Generation of a Pre-Optimized Surface Mesh From a CSG Model

Abstract

Mesh generation remains a tricky and time-consuming step between design and finite element analysis. This paper proposes a new method to automatically build an optimized triangular mesh where density is adapted to finite element analysis. To achieve this goal, the method makes the most of both geometric data and information provided by the functional analysis of the product. This information determines where the mesh must be coarse or fine and, combined with a geometric model, allows to directly build an optimized triangular mesh meeting simulation requirements. For the geometric model, a CSG representation is used instead of a BRep because maintaining geometrical and topological consistency in a BRep raises important and sometimes unresolved issues. Moreover, it is possible to assign an importance regarding finite element analysis to each node of the CSG, which allows to determine the mesh density and nodes blending strategy. The mesh of Boolean nodes in the CSG is directly computed from the meshes of the operands. The paper highlights and solves the problems induced by this approach, that is triangles quality where meshes merge and errors introduced by approximating surfaces intersection with meshes intersection. A new meshing technique based on the advancing front method and taking into account triangles size variation between areas having very different densities, is presented.