Revisiting RANS turbulence modelling used in built-environment CFD simulations

Abstract

The strong commitment of European Union (EU) to energy efficiency and the increasing energy prices will put pressure on the building sector to find solutions that are simultaneously very low-level or nearly zero energy-consuming, efficient, thermally comfortable and disease free. In the pursuit of such solutions, Computational Fluid Dynamics (CFD) has been used, with great success, in predicting and optimizing built-environment flows. Nonetheless, it is known that the choice of the turbulence model and the way in which it treats the near wall region, influences the quality of the yielded results. A set of experimental three-dimensional particle image velocimetry (3D PIV) data is used to assess the predictive ability of six turbulence models, commonly used in built-environment simulations, together with two new variants of the turbulence model k−ε−v‾2−f. The phenomena variables, inside a 1:30 lab-scale room, with an emulated occupant were computed for two different ventilation strategies, displacement and mixing. Only the k-ε RNG VisEff and the original ‘code-friendly’ variant of the k−ε−v‾2−f (LKM) coherently described the two simulated flows. Furthermore, it was not unequivocal that obeying the dimensionless wall distance (Y+) less than 1 rule, for the first grid node, guaranteed the enhancement of the computed results. The integration of the Standard Wall Functions (SWF) with the k−ε−v‾2−f (LKM) turbulence model proved to yield more accurate and less grid-dependent results than the stand-alone k−ε−v‾2−f (LKM) model, showing, simultaneously, a predictive ability similar to that of the k-ε RNG VisEff model, despite the lower computational complexity of the latter.