DOE OSTI2024
Although their adoption by practitioners has been limited, the introduction of similarity indices in criticality safety was a major step forward in reducing the reliance on expert judgement in discerning applicable experiments for the validation of new appliations in criticality safety analyses. Similarity indices have been successfully employed in bias trending and data assimilation techniques, but it is often unclear which acceptance criteria should be used. In their 2004 paper, Broadhead et al. specify the most widely used similarity parameter, ck, as an acceptance cutoff at 0.9. (Broadhead et al., ”Sensitivity and Uncertainty-Based Criticality Safety Validation Techniques,” Nucl. Sci. Eng. 146, 340–366, 2004). Experiments with a ck < 0.9 are often not considered applicable for code validation. This heuristic is based on quantitative studies and engineering judgement, but in some cases, experiments with ck < 0.9 can be used to accurately estimate computational bias. This suggests that further analysis is needed to determine what components of ck are driving applicability and accuracy in bias estimation. For cases in which applicable experiments may not be available (as is the case with UF6 transport canisters), understanding what distinguishes experiments in providing adequate bias estimates aside from just the similarity index is particularly necessary. To further the goal to better interpret ck values, several visualization tools were developed to assist in the investigation of which components of ck are driving applicability.