| dc.description.abstract | An emerging trend in the nuclear community is to utilize RANS based turbulence models to supplement thermal-hydraulic system codes in the design of Generation-IV (Gen-IV) power reactors. Prior to full integration into the design process for Gen-IV reactor concepts, RANS models must undergo thorough validation studies to justify their applicability in both normal and accident conditions. Several previous numerical studies have raised concerns regarding the performance of RANS models within buoyant flow regimes indicative of a loss of flow accident (LOFA). To address these concerns, the current research performs a detailed assessment of 5 different RANS based turbulence models against benchmark experimental data designed to replicate the transient conditions of a LOFA along a heated vertical plate. Boundary conditions and system response quantities for the numerical model are supplied from the experiment every 0.2 seconds during the 18.2 second transient. ASME standards are used to quantify the numerical uncertainties while the input uncertainties are handled using a Latin Hypercube Sampling (LHS) method based on the steady-state conditions (t=0 s). Qualitative comparisons between numerical and experimental results at several downstream locations are supported using a validation metric based on the statistical disparity between the respective empirical and cumulative distribution functions. Overall, the RANS based turbulence models are unable to accurately predict the system response quantities supplied through the experiment, most notably during the flow reversal where the buoyancy force is dominant. In comparison to the other RANS models, the Abe-Kondoh-Nagano (AKN) k-ϵ variant is the most consistent with the experimental results and is selected for additional assessment and development.
To further assess the AKN models applicability in buoyant flows, a comprehensive analysis is conducted against well trusted Direct Numerical Simulation (DNS) databases for mixed convection flows characterized over a range of Richardson numbers. The analysis suggests that at intermediate Richardson numbers, the AKN model is insensitive to changes in buoyancy, resulting in grossly overpredicted values of Nusselt number and skin-friction coefficient. To correct this, a new source term is derived to increase the turbulent dissipation rate as a function of local buoyancy related flow variables to improve the prediction of the turbulent viscosity. Using the previously mentioned validation metric, the modified AKN model is shown to greatly improve predictions at the intermediate Richardson numbers while maintaining the integrity of the original model. | en |
