Quantifying and Counteracting the Impact of Erroneous and/or Incomplete Information via High-Order Predictive Modeling: Methodology and Illustrative Applications to Nuclear Reactor Models

This work highlights the major benefits of applying the “Fourth-Order Best-Estimate Predicted Results with Reduced Uncertainties Predictive Modeling” (4th-BERRU-PM) methodology to quantify and remedy the impact of possibly faulty computational or experimental information. After discussing the importance of assessing the consistency/inconsistency of the input information prior to applying the 4th-BERRU-PM methodology, the best-estimate results predicted by the BERRU-PM methodology are illustrated for a simple model of a nuclear reactor system comprising a neutron source distributed within a moderating material comprising 99.9% pure carbon and boron impurity (0.1%). For a correctly computed response, the 4th-BERRU-PM methodology predicts a best-estimate response value that is between the values of the computed and measured response, along with a reduced predicted standard deviation for the predicted response. When the boron impurity is mistakenly omitted, the 4th-BERRU-PM methodology yields a calibrated value for the reactor’s source and a non-zero calibrated value for the reactor’s absorption cross sections, which both consistently indicate the existence of an omitted absorbing material to be accounted for. Using the OECD/NEA Polyethylene-Reflected Plutonium Metal Sphere reactor physics benchmark, this work also illustrates the application of the 4th-BERRU-PM methodology to quantify the impact of high-order sensitivities on resolving apparent discrepancies between actually consistent computational and experimental information.