A hybrid RANS/LES model for simulating time-dependent cloud cavitating flow around a NACA66 hydrofoil

Cloud cavitating flow is highly turbulent and dominated by coherent large-scale anisotropic vortical structures. For the numerical investigation of such a class of flow, large eddy simulation (LES) is a reliable method but it is computationally extremely costly in engineering applications. An efficient approach to reduce the computational cost is to combine Reynolds-averaged Navier-Stokes (RANS) equations with LES used only in the parts of interest, such as massively separated flow regions. A new hybrid RANS/LES model, the modified filter-based method (FBM), is proposed in the present study which can perform RANS or LES depending on the numerical resolution. Compared to the original FBM, the new method has three modifications: the state-of-the-art shear stress transport (SST) model replaces the k-epsilon model as a baseline RANS model. A shielding function is introduced to obviate the switch from RANS to LES occurring inside the boundary layer. An appropriate threshold controlling the switch from RANS to LES is added to achieve an optimal predictive accuracy. The new model is assessed for its predictive capability of highly unsteady cavitating flows in a typical case of cloud cavitation around a NACA66 hydrofoil. The new model results are compared with data obtained from the Smagorinsky LES and SST model based on the same homogeneous Zwart cavitation model. It is found that the modified FBM method has significant advantages over SST model in all aspects of predicted instantaneous and mean flow field, and its predictive accuracy is comparable to the Smagorinsky LES model even using a much coarser grid in the simulations.