Choked-flow model parameter uncertainty determination using hierarchical calibration

Nowadays Best Estimate Plus Uncertainty (bepu) methods are applied in nuclear safety investigations with thermal-hydraulics system codes. Their applications remain challenging because (1) many semi-empirical parameterized models are relevant in a given accidental scenario, (2) model parameters probability density functions are often only determined by expert judgement, and (3) models exhibit discrepancies. Inverse uncertainty quantification methods, such as Bayesian calibration, have recently been used to inform model parameter probability density functions based on Separate Effect Test Facility (setf) experimental data. In this paper we present the application of hierarchical Bayesian calibration to choked-flow experimental data from two setfs, provided in the context of the oecd/nea atrium project. This calibration can accommodate boundary conditions varying from one setf to another and account for code inadequacies by inflating the model parameter uncertainties. As such and contrary to standard Bayesian calibration, the method yields a unique joint posterior probability density function validated on similar experiments and applicable for forward uncertainty quantification of an accidental scenario.