Adaptation of the Polynomial Chaos Method for the Uncertainty Analysis of Critical Experiments

Abstract

Improvement of computational methods used for the uncertainty analysis of critical experiments allows the accuracy of the results to be more realistically estimated, which is important for further improvement of computational codes and cross section libraries used to justify nuclear safety of the research on uses of nuclear energy. Standard uncertainty estimation methods (Monte Carlo, sensitivity analysis, etc.) have disadvantages arising from considerably growing needs for computational resources to be used for nonlinear problems with a lot of measured parameters that contribute to the total uncertainty of a critical experiment. This has provoked increasing interest in spectral methods for uncertainty analysis, in particular, the polynomial chaos method. In this work, the necessity of the regression analysis is substantiated for approximations of random variables by the polynomial chaos method to estimate uncertainties and confidence levels of the results of critical experiments.