A “void-trimming” methodology of generating shrink -wrapped mesh for component-based complex “dirty” geometry
The geometric surface model generated by common CAD tools is often “dirty” (cracks, small gaps, small holes, surface penetration, inconsistent surface orientation, bad edge-face connectivity, etc.). Also, problems of component overlapping, island components, and patch duplication exist in a component-based system. The process of traditional geometric healing and repairing methods is time-consuming (weeks or months), and often time fails when dealing with a complex “dirty” geometric model.
In this dissertation, a new methodology based on “void volume trimming” is presented to resolve problems stated above. The meshing process starts from generating a Cartesian volume mesh using the 2N tree (instead of the traditional Octree) data structure. With this structure, several mesh adaptation methods based on geometric features coupled with a smoothing algorithm between neighbor cells are developed to generate the preferred mesh sizes at desired regions while ensuring the gradual transition between dense and coarse meshes. In the process of constructing surface mesh for “dirty” geometric components, an effective “surface orientation free” algorithm is proposed. For resolving of “mesh leak” at cracks and small gap regions, the continuous “intersecting cell” set is used instead of geometric surfaces as the domain bound. The major contribution of this dissertation is the development of “void volume trimming” algorithm. With this methodology, the watertight feature can be promised, and the axis-aligned surface mesh is gradually adjusted to be geometric aligned while maintaining high mesh quality. Meanwhile, the surface mesh is pushed towards the geometry for satisfaction of mapping criteria.
The constrained smoothing algorithm presented in this dissertation further improves the mesh quality while shrinking the surface mesh closer to geometry components. At the same time, the use of the SPP (Shortest Path Projection) algorithm coupled with the ADT (Alternating Digital Tree) data structure has been shown that it is efficient when generating body-fitted surface meshes for complex “dirty” geometries while maintaining high performance. The present critical feature preservation method has shown its capability of capturing the detailed features, while the introduced patch mapping method can topologically maintain the geometric model property. Case studies and application results have demonstrated that the current methodology is efficient for handling the component-based complex “dirty” geometric model.
0984: Computer science