节点文献
面向快速原型制造的反求工程关键技术研究
Study on Key Techniques of Reverse Engineering oriented to Rapid Prototyping Manufacturing
【作者】 孙玉文;
【导师】 刘健;
【作者基本信息】 大连理工大学 , 机械制造, 2000, 博士
【摘要】 快速原型制造(RP&M)是当代制造业的新兴技术,属前沿课题。反求工程与快速原型制造结合,将从根本上改变传统的产品设计、制造模式。但是,当前的反求技术还无法达到与快速制造技术的匹配。本文在探讨了面向RP&M的反求工程的科学体系与框架的基础上,从数据预处理、网格划分、光滑曲面构造、STL文件信息处理等方面进行了系统、深入地研究,以期实现RE/RP&M的有机集成,从而提高快速产品开发的能力。 首先,根据RP技术中STL文件的特点,有针对性地提出了离散数据的预处理和网格划分的具体方法。基于神经网络的数据修补,自动数据分块、以及自适应卡尔曼滤波等多种技术手段的使用,解决了当前在数据预处理方面存在的难点问题。云状数据的基于物理的网格划分和选点三角剖分等方法的提出,进一步丰富和发展了网格划分的内容,显著地提高了网格的生成质量。 其次,以现代微分几何学为工具,系统研究了光滑曲面的C~1和G~1构造方法:指出了初始网格的细分标准,并就此给出了均匀细分和选择性细分两种方式;指明了细分中细分数与反求容差的量化关系,提高了细分的科学性和合理性;运用优化理论,提出了曲面质量迭代提高以达到准C~2连续曲面的方法,极大地改善了曲面的光顺效果。 此外,通过对现有的多分辨率造型方法的分析,首次在反求工程中引入同心超面、星形邻域等概念,并提出了一种基于局部拓扑结构识别的网格简化新算法。这不仅有效地消除了网格模型中存在大量的冗余网格,减少了STL文件的长度,也使RP&M中的造型显示、支撑添加、切片等后续处理更为方便、快捷。 最后,对反求技术和快速原型技术的集成问题进行了深入研究;指出了表面模型用于RP&M中存在的问题;给出了STL文件拓扑重建的方法和数据结构;讨论了STL文件的错误校验和修复技术;在此基础上,进行了快速成型机的接口文件生成和实体成型加工实验,验证了本文所提技术路线的正确性和科学性,为实现RE/RPM的一体化起到了积极地推动作用。
【Abstract】 RP&M (Rapid Prototyping Manufacturing) is a newly developing technology in the field of manufacturing industry of the time. Reverse Engineering (RE), combined with RPM, may fundamentally change the traditional mode of production design and manufacturing. But, the present technology of RE doesn match the rapid prototyping technology at all. In this paper, the system configuration of RE based on RP&M (RERP) is proposed. It is expounded comprehensively from the aspect of data preprocessing, triangulation, smooth surface construction, STL file information processing , etc. Once realizing the organic integration of RE/RPM, it can greatly improve the ability of Rapid Product Development (RPD). Firstly, according to the characteristic of STL file, the scattered data preprocessing and methods of triangulation are presented. Data repairing based on neural network, automatic boundary point recognition, adaptive kalman filtering, etc, various kinds of technical means are utilitied for solving the difficult problems in data preprocessing. Some methods, such as physically-based triangulation and selective triangulation, are investigated enriching and developing the content of triangulation and improving the grid quality greatly. Secondly, by means of differential geometry, some methods of smooth surface construction are studied systematically. Given the subdivision criterion of initial grid, two techniques of averaged subdivision and selective subdivision are discussed. Resort to the quantitative relationship between subdivision number and allowed error of RE, subdivision becomes more scientific and reasonable. Using the optimum theory. an iterative method of improving surface quality so as to gain quasi-C2 surface is proposed. This scheme is capable of yielding smoother surfaces, if the fairing algorithm is applied one after the other. In addition, there are lots of redundant grids in polyhedral model gained from RE. It is unbeneficial for later treatment of RP&M. Through analyzing all kinds of methods of multi-resolution model, concepts of concentric superface and star neighbourhood are first presented in RE. A new model simplification algorithm, based on local geometry and topology recognition. is developed. It is not only reduce the length of STL file, but also bring convenience for displaying graphics and adding support. Finally, the integration of RP technology and RE is deep investigated. Some problems related with surface model and RP technology is pointed out. The method and data structure of reconsfructing the topology information of STL file are proposed. Also addressed in the paper are issues of the mistake check methods and repairing techniques of STL file. Then an experimental procedure of creating interface file and making the seat prototype of a bike is summarized in detail. The result testifies the feasibility and practicability of the technical line presented in this paper. All of the above-mentioned measures, impact the development for realizing the integration of RE/RPM.