Semantic-Based Assembly Model Retrieval

As a basis for model design reuse, CAD model retrieval is important for shortening the product development cycle and reducing the product cost. Compared with part model retrieval, researches about assembly model retrieval are preliminary. Results of current methods are far from desired assembly models, so they can't meet requirements of assembly model reuse. Semantic-based assembly model retrieval is studied in this dissertation for effectively supporting assembly model reuse.The main contents of the dissertation are presented as follows:(1) A hierarchical representation containing semantic of mechanism method of assembly model is presented, which supports different requirements oriented retrieval. Firstly, the partial regions and hierarchical representation are defined. Secondly, the algorithm of extracting semantic information of mechanism in assembly model and building SAG (an abbreviation of semantic assembly graph) are described. Lastly, combining with spectral graph theory and invariants of SAG, SAG is converted to assembly signature vectors, and then the hierarchical representation of assembly model is completed. Hierarchical representation is to support the global retrieval of desired assembly model through a fuzzy query; partial regions are represented to support the retrieval of independent regions in complex models; assembly signature vectors deduced from SAG are used for similarity assessment, effectively improving the efficiency of the assembly model retrieval.(2) A hierarchical-representation-based retrieval method of assembly model is presented. Firstly, According to requirements of design reuse, two retrieval modes are devised. In order to avoiding the amount of time wasted caused by one-by-one matching, m-tree indexing is used for improving the efficiency of retrieval based on the hierarchical representation. According to two retrieval modes, corresponding matching is realized and assembly models satisfying the retrieval requirements of users are returned by the matching results.Based on the above researches, a prototype system ZD-SBAMRS is developed. Meanwhile, some experiments are given to show the validity and the efficiency of the proposed approaches.