A Study Of Combinatorial Assembly Model Retrieval Methods Incorporating Semantic Information

With the popularization of "smart manufacturing" in manufacturing enterprises,a large number of 3D part models have been produced.The rapid and effective reuse of 3D models is important for improving the efficiency and accuracy of product design and expanding the market competitiveness of enterprises.Traditional 3D model retrieval systems are based on individual parts for similarity retrieval,which is difficult to apply in the face of multi-featured combined assembly model retrieval problems,limiting the scope of 3D model retrieval and reuse.Therefore,this paper proposes a combined assembly model retrieval method incorporating semantic information,which effectively solves the difficult problem of retrieving multi-featured combined assembly models in the field of 3D model retrieval under the trend of smart manufacturing and provides technical support for the reuse of assembly models.The main research contents of this paper are as follows:(1)Analysis definition and problem of combinatorial assembly model retrieval.A combined assembly model is one in which parts with different assembly characteristics are assembled together under certain assembly conditions,and the part model occupies a specific spatial location and topology in the assembly model,so that the different assembly characteristics can be adapted to the performance and functional requirements of the product design.Most of the existing model retrieval techniques are based on single assembly interface retrieval for a single part,and do not take into account multiple interfaces and dual input model matching retrieval,and are not able to meet the multiple problems encountered in the production process.(2)Evaluate Similarity of assembly models incorporating semantic information.By analyzing the semantic attribute information and constraint relationships between the component models involved in the assembly,the relevant semantics are classified and thus formally expressed;the calculated values of the semantic similarity of the assembly of the models involved in the retrieval are obtained by empirical calculation,and the similarity of the component models involved in the assembly is evaluated by using the optimal matching of the bipartite graph.The fuzzy search before geometry retrieval is completed to exclude the part models that do not match the assembly intention and reduce the size of the pending retrieval for geometry retrieval,thus improving the efficiency of the assembly model retrieval.(3)Research on the combined assembly model retrieval method based on diagram isomorphism.With the scope of semantic retrieval results,accurate retrieval of combined assembly models is carried out from a geometric perspective.Considering the characteristics of the assembly model interface features fitting each other,a conjugate subgraph is established.Using attribute adjacency diagrams to represent the assembly and attribute relationships between assembly models,the principle that the structure of attribute adjacency diagrams between assembly interfaces is the same but the attributes(concavity)are opposite,the corresponding algorithms are used to make diagram isomorphism judgements.Retrieval methods for single-input models and dual-input models are investigated according to the differences in input models.(4)A prototype system for retrieval of combinatorial assembly models incorporating semantic information.Based on the general 3D model retrieval system architecture,the main functional modules of the model retrieval system are identified,and the overall architecture of the system is designed and a prototype system is constructed to verify the validity and rationality of the theoretical study.The results show that the combined assembly model retrieval method incorporating semantic information can effectively solve the retrieval problems of combined assembly models such as single-input and dual-input models,and extend the scope of 3D model retrieval applications.