Fast and Accurate Prediction of Vehicle Aerodynamics Using ANSYS Mosaic Mesh 2020-01-5011
ANSYS Mosaic™ meshing technology automatically conformally connects different types of meshing elements with general polyhedron elements. Poly-Hexcore, the first application of the Mosaic technology, fills the volume region with octree hexahedra, boundary layer with the high-quality layered poly-prism and conformally connect these two meshes with general polyhedron elements. A generic vehicle model - the DrivAer car, developed at the Technical University of Munich (TUM) in cooperation with the Audi AG and the BMW group, is used to validate the Mosaic meshing technology to predict complex aerodynamic phenomenon around the car. The Mosaic meshing technology is parallel scalable on high-performance computing (HPC) platform. For example, on ANSYS HPC (CPU: 2x Intel(R) Xeon(R) Gold 6142 CPU @ 2.60 GHz, RAM: 192 GB (6 GB/core)), approximately 63 million element mesh is 6.5 times faster, when scaled from serial compute core to 32 parallel compute cores. Furthermore, compared to conventional Hexcore, Mosaic Poly- Hexcore shows ~40% reduction in the total element count, which consequently speed up the ANSYS Fluent solver by ~14%. However, the latest developments in the ANSYS Fluent solver for Mosaic Poly- Hexcore mesh showed further increase in the solver speedup and this will be covered in a separate paper. Next, comparing aerodynamic force coefficients between the computational Scale Resolved Simulation (SRS) and the wind tunnel measurements, shows, the Mosaic Poly-Hexcore (one drag count difference with measurements) predict more accurate results than the Hexcore mesh (six drag count difference with measurements). Further, Mosaic Poly-Hexcore predict accurate flow behavior and pressure distribution over the car, especially at rear-windshield compared to the Hexcore mesh, which shows early flow separation due to the adverse pressure gradient and leads to non-uniform/non-symmetric flow pattern and pressure distribution compared to the experimental visualizations.