The prediction of cavitation is of large interest for the design of ship propellers. Cavitation influences the propeller efficiency and causes undesired phenomena such as noise, vibrations and erosion.

While potential flow solvers are routinely used for propeller design, viscous CFD calculations are used when a higher fidelity is needed. Within a finite volume framework, the flow domain is subdivided into grid cells and the governing equations are solved numerically in every cell.

Essentially two approaches are possible to discretise the domain, with structured or unstructured meshes. In the former case, it is always possible to construct a mapping function between the physical grid and a uniform cartesian grid. In the unstructured mesh there is not such a correspondence. A structured mesh is easier to handle for a numerical solver, but for complex geometries it becomes difficult to generate. The size and quality of the mesh affect the quality of the flow solution.

The objective of this work is to evaluate the effect of two different grid types when they are used to simulate cavitating flow using an unsteady Reynolds averaged Navier-Stokes (RANS) solver. The test case chosen is a 2D NACA0015 profile at 6 degrees angle of attack, confined in a water tunnel. Many numerical studies on the same test case are found in literature, e.g. Hoekstra (2011), Yakubov et al. (2015), and experimental tests, Arndt et al. (2000). After some preliminary wetted flow computations, which aim to investigate the numerical uncertainty, the main part of this work is a comparison of cavitating flow dynamics predicted with a structured and an unstructured mesh.

Download this paper or see the paper overview.

For more information contact Tom van Terwisga.