Influence of the turbulence model for channel flows with strong transverse temperature gradients

The effects of a strong transverse temperature gradient on a turbulent Poiseuille flow are studied numerically using Reynolds-averaged Navier-Stokes (BANS) models. Such a situation is very common for numerous industrial applications. Since a large majority of industrial computations are based on the BANS approach, the aim of the present work is to investigate the ability of different BANS models to reproduce the main physical phenomena at the origin of the asymmetry of the flow and thermal fields. Comparison are performed with available direct numerical simulations (DNS) or large eddy simulation (LES) databases. With the prospect of future application of the models in the industrial context, models based on the widely used eddy-viscosity and simple gradient diffusion (SGDH) hypotheses are compared to more elaborate second-moment closures for the Reynolds stress and turbulent heat flux. The aim is to determine the closure level necessary to reproduce the influence of strong temperature gradients on the turbulent flow, for a wide range of wall-temperature ratios. Eddy-viscosity models prove able to correctly reproduce the asymmetry of the flow and the tendency toward relaminarization close to the hot wall, which are mainly due to the strong variations of the physical properties (namely the molecular viscosity and the density). Discrepancies in the predictions of the different closure levels only appear for the highest temperature ratios. Unfortunately, reliable reference data are lacking for these configurations, which calls for future DNS or refined LES studies.