Steps Towards Automated Robust RANS Meshing

The creation of high-quality discretizations for use in viscous flow simulations remains a challenging task. Even with modern software tools and substantial human effort, the application of state-of-the-art mesh generation algorithms in the presence of geometric features such as concave corners may still result in inadequate local mesh configurations, which can severely affect the resolution of important flow features. To address such issues, mesh generation tools for hybrid unstructured grids often expose a considerable number of algorithm configuration parameter. The resulting flexibility does indeed enable the creation of sufficiently resolved hybrid meshes, although the process often requires a very considerable amount of time even for an experienced user. In a production environment where a large number of detailed simulations of single aircraft configuration are performed, the cost in terms of man-hours may be acceptable. For other applications with requirements for short turn-around time, a more automated approach is desirable. Since an automatic mesh generation procedure cannot rely on user intervention for the resolution of geometric complications or edge cases, a robust strategy for the handling of the surface geometry en- countered in realistic aircraft configurations must be implemented.The approach presented here is based on a segregated prismatic/tetrahedral mesh generation procedure, and aims to achieve robustness by means of local geometric modifications. Criteria chosen and algorithmic modifications make use of similar principles as in earlier work, but are adapted for the specific requirements of mesh generation for aircraft configura- tions. An existing set of open-source tools is exploited for mesh data structures, file format support, surface mesh generation and tetrahedral volume meshes.The mesh generation strategy presented is based on four phases, starting with the creation of a sufficiently resolved surface mesh. In a second step, the envelope mesh of the prismatic boundary layer mesh is determined; the robustness of this stage is the primary contribution of the present work. Thirdly, tetrahedral elements are generated to fill the volume between the envelope of the prismatic layer and the farfield boundaries, and finally, pentahedral elements are grown between adapted wall and envelope mesh.The algorithm implemented into existing open source libraries was applied to two applications presented in this study, a fairly simple wing-body-stabilizer configuration typical for a tran- sonic transport aircraft (CRM) and a rather complex, detailed geometry of a delta wing fighter prototype (F-16XL). RANS solutions converged to engineering accuracy are found to yield solutions in close agreement with meshes produced by a well established grid generator for the EDGE flow solver provided that comparable resolutions are used for both the prismatic layer and the tetrahedral domain.When comparing mesh generation timings, an interesting observation was made. For the common situation where parallel CFD solutions are performed on a compute cluster, the analyst may be evaluating post-processed results of a simulation based on a mesh created with the method presented in this paper before a serial advancing front mesh generation has even been completed.Obviously, this does not mean that there is no need for high-quality advancing-front mesh generation tools. A substantial proportion of relevant geometries and flight conditions likely require more detailed control over mesh generation parameters than is available in a hybrid Delaunay method. However, for routine solutions where serial mesh generation time is a bottleneck, the libraries including the present method can be used to accelerate the turnaround time considerably.