High-Density Mesh Flow Computations with Pre-/Post-Data Compressions

In this paper, data compressions of pre-/post-files of large-scale flow computations are discussed within a framework of the Building-Cube Method (BCM), that was proposed aimed for DNS around real geometries with an expectation of near-future high-performance computers. In the BCM, a flow field is described as an assemblage of building blocks of cuboids, named ‘Cube’. Each cube is a sub-domain of the flow computation and an equallyspaced Cartesian mesh is used in it. All cubes have the same number of grid points and the geometrical size of each cube is determined by adapting to the geometry and the flow features so as to fit the grid spacing to the local flow scale. The grid data are compressed using the run-length technique, combined with the cube framework and the simple treatment of the wall boundaries. The solution data of each time frame of an unsteady flow are compressed by a technique similar to the vector quantization method proposed for image-data compressions. The present method is applied to two airfoils whose results reveal the detailed vortex motions in the boundary layers and shows the importance to use the isotropic, high-density mesh in the boundary layer.