In viscous flow simulations, the techniques based on prismatic–tetrahedral hybrid grids represent the best trade-off between ease of use and viscous accuracy. A major performance bottleneck of executing these techniques in real-world examples is how to create a high-quality hybrid grid fully automatically. Automatic hybrid grid generation is key to set up an efficient and scalable viscous flow simulation workflow. To achieve this, its three basic procedures must be automated, as listed below, (1) preparing a valid input geometry, (2) computing an appropriate sizing function, and (3) creating a high-quality grid with the geometry and sizing function as inputs.
In this report, a few new technologies on the three procedures are introduced. In the geometry-preparation procedure, the focus is on how to create a valid surface triangulation on geometries with various errors. A new top-down meshing technique is proposed by first creating a volume mesh and then creating the surface triangulation. Feature preserving are key to the success of this technique and our recent work on this issue will be detailed. In the procedure of sizing function computation, both prior (geometry-based) and posterior (solution-based) techniques are presented, along with a novel sizing-function smoothing algorithm based on nonlinear programming. In the procedure of grid generation itself, a new algorithm is proposed to create high-quality prismatic boundary layer grid. In this algorithm, instead of computing marching directions of layer grids by using heuristic geometric rules, Laplacian-type PDEs are solved by using the boundary element method to provide a marching direction field with smooth transitions.
The above efforts enabled us to set up robust and automatic meshing pipelines for various simulations, including but beyond viscous flow simulations. Grid examples with configurations of a complication level experienced in industry will be selected to demonstrate the capability of the developed technologies.