Back to overview


Type of publication Peer-reviewed
Publikationsform Proceedings (peer-reviewed)
Author Hursin Mathieu, Siefman Daniel, Perret Gregory, Pautz Andreas,
Project Development of a Methodology for Nuclear Data Assimilation in Reactor Physics employing the PROTEUS Experimental Data Base
Show all

Proceedings (peer-reviewed)

Title of proceedings ANS Best Estimate Plus Uncertainty International Conference (BEPU 2018)
Place Lucca, Italy

Open Access


SHARK-X is a set of Perl-based tools built around the lattice code CASMO-5 and developed at the Paul Scherrer Institut. SHARK-X has been used to re-analyze the reactivity worth measurements performed in the zero power research reactor PROTEUS during the LWR-PROTEUS Phase II experimental program. The present paper presents two approaches to assess the representativity of the LWR-PROTEUS Phase II experiments to a typical spent fuel pool. The representativity indices computed using random samples (SS-route) or obtained from sensitivity coefficients (DP-route) produce consistent results provided that at least 500 samples are used. The discrepancies between the DP- and SS-routes are attributed to the calculation of the sensitivity coefficients required by the DP-route. Because the cost of the SS-route is lower than the one of the DP-route when considering a large number of uncertain inputs, it is the superior approach going forward. Next, various methods to perform global sensitivity analysis are implemented and tested in SHARK-X to obtain sensitivity information from the SS-route. Sobol and FAST-RBD methods provide accurate sensitivity indices even for non-linear functions, although the cost of estimating the Sobol indices appears prohibitive for the number of uncertain inputs typically involved in the analysis of the LWR-PROTEUS Phase II experiments. The performance of the various methods is illustrated through the sensitivity analysis of a depletion calculation on LWR fuel considering perturbations in the 238U capture cross sections. It shows that even though a certain degree of non-linearity in the k-inf of the model is observed at EOC, it is not enough to be visible with the sensitivity indices based on squared Pearson correlation coefficient. As this approach appears to be the best compromise between accuracy, ease of implementation and computational cost, it will be used for the sensitivity analysis of the LWR-PROTEUS Phase II experiments. 1.