Cavitating tip vortices of marine propellers are the main contributors to underwater radiated ship noise due to their dynamic behavior and broadband sound spectrum. Experimentally, it is challenging to get a detailed insight into the vortex structure which makes it difficult to understand the noise generating mechanisms. Although Computational Fluid Dynamics (CFD) could offer a solution to this, knowledge regarding the modelling of the inception and dynamics of tip vortex cavitation is insufficient to obtain reliable numerical results in relation to noise predictions. To this end, an extensive verification and validation study in the context of transition and turbulence modelling of an elliptical planform with NACA662−415 cross-section (`Arndt’ wing) is carried out. The finite length of the lifting surface causes the generation of a tip vortex without the additional complexity of secondary motion (propeller rotation). CFD simulations are carried out using the community based open-usage code ReFRESCO.

An accurate prediction of propeller hull interaction is an important step in the design of a new vessel. The prediction of full-scale flow phenomena, which eliminates scale effects, is becoming available due to increasing computational power. However, the complexity of full-scale CFD calculations combined with a lack of validation data results in unknown uncertainties. This study contributes to the uncertainty estimation for full-scale calculations by answering the question With what uncertainty can we currently numerically predict resistance and propeller power on full-scale Reynolds numbers?. The resistance and propeller flow predictions are done for the general cargo vessel MV-regal for three cases; a double body, a free surface resistance simulation and an open water propeller simulation. The simulations are performed for the design speed of 14 knots, resulting in a full-scale Reynolds number of 푅푒 = 1.12 ⋅ 10ዃ. The discretization error is determined by the grid refinement study as presented by Eça and Hoekstra for the propulsion parameters; resistance coefficients and the wake factor. For each case the flow field is analysed, an uncertainty assessment is made and the results are compared to a group of numerical results for the same simulation performed by 20 participants of a case study organised by Lloyd’s register on the same vessel as is considered in this thesis. Modelling the boundary layer of a flat plate on model and full-scale Reynolds numbers encourages the use of unstructured grids for full-scale Reynolds numbers. The uncertainties as predicted by the grid refinement study, vary between 0.6 and 24.5 percent for the friction coefficient. For the 푅푒 = 10^7 the values are compared to a structured grid study, which had a better trend over the grid refinement series. Comparison to theoretical friction coefficient calculations confirmed the absolute friction result. The double body simulation, performed on the full-scale number 푅푒 = 1.12 ⋅ 10^9, demonstrated the use of the unstructured grids on the full-scale Reynolds numbers. The iterative error had to be closely monitored in order to get a stable solution. While the iterative errors had the same order of magnitude, the uncertainties as predicted by the grid refinement study for the propulsion parameters varied between 1.5 and 140 percent. This is an unacceptable large scatter in uncertainty which calls for another method to determine the uncertainty. The free surface simulation added complexity, by modelling the free surface resistance of the vessel. The order of convergence is lower for the free surface simulation, which creates a higher iterative error in the uncertainty assessment. The discretization uncertainty prediction varies between 2.7 and 23.5 percent. The open water simulation, which is based on a hybrid grid of structured and unstructured grids, showed for the monitored parameters a sufficiently low iterative error and a low discretization uncertainty (between 0.1 and 3.4 percent) for the thrust and torque coefficients. Although all different cases showed mixed results for the discretization error, the absolute values are well within the range of results from the test group of 20 participants. This is a good starting point of the repeatability of the flow parameters. It is noted that the current uncertainty estimation is larger than the difference between two grids.