A set of large eddy simulations for the free surface backward facing step (FSBFS) are carried out to study wave formation behind the step. The volume-of-fluid ghost fluid method is employed to capture the free surface. Previous studies have indicated that the wave physics depend on the step draught-based Froude number (Fr). For small Fr, the rear face of the step (transom) becomes wet, while for large Fr, the wave separates smoothly from the transom. Close to a critical Fr separating wet and dry transoms, both conditions may occur. Here, we study wet, critical, and dry conditions based on the Fr classification with three different inflow boundary layer profiles ( ). For Fr =1.75 (wet conditions), we observe a weak dependence on the ReL. A proper orthogonal decomposition of the velocity field at Fr =1.75 shows a coherent vortex street forming beneath the free surface. At Fr =2.66 (critical conditions), we observe that an increase in the ReL results in a decrease in the wavelength and pronounced gas entrainment due to wave breaking. For Fr =3.17 (dry conditions), we also observe shorter wavelength at increased ReL. Further, in the dry conditions, a breaking wave is noticed to occur at higher ReL, while breaking waves are not observed for the smallest studied ReL. Based on the results, we conclude that the wave shape for FSBFS cannot be characterized by the Froude number alone.
The performance of various turbulence modelling methods for simulating cavitating flow over a hydrofoil was investigated. VOF mixture modelling was applied for the multiphase flow, along with a standard two-equation turbulence model, a hybrid RANS-LES method, and a wall-modeled LES approach. The simulations were conducted in a numerical cavitation tank with experimental data available for a range of Reynolds numbers and cavitation conditions. A Reboud damping for eddy viscosity was applied (hereafter referred to as SST-R). A less common approach, incorporating interfacial turbulence damping based on physical arguments regarding the wall-like behavior of phase interfaces, was also applied (referred to here as SST-D). Our results indicate that the standard RANS method fails to predict the breakdown of lift with decreasing cavitation numbers, a phenomenon observed in the experiments and in earlier studies. Incorporating turbulence damping at the cavity interface or directly on the eddy viscosity improves predictions for both URANS and hybrid RANS-LES methods. Both the SST-D and SST-R agreed well with available experimental data, and the LES method consistently provided accurate results across all numerical grids.