Computational Grid Generation
It is well understood that automatic discretization (boundary-conforming computational grid generation) is the main bottleneck of the entire computational aerodynamics Typically, it takes more time to generate an acceptable new grid for a new realistic 3-D aerodynamic configuration than it takes to predict the 3-D flow field around it. Since any shape inverse design and optimization effort implies repetitive generation of the 3-D grid, it is quite clear that developing a user – friendly, fast and reliable 3-D computational grid generation code is an extremely important issue. The existing automatic 3-D grid generation codes accept surface grid coordinates as an input and then generate the coordinates of field grid points. This approach is not reliable in the sense that there is no guarantee that some of the generated grid cells will not fold-over, or become ncc – dlc-likc. or that excessively large cells will not neighbor cxccssivcly-small gnd cells, or that the grid will not be sufficiently clustered in the regions of interest, or dial the grid will not become excessively non-orthogonal. Such problems cause significantly slower convergence of the analysis iterative algorithms, they create significantly larger computational errors, cause numerical instability, and often lead to outright divergence of the iterative process.
Any grid generation algorithm must be computationally efficient in terms of storage and execution time and able to accept even incomplete initial grids, that is. the grids that have only surface grid points resulting from a solid modeling software package (for example. CAT1 A). One technique capable of generating reliable 3-D gnds in an a posteriori fashion is the grid optimization method which has been applied to 2-D  and 3-D  structured and block-structured [2991 and to 2-D triangular unstructured gnds including automatic solution – adaptive clustering [300). This fully 3-D robust computational grid post-processing algorithm checks the enure onginal grid for possible negative Jacobian* (existence of fold-over grid cells), untangles the gnd lines, and optimizes the grid in a sense that it becomes maximally locally orthogonal and smoothly clustered in the regions of interest.
Notice that this grid generation procedure is especially suitable to the tightly coupled approached mentioned above. In this case, the grid from the flow field smoothly blends with the grid in the solid structure and the interior of the aircraft while clustering at the interface surfaces.