The construction of a quantitative comparison of upper subcritical methods for novel neutronic systems

An Upper Subcritical Limit (USL) comparison study was conducted by constructing 167 44-group ENDF/BVII.1 covariance data sets and using that covariance data to construct three randomly perturbed the ENDF/BVII.1 cross section libraries. A perturbed library consisting of 165 cross section files, both fuel and non-fuel was created. Additionally, two libraries consisting of only 20 perturbed fuel cross section files and 145 non-fuel cross section files were created. These randomly perturbed cross section libraries were used to calculate three k_calc values per application case in this study. The unperturbed ENDF/BVII.1 cross section library was used to calculate a k_exp value for each application and thereby construct a reference USL value for each application case as a means to compare the efficacy of the Whisper, TSURFER, and USLSTATS USL estimation methodologies. Using the maximally perturbed cross section library and the perturbed non-fuel cross section libraries to calculate k_calc values for application cases produced results approximately 50% smaller than expected. This aberration was traced to the magnitudes of the variance data of ¹⁶O absorption cross section being small and creating exaggerated weighting factors when using Cholesky decomposition to decompose the covariance matrix in the random perturbation process. This was not seen in better characterized cross section data associated with randomly perturbed fuel isotope cross section library; thus the USL comparison study was conduction only with the randomly perturbed fuel isotope cross section library. This finding suggests other matrix decomposition methods or a potential weighting factor dampening filter should be prescribed for less characterized covariance data. (author)