Uncertainty quantification of thermal-hydraulic CFD simulation using hybrid Monte-Carlo and polynomial chaos method

The use of Computational Fluid Dynamics (CFD) simulations for thermal-hydraulic system analysis is continuously increasing, as the computation platforms and the numerical models are improving. In order to use those simulations for thermal-hydraulic analysis of nuclear power plants, Uncertainty Quantification (UQ) must be performed, either for the input parameters or for the numerical model coefficients. This paper presents a UQ method for CFD simulations using polynomial chaos (PC) with a Non-Intrusive Point Collocation method (NIPC). Since PC does not constitute a complete UQ methodology, and can only obtain the appropriate moments for the output (mean, standard deviation, etc.), a NIPC based surrogate model was developed in order to use hybrid Monte-Carlo and NIPC methods to generate the cumulative density function (CDF) and the probability density function (PDF) for output results. Our surrogate code was benchmarked against the Ishigami function and was applied for stochastic-CFD simulations of the T-junction turbulent mixing of water. Those CFD simulations were accompanied by NIPC and surrogate code to quantify the uncertainty and sensitivity of k-ω SST turbulence model due to uncertainty in the closure coefficients values.