Variability in spent fuel inventory data

A new methodology has been developed to assess how representative a collection of nuclide mass estimates is of real-world spent fuel. The analysis is approached as an “applicability range” evaluation, which quantifies the fraction of the historical population that is represented by one mass estimate. The new methodology is applied to a database containing spent fuel inventory estimates for operating U.S. reactors and uses historical assembly designs as a reference. The evaluation consists of two major steps: the implementation of a sampling and randomization scheme and the calculation of the applicability range scores. The mass estimates are assigned scores for each of the assembly designs they should represent, and the scores are averaged together to find overall applicability range scores for the database estimates. The results showed that the estimates for newer, PWR assembly classes had much higher scores than estimates for older or BWR assembly classes. The applicability range methodology can be extended beyond the database analyzed in the present work to any study that needs to use one value to represent a population with broad variations and limited knowledge of the underlying distributions.