An efficient LES framework for cavitation shedding in high-Reynolds-number internal flows relevant to marine hydraulic systems

Unsteady cavitation in high-Reynolds-number internal flows can generate periodic vapor cloud shedding and associated pressure pulsations, which are of particular concern in marine hydraulic systems, where flow-induced loads and vibration may affect component integrity. Although Large Eddy Simulation (LES) is widely recognized as a suitable tool for capturing such complex unsteady phenomena, accurate prediction of cavitation shedding frequencies – especially at high frequencies – typically requires very fine meshes and significant computational effort. This work presents an efficient LES-based approach for predicting cavitation shedding in high-Reynolds internal flows, with emphasis on achieving a balance between accuracy and computational affordability. The framework combines an incompressible two-phase formulation using the Schnerr–Sauer cavitation model with LES and the WALE subgrid-scale model. Accuracy in frequency prediction is enhanced through three complementary numerical elements: (i) a precursor-LES-based inlet turbulence specification imposed via the Synthetic Eddy Method, (ii) a multistage initialization strategy to accelerate convergence of unsteady cavitating regimes, and (iii) a mesh refinement methodology guided by the spatial distribution of the Taylor microscale. The approach is validated against experimental measurements of partial cavitation in a converging–diverging Venturi nozzle operating in a re-entrant-jet-dominated regime. The simulations reproduce the main cavitation dynamics and capture both the low-frequency cloud shedding cycle and the higher-frequency pulsations associated with partial cavity detachments. The predicted high-frequency shedding component (≈300 Hz) shows excellent agreement with the experimental value (298.65 Hz), while the total mesh size (∼18.5 million cells) remains substantially lower than that reported in comparable LES studies. The results indicate that reliable prediction of cavitation shedding frequencies in high-Reynolds internal flows can be achieved without resorting to excessively large computational grids. The proposed approach therefore provides a computationally efficient strategy with potential relevance for marine hydraulic systems applications where accurate characterization of cavitation-induced pressure fluctuations is required.