基于改进豪斯多夫距离的非参数轮廓变点识别  

A Change Point Detection Method Based on Modified Hausdorff Distance in Nonparametric Profiles

作　　者：聂斌[1] 孙会东[1] 李佩[1] 杜文超[1] 

机构地区：[1]天津大学,天津300072

出　　处：《中国机械工程》2015年第8期1029-1034,共6页China Mechanical Engineering

基　　金：国家杰出青年科学基金资助项目(71225006;7141123);国家自然科学基金资助项目(71102140)

摘　　要：轮廓监控中的变点识别问题是统计过程控制的重要研究内容。以非参数轮廓数据为研究对象,运用图像特征识别的豪斯多夫距离测量了样本轮廓之间的特征差异,设计了改进方法,提出了基于二维空间的改进豪斯多夫距离算法,以识别非参数轮廓变点。大量的仿真与论证表明,改进方法在识别变点位置和稳定性方面具有优异的性能。Change point identification in profile monitoring is an important research topic in statis- tical process control. Herein,the profile had nonparametric characteristics. The proposed method was based on Hausdorff distance, and could be used to measure difference between profiles. A modified Hausdorff distance algorithm was proposed to identify nonparametric profile change point. The com- parison results of simulation study show that when there exists local changes in nonparametric pro- file,the modified algorithem has advantages in locating change points and performance stability.

关 键 词：非参数轮廓 变点识别 豪斯多夫距离 T2统计量 

分 类 号：TH165[机械工程—机械制造及自动化]