Content area
Abstract
This work is motivated by the aim to support numerical methods to solve partial differential equations. Among them are finite element and finite volume methods. For these, a given domain must first be subdivided into many simple cells. The quality of the subdivision will tremendously affect the accuracy and convergence of the method. A boundary conforming Delaunay mesh is a partition of a polyhedral domain into Delaunay simplices and all boundary simplices satisfy the Gabriel property. It is important in Voronoi-box based finite volume schemes. It allows to carry important qualitative properties from the continuous to the discrete level. In this work, we study meshing problems for the generation of three-dimensional good quality boundary conforming Delaunay meshes. A simple and efficient algorithm to generate boundary conforming Delaunay meshes is Delaunay refinement. Shewchuk's algorithm is reanalyzed, achieving new results on the termination (angle) condition, on the bounds on characteristics of the tetrahedral shape and on the mesh size distribution. In practice, it is observed that this algorithm usually greatly outperforms the proven estimates. Next we propose an adaptive Delaunay refinement algorithm which guarantees the termination for all valid inputs. It is known that not all polyhedra can be tetrahedralized without using additional points (so-called Steiner points). The three-dimensional boundary conforming Delaunay mesh generation problem has many difficulties. An algorithm to construct constrained Delaunay tetrahedralizations is presented. This algorithm adds few Steiner points on the domain boundary. The correctness is proven for an arbitrary valid polyhedral domain. The complexity issues are discussed. It is practically efficient.