一种剔除测量异常点的循环逼近算法的设计技术及其工程应用

A Circular Approximation Algorithm for Eliminating Outliers and Its Engineering Application

对基于"裹圆"结构的少半圆类零件进行反求设计时,三点定圆法会使测量误差放大,且圆心角越小,方法系数增长越剧烈,通常采用密集采点,然后对采集到的数据进行处理的方法评价圆弧,而剔除影响测量精密度的异常点是关键问题。提出一种操作性强、精度可靠、工程应用性强的循环逼近算法,此算法假设每一个测量值均为测量真值,得到一个标准公差数列,根据剔除测量点前后的标准公差最大值的相对误差判别该点是否剔除,逐步逼近设定的相对误差值,直至剔除前后标准公差数列基本无变化。此算法简单可行、实时性好,适合工程应用。

In the reverse design of the semicircle parts based on the ＂wrap round＂ structure,the method of threepoint fixing circle will enlarge the measurement error,and the smaller the circle angle,the increase of the methodcoefficient is stronger.And usually,the intensive mining point is used,then the arc is evaluated by the method ofprocessing the collected data,and the anomaly points that affect the precision of the measurement are eliminated,it is the key problem. In this paper,a kind of cyclic approximation algorithm with strong operational,reliable andpractical engineering application is proposed. This algorithm assumes that each measurement value is the true val-ue of measurement,and gets a standard tolerance series. It is judged whether the point is eliminated or not,ac-cording to the relative error of the maximum standard tolerance before and after the elimination of the measurementpoint,and gradually approximated the set relative error value until the standard tolerance sequence is basically notchange before and after the elimination. This algorithm is simple,feasible and real-time. It is suitable for engi-neering application.