Publications
Bayesian calibration of empirical models common in MELCOR and other nuclear safety codes
In modern scientific analyses, physical experiments are often supplemented with computational modeling and simulation. This is especially true in the nuclear power industry, where experiments are prohibitively expensive, or impossible, due to extreme scales, high temperatures, high pressures, and the presence of radiation. To qualify these computational tools, it is necessary to perform software quality assurance, verification, validation, and uncertainty quantification. As part of this broad process, the uncertainty of empirically derived models must be quantified. In this work, three commonly used thermal hydraulic models are calibrated to experimental data. The empirical equations are used to determine single phase friction factor in smooth tubes, single phase heat transfer coefficient for forced convection, and the transfer of mass between two phases. Bayesian calibration methods are used to estimate the posterior distribution of the parameters given the experimental data. In cases where it is appropriate, mixed-effects hierarchical calibration methods are utilized. The analyses presented in this work result in justified and reproducible joint parameter distributions which can be used in future uncertainty analysis of nuclear thermal hydraulic codes. When using these joint distributions, uncertainty in the output will be lower than traditional methods of determining parameter uncertainty. The lower uncertainties are more representative of the state of knowledge for the phenomena analyzed in this work.