Transformation of VRML-files into graph structures in order to detect similarities and build clusters

Abstract

This paper presents a method for the automatical detection of similarities in CAD-models. The main concept is the possible transition of proprietary engineering parts of all kinds of software into the non-native VRML-file format that completely contains the geometrical information. The whole procedure can be divided in three subsections. At first an information extraction takes place. The geometry of the transferred VRML-model gets analyzed where especially the surface structure is extracted for the further processing in the next steps. The algorithm recognizes the shape of every involved face and delivers this information to the conditioning in the second step. There the gained data is used for the generation of graph-structure-like fingerprints that display all surface types as well as their in between connections. These can be considered as signatures, which contain the most important information in a condensed way. In the further steps it is intended to compare the CAD-parts with each other in order to find similarities, build clusters and form groups of topologically related geometries. For a powerful algorithmic comparison a further simplification is necessary. This is implemented by a translation of the graphical fingerprint into a text format and the removal of all smaller and non determining surfaces. Thus, topologically identical and also similar CAD-parts can be recognized and sorted into the same cluster. Due to the fact that the whole procedure is based on the analysis of text, it is qualified for a comparison of several engineering parts as well as for huge amounts of data, like for instance in construction companies.