Numerical methods based on CFD (Computational Fluid Dynamics) are rapidly expanding their application range, especially with recent remarkable developments in computer performance. Currently, the RANS (Reynolds average Navier-Stokes equation) model, which is a time-averaged model of turbulent flow, is generally used for ship performance estimation. The estimation accuracy of a turbulence model using numerical methods can be improved by adjusting the model with more accurate and detailed flow field data. However, it is still difficult to obtain detailed flow field data through experiments on actual ships and scale models. In a LES (Large Eddy Simulation) model, which is used to spatially filter turbulent flow, if the grid resolution (filter size) is fine, the limitations imposed by the model itself are reduced. LES simulations performed with a sufficiently grid resolution are known to achieve the same estimation accuracy as DNS (Direct Numerical Simulation) but with a lower computational load.