Analysis feature recognition and mixed-dimensional model reconstruction from finite element analysis mesh model

Abstract

Instead of extracting mid-surfaces from computer-aided design model, an automatic dimensional reduction approach is proposed, which simplifies the finite element analysis model into the mixed-dimensional model. The input finite element analysis model is first decomposed into a set of locally prominent cross-sections. Each prominent cross-section is digitalized as a d-dimensional point and clustered in the embedded space. By the clustered points, both long-slender and thin-wall regions are detected and recognized with the help of aspect ratio. Further the identified features are reduced into skeletons and mid-surfaces, respectively, and the elements of lower dimensionality are generated simultaneously. This dimensional reduction procedure is general and feasible. In this case, multi-resolution mesh models could be created without being transformed back into computer-aided design software, which is essential to the multi-disciplinary simulation. Finally, the simplification degree tests show that the nodes and elements are largely decreased; furthermore, the proposed approach is more effective than the traditional manual method on time consuming.