing more than 2000 experiments from the International Criticality Safety Benchmark Evaluation Project handbook (ICSBEP) and French experimental programs. In most cases, even if a correlation is identified between the experiments, the value of this correlation might not be known. Validation tests of the RIB tool, in particular by arbitrary applying strong correlations between integral experiments (>0.9), point out some unrealistic results: a strong deviation of the bias with significant reduction of the uncertainty due to ND. To check the RIB tool implementation, transposition method equations have been implemented in Matlab. Another tool developed at CEA and dedicated to ND evaluation, CONRAD (Code for Nuclear Reaction Analysis and Data Assimilation), has also been used to estimate bias and uncertainty due to ND through the transposition method using integral experiments. Results obtained with these three tools from several combinations 'Application case/integral experiments', have been compared, confirming the RIB observed tendencies. Focusing on the post-adjustment ND, it seems that such phenomenon is associated to a strong variation of the cross sections which may be due to Peelle's pertinent puzzle effect. This paper describes the different tools and presents the results obtained for the tested combinations 'Application case/integral experiments' in function of the experimental correlation factor. Some potential explanations of the observed results, using strong experimental correlation factors, are discussed. (authors)