Evaluation of BWR core attributes uncertainties due to multi-group cross-section uncertainties

This paper outlines the current work on the propagation of multi-group cross-section uncertainties to BWR core attributes uncertainties. The ORNL PUFF-III code propagates the ENDF point-wise cross-section uncertainties to the multi-group level. Our work propagates the multi-group cross-section uncertainties through the subsequent nuclear calculations in a computationally efficient manner. This includes the propagation of multi-group uncertainties through lattice physics codes to generate lattice-averaged few-group cross-section uncertainties. This is followed by reactor calculations to propagate few-group cross-section uncertainties to uncertainties in evaluated core attributes, e.g. core reactivity, nodal power, and reactivity coefficients. In this work, all crosssection uncertainties are assumed to follow normal Gaussian distributions, and only the first and second moments of the uncertainty distributions, i.e. the means and covariances, are propagated through the calculations. The critical issue in this approach is the computational burden of the sensitivity analysis-type calculations required for the uncertainty analysis. This paper outlines a direct method of propagating cross-sections uncertainties through each computational model in an accurate and computationally efficient manner based on the singular value decomposition of the cross-sections covariance matrices. Core attributes' uncertainties are determined for two different evaluations of the multi-group uncertainty libraries provided in the ORNL SCALE code package. The core attributes' uncertainties are then used to adjust few-group cross-sections to improve the agreement between two different core simulator models by a process referred to as Adaptive Simulation. Posterior few-group cross-section covariance matrices and core attributes covariance matrices are calculated along with the adapted solution.