Performance of various RANS eddy-viscosity models for turbulent natural convection in tall vertical cavities

Abstract

The present study is dedicated to the identification of turbulence models that are accurate and numerically economic for computing the natural air-flow and heat transfer by convection in tall cavities with differentially heated vertical walls. The eddy-viscosity models (EVM) are among the simplest to implement and the most economical to treat this problem. This study evaluated the dynamic, thermal and computational performances of twenty EVM turbulence models with one, two or three-equation closure. All the models were first implemented in several in-house codes using the finite volume method. The predictions of the retained models in terms of profiles of velocity, temperature and vertical velocity fluctuations in the cavity have been compared with those of experimental or numerical studies. The obtained results were used to identify the turbulence models that are accurate and numerically economic in predicting natural convection in vertical cavities with a high aspect ratio. The EVM models with three-equation (v2-f and ζ-f) provide the most accurate mean and fluctuating quantities, followed by the k-e RNG (ReNormalization Group) and k-ω SST (Shear Stress Transport) models. The computing time of these four models is higher than that of the 2L (two-layer) and q-ω models, which provide fairly accurate results especially for the mean heat transfer between the vertical active walls. The other one-equation (Spalart and Allmaras model) and two-equation (k-e, k-ω and hybrid models) turbulence models tested in this work, have a high computing time and/or predictions that are not sufficiently precise simultaneously for both velocity and temperature fields.