An evaluation of recently developed RANS-based turbulence models for flow over a two-dimensional block subjected to different mesh structures and grid resolutions

Understanding, analyzing and predicting meteorological phenomena related to urban planning and built environment are becoming more essential than ever to architectural and urban projects. Recently, various version of RANS models have been established but more validation cases are required to confirm their capability for wind flows. In the present study, the performance of recently developed RANS models, including the RNG k-ε, SST BSL k-ω and SST γ−Reθ, have been evaluated for the flow past a single block (which represent the idealized architecture scale). For validation purposes, the velocity streamlines and the vertical profiles of the mean velocities and variances were compared with published large-eddy simulation (LES) and wind tunnel experiment results. Furthermore, other additional CFD simulations were performed to analyze the impact of regular/irregular mesh structures and grid resolutions based on selected turbulence model in order to analyze the grid independency. Three different grid resolutions(coarse, medium and fine) of Nx × Ny × Nz = 320 × 80 × 320, 160 × 40 × 160 and 80 × 20 × 80 for the computational domain and nx × nz = 26 × 32, 13 × 16 and 6 × 8, which correspond to number of grid points on the block edges, were chosen and tested. It can be concluded that among all simulated RANS models, the SSTγ−Reθ model performed best and agreed fairly well to the LES simulation and experimental results. It can also be concluded that the SSTγ−Reθ model provides a very satisfactory results in terms of grid dependency in the fine and medium grid resolutions in both regular and irregular structure meshes. On the other hand, despite a very good performance of the RNG k-εmodel in the fine resolution and in the regular structure grids, a disappointing performance of this model in the coarse and medium grid resolutions indicates that the RNG k-εmodel is highly dependent on grid structure and grid resolution. These quantitative validations are essential to access the accuracy of RANS models for the simulation of flow in urban environment.