Design Reuse Oriented Discovery Of Multisource Information In 3D Assemblies

Design reuse is an effective way to accelerate the design process and improve the design efficiency,and model retrieval,typical structure discovery,function structure relationship discovery are the typical discovery methods to obtain the design resource to support design reuse.With the development of MBD-based 3D product design system,the information carried by the model keeps increasing in type,quantity and source,and the CAD model has been developed as a collection of multisource information.In this case,how to properly describe the design intent and obtain the reusable design resource considering the influential information has been a major obstacle that impedes the reuse.Therefore,carrying out methods of design resource discovery with the consideration of multisource information is a feasible solution to this problem.Currently,most of the work on this topic is centered on part,mainly focuses on retrieving and reusing the part by taking advantage of model retrieval technology.Assembly is a combination of parts,and it is also a comprehensive expression of function,connection,structure and other design information.Therefore,discovery of the design resource from assembly obtains more valuable design resource that fits the design requirement better,and shows more important practical significance.In this paper,the design reuse oriented discovery of multisource information in 3D assemblies has been proposed,which first builds a description model of 3D assembly information,and then carries out the assembly retrieval,common structure discovery and function-structure relationship mining.The main work of this paper is as follows:(1)A multisource information description model for assembly is built.First,the formation of the multisource information in assembly is analyzed,and the definition and characters of multisource information are summarized.Then,the description model is built using UML,in which the function information,part information,assembling information and organization information is described with the proposed normalized description method.Finally,some description rules are set based on the analysis of the relationship between information to ensure the completeness and accuracy of the description.(2)An assembly retrieval approach that considers multisource information in assembly is proposed.First,an assembly description space is established by taking the part information as dimensions to describe assembly information.Based on that,an assembly is represented as a point set,and the comparison of assemblies is transformed into the comparison of point sets.Afterwards,two kinds of distance measurement methods are presented to calculate the distance between shape attributes and the distance between semantic attributes,respectively,and the distance between points is obtained by merging the distances in all dimensions.Finally,the many-to-many match method is employed to match the point sets and then comprehensively evaluate the differences between assembly models,which obtains the retrieval results considering multisource information.(3)A discovery method is developed to obtain the common structure in assemblies with the consideration of local difference in parts.First,to achieve the double mapping between assembly and descriptor,a graph-based assembly descriptor is developed.In the descriptor,the vertices and edges represent the parts and relationship between parts,respectively,and the attributes of vertices and edges are also defined.Then,a clustering step is introduced into the discovery to cluster the assembly descriptor,which obtains the similar parts and relationship in assemblies and achieves the goal of ignoring the local difference in parts.Finally,a frequent subgraph mining is employed to obtain the frequent subgraphs in the clustered descriptors,and the frequent subgraphs correspond to the required common structures in assemblies.(4)A statistical method is proposed to obtain the function-structure relationship in assemblies.First,the strategy of mining the function-structure relationship is proposed,which mines the function-structure relationship in assemblies to directly obtain the mapping results and avoid the complex analysis.Then,a statistical strategy is proposed that treats the assemblies we gather as the sample,and the maximum likelihood estimation(MLE)is employed to estimate the probabilistic function-structure relationship,i.e.the conditional probability of the appearance of a structure given a function in the assemblies.Finally,the core functions set in assemblies is obtained with mining,and the typical structures set that provides the core functions set is achieved based on the probabilistic function-structure relationship.Some assemblies in mechanical field are selected as example to validate the effectiveness of the proposed methods through assembly multisource information modeling,assembly retrieval,common structure discovery and function-structure relationship mining.Meanwhile,a prototype system is developed based on the research,and the functions in the prototype system are achieved to evaluate the feasibility of the study and the system.