On the computational efficiency of LES and hybrid RANS-LES models in building aerodynamics

Large-eddy simulation (LES) has proven to offer superior accuracy in regards to predicting surface pressures compared to the Reynolds-averaged Navier Stokes (RANS) models. However, the primary impediment is the high computational cost associated with LES. The authors attempt to investigate the computational cost and accuracy by employing different sub-grid scale (SGS) models in LES and hybrid RANS-LES models. One of the prerequisites of accurate pressure estimations is to ensure a horizontally homogeneous empty computational domain. This study aims to compare the computational competence qualitatively and quantitatively using an empty domain in regards to the ability to maintain horizontal homogeneity. The Wall-adapting eddy viscosity (WALE) SGS model in LES exhibits a significant reduction in computational time. Moreover, the application of detached eddy simulation (DES) and its modified versions manifest encouraging results in reducing computational time and retaining accuracy