基于统计学习的自由曲面轮廓度误差评定与不确定度研究

发布时间：2019-01-24 08:58

【摘要】：随着自由曲面类零件在汽车制造、航空航天、造船、模具加工等许多工业领域的广泛应用,对此类零件的加工精度也提出了越来越严格的要求。目前,如何实现自由曲面零件的高效率、高精度的检测,已成为当前产品表面质量检测的重要研究内容。为了获取自由曲面的加工质量信息,本文重点研究了自由曲面轮廓度误差的评定模型、算法以及误差不确定度估计方法。首先,依据形状几何误差经典定义,综合刚体运动学中的齐次坐标变换理论和最小区域准则建立了自由曲面轮廓度误差评定的数学模型。该模型的优点是利用最小二乘法对测点进行预定位处理,并将预处理后的有效测点作为精定位时的初始测量数据,进而为精定位算法的实施锁定了高质量的初始位姿。其次,研究了基于序列二次规划算法(SQP)的自由曲面轮廓度误差评定方法。针对仅已知CAD模型的曲面,提出了计算测点到自由曲面最小距离STL(STeror Lithography)模型方法;针对已知参数表达式的曲面,采用分割逼近法计算测点到该曲面的最小距离。然后,采用序列二次规划算法,并结合距离函数的线性微分运动模型,实现测点与理论曲面的最佳配准。此外还利用MATLAB对误差评定算法进行了仿真实验,试验的对比结果表明该轮廓度误差评定算法的有效性。最后,分析了影响自由曲面轮廓度误差的不确定度因素,并依据测量不确定度表示指南中的不确定度传播律,建立了曲面轮廓度误差不确定度估计模型。最后,通过搭建实验平台对上述模型和算法进行了实验验证。
[Abstract]:With the wide application of free-form parts in automobile manufacture, aerospace, shipbuilding, die processing and so on, more and more strict requirements have been put forward for the machining accuracy of these parts. At present, how to realize the high efficiency and high precision detection of free-form surface parts has become an important research content of product surface quality inspection. In order to obtain machining quality information of free-form surface, the evaluation model, algorithm and error uncertainty estimation method of free-form surface contour error are studied in this paper. Firstly, according to the classical definition of shape geometric error, a mathematical model for evaluating the contour error of free-form surface is established by integrating the homogeneous coordinate transformation theory and the minimum region criterion in the kinematics of rigid body. The advantage of this model is that it uses the least square method to prelocate the measured points, and takes the pre-processed effective measuring points as the initial measurement data for precise positioning, thus locking the high quality initial pose for the implementation of the precision localization algorithm. Secondly, a method for evaluating the contour error of free surface based on sequential quadratic programming (SQP) is studied. For the surface with only known CAD model, a STL (STeror Lithography) model method is proposed to calculate the minimum distance from the measuring point to the free surface, and the partition approximation method is used to calculate the minimum distance from the measured point to the surface for the surface with known parametric expressions. Then, using the sequential quadratic programming algorithm and the linear differential motion model of the distance function, the optimal registration between the measured points and the theoretical surface is realized. In addition, the error evaluation algorithm is simulated by MATLAB, and the comparison results show that the algorithm is effective. Finally, the factors of uncertainty affecting the contour error of free-form surface are analyzed, and a model for estimating the uncertainty of contour error of curved surface is established according to the uncertainty propagation law in the guide to express the uncertainty of measurement. Finally, the above models and algorithms are verified by building an experimental platform.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.72;O186.11

【参考文献】