【作者】 龚友平；

【导师】 童水光； 金涛；

【作者基本信息】 浙江大学 ， 化工过程机械， 2006， 博士

【摘要】 逆向工程是实现产品再创新开发及快速制造的重要技术之一，在汽车、摩托车、飞机、家电、模具等行业的改型设计与创新具有广阔的应用前景。作为逆向工程中的关键技术——模型重建的理论方法一直是逆向工程的研究热点。本文围绕逆向工程中基于特征约束的模型重建理论与方法展开相关的研究和实践。基于特征约束建模的思想就是从测量数据中提取出特征参数，并推断出特征之间的约束关系，从而实现设计意图的重现，进行再设计，实现产品的创新。本文提出基于特征约束的模型重建框架结构，重点研究了基于测量点集的区域分割，截面轮廓特征信息的获取，二次曲面特征识别和重构，并提出有参自由曲面重建的思想和重建策略。在截面轮廓特征单元和二次曲面轮廓特征单元识别过程中，研究和总结了这些特征单元之间的约束关系自动识别和推理规则，并用识别的约束对截面轮廓特征单元和二次曲面单元进行约束优化拟合，本文提出用同伦法来求解约束优化拟合过程。测量点集预处理是后续特征约束建模的基础，因此，这里进行了测量数据前处理的相关研究。为了能准确、快速实现点云的区域分割，提出基于计算局部曲率信息进行数据区域分割的方法；为了满足后续有参自由曲面模型重建所需要的截面轮廓信息，采用截面切片的方法来获取要截面轮廓信息，在进行切片处理时，提出一种类似于移动最小二乘的细化算法。整个细化算法不对测量数据进行局部坐标变换，迭代步长由点云束密度控制迭代步长，最后得到的近似轮廓点集。对细化后的点集进行排序，冗余数据的删除以及特征点的提取处理，为后续的截面轮廓特征单元的识别和重构奠定基础。截面轮廓的生成是实现后续有参自由曲面特征重建的关键，同时对后续模型重建中曲面边界的确定也起着重要的作用，本文主要研究了根据平面轮廓数据点提出了一种新的基于约束的草图轮廓重建思路。首先分步法实现了以最少的轮廓特征单元拟合二维轮廓特征，然后根据拟合特征单元的信息，对轮廓特征之间的约束进行了识别，并建立了基于约束的二维轮廓特征优化拟合模型，最后由同伦法实现了二维轮廓特征的约束优化拟合技术。提出了基于约束识别的规则二次曲面特征拟合的方法和基于神经网络直接由测量数据实体特征直接提取的技术。在进行约束优化拟合二次曲面之前，采用最小二乘法获取了平面和球面的初始特征参数，用“代数法”确定柱面和锥面的初始参数；根据二次曲面初步参数，识别二次曲面间的约束，建立基于约束的二次曲面拟合模型，首次提出用“同伦法”来求解基于约束的二次曲面特征的优化拟合问题。针对一些机械零件中比较常用的功能零部件，如凸台，槽等标准特征，则采用神经网络方法直接由测量数据中将其识别构成实体特征。自由曲面是工业产品，尤其是飞机、汽车、摩托车等行业产品外形的重要特征。本文提出有参自由曲面特征的重建策略，即将截面轮廓和一些方向矢量表达作为自由曲的参数，选取适当的曲面重构方式，实现自由曲面的重建工作。本文研究了放样、扫掠、旋转、网格和融合等典型曲面的重构方式以及构建过程中所需要的参数，实现了自由曲面的参数化重构。为了解决曲面精度评价问题，提出一种基于切平面法快速求解点到自由曲面的算法。应用上述理论研究成果，在教育部博士点基金项目“基于特征约束的模型重建方法理论研究”(项目编号：2002033062)的支持下，研制开发了基于特征约束的模型重建系统的原型系统，为特征约束重建理论方法在实际应用提供理论支持和实践参考。更多还原

【Abstract】 Reverse Engineering (RE) is the important techniques in product innovation and fast manufacturing, also has vast applied foreground in modified design and creative design in automobile, motorcycle, airplane, appliance, molding tool etc. As the key technique within Reverse Engineering—the theories method of the model reconstruction has been the research heat. The paper carries on related work around model reconstruction theories and methods based on features in Reverse Engineering.The thought of model reconstruction based on feature and constraints is to extract parameter from measure data, infer the constraints of these features, and the design idea will be obtained, product can be re-design and the innovation can be carried out. The paper puts forward the model reconstruction frame structure, research in the segmentation of point cloud data, the cross-section profiles information achievement, the recognition and reconstruction of quadratic-surface (QS), puts forward the thought of para-free form surface (PFFS) and some strategies of these surfaces reconstruction. In the process of profiles feature elements and quadratic-surface feature elements recognition, some auto-recognition and inference rules in these feature elements are summarized and researched. Also, these constraints are used in fitting process of these feature elements, and the paper put forward homotopy method to solve theses fitting process.The point data pre-processing is the foundation of model reconstruction; problems related to data cloud processing are focused on. To segment point cloud quickly and exactly, the paper puts forward to compute local curvatures information, then segment these point cloud based on the information. To satisfy the information of PPFS reconstruction, slicing method is used to get profiles information. In the processing of slicing data, a method likes Moving Least-Squares is used to simplified these slicing data. The method don’t make coordinate conversion of these point data, the iteration step size is decided by density of slicing point cloud, at last, a point set satisfied profiles accuracy standard has been generated. After point data simplifying, there are some other processing method about point cloud: point set ordering, redundant point eliminating, and feature point extracting, which make solid foundation to profile feature elements recognition.The profiles generation is the key of PFFS reconstruction, at the same time, it play great role in confirm the boundary of reconstructed surface. The paper researched a new idea to reconstruct sketch profiles from slicing point data. At first, make use of sub-step method to get two-dimension profiles by minimal profiles feature element, then according to the feature information of fitting elements, the constraints of these elements have been recognized, and a optimal fitting model to two-dimensions profiles is created, the paper firstly solves the fitting model by homotopy method.Quadric surface fitting method based on feature parameters and constraints and solid features extracting by neural net method are used in accurate reconstruction of CAD model. Before optimal fitting QS, least squares method (LSM) is used to getplane and sphere surface parameters, algebra method is used to get cylinder and cone surface parameters, according to these preliminary parameters, constraints between these quadric surfaces can be recognized, and optimal fitting model based on constraints is constructed. Here presents the homotopy method to solve fitting problems. To some standard feature such as slot, boss, the neural net method is used to recognize as solid feature.Free form surface is most important feature to shape of product such as plane, automobile, motorcycle. The paper put forward a strategy to reconstruct the free form surface—Parameterized Free Form Surface (PFFS), here defines some surface construct by profiles and guiding vector etc as PFFS, and these profiles and guiding vector are parameters of the surface. The paper performs the strategy to reconstruct the PFFS is: firstly, getting parameters of the surface, then selecting suitable constructing method, at last, computing the error between points cloud to surface. The paper researches lofting, sweeping, revolving, net and blending method. To evaluate if the surface is suitable to the point cloud, a quick computing distance from point cloud to surface—tangent plane method has been carried out.Supported by Educational Part Doctor Found (The Theory and Method in Model Reconstruction based features and constraints, project number: 2002033062), the proposed methodology has been approached in Model Reconstruction System Based on Feature and Constraints prototype system.更多还原