Sensitivity Analysis and Data Assimilation in a Subcritical Plutonium Metal Benchmark

Abstract

This work presents the application of first-order adjoint sensitivity analysis, uncertainty quantification, and data assimilation to a subcritical plutonium benchmark experiment using a modified version of the discrete ordinates radiation transport code Denovo. Previous Monte Carlo simulations of this benchmark saw a consistent overprediction of the mean and variance of the measured neutron multiplicity distribution. It was observed that a small scalar reduction in the value of the 239Pu-induced fission neutron multiplicity was capable of significantly reducing the discrepancies. This work extends those results by computing first-order sensitivities to each nuclide, reaction type, energy, and material region in the benchmark. The sensitivities are then used in a data assimilation methodology to simultaneously calibrate all responses and multigroup nuclear data. The resulting best-estimate values for the energy group differential multiplicity ν̄(Eg) are 1σ to 2σ less than the nominal values found in ENDF/B-VII for energies less than ˜1.5 MeV.