设为首页 联系我们 加入收藏

当前位置: 网站首页 期刊分类目录 2025第5期 正文

基于动态局部-全局主成分分析的故障检测

作者:时间:2025-11-01点击数:



全文下载: 202505017.pdf


文章编号: 1672-6987(2025)05-0133-09 DOI: 10.16351/j.1672-6987.2025.05.017

郭金玉, 王乐, 李元(沈阳化工大学 信息工程学院, 辽宁 沈阳 110142)

摘要: 为了有效提取数据的动态、局部和全局信息,提出一种基于动态局部-全局主成分分析(DLGPCA)的故障检测方法。首先,对原始数据矩阵构造一个具有时滞变量的增广矩阵,将增广矩阵标准化。然后,将局部保持投影引入到常规主成分分析中,计算投影矩阵,将高维空间线性压缩为一个低维空间,使低维空间具有和原始数据空间相似的全局结构,并且保留了相似的局部近邻结构。最后,通过计算T2SPE统计量监控样本状态。将该方法应用于数值例子和田纳西-伊斯曼过程。仿真结果表明,与常规的故障检测方法相比,DLGPCA既能有效地检测多变量动态过程中的故障,又能捕捉到隐藏在观测数据中的局部和全局信息,显示出了优越的过程监控性能。新算法解决了传统算法中具有时间相关性的数据信息获取不全面的问题,为提高传统算法在动态工业过程故障检测中的性能提供了参考。


关键词: 故障检测; 局部保持投影; 主成分分析; 局部-全局主成分分析; 动态建模


中图分类号: TP 277        文献标志码: A


引用格式: 郭金玉, 王乐, 李元. 基于动态局部-全局主成分分析的故障检测[J]. 青岛科技大学学报(自然科学版), 2025, 46(5): 133-141.


GUO Jinyu, WANG Le, LI Yuan. Fault detection based on dynamic local-global principal component analysis[J]. Journal of Qingdao University of Science and Technology(Natural Science Edition), 2025, 46(5): 133-141.

Fault Detection Based on Dynamic Local-Global Principal Component Analysis

GUO Jinyu, WANG Le, LI Yuan(College of Information Engineering, Shenyang University of Chemical Technology, Shenyang 110142, China)

Abstract: A fault detection method based on dynamic local-global principal component analysis (DLGPCA) is proposed to effectively extract dynamic, local and global information of data. Firstly, an augmented matrix with time-delay variables is constructed for the original data matrix, and the augmented matrix is normalized. Secondly, locality preserving projections (LPP) is introduced into conventional principal component analysis (PCA), a projection matrix is calculated and a high-dimensional space is linearly reduced into a low-dimensional space. The low dimensional space has the global structure similar to the original data space, and keeps the similar local neighbor structure. Finally, T2 and SPE statistics are calculated to monitor the sample state. The proposed method is applied to a numerical example and the Tennessee-Eastman process. Simulation results show that compared with the conventional fault detection method, DLGPCA can not only detect faults in multi-variable dynamic processes effectively, but also capture local and global information hidden in the observed data, showing superior process monitoring performance. The new algorithm solves the problem of incomplete acquisition of time-dependent data information in traditional algorithms, and provides a reference for improving the performance of traditional algorithms in fault detection of dynamic industrial process.


Key words: fault detection; locality preserving projections; principal component analysis; local-global principal component analysis; dynamic modeling

收稿日期: 2024-11-18

基金项目: 国家自然科学基金项目(62273242); 辽宁省教育厅项目(LJ2019007).

作者简介: 郭金玉(1975—), 女, 副教授.


Copyright © 2011-2017 青岛科技大学学报 (自然科学版)