大尺寸测量的不确定度分析方法探析

张福民 曲兴华 叶声华

（天津大学精密测试技术及仪器国家重点实验室 天津 300072）

摘要：常规尺寸的不确定度评价方法不适用于大尺寸测量，尤其是特定拟合任务中的不确定度评价。针对大尺寸测量的特殊性，研究了基于蒙特卡罗法的不确定度评价方法。通过随机仿真各项测量误差源，得到测量结果的仿真样本，并用计算机可视化直观表示该样本，形成离散不确定度点云，从而评价大尺寸特定测量对象的不确定度。以激光跟踪仪测量大型圆形截面工件为例，给出测点对称、均匀分布和半径约束等优化测量思想，减小测量不确定度。最后，将该方法应用于激光跟踪仪测量隧道构件的实例中，结果表明，常规圆心拟合不确定度为2.5525mm，加入半径约束优化测量方案后，不确定度减小至0.0326mm。仿真和实际实验表明，蒙特卡罗评价和离散点云表示法可准确、直观地评价特定大尺寸测量对象的不确定度，制定的最优测量方案可提高测量精度。

关键词：大尺寸测量 不确定度 蒙特卡罗法 面向对象 激光跟踪仪

中图分类号：TB92 文献标识码：A

收稿日期：2008-01-21；修订日期：2008-04-08。

基金项目：国家自然科学基金资助项目（No.50575158；No.60723004）；高等学校博士学科点专项科研基金资助项目（No.20060056015）。(www.xing528.com)

Uncertainty estimation of large-scale measurement for special fitting task

ZHANG Fumin QU Xinghua YE Shenghua

（State Key Laboratory of Precision Measuring Technology and Instruments，Tianjin University Tianjin 300072）

Abstract：A new uncertainty estimation method was researched based on Monte Carlo evaluation，for some uncertainties of large-scale measurement could not be anslyzed by conventional methods，especially for special fitting task.In proposed estimation，the simulated sample was obtained by simulating each measuring error source randomly and denoted as discrete point-clouds by computer vision，so as the uncertainty of given measuring object could be evaluated.By taking analyzing large-scale circular section part by laser tracker for example，the optimized measurement con-ceptsincluding point symmetry，equal distribution and radius constraint were given.Finally，the op-timized evaluation method was used in measuring practical tunnel components by laser tracker，re-sults show that the uncertainty is decreased to 0.0326mm after radius constraint optimization，which is priority to average fitting uncertainty of circle center of 2.5525mm by traditional method.It is proved that Monte Carlo evaluation and discrete point-cloud representation can evaluate accurately and intuitively the uncertainty for largescale object and the optimum sampling strategy can improve the measuring precision.

Keywords：large-scale measurement uncertainty Monte-Carlo method special fitting task la-ser tracker