多元质量控制的鲁棒主成分分析方法  

Robust Principal Component Analysis Method for Multivariate Quality Control

作　　者：刘宗宝 王木子 刘更 张力 LIU Zongbao;WANG Muzi;LIU Geng;ZHANG Li(Beijing Institute of Computer Technology and Application,Beijing 100854)

机构地区：[1]北京计算机技术及应用研究所,北京100854

出　　处：《计算机与数字工程》2022年第11期2382-2385,共4页Computer & Digital Engineering

摘　　要：主成分分析(PCA)对异常数据点很敏感,基于PCA的多元质量控制在异常点存在情况下容易失效。针对该问题,论文基于鲁棒主成分分析理论,提出了基于RobustPCA的多元质量控制方法。该方法对观测数据进行鲁棒主成分分析,然后进行特征值分解,最后绘制休哈特控制图以判断生产过程各时刻的状态。仿真结果和实际数据表明,论文提出的方法可有效应对质量观测数据存在异常值的情况,相比基于PCA的质量控制方法具有更强的鲁棒性。Principal component analysis(PCA)is sensitive to outliers.Mltivariate quality control based on PCA performs poorly with respect to grossly corrupted observations.In this paper,a new quality control method is proposed based on robust principal component analysis(RobustPCA).The proposed method analyzes RobustPCA to observations,the output is then analyzed by eigenvalue decomposition.Finally Hotelling T2control chart is presented for process understanding and monitoring.Simulation and a typical dataset are used to illustrate the practical benefits of using RobustPCA in quality control,which is more robust than conventional PCA methods.

关 键 词：质量控制 鲁棒主成分分析 异常点 特征值分解 

分 类 号：V419.3[航空宇航科学与技术—航空宇航推进理论与工程]