Influence of the inlet boundary conditions on the LES performance for the cavity flow benchmark

A cavity flow consists of one air inlet and one outlet slot. The inlet slot is positioned in the upper left corner of the cavity, whereas the outlet slot is located in the lower right. This cavity flow is representative of mixing ventilation. The literature shows that the prevalent two-equation RANS turbulence models can reproduce the measured velocity field in the cavity for the transitional and fully turbulent flow regimes. However, a single turbulence model cannot perform equally well at these two flow regimes: k-ε models perform well for the fully turbulent regime, while the k-ω models perform best in the transitional regime. In general, this dependence on the flow regime can make the use of RANS less reliable during the ventilation design phase. By definition, LES is expected to be suited for transitional and fully turbulent flow regimes. Therefore, it is worth investigating and comparing the performance of LES and RANS on two isothermal cavity flow benchmarks, which differ by their geometry (i.e., the aspect ratio of the room) and flow regimes. Simulations are performed on structured grids using the Dynamic Smagorinsky subgrid-scale (SGS) model for LES and the standard k-ε, standard k-ω and BSL k-ω turbulence models for RANS. In addition, the performance of DES is also investigated and compared using the Spalart-Allmaras and realizable k-ε DES. Results show that the LES using the Dynamic Smagorinsky is indeed able to reproduce the velocity field for both flow regimes, making the model more universal than RANS. However, results are strongly dependent on the turbulence level at the inlet. In addition, it is shown that spatial-developing synthetic turbulence at the inlet gives comparable results as a separate LES simulation, leading to simpler and less computationally expensive simulations. Regarding DES, the realizable k-ε DES gives fairly good results for the fully turbulent case. However, the DES has numerical stability issues when adding considerable synthetic turbulence at the inlet and suffers from the depletion of turbulent structures under the transition from RANS to LES. In conclusion, LES can be more universal to predict the ventilation performance of mixing ventilation in buildings, but it requires a good knowledge of the turbulence intensity at the air inlets, which may not be a straightforward task during design. Regarding DES, it has the potential to decrease computational costs compared to LES, but it requires further research for the cavity flow.