WHISPER S/U benchmark analysis of metal-water critical mass curves

The purpose of this paper is to discuss the application of Sandia National Laboratories' (Sandia's) generic standard set of ICSBEP benchmarks, the set of benchmarks shipped with the Whisper-1.1 software (LANL), and the combined set of benchmarks in support of criticality calculations with MCNP-6.2. Sensitivity and uncertainty (S/U) analysis was performed with Whisper-1.1 software to determine similarity between the benchmarks and the computational models of spherical fissile metal-water mixtures that are formulated to compose the critical mass curves of ²³⁵U and ²³⁹Pu. Relationships between the chosen benchmark set and the observed results from the Whisper S/U software are discussed, including discussions of the correlation coefficient, ck, software weighting factors, and the upper subcritical limit (USL) of the effective neutron multiplication factor, k_eff. The results indicate that all studied benchmark suites are sufficient to support reasonable criticality limits along the critical mass curves for ²³⁵U and ²³⁹Pu. The Sandia benchmarks are strongly correlated to the models that compose the critical mass curves in most of the thermal range and near the fast/metal region for both nuclides. As expected, the number of benchmarks with very high similarity decreases near the minimum concentration necessary to achieve criticality and in the very poorly moderated region. This trend is shown for ²³⁵U in Figures 3-5, with the contribution from various benchmark system types detailed. To improve similarity in regions that have lower values of c_k (e.g., <0.9), a greater number of applicable benchmarks is required. Figure 6 shows the average ck of Whisper-1.1 benchmarks when using the generic set of Sandia ICSBEP benchmarks, the ICSBEP benchmarks shipped with Whisper-1.1 (LANL), and a combined suite of benchmarks from both sources, with and without specific benchmarks that fail statistical tests (e.g., chi-squared) excluded for both ²³⁵U and ²³⁹Pu curves. Figure 6 shows the incremental improvements in average c_k that occur when benchmark suites are combined. These improvements are most noticeable in regions with lower values of c_k. While c_k improves with the addition of more benchmarks, the Whisper-calculated USL value does not necessarily improve, as shown in Figure 7 for ²³⁵U and Figure 8 for ²³⁹Pu. In addition to providing the Whisper-calculated USL for different benchmark suites, Figure 8 highlights which individual benchmarks have the largest effect on the calculated baseline USL. A noticeable increase in USL is observed when these highlighted benchmarks are removed from the suite. (authors)