Towards the development of upper subcriticality limits on the basis of benchmark criticality calculations

This paper concerns the assessment of standard point-wise neutron data libraries for criticality safety evaluations in units of the effective neutron multiplication factor, keff, the aim being to establish a methodology for the analysis of storage pools containing fuel assemblies discharged from the Swiss Light Water Reactors. The selected approach is based on using the Monte Carlo code MCNPX (version 2.4.0 was applied in the study at hand) and a modern standard point-wise neutron data library officially distributed by OECD/NEA databank. The approach is oriented towards meeting the broadly accepted general requirements to establish subcriticality, such as those formulated in the ANSI/ANS-8.1-1998 and ANSI/ANS-8.17-2004 Standards. In the above perspective, the results of the assessment of the standard neutron data libraries JEF-2.2 and JENDL-3.3 for criticality safety analysis of UO2 light water reactor fuel assemblies immersed in water, are provided and discussed. The assessment has been performed on the basis of a suite of low-enriched uranium thermal compound benchmarks selected from the International Handbook of Evaluated Criticality Safety Benchmark Experiments. Special emphasis is given to the appraisal of the applicability of the Gaussian distribution in the approximations and posterior analyses of the calculated benchmark results that are necessary to establish the keff safety margins. Such application has been found to be justified, on the one hand, by the observation of a very close agreement between parametric and non-parametric evaluations of the analyzed samples; on the other hand, one indeed expects the values of an ideal (cluster-less) set of criticality benchmarks with similar physical properties to follow a normal distribution. Nevertheless, we presume that similar deficiencies in the specification of the configurations belonging to one or similar series of experiments could cause unspecified experiment/evaluation-related systematic errors in the benchmark evaluations and consequently in the -values. Therefore, even distribution-free estimates will not be rigorous because they also rely on the assumption of random sampling of the population. Moreover, we suggest that the question of normality appears to be of minor importance compared to other hypotheses and to the conservative approximations typically assumed for criticality safety evaluations.