针对用子孔径干涉像差修正拼接法测量非球面的拼接精度问题,提出了一种基于t分布统计的拼接精度分析评定方法。根据测量原理及拼接模型特点,利用残差分析对模型进行回归诊断,使用因变量预测算法和t分布统计模型对拼接精度进行估计与分析。实验结果显示:按t分布评定拼接结果的扩展不确定度为0.362 3λ(0.229 μm),而常规的系统误差检验方法评定拼接结果的扩展不确定度为0.234 μm,两种方法结果基本一致,表明基于t分布统计的方法在保证评定结果准确度的前提下,克服了比较分析法的不确定性,解决了正态分布统计法中置信系数k不能反映子样标准差可靠性对置信概率的影响问题。

A method for analyzing and evaluating stitching precision based on t distribution statistic was presented aiming at the problem of stitching precision of aspheric surface analyzed by sub-aperture interferometry based on correcting aberration. The residual error was used to implement the regression diagnosis according to the principle of stitching measurement and the feature of stitching model. Then, based on t distribution statistic,the forecast dependent variable in the regression analysis was used to assesse and analyze the stitching precision. The uncertainty in experiment result was analyzed, analyzed result indicates that the expansion uncertainty based on t distribution statistic is 0.362 3λ(0.229 μm), and the expansion uncertainty by the conventional precision evaluating method is 0.234 μm. The results of the two precison evaluating methods are idential basically,which shows the new precison evaluating algrithom overcomes the uncertainty of comparative analyzing method on the premise of ensuring the precision,and also solves the problem that the confident coefficient according to normal distribulion can not reflect the effect of sample root-mean-square reliability on the fiducial probability.