Along the Management and Uncertainty of Severe Accidents (MUSA) project, attention was paid to the unfolding of uncertainty analysis when dealing with severe accident (SA) scenarios. While the quantification of the uncertainty linked to SA simulations’ results was the main focus of the project, some efforts were also addressed to the identification of the variables being the root of it. To this end, a complementary sensitivity analysis was deemed to be of high importance. Following this path, the present paper reports the advancements made in the attempt to enhance and optimize the sensitivity analysis process. More commonly used sensitivity analysis techniques, such as correlation coefficients or simple regression, are complemented by more advanced techniques through the integration of feature selection algorithms. As a further step, a testing phase is foreseen; in particular, the selected sensitivity methods are applied against a SA scenario, namely, an unmitigated station blackout in a pressurized water reactor. Outcomes according to the different techniques are reported and compared, with a certain level of agreement being shown. The analysis also highlighted the need to support the application of sensitivity methods with expert judgment to corroborate the physical consistency of the obtained results.
Techniques for Sensitivity Analysis in Severe Accidents: A Comparative Study
Michela Angelucci
;Sandro Paci
2026-01-01
Abstract
Along the Management and Uncertainty of Severe Accidents (MUSA) project, attention was paid to the unfolding of uncertainty analysis when dealing with severe accident (SA) scenarios. While the quantification of the uncertainty linked to SA simulations’ results was the main focus of the project, some efforts were also addressed to the identification of the variables being the root of it. To this end, a complementary sensitivity analysis was deemed to be of high importance. Following this path, the present paper reports the advancements made in the attempt to enhance and optimize the sensitivity analysis process. More commonly used sensitivity analysis techniques, such as correlation coefficients or simple regression, are complemented by more advanced techniques through the integration of feature selection algorithms. As a further step, a testing phase is foreseen; in particular, the selected sensitivity methods are applied against a SA scenario, namely, an unmitigated station blackout in a pressurized water reactor. Outcomes according to the different techniques are reported and compared, with a certain level of agreement being shown. The analysis also highlighted the need to support the application of sensitivity methods with expert judgment to corroborate the physical consistency of the obtained results.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
