Description: Cascade Technologies has developed a massively parallel Voronoi-based mesh generation tool called ``Stitch''. Given a water-tight surface triangulation of arbitrary complexity and a set of generating points (effectively cell centers), Stitch can efficiently compute the 3D clipped Voronoi diagram and output a mesh. This approach was demonstrated to be fast and scalable, as well as robust to complex geometry: an 11.4B cell mesh of the NASA HL-CRM was generated in 22 minutes on 43200 cores. Because the generation of the Voronoi mesh is already accomplished, the technical objectives of this proposal are focused on investigating and optimizing the details of how the generating points are spatially distributed. The quality of Stitch's meshes relative to decisions regarding the point cloud specification have yet to be fully characterized. Obviously the point cloud, essentially the spatial discretization of the domain, will have a significant impact on solution accuracy. We will asses sensitivity with respect to several aspects of the spatial arrangement of the point cloud. Within the Voronoi-paradigm uniform isotropic cells, which can impart higher-moment conservation properties appropriate for turbulence, can be achieved by arranging the generating sites with a structured lattice. The different unit cell topologies based on choice of lattice will be investigated. Additionally understanding these lattices at resolution transitions is of interest. Another focus is further development of anisotropic body-conformal mesh generation. Using only isotropic cells can introduce large cell counts in regions which could leverage anisotropic arrangement and retain fidelity for both wall-resolved and wall-modeled simulations. A final focus of this work will be to identify potential solvers that can benefit from and utilize the Voronoi meshing technology. This will characterize potential customers and the market for Stitch as a stand-alone product.

Benefits: As a stand-alone product Stitch will provide high-quality finite volume meshes, which would benefit any solver utilizing polyhedral elements. One example could be NASA's LAVA flow solver. Stitch's ability to quickly and scalably generate uniform and isotropic cells as well as locally anisotropic topology make it an ideal candidate for scale-resolving simulation tools. CharLES, Cascade's low-dissipation multi-physics LES solver, currently leverages these meshes for accurate multi-physics and multi-scale simulations, which could benefit also NASA.

Consumers utilizing CharLES (e.g., gas-turbine sector of General Electric, aero-thermodynamics research at Honda, and fuel-injection research at Bosch) would benefit from the mesh efficiencies gained through this work. Furthermore, CharLES has been successfully applied to analyze problems of flow separation and control, emissions, aero-acoustics, fan noise, and aerodynamic performance.