BEPU robustness analysis via perturbed law-based sensitivity indices

The “best-estimate plus uncertainty” (BEPU) methodology is the term used in nuclear engineering community when dealing with uncertainty quantification issues realistic numerical simulation models. One of most critical hypothesis these studies choice probability distributions uncertain input variables which are propagated through model. Bringing stringent justifications to BEPU approach, especially a safety study, requires quantifying impact potential on variable distribution. To solve this problem, paper deepens robustness analysis based “Perturbed Law-based sensitivity Indices” (PLI). PLI quantifies perturbation an distribution quantity interest (as quantile model output or margin) study. mathematical formalism applied two particular quantities interest: and superquantile. For both quantities, can be easily computed using unique Monte-Carlo sample containing inputs output. Numerical tests developed order define validity criteria PLI-based analysis. practical use method illustrated thermal-hydraulic computer experiments, simulating cold leg Intermediate Break Loss Of Coolant Accident (IBLOCA) pressurized water reactor.