Development of high precise similar sub-part search technique for automatic mesh generation reusing proven models

We proposed an advanced mesh generation technique that reuses the proven analysis models by similar sub-part search. The purpose of this research is to reduce interactive mesh improvement work time and to comply with the mesh specifications. It is our feature technique to search sub-parts similar to the archived sub-parts of proven models from new design CAD. In this technique, the similar sub-parts are retrieved from a CAD model described by boundary representation and made correspondence relation surface pairs of a retrieval model (proven model) and a target model (CAD of new design). A similarity score is based on the attributed graphs of a retrieval model and a target model. And, this score is calculated by a geometrical similarity and topological similarity. If there is such search noise, unreusing sub-parts must be canceled by interactive operation. This is an obstacle to automation. Therefore, with the goal of eliminating search noise, we developed a technique to improve the accuracy of similar partial shape retrieval. By quantifying the geometric relationships such as concentric circles and face-to-face distance as similarity, high precision could be realized. It was applied to a door switch cover part of an automobile which is a verification model, and it was confirmed that search noises disappeared.