Assessing Engineering Turbulence Models in Buoyant, Non-Unity Prandtl Flow

Understanding the impact of model error in turbulent CFD predictions is a key step to building confidence in and promoting the use of CFD in the nuclear industry. This paper reviews current engineering methods for turbulent heat transfer modeling and considers the most widely used turbulence models to buoyant non-unity Prandtl flow cases. The physical challenges associated with these flow conditions are discussed, and a test case is used to explore the discrepancy between the CFD and direct numerical simulations. Key findings include that the engineering models tend to overpredict the Nusselt number by 2%-38% in both high and low Prandtl flow. Of the models surveyed, the k-ε model demonstrates strongest turbulent momentum transfer modeling performance, but this does not result in superior heat transfer predictions. We also demonstrate that buoyancy-induced laminarization may not suppress heat transfer in low-Prandtl flow due to the relatively minor role of turbulent heat transfer in these flows.