Comparisons of Turbulence Modelling for Vertical Natural Convection Boundary Layer

Abstract: In this research, we appraise the performances of three Reynolds-averaged Navier Stokes (RANS) models and two subgrid-scale (SGS) stress models for the simulations of the natural convection boundary layer (NCBL). Simulations with Pr = 0.69 and Pr = 5.48 were carried out. The results demonstrate that although the use of enhanced wall treatment provides a better estimation of the temperature profile for the RANS models, the RANS models tend to underpredict the Reynolds shear stress by 20% to 50% and fail to accurately predict the turbulent heat flux. Among RANS models, the k-w SST performs generally better than the k?? models since it shows an obvious transition. The SGS models are shown to be able to predict the overall heat transfer and the turbulent statistics in good agreement with the experimental data. The Dynamic Smagorinsky (DS) and the wall adaptive local eddy (WALE) SGS models show no observable differences in heat transfer, and turbulence statistics. Keywords: natural convection, turbulent boundary-layer flows, turbulence modelling