Large Eddy Simulation of Yacht Aerodynamics

Abstract

Accurate modelling of the flow field around the sails of a yacht is important for the pre-diction of the forces generated by the sails, which is needed to determine the performance of the yacht. A methodology using Large Eddy Simulation is described for modelling upwind and downwind sails. The important aspect of this approach is the ability of the Large Eddy Simuation to resolve the three-dimensional, turbulent, unsteady structures that develop in the boundary layer, in the regions of flow separation, and in the wake, rather than modelling them. The reproduction of two experiments, involving nominally two-dimensional flow, and characterised by leading and trailing edge separation, showed the ability of the model to reproduce flow separation, transition to turbulence, relam-inarisation of the flow, and helped to determine the resolution requirements for LES in time and space. The method has been subsequently been applied to model upwind sailing, by reproducing a wind tunnel experiment that has previously been investigated using other models. The method predicted the experimental data with great accuracy than the computational results obtained by other authors, by reproducing the turbulent structures developing at the leading edge of the sail. Finally, the model has been applied to downwind sailing, by reproducing a wind tunnel experiment. The simulation of un-steady turbulent structures developing in the boundary layer allowed the computation of the leading edge and trailing edge separation, the tip vortices, and the wake. As a result, the model’s prediction of the pressure distribution reproduced the experimental data, improving the accuracy of the computated force on the sails. However, LES resulted particularly sensitive to the mesh resolution, mesh quality and to the time discretiza-tion, introducing errors if the minimum requirements were not met by the computational model. As a result, although the accuracy of the simulation improved in comparison with other computational models, the required spacial resolution was not met and the LES model resulted in a significantly increase of computational cost.