Characterization of non-isothermal flows typical of built environments in a laboratory scale model. Part II – Numerical predictions with CFD

Abstract

CFD predictions of built environment flows require turbulence modelling. RANS-based turbulence models (standard and RNG k–ε) have been widely used due to their ability to capture the main features of the time-averaged variables of such flows without demanding unaffordable computer requirements. Herein, the experimental results presented in Part I are used to validate numerical predictions of non-isothermal flows performed with the standard, the RNG with turbulent viscosity formulation and, for the first time, the RNG with differential effective viscosity formulation k–ε models. Two office geometries are simulated for mixing and displacement ventilation strategies. For mixing the RNG k–ε model with the differential effective viscosity provides the most accurate results for both U and V velocity components, temperature, turbulence intensity and PPD comfort index. However, the W velocity component and the DR comfort index are better predicted by the high-Reynolds number models. For displacement there is not a remarkable improvement in results of the velocity, temperature, turbulence intensity and comfort indexes generated by the RNG k–ε with the differential effective viscosity calculation, when compared with those obtained with the turbulent viscosity formulation. Both models yield much more accurate results for temperature and velocity than those given by the standard k–ε model. In opposition, the former yields the best results for the DR comfort index. From the analysis of the turbulence intensity profiles for both ventilation strategies it is possible to infer the need to improve the buoyancy model when using low-Reynolds number turbulence models.