【作者】 刘元朋；

【导师】 张定华；

【作者基本信息】 西北工业大学 ， 航空宇航制造工程， 2006， 博士

【摘要】 航空发动机管路设计质量直接影响其可靠性和维修性。由于布局空间限制严格，管道层叠交错、形状复杂，管路设计及其数据管理一直以实物样件为信息介质。随着航空发动机设计试验技术向数字化方向发展，对管路数字化设计中的反求建模新技术提出了迫切需求。基于特征的反求建模是反求工程研究领域近年来发展起来的一项具有广阔应用前景的新技术。将其应用于航空发动机管路设计和数据管理中的三维建模，不仅可以提高设计质量和效率，而且也可以直接驱动管路的数字化制造与检测，大大缩短研制周期，降低研制费用。 本文针对航空发动机管路的反求建模问题，系统地研究了基于特征的反求建模理论和方法。提出了基于特征的航空发动机管路反求建模系统框架；并围绕其中的多视数据配准、数据分割、特征提取和管路特征建模等关键技术进行了深入研究；在此基础上，设计并开发出一个软件原型系统，并通过计算机仿真和实际管道的测量与重构实验，对本文的理论、方法和系统进行了实例验证。论文的主要研究成果包括： 1．提出了基于特征的航空发动机管路反求建模系统框架。分析了航空发动机管路中管道设计的特点及其技术发展状况；研究了基于特征的反求工程技术，提出了基于曲面特征和截面特征相结合的管路反求建模方法。分析了反求建模方法的实现流程及其关键技术，进而提出了基于特征的管路反求建模系统框架及其实现方法。 2．分析了曲面微分几何属性，提出了管路测量数据的多视数据配准算法。总结了曲面的微分几何属性，并对微分几何属性的估计技术及其实现方法进行了分析。针对管路测量数据的多视数据配准问题，提出了一种粗配准和精配准相结合的多视数据配准新算法。粗配准可有效地缩小模型坐标系间的差异，并能确定精配准模型参数的边界范围。在粗配准的基础上，利用L-BFGS-B算法对带边界约束的精配准模型进行优化求解。 3．提出了数据分割的理论以及管路数据的分割方法。从逻辑学角度对数据分割的一般机理进行了分析，在此基础上，结合数据分割方法、计算机图形学和集合论等知识对数据分割机理中具体特点及其分析过程进行了总结。利用数据分割理论研究的成果，对管路的分割方法进行了研究，提出了一种管路测量数据的区域增长分割方法。 4．提出了新的特征拟合算法。针对管路特征的提取问题，总结了管路特征提取过程中拟合问题的求解流程及方法，分析了特征约束关系的类型以及约束模更多还原

【Abstract】 Pipeline system with complex layout and high design criterion play an important role in the aeroengine. Traditional design and data management is based on practical mockup for aeroengine pipeline system due to no original drawings or documentations. Reverse engineering is a process of constructing CAD model of a physical part from the measuring point data so that subsequent modifying, machining and inspecting processes may be implemented. The CAD model of aeroengine pipeline system reconstructed by reverse engineering can provides enormous benefit in improving the quality and efficiency of design, manufacture, analysis and data management. Research on such a typical system is valuable in the idea and technology to the similar complex industrial environments.In this dissertation, the concept and traditional methods of pipeline system is presented firstly. The basic theories including the definition, classification and representation of geometric feature based reverse engineering are studied systematically. The key techniques for CAD model reconstruction of aeroengine pipeline system are investigated and developed, including multi-view data registration, feature extraction, solid model reconstruction and theory of data segmentation. Main results achieved in dissertation are summarized as follows: Flowchart of modeling techniques based on feature for aeroengine pipeline system: First, the concept and characteristic of pipeline are analyzed. It outlines the present study status of reconstruction method for pipeline in domestic and abroad. The boundary surface of the pipelines is mainly composed of geometric regular surfaces, such as quadric surfaces, sweep surfaces and loft surfaces. Regular surfaces are extracted, which allow the most important feature parameters of the pipelines to be calculated. We present in this paper a method to obtain CAD models of the pipelines from measuring data. The models are based on two classification geometric features, surface-feature and section-feature, which are enough to reconstruct most parts of the pipeline system. The flowchart, architecture, and key techniques related to this methodology are studied. Multi-view data registration technique: Differential geometry properties of the surface are also essential elements of shape recovery methods. In this paper, we recall a few key notions of differential geometry firstly. Some estimation methods of local surface geometry properties for point data also are discussed and implemented. Afterwards, a multi-view registration method is presented, which is done by rough registration and accuracy registration. Rough registration can reduce the difference between models availably and make preparations for accuracy registration.In the process of accuracy registration, The registration algorithm based on L-BFGS-B are proposed.The experiments show this algorithm has better registration accuracy than traditional registration algoritms. Analysis and practical applications of mechanism for data segmentation:Though data segment technique has been researched for sever decades, there has been so lacking available theory to support segmentation algorithms. The mechanism on data segmentation process is investigated according to the principle of normal更多还原