Finite-element mesh generation from constructive-solid-geometry models (Q1331426)

Evidently this is a reformulation of the first author's PhD thesis (Delft 1992, of the same title). Starting point is a constructive solid geometry, in which three- dimensional solid objects are modelled as a collection of primitive objects with planar or curved surfaces. An approach to automatic finite element mesh generation from that representation is described. The main goal is to show a two-step method, because a necessary intermediate step is the conversion of the constructive solid geometry model into a boundary representation, in which objects are modelled through their boundary element faces, edges, and vertices. Special attention is paid to a high accuracy in the calculation of the boundary representation, e.g. by the use of a Bernstein basis (instead of the common power basis) for parametric polynomial curves and surfaces. The mesh generation again is divided into two steps. First, a boundary confirming triangulation of the faces consisting of well shaped triangles can be interpreted as a polyhedral approximation of the solid. In the second step, this polyhedral approximation -- instead of the object itself -- is then meshed with tetrahedra, which entails the insertion of nodes in the interior of the solid. The whole approach for the derivation of a finite element mesh looks promising, although not all details are implemented until now. Remarkable the carefully compiled bibliography.