基于部分抽样检验的在线异常监控方法

doi:10.16381/j.cnki.issn1003-207x.2022.1292

中国管理科学 ›› 2025, Vol. 33 ›› Issue (8): 123-130.doi: 10.16381/j.cnki.issn1003-207x.2022.1292

• • 上一篇

基于部分抽样检验的在线异常监控方法

韩雄¹, 杨扬¹, 邓晓春¹, 缑建杰¹, 郭捷²(), 张晨³

^1.成都飞机工业（集团）有限责任公司，四川成都 610404
^2.南京航空航天大学经济与管理学院，江苏南京 210016
^3.清华大学工业工程系，北京 100084

收稿日期:2022-06-13 修回日期:2023-05-16 出版日期:2025-08-25 发布日期:2025-09-10
通讯作者: 郭捷 E-mail:guojie1144098@163.com
基金资助:
国家自然科学基金项目(72271138);国家自然科学基金项目(71932006);国家自然科学基金项目(71901131);北京市自然科学基金项目(9222014);航空科学基金项目(2020Z063058001)

Partially Sampling Inspection Process Based Online Change Detection

Xiong Han¹, Yang Yang¹, Xiaochun Deng¹, Jianjie Gou¹, Jie Guo²(), Chen Zhang³

^1.Chengdu Aircraft Industrial （Group） Co. ，Ltd，Chengdu 610404，China
^2.College of Economics and Management，Nanjing University of Aeronautics and Astronautics，Nanjing 210016，China
^3.Department of Industrial Engineering，Tsinghua University，Beijing 100084，China

Received:2022-06-13 Revised:2023-05-16 Online:2025-08-25 Published:2025-09-10
Contact: Jie Guo E-mail:guojie1144098@163.com

摘要/Abstract

摘要：

在生产系统中，管理者可以通过抽样检验的方式对系统中存在的异常进行实时监控。考虑到产品所需要检验的特性（变量）很多，而检验资源通常是有限的，例如人力或检验仪器，因此，在每一个时刻只能选择一部分特性进行检验，导致只能得到部分特性的不合格数。本文假设每个特性的不合格数服从二项分布，提出一个针对二项分布的高维数据流的基于部分抽样的在线异常监控方案。首先，本文将高维数据分解为平滑的正常信号和稀疏的异常信号。为了对变量间的相关性进行建模，正常信号被分解为背景基函数乘以相应的系数。对于稀疏异常参数的估计，假设其服从 spike-slab 分布，分布中的参数通过变分贝叶斯估计得到。然后，构造基于似然比检验的监控统计量对系统异常进行监控。最后，通过将监控统计量作为多臂老虎机问题中的收益函数，本文构造了一种基于 Thompson 采样的变量选择策略，很好地平衡了变量搜索的深度和广度，从而达到最小化异常监控延迟时间的目的。

关键词: 在线学习, 部分观测, 抽样检验, 二项分布, 稀疏异常

Abstract:

In manufacturing systems， practitioners rely on sampling inspection to detect real-time changes within the system. However， due to the large number of categories （variables） that need inspection and the limited availability of inspection resources such as human labor or instruments， only a subset of categories can be inspected at each time point. As a result， only partial observations of the defective numbers for each category can be obtained. To enable a prompt system change detection， it requires not only a powerful change detection scheme that can deal with partially observable data， but also an adaptive variable selection strategy to identify which set of variables to be observed for the next time point such that the change information can be reserved maximally. The challenge of online change detection is addressed for high-dimensional data streams following a binomial distribution， based on a partially sampling inspection process. First， high-dimensional data is decomposed into smooth normal signals and sparse abnormal signals. The normal signals are represented as a linear combination of basis functions multiplied by corresponding coefficients， capturing the correlations between variables. The anomalous parameter is modeled using a spike-slab distribution and variational Bayesian inference is employed to estimate the distribution parameters. Next， a likelihood ratio test is constructed as the detection statistic for detecting system changes. Furthermore， combinatorial multi-armed bandit （CMAB） algorithms are leveraged by treating the test statistics as the reward function. Specifically， a variable selection policy based on Thompson sampling is proposed， enabling the selection of the most anomalous categories for inspection at each time point and minimizing change detection delay. Through experimental evaluations， the results highlight its potential to improve the efficiency and accuracy of defect detection in manufacturing systems while considering the constraints of limited inspection resources.

Key words: online learning, partial observations, sampling inspection, binomial distribution, sparse change

中图分类号:

F402.4

韩雄, 杨扬, 邓晓春, 缑建杰, 郭捷, 张晨. 基于部分抽样检验的在线异常监控方法[J]. 中国管理科学, 2025, 33(8): 123-130.

Xiong Han, Yang Yang, Xiaochun Deng, Jianjie Gou, Jie Guo, Chen Zhang. Partially Sampling Inspection Process Based Online Change Detection[J]. Chinese Journal of Management Science, 2025, 33(8): 123-130.

图/表 6

图1

图2

表1

不同方法的准确度百分数的比较"

$ξ$	BBCD	TRAS	SASAM	NAS	R-SADA
1	92.3	61.9	64.0	91.2	87.8

表1

图3

图4

图5

参考文献 26

[1]	Zhang C， Yan H， Lee S， et al. Weakly correlated profile monitoring based on sparse multi-channel functional principal component analysis［J］. IISE Transactions， 2018， 50（10）：878-891.
[2]	梁海玲，白森，李坚. 基于鲁棒稀疏PCA的工业异常检测［J］. 科学技术与工程， 2022， 22（15）：6164-6171.
	Liang H L， Bai S， Li J. Industrial anomaly detection based on robust sparse PCA［J］. Science Technology and Engineering， 2022， 22（15）： 6164-6171.
[3]	Yan H， Paynabar K， Shi J J. Anomaly detection in images with smooth background via smooth-sparse decomposition［J］.Technometrics， 2017，59（1）：102-114.
[4]	覃凤婷，杨有龙，仇海全.基于稀疏子空间的局部异常值检测算法［J］.计算机工程与应用，2020，56（19）：152-159.
	Qin F T， Yang Y L， Qiu H Q. Sparse subspace-based method for local outlier detection［J］.Computer Engineering and Applications， 2020， 56（19）：152-159.
[5]	Liu K B， Mei Y J， Shi J J. An adaptive sampling strategy for online high-dimensional process monitoring［J］. Technometrics， 2015， 57（3）：305-319.
[6]	Xian X C， Wang A D， Liu K B. A nonparametric adaptive sampling strategy for online monitoring of big data streams［J］. Technometrics， 2018，60（1）：14-25.
[7]	Wang A D， Xian X C， Tsung F， et al. A spatial-adaptive sampling procedure for online monitoring of big data streams［J］. Journal of Quality Technology， 2018， 50（4）：329-343.
[8]	Xian X C， Zhang C， Bonk S， et al. Online monitoring of big data streams： A rank-based sampling algorithm by data augmentation［J］.Journal of Quality Technology， 2021，53（2）：135-153.
[9]	Meier L， van De Geer S， Buhlmann P. High-dimensional additive modeling［J］. The Annals of Statistics， 2009， 37（6B）：3779-3821.
[10]	Anandkumar A， Ge R， Hsu D， et al. Tensor decompositions for learning latent variable models［J］. Journal of Machine Learning Research， 2014，15（1）：2773-2832.
[11]	Yao F， Müller H G， Wang J L. Functional data analysis for sparse longitudinal data［J］. Journal of the American Statistical Association， 2005，100（470）： 577-590.
[12]	Yu K， Wu X D， Ding W， et al. Scalable and accurate online feature selection for big data［J］. ACM Transactions on Knowledge Discovery from Data， 2016， 11（2）：1-39.
[13]	Cai D， He X F， Han J W， et al. Graph regularized nonnegative matrix factorization for data representation［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2011， 33（8）：1548-1560.
[14]	Wood S N， Goude Y， Shaw S. Generalized additive models for large data sets［J］. Journal of the Royal Statistical Society： Series C （Applied Statistics）， 2015， 64（1）：139-155.
[15]	Ba S， Joseph V R. Composite Gaussian process models for emulating expensive functions［J］. The Annals of Applied Statistics， 2012，6（4）：1838-1860.
[16]	Zhang L M， Wang K B， Chen N. Monitoring wafers’ geometric quality using an additive Gaussian process model［J］. IIE Transactions， 2016， 48（1）：1-15.
[17]	Yan H， Paynabar K， Shi J J. Real-time monitoring of high-dimensional functional data streams via spatio-temporal smooth sparse decomposition［J］. Technometrics， 2018， 60（2）：181-197.
[18]	Montgomery D C. Introduction to statistical quality control［M］.Hoboken， NJ， USA： John Wiley & Sons， 2009.
[19]	Mitchell T J， Beauchamp J J. Bayesian variable selection in linear regression［J］. Journal of the American Statistical Association，1988， 83（404）：1023-1032.
[20]	Zhao N， Zhang H Y， Clark J J， et al. Composite kernel machine regression based on likelihood ratio test for joint testing of genetic and gene-environment interaction effect［J］. Biometrics， 2019， 75（2）：625-637.
[21]	Mansouri M， Hajji M， Trabelsi M， et al. An effective statistical fault detection technique for grid connected photovoltaic systems based on an improved generalized likelihood ratio test［J］. Energy， 2018， 159：842-856.
[22]	Vincent F， Besson O. One-step generalized likelihood ratio test for subpixel target detection in hyperspectral imaging［J］. IEEE Transactions on Geoscience and Remote Sensing， 2020， 58（6）：4479-4489.
[23]	Bubeck S， Wang T Y， Viswanathan N. Multiple identifications in multi-armed bandits［C］// Proceedings of International Conference on Machine Learning， Atlanta， USA， June 16-21，ACM 2013： 258-285.
[24]	Zhuang H L， Wang C， Wang Y F. Identifying outlier arms in multiarmed bandit［C］//Proceedings of Advances in Neural Information Processing Systems， Long Beach， USA， December 4-9 ，The MIT Press， 2017：30.
[25]	Durand A， Gagne C. Thompson sampling for combinatorial bandits and its application to online feature selection［C］// Proceedings of Workshops at the Twenty-Eighth AAAI Conference on Artificial Intelligence， Québec City， Canada， July 27-31 ，AAAI， 2014：181.
[26]	Kaufmann E， Korda N， Munos R. Thompson sampling： An asymptotically optimal finite-time analysis［C］// Proceedings of International Conference on Algorithmic Learning Theory， Lyon， France， October 29-31 ，PMLR， 2012： 199-213.

基于部分抽样检验的在线异常监控方法

Partially Sampling Inspection Process Based Online Change Detection

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 6

参考文献 26

相关文章 3

Metrics

本文评价

推荐阅读 0

[1]	房超,胡雅静,郑维博,冯耕中. 基于在线学习的收益信息不确定下新产品开发项目组合动态选择策略[J]. 中国管理科学, 2024, 32(6): 151-162.
[2]	张永, 龙婉容, 杨兴雨, 张卫国. 基于在线算法的改进指数梯度投资组合策略[J]. 中国管理科学, 2022, 30(9): 49-60.
[3]	张斌, 华中生. 供应链质量管理中抽样检验决策的非合作博弈分析[J]. 中国管理科学, 2006, (3): 27-31.