General and Partial Shape Matching Approaches on Feature-Based CAD Models to Support Efficient Part Retrieval

Abstract

Presently, mechanical design activities are almost fully digitalized and a large percentage of CAD models that are digitally created can be reused to facilitate new designs. Conventional alphanumeric-based part retrieval methods are error-prone. In order to support users to efficiently locate desirable parts to reuse, two novel knowledge-based approaches for feature-based model retrieval are proposed in the paper, namely general shape matching and partial shape matching, to assess CAD models by measuring their overall similarity without details and local similarity of sub-parts, respectively. The first approach simplifies feature-based CAD models from fully detailed to less detailed, and progressively simplified shapes are characterized by general shape descriptors. When a 3D query model that represents the general shape of a desired part is sketched, the approach compares the query model against pre-generated general shape descriptors so that all parts similar to the 3D query in overall shapes are retrieved; therefore the general shape matching is achieved. The second approach extracts reusable sub-parts from feature-based CAD models in an unsupervised way and indexes them using partial shape descriptors. Based on the partial shape descriptors indexed, designers could retrieve all existing parts sharing a desirable sub-part by sketching the shape of the sub-part; the partial shape matching is thereby realized to facilitate the part retrieval on local similarity. The approaches proposed in this paper are implemented on a prototype system and tested by hundreds of mechanical parts. The preliminary results show that the proposed approaches are feasible and promising.