【作者】 何雪明；

【导师】 李成刚； 胡于进；

【作者基本信息】 华中科技大学 ， 机械设计及理论， 2010， 博士

【摘要】 逆向工程作为一种产品创新设计的重要手段,在机械、航空、航天、医疗等领域得到了越来越多的重视和广泛应用,是随着计算机科学、计算几何和测量技术的迅速发展而成熟起来的一门新兴综合性学科与技术。采用逆向工程技术,不但可以缩短产品开发周期,还可以较快地赶上或超过世界先进生产水平,综合运用CAD/CAE/CAM以及RPM(快速成型制造),找到原型产品的缺点并加以改进,设计生产出更优质的产品。对具有自由曲面特征三维散乱点云的直接加工重构技术更是当前逆向工程研究的热点,也是其中的重点和难点。本文就是基于双映射和人工神经网络,针对采自具有自由曲面特征机械产品数据点云的直接加工一些关键问题和技术进行了深入的探讨和研究,具体内容如下:(1)基于曲线曲面几何理论,在以接触式三坐标测量机为载体测量具有自由曲面特征机械产品时,首次提出了为保证测量精度和测量速度的曲率连续自适应测量法,建立了向前探测步长和侧向跨距模型,模型具有较好的通用性和鲁棒性,演绎了高阶多项式和Bezier参数形式的曲率连续自适应测量方法,为具有自由曲面特征机械产品数据点云的直接加工搭建了较好的平台。(2)对经三坐标测量机或激光扫描仪通过多次装夹多次采集得到的数据点云,提出借助人工神经网络来统一和对齐无序数据点云的方法,详细比较了目前较典型的五大类16种神经网络结构的训练和学习功能,寻找分段和最快排序的最优人工神经网络模型,选择运行函数、各种系数、初值和隐含层神经元个数等参数,确定了快速准确识别并对齐各数据点云块的神经网络模型,提升了对齐的效率和精准性。(3)通过建立点云密度的计算模型,确定了对齐点云的切片参数和方向,探索最佳的交互切片方式,详细研究了适合无序数据点云的存储结构,基于正交投影理论建立了快速搜索并定位切片内与当前点最邻近点的双映射排序模型和方法,在数据k-邻域理论基础上获得了近点矢量趋势计算法完善了双映射排序算法,完成切片内数据点的正确序列,分析了相邻截面线之间的剖分方法,基于较少数据点截面线的三角化法构建了散乱点云三维空间无空洞无缠绕的拓扑结构。(4)在三角面片上直接安排数控加工,将无干涉刀具路径的生成与干涉检查两个相互耦合的过程相对独立开来,建立了自由曲面数控加工过程中刀具与工件的干涉检查模型,演绎了适合球头刀、平底刀和圆环刀的干涉检查方法,实现了相应的全干涉检查,检测出了曲面加工中的局部和全局干涉,采用等残留高度法生成了无干涉的刀具路径,后置处理成G代码,使得整个数控加工有序、精准而快速。最后结合OpenGL开放式图形库采用VC++开发了基于双映射和人工神经网络的数据点云直接加工原型系统,三角面片后跳过传统曲面重构直接安排数控加工,保证了反求产品的质量、精度,集成并加快了反设计及数控加工的一体化过程。更多还原

【Abstract】 As an important approach in creative product design, reverse engineering(RE) has been aroused more and more attention and used widely in many areas such as machinery, aviation, aerospace and medical treatment etc. Reverse engineering is a novel integrated subject and technology coming up with the development of computer science, computational geometry and measuring technology. By using reverse engineering technology, the product development cycle can not only greatly been shorten. It is more importantly that the level of the world’s advanced production technology can rapidly been catch up or exceeded as the technology secrets of the foreign advanced equipments and products were opened and the weaknesses of prototype product were found out. Together with CAD/CAE/CAM and the rapid prototyping and manufacturing (RPM), the better quality products can be designed and manufactured, and then the market will be seized quickly. Research on direct processing reconstruction of data point cloud digitized from sculptured surfaces is a focus and key one.Aiming to the features in this dissertation, some of the key issues involved in direct processing of data point cloud are studied and explored in depth digitized from the mechanical product characterized by sculptured surfaces based on dual mapping and artificial neural network, and the main research works and conclusions include:(1) Based on the geometric theory of curves and surfaces, the adaptive curvature continuous measurement method is proposed to ensure the measurement accuracy and speed when the mechanical product characterized by sculptured surfaces is measured on the contact coordinate measuring machines(CMM). The model of detection forward step and side span of the probe is presented and has good versatility and robustness. The adaptive curvature continuous expressions of higher order polynomial form and Bezier form are derived so that a better platform of systematic direct processing reconstruction of data point cloud digitized from sculptured surfaces is built.(2) The artificial neural network method is brought up to align and unify disorder data point cloud collected several times on CMM or laser scanners through repeatedly clamped. After a meticulous comparison of neural network structure is made in training and learning function among current five typical classes of 16 neural networks, the optimal segmentation and sequencing artificial neural networks are found and run functions, coefficients, initial value and the number of neurons in the hidden layers are selected. The neural network model is determined so quickly and accurately to identify and align several of data point clouds that the efficiency and the accuracy are enhanced.(3) Through the calculation model of point cloud density is established, the slicing parameters and direction are identified on data point cloud aligned and the best interactive slicing way is explored. Appropriate data structures are discussed in detail to store unordered data point cloud. Based on the theory of orthogonal projection, the algorithm model and method of dual mapping sequence are set up to search and locate quickly the nearest points of the current point studied in the slice, and then the calculation method of point vectors is acquired to improve the above sort algorithm in view of k-neighborhood theory. After the right sequence is finished within the data slice point and the subdivision method of adjacent section lines is analyzed, the overall 3D topology of scattered data point cloud is constructed by the triangular of the cross-section line subject to less data points without inanition and winding.(4) CNC machining is directly arranged on the triangular facets. Two mutual coupling processes are separated relatively in the generation of interference-free tool path and the interference detection. The model of interference detection is established in the process of CNC machining. The patterns and algorithms are derived suitable for ball-end cutter, flat cut and ring cut, and then the corresponding interference inspection procedures are achieved to check out the local and the global interference when machining on sculptured surfaces. Due to the method of the constant scallop height, the interference-free tool path is accomplished. The tool path is post-processed into the G code program so that sculptured surfaces are machined in order, accurate and fast.Finally based on double mapping and artificial neural network, a direct processing reconstruction system is developed for data point cloud digitized from sculptured surfaces by the combination of OpenGL graphics library using VC++. After the establishment of triangular facets, the CNC direct machining is arranged to skip the traditional surface reconstruction. The integration process of reverse design and CNC machining is integrated and speeded up in order to ensure the product quality and accuracy.更多还原