Uncertainty Propagation Methodology and Nuclear Data Sensitivity Analysis in 3D Heterogeneous Problems

Abstract

This paper presents a 3D uncertainty propagation methodology and its application to the case of a small heterogeneous reactor system (“slab” reactor benchmark). Key neutron parameters (keff, reactivity worth, local power, …) and their corresponding cross-section sensitivities are derived by using the French calculation route APOLLO2 (2D transport lattice code), CRONOS2 (3D diffusion code) and TRIPOLI4 (3D Monte-Carlo reference calculations) with consistent JEF2.2 cross-section libraries (punctual or CEA93 multigroup cross-sections) and adapted perturbation methods (the Heuristically-based Generalized Perturbation Theory implemented in the framework of the CRONOS2 diffusion method or the correlation techniques used in Monte-Carlo simulations). The investigation of the slab system underlined notable differences between the 2D/3D computed sensitivity coefficients and consequently a priori uncertainties (when sensitivity coefficients are combined with covariance matrices the discrepancies rise up to 20% due to thermal and fast flux variations). In addition, the induced local power effect of nuclear data perturbations (JEF-2.2 vs. Leal-Derrien-Wright-Larson 235U evaluation) had been be correctly estimated with the standard 3D CRONOS2 depletion calculations. For industrial applications (PWR neutron parameters optimization problems, R&D studies dealing with the design of future fission reactors, …), the same calculation route could be advantageously applied to infer the target accuracies (knowing the required safety criteria) of future nuclear data evaluation (JEFF-3 data library for instance).