Material Composition Interpolation for High-Assay Low-Enriched Uranium Fuel Macroscopic Cross-Section Estimation Using Gaussian Process

Material composition interpolation was investigated for the rapid generation of macroscopic cross-section libraries for core design, specifically focusing on pressurized water reactor fuel assemblies containing high-assay low-enriched uranium fuel pins. Rapid cross-section-library generation can accelerate the deployment of high-enriched fuels in commercial light water reactors. A Gaussian process interpolation method was used with adaptive sampling in order to minimize the number of required lattice physics calculations. The interpolation method was applied to a two-dimensional parameter space defined by fuel enrichment and Gd₂O₃ concentration. The accuracy of the interpolated cross-section libraries was assessed by directly comparing them with those generated by lattice physics calculations at the interpolated material compositions. A high-quality approximate cross-section library with an error of less than 0.1% was generated using 15 sample points. The reactivity uncertainty propagation was estimated to have a standard deviation of less than 23 pcm, with the actual maximum reactivity error observed in PARCS depletion calculations reaching 53 pcm.