Shape Structure Analysis Methods For CAD Models

Design reuse as a significant method for enhancing core competence of enterprises plays an important role in mechanical design. With the wide application of digital design, which has gradually become the main design approach, a large number of 3D digital models have been accumulated in the enterprises. The reutilization of these models is an efficient way to improve the design efficiency, accelerate design process and shorten development period. Since shape structure informations constitute the major engineering information of digital models, shape structure analysis of 3D models is essential for achieving model reusability. Therefore, in this dissertation, based on the systematical investigation on the shape structure of digital models, reuse-oriented shape structure analysis methods for 3D CAD models are put forward.According to the analysis of the existent research work regarding these issues, a shape analysis principle based on space translation is presented. Generally, graph-based methods are often accepted for the CAD models analysis problems, for example, face-relation graph (FRG) and volume-relation graph (VRG). Nodes and arcs of FRG respectively represent the face information and the connection information of two faces in a CAD model, but those of VRG are respectively represent volume information and the connection information of two volumes. Considering the curve and surface informations of FRG, a point-property method has been proposed to describe the shapes of curve and surface with a set of vectors obtained from space translation. On the other hand, a code-based method is developed to describe volume's shape characteristics in VRG, which also utilizes vectors to represent shape information.Usually, existant design structures are usually adopted in part design process of mechanical design which leads to repeated appearance of the common local design structures in many CAD models. For these local structures are important reusable design resources, a method of recognition and extraction of these structures among multiple models is presented. From the viewpoint of the same structure with the same shape feature, extraction of local design structures is just equal to that of local structures with same shape feature. Using the FRG and its corresponding shape description of CAD models, this problem is turned into common subgraphs extraction of multiple FRGs. Moreover, subgraphs and its relations in VRGs are described with the constraints. After detection methods of constraint equivalents are given, the common design structures can be effectively extracted from multiple models.An approach of uniform representation and integration for multiple models based on the design pattern recognition is presented. Through getting design patterns from plenty of part models, these independent models can be described as the combination of design patterns to attain the object of uniform representation and integration, which is also helpful for model resource management, model database compression and rapid modularization design. In this method, a cutting-loop shrinking algorithm is first adopted to general VRG representation of all the CAD models. Then, by means of the principle of Minimum Description Length (MDL) in the theory of Data Mining, an optimization model based on the graph MDL is proposed for design pattern recognition. Finally, the MDL optimization problem is solved to get the design patterns with a greedy search strategy to different node merging options.Considering many mechanical parts have the containment relations, a method of containment relation analysis for models is given based on the investigation of the hierarchical structural relationships for the organization of multiple models. In this method, CAD model is also represented with graph and the shape information obtained from the boundary segmentation for models. Thus, the model containment is just a problem of graph containment. Then, the containment relations of models are determined by means of graph matching.To validate the ideas and methods proposed in this dissertation, the general architecture of a prototype system for shape structure analysis for CAD models has been developed using Visual C++6.0, geometry engine ACIS and display engine HOOPS. The system is comprised of the modules of common structure recognition, design pattern extraction, containment analysis and so on. The experimental examples show the correctness and effectiveness of the research results.