This paper presents CFD validation efforts for the high-speed, multihull Delft Catamaran 372 advancing in calm water with steady drift angles. Available experimental data include hydrodynamic loads (from BSHC), sinkage and trim measurements (from BSHC and CNR-INSEAN), and stereo-PIV measurements on several transverse planes (from CNR-INSEAN). Three organizations from three countries have conducted unsteady Reynolds averaged Navier Stokes (URANS) or Detached Eddy Simulations (DES) by using their own codes: CNR-INSEAN using Xnavis; IIHR, The University of Iowa using CFDShip-Iowa and CNRS/ECN using ISIS. Computations have been made using different grid strategies (structured grid with overlap, unstructured grid with or without an automatic mesh refinement (AMR)), several turbulence models (the isotropic one equation Spalart-Allmaras model, the anisotropic two equations Shear Stress Transport (SST) k and Explicit Algebraic Reynolds Stress Model (EARSM) models, and Detached Eddy Simulations) and different free surface approaches (single phase level set (LS) and volume of fluid (VoF)). Comparisons are made in terms of hydrodynamics loads, local quantities in separated vortex cores (i.e. axial vorticity, axial velocity, position of the vortex and turbulent kinetic energy (TKE)), planar data (velocity field, axial vorticity field and TKE on selected planes) and wave patterns. Discussions and comparisons on the onset and progression of the separated vortical structures are presented. In general computational results have shown that the main features of the separated flow field are well captured by all the computations. No large differences between submissions can be inferred from the cross planar fields; effects of grid resolution and turbulence model have been investigated by the vortex analysis. Grid resolution effects are dominant in the onset region; only very refined grids (total of order of 100M of grid points or using an AMR technique) are able to provide field quantities closer to EFD data. Turbulence model effects are dominant in the progression of the main vortices, with DES simulations and RANS based on non-isotropic models proved to be superior (stronger and less dissipated vortices) than one equation isotropic models. In terms of hydrodynamic loads, a general good agreement between the submissions has been observed (standard deviation on the resistance prediction of about 3.5%). Similar differences between CFD have been seen for the lateral force prediction, whereas larger discrepancy is observed for the yaw moment estimation (about 7%). The comparison with measured data reveals a rather large overall error (of the order of 15%). However, EFD values where collected for different conditions and a larger model; reference values have been obtained by an interpolation procedure on speed/drift plane, whereas no Reynolds number correction was considered. Differences due to grid resolution and the model adopted for the description of the free surface have been also investigated; very fine grids are required to capture wave breaking and rebounds phenomena. Both single phase LS and VoF approaches are able to capture wave induced vortices, some differences are seen in their progression; the VOF method shows the formation of foam close to the free surface region, whereas the LS provides a long life vortices.
Validation of Turbulence Models for Analysis of a Single-Phase Turbulent Natural Convection
