Research On The Reconstruction Method Of CAD Model For Conventional Engineering Parts Based On ATOS Measuring System

3D CAD technology has been widely used in the manufacturing industry with the development of computer graphics, computer aided geometric design, high performance computing and other related technologies. The CAD model is the basis of product development, manufacturing, re-design and maintenance. Therefore, it is necessary to acquire the CAD model by conventional engineering or reverse engineering. Reverse engineering of physical objects refers to the process of creating engineering design data (including the CAD model) from existing parts by using many digitized technologies and CAD technology. The thesis addresses the reconstruction theory and methods of CAD models of conventional engineering parts based on the ATOS (Advanced Topometric Sensor) measuring system. The structured light data acquisition method is prevailing in current commercial measuring systems because of it ability to catch edges and high curvature corners, which is very important in reverse application The ATOS system is typical in the measuring system base on the structured light method. The working principle of ATOS is described and discussed here. And several key technologies of ATOS, triangulation, fringe projection and phase shifting, are investigated in detail. Then, the process and main steps of acquiring point clouds of the measured objects are discussed. The principles of placing the reference points on the measuring object are proposed.The original point clouds acquired by ATOS are highly dense and scatted, which include large number of redundancies and noises. Thus it is necessary to delete those redundancies and noises to reduce the compute cost of the downstream process. Data preprocessing refers to the process of polygonization, smoothing, decimation and registration. The flow and main steps of data preprocessing are described here. Several specific cases, for example the sampling method based on chordal deviation, the feature based registration, are discussed.The surface of conventional engineering part is composed of simple analytical surfaces and four-boundary ruled surface primarily. Therefore a point cloud segmentation method based on those feature surfaces is proposed. Several algorithms to identify the simple analytical surface feature are investigated and discussed. The theory of fitting scattered point to B－Spline and Bézier surfaces is introduced and a fitting method based on four boundary and point set is also discussed. Moreover, anew method to recover the fix-radius rolling ball blend is proposed and discussed in detail.The reconstruction model should satisfy the following requirements: it is geometrically accurate; it can represent smooth, curvature continuous surface as well as sharp edge, common in manufactured parts. The thesis uses the profile tolerance as the indicator to evaluate the dimensional error between the reconstruction model and the physical object. Then the algorithm for computing the profile tolerance is discussed. Several methods to evaluate the surface smoothness, such high line and reflection line, etc. are also investigated and discussed.At last, the theory and methods discussed above were applied in the project "The Construction of the Virtual Mock-up of the Assemblies of the Main Floor and the Rear Floor of a Car" successfully. The CAD model of the rear floor reconstructed by those methods satisfy the requirements of the customer.