Article(id=1325495688502575310, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325495515563041259, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202550364, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1748534400000, receivedDateStr=2025-05-30, revisedDate=1760976000000, revisedDateStr=2025-10-21, acceptedDate=1761840000000, acceptedDateStr=2025-10-31, onlineDate=1793952398657, onlineDateStr=2026-11-06, pubDate=1772294400000, pubDateStr=2026-03-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1793952398657, onlineIssueDateStr=2026-11-06, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1793952398657, creator=13701087609, updateTime=1793952398657, updator=13701087609, issue=Issue{id=1325495515563041259, tenantId=1146029695717560320, journalId=1324759914146529478, year='2026', volume='63', issue='3', pageStart='539', pageEnd='823', issueExtLink='null', onlineDate='null', pubDate='1772294400000', pubDateStr='2026-03-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1793952357425, creator='13701087609', updateTime=1793952549954, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1325496323193057998, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325495515563041259, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1325496323193057999, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325495515563041259, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=615, endPage=626, ext={EN=ArticleExt(id=1325495688724873423, articleId=1325495688502575310, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Time-Series Anomaly Detection Method Based on Time-Frequency Enhanced Contrastive Learning, columnId=1325495518645859037, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on Data Storage and Processing for New Quality Productive Forces, runingTitle=null, highlight=null, articleAbstract=

The rapid development of emerging industries such as artificial intelligence, big data, and cloud computing has led to the widespread application of large-scale cloud computing centers, which continuously generate massive amounts of operational data during operation. Efficient and accurate anomaly detection is a key part in ensuring intelligent operation and maintenance as well as the system reliability. Due to factors such as equipment aging, sudden environmental changes, and dynamic loads, operational data often contain problems such as data noise pollution, abnormal signal coupling, and dynamic drift in equipment status. Current mainstream anomaly detection methods typically rely on assumptions of data purity, identifying anomalies by learning normal data features and evaluating reconstruction errors. However, industrial-grade intelligent systems face inherent challenges such as sensor noise contamination, dynamic state drift, and coupled anomaly signals due to their complex topological structures and heterogeneous operating environments, all of which degrade data quality and compromise detection efficacy. To address these issues, we propose a time-series anomaly detection method based on time-frequency enhanced contrastive learning(TF-ECL). It introduces a dual-view enhancement mechanism to implement controllable augmentations in temporal and frequency domains, thereby improving model noise immunity based on cross-view contrastive learning. Meanwhile, based on the differential representations of abnormal signals in time-frequency space, a dual-domain collaborative detection mechanism is developed to achieve abnormal localization using time-frequency contrastive errors. Experimental results from multiple real-time data sets demonstrate that compared with previous multidimensional time-series anomaly detection methods, TF-ECL achieves robust and excellent performance, with metrics such as F1 score, VUS-PR, and VUS-ROC surpassing or matching those of current state-of-the-art (SOTA) methods.

, authors=Lun Ge1, 2, Yijie Wang1, 2, *, Ying Zhang3, Xiaohui Zhou1, 2, authorsList=Lun Ge, Yijie Wang, Ying Zhang, Xiaohui Zhou, authorCompany=null, correspAuthors=Yijie Wang, authorNote=

born in 1986. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

born in 1971. PhD, professor, PhD supervisor. Her main research interests include big data analysis, cloud computing, and distributed storage

born in 1983. PhD candidate. Her main research interests include anomaly detection, data mining, and space trajectory intent recognition

born in 2000. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1325495692009013472, articleId=1325495688502575310, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=基于时频增强的对比学习时序异常检测方法, columnId=1325495520386495199, journalTitle=计算机研究与发展, columnName=面向新质生产力的数据存储与处理专题, runingTitle=null, highlight=null, articleAbstract=

人工智能、大数据、云计算等新兴产业的飞速发展,催生了大规模云计算中心的广泛应用。这些云计算中心在运行过程中持续产生海量级运维数据,高效、准确的异常检测是保障智能运维和系统可靠性的关键环节。由于设备老化、环境突变、动态负载等因素,运维数据往往包含数据噪声污染、异常信号耦合、设备状态动态漂移等问题。当前主流的异常检测方法通常依赖数据纯净性假设,通过对正常数据的特征学习并评估重构误差来实现异常判别,而数据质量的降低会严重影响异常检测效果。针对上述问题,提出基于时频增强的对比学习时序异常检测方法(TF-ECL),通过构建时频双视图增强机制,在时域和频域实施可控增强,基于时频多视图对比学习的方法提升了模型抗噪能力。同时,利用异常信号在时频空间的表现差异性,构建双域协同检测机制,利用时频跨域对比误差进行异常定位。在多个真实时序数据集的实验结果表明,相较于之前的多维时间序列异常检测方法,TF-ECL的F1分数,VUS-PR,VUS-ROC等指标都优于或达到现有的SOTA方法水平,表现出稳健而优异的性能。

, authors=葛伦1, 2, 王意洁1, 2, *, 张莹3, 周小晖1, 2, authorsList=葛伦, 王意洁, 张莹, 周小晖, authorCompany=null, correspAuthors=王意洁, authorNote=

1986年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

1971年生。博士,教授,博士生导师。主要研究方向为大数据分析、云计算、分布式存储

1983年生。博士研究生。主要研究方向为异常检测、数据挖掘、空间轨道意图识别

2000 年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

, correspAuthorsNote=
王意洁()
, copyrightStatement=版权所有 © 2026 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=BJPmlT3p63CmMWc7woOvdw==, magXml=s8riT+Xwuhz6SOVlZW43Ww==, pdfUrl=null, pdf=qMvNq7zpfhIC+Q4rf0EXRQ==, pdfFileSize=1567163, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=3ZRq4eGGuaLCGdy5pw3m/A==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=IgdiOJmZtt+bTLBv9H8Ocg==, mapNumber=null, fund=null)}, authors=[Author(id=1325495692592021739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=gelun8686@163.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495692713656558, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692592021739, language=EN, stringName=Lun Ge, firstName=Lun, middleName=null, lastName=Ge, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"

born in 1986. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

"}, bioImg=null, bioContent=

born in 1986. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495692789154031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692592021739, language=CN, stringName=葛伦, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
2国防科技大学计算机学院 长沙 410073, bio={"img":"L6f9pEzlo5eWmgA6mVB3Gg==","content":"

1986年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

"}, bioImg=L6f9pEzlo5eWmgA6mVB3Gg==, bioContent=

1986年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)])]), Author(id=1325495692877234417, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wangyijie@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1325495692961120500, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692877234417, language=EN, stringName=Yijie Wang, firstName=Yijie, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"

born in 1971. PhD, professor, PhD supervisor. Her main research interests include big data analysis, cloud computing, and distributed storage

"}, bioImg=null, bioContent=

born in 1971. PhD, professor, PhD supervisor. Her main research interests include big data analysis, cloud computing, and distributed storage

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495693061783797, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692877234417, language=CN, stringName=王意洁, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
2国防科技大学计算机学院 长沙 410073, bio={"img":"NI2rBT+0Q7FSe8ezGbG4Ag==","content":"

1971年生。博士,教授,博士生导师。主要研究方向为大数据分析、云计算、分布式存储

"}, bioImg=NI2rBT+0Q7FSe8ezGbG4Ag==, bioContent=

1971年生。博士,教授,博士生导师。主要研究方向为大数据分析、云计算、分布式存储

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)])]), Author(id=1325495693141475575, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495693229555961, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693141475575, language=EN, stringName=Ying Zhang, firstName=Ying, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3School of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310058, bio={"content":"

born in 1983. PhD candidate. Her main research interests include anomaly detection, data mining, and space trajectory intent recognition

"}, bioImg=null, bioContent=

born in 1983. PhD candidate. Her main research interests include anomaly detection, data mining, and space trajectory intent recognition

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495693309247738, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693141475575, language=CN, stringName=张莹, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3浙江大学航空航天学院 杭州 310058, bio={"img":"zlFnA2t9nNrng9NPgR7J+g==","content":"

1983年生。博士研究生。主要研究方向为异常检测、数据挖掘、空间轨道意图识别

"}, bioImg=zlFnA2t9nNrng9NPgR7J+g==, bioContent=

1983年生。博士研究生。主要研究方向为异常检测、数据挖掘、空间轨道意图识别

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692478775527, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=3, ext=[AuthorCompanyExt(id=1325495692487164136, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692478775527, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3School of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310058), AuthorCompanyExt(id=1325495692495552745, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692478775527, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浙江大学航空航天学院 杭州 310058)])]), Author(id=1325495693409911036, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495695112798464, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693409911036, language=EN, stringName=Xiaohui Zhou, firstName=Xiaohui, middleName=null, lastName=Zhou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"

born in 2000. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

"}, bioImg=null, bioContent=

born in 2000. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495695209267457, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693409911036, language=CN, stringName=周小晖, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
2国防科技大学计算机学院 长沙 410073, bio={"img":"VzRvj3Pt5kn3I0kiDxo7PA==","content":"

2000 年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

"}, bioImg=VzRvj3Pt5kn3I0kiDxo7PA==, bioContent=

2000 年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)])])], keywords=[Keyword(id=1325495695322513666, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, orderNo=1, keyword=multivariate time series), Keyword(id=1325495695393816835, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, orderNo=2, keyword=time-series anomaly detection), Keyword(id=1325495695465120004, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, orderNo=3, keyword=contrastive learning), Keyword(id=1325495695553200389, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, orderNo=4, keyword=time-domain augmentation), Keyword(id=1325495695649669382, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, orderNo=5, keyword=frequency-domain augmentation), Keyword(id=1325495695725166855, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, orderNo=1, keyword=多维时间序列), Keyword(id=1325495695800664328, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, orderNo=2, keyword=时序异常检测), Keyword(id=1325495695867773193, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, orderNo=3, keyword=对比学习), Keyword(id=1325495695934882059, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, orderNo=4, keyword=时域增强), Keyword(id=1325495696048128268, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, orderNo=5, keyword=频域增强)], refs=[Reference(id=1325495699613286685, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Pang Guansong, Shen Chunhua, Cao Longbing, et al. Deep learning for anomaly detection: A review[J]. ACM Computing Surveys, 2021, 54(2): 1−38, articleTitle=null, refAbstract=null), Reference(id=1325495699722338590, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Kim D, Park S, Choo J, et al. When model meets new normals: Test-time adaptation for unsupervised time-series anomaly detection[C]//Proc of the 38th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2024: 406−413, articleTitle=null, refAbstract=null), Reference(id=1325495699793641759, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Zexin, Pei Changhua, Ma Minghua, et al. Revisiting VAE for unsupervised time series anomaly detection: A frequency perspective [C]//Proc of the 24th ACM Web Conf. New York: ACM, 2024: 3009−3017, articleTitle=null, refAbstract=null), Reference(id=1325495699869139232, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Iglesias G, Talavera E. Data augmentation techniques in time series domain: A survey and taxonomy[J]. Neural Computing and Applications, 2023, 35(14): 1123−1145, articleTitle=null, refAbstract=null), Reference(id=1325495699940442401, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Tian, Ma Ziqing, Wen Qingsong, et al. FEDformer: Frequency enhanced decomposed Transformer for long-term series forecasting[C]//Proc of the 39th Int Conf on Machine Learning. New York: ACM, 2022: 78−85, articleTitle=null, refAbstract=null), Reference(id=1325495700011745570, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Su Ya, Zhao Youjian, Niu Chenhao, et al. Robust anomaly detection for multivariate time series through stochastic recurrent neural network[C]//Proc of the 25th ACM SIGKDD Int Conf on Knowledge Discovery and Data Mining. New York: ACM, 2019: 2828−2837, articleTitle=null, refAbstract=null), Reference(id=1325495700087243043, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Jiehui, Wu Haixu, Wang Jianmin, et al. Anomaly Transformer: Time series anomaly detection with association discrepancy[C/OL]//Proc of the 10th Int Conf on Learning Representations. 2022[2025-06-01]. https://openreview.net /pdf?id=LzQQ89U1qm_, articleTitle=null, refAbstract=null), Reference(id=1325495700154351908, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Haixu, Hu Tengge, Liu Yong, et al. TimesNet: Temporal 2D-variation modeling for general time series analysis[C/OL]//Proc of the 11th Int Conf on Learning Representations. 2023[2025-06-01]. https://openreview.net/pdf?id= ju_Uqw384Oq, articleTitle=null, refAbstract=null), Reference(id=1325495700238237989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Yutong, Xu Hongzuo, Pang Guangsong, et al. Self-supervised spatial-temporal normality learning for time series anomaly detection[C]//Proc of the 2024 European Conf on Machine Learning and Principles and Practice of Knowledge Discovery in Databases. Berlin: Springer, 2024: 183−195, articleTitle=null, refAbstract=null), Reference(id=1325495700313735462, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=周小晖, 王意洁, 徐鸿祚, 等. 基于融合学习的无监督多维时间序列异常检测[J]. 计算机研究与发展, 2023, 60(3): 496−508, articleTitle=null, refAbstract=null), Reference(id=1325495700380844327, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Xiaohui, Wang Yijie, Xu Hongzuo, et al. Unsupervised multivariate time series anomaly detection via fusion learning[J]. Journal of Computer Research and Development, 2023, 60(3): 496−508 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495700460536104, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=袁子淇, 孙庆赟, 周号益, 等. MNDetector: 基于多层网络的异常访问检测方法[J]. 计算机研究与发展, 2025, 62(3): 765−778, articleTitle=null, refAbstract=null), Reference(id=1325495700540227881, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Yuan Ziqi, Sun Qingyun, Zhou Haoyi, et al. MNDetector: Anomaly access detection method based on multiplex network[J]. Journal of Computer Research and Development, 2025, 62(3): 765−778 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495700615725354, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Ruff L, Vandermeulen R, Goernitz N, et al. Deep one class classification[C]//Proc of the 35th Int Conf on Machine Learning. New York: ACM, 2018: 4393−4402, articleTitle=null, refAbstract=null), Reference(id=1325495700691222827, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Roy A, Sh Juan, Li Jia. GAD-NR: Graph anomaly detection via neighborhood reconstruction[C]//Proc of the 17th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2024: 576−585, articleTitle=null, refAbstract=null), Reference(id=1325495700754137388, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Xiao, Zhao Xuejiao, Shen Zhiqi. A generalizable anomaly detection method in dynamic graphs[C]//Proc of the 2025 AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025: 481−495, articleTitle=null, refAbstract=null), Reference(id=1325495700829634861, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=吴彦文, 谭溪晨, 葛迪, 等. 结合重构和图预测的多元时序异常检测框架[J]. 计算机工程与应用, 2024 , 60 (13): 301−310, articleTitle=null, refAbstract=null), Reference(id=1325495700900938030, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Yanwen, Tan Xichen, Ge Di, et al. Combining reconstruction and graph prediction for multi-dimensional time series anomaly detection framework[J]. Computer Engineering and Applications, 2024, 60 (13): 301−310 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495700997407023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Chen, He Shibo, Zhou Qihang, et al. Large language model guided knowledge distillation for time series anomaly detection[C]//Proc of the 33rd Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2024: 451−465, articleTitle=null, refAbstract=null), Reference(id=1325495701081293104, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Jeong Y, Yang E, Ryu J H, et al. AnomalyBERT: Self-supervised transformer for time series anomaly detection using data degradation scheme[C/OL]//Proc of the 11th Int Conf on Learning Representations. 2023[2025-06-01]. https://arxiv.org/pdf/2305.04468v1.pdf, articleTitle=null, refAbstract=null), Reference(id=1325495701160984881, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Jiao Yang, Yang Kai, Song Dongjing, et al. TimeAutoAD: Autonomous anomaly detection with self-supervised contrastive loss for multivariate time series[J]. IEEE Transactions on Network Science and Engineering, 2022, 9(3): 1604−1619, articleTitle=null, refAbstract=null), Reference(id=1325495701228093746, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Kim H, Kim S, Min S, et al. Contrastive time-series anomaly detection[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(12): 1−14, articleTitle=null, refAbstract=null), Reference(id=1325495701299396915, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Darban Z Z, Yang Yiyuan, Webb C I, et al. DACAD: Domain adaptation contrastive learning for anomaly detection in multivariate time series[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 37(8): 4485−4496, articleTitle=null, refAbstract=null), Reference(id=1325495701383282996, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Yiyuan, Zhang Chaoli, Zhou Tian, et al. DCdetector: Dual attention contrastive representation learning for time series anomaly detection[C]//Proc of the 29th ACM SIGKDD Conf on Knowledge Discovery and Data Mining. New York: ACM, 2023: 571−586, articleTitle=null, refAbstract=null), Reference(id=1325495701462974773, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhong Zhijie, Yu Zhiwen, Yang Yiyuan, et al. PatchAD: A lightweight patch-based MLP mixer for time series anomaly detection[J]. IEEE Transactions on Big Data, 2025, 1(15): 234−245, articleTitle=null, refAbstract=null), Reference(id=1325495701538472246, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Xie Yongzheng, Zhang Hongyu, Babar M A , et al. Multivariate time series anomaly detection by capturing coarse grained intra and inter variate dependencies[C]//Proc of the 2025 World Wide Web Conf. New York: ACM, 2025: 697−705, articleTitle=null, refAbstract=null), Reference(id=1325495701613969719, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Fang Yuchen, Xie Jiadong, Chen Lu, et al. Temporal-frequency masked autoencoders for time series anomaly detection[C]//Proc of the 40th IEEE Int Conf on Data Engineering. Piscataway, NJ: IEEE, 2024: 158−172, articleTitle=null, refAbstract=null), Reference(id=1325495701685272888, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Kim T, Kim J, Tae Y, et al. Reversible instance normalization for accurate time-series forecasting against distribution shift[C/OL]//Proc of the 10th Int Conf on Learning Representations. 2022[2025-06-01]. https://openreview.net/ forum?id =cGDAkQo1C0p, articleTitle=null, refAbstract=null), Reference(id=1325495701760770361, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]//Proc of the 31st Int Conf on Neural Information Processing Systems. Cambridge, MA: MIT, 2017: 5998−6008, articleTitle=null, refAbstract=null), Reference(id=1325495701848850746, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Hongzuo, Wang Yijie, Pang Guangsong, et al. RoSAS: Deep semi-supervised anomaly detection with contamination-resilient continuous supervision[J]. Information Processing and Management, 2023, 60(5): 1059−1072, articleTitle=null, refAbstract=null), Reference(id=1325495701932736827, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Xiang, Zhao Ziyuan, Tsiligkaridis T, et al. Self-supervised contrastive pre-training for time series via time-frequency consistency[C]//Proc of the 36th Conf on Neural Information Processing Systems. Cambridge, MA: MIT, 2022: 431−446, articleTitle=null, refAbstract=null), Reference(id=1325495702004039996, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Peng Furong, Luo Jiachen, Lu Xuan, et al. Cross-domain contrastive learning for time series clustering[C]//Proc of the 38th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2024: 8921−8929, articleTitle=null, refAbstract=null), Reference(id=1325495702092120381, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Xinlei, He Kaiming. Exploring simple siamese representation learning[C]//Proc of the 2021 IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 15745−15757, articleTitle=null, refAbstract=null), Reference(id=1325495702192783678, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Yufeng, Pan Jialu, Liu Wanwei, et al. Kullback-Leibler divergence-based out of distribution detection with flow based generative models[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 36(4): 1683−1697, articleTitle=null, refAbstract=null), Reference(id=1325495702280864063, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Sarfraz M, Chen Meiyen, Layer L, et al. Position: Quo vadis, unsupervised time series anomaly detection?[C]//Proc of the 41st Int Conf on Machine Learning. New York: ACM, 2024: 187−196, articleTitle=null, refAbstract=null), Reference(id=1325495702381527360, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=John P, Paul B, Themis P, et al. Volume under the surface: A new accuracy evaluation measure for time-series anomaly detection[J]. Proceedings of the VLDB Endowment, 2022, 15(11): 2774−2787, articleTitle=null, refAbstract=null), Reference(id=1325495704076026177, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Zong Bo, Song Qi, Min M R, et al. Deep autoencoding gaussian mixture model for unsupervised anomaly detection[C/OL]//Proc of the 6th Int Conf on Learning Representations. 2018[2025-06-01]. https://openreview.net/ forum? id=BJJLHbb0-, articleTitle=null, refAbstract=null), Reference(id=1325495704155717954, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Tuli S, Casale G, Jennings N R, et al. TranAD: Deep transformer networks for anomaly detection in multivariate time series data[C]//Proc of the 48th Int Conf on Very Large Data Bases. New York: ACM, 2022: 338−351, articleTitle=null, refAbstract=null), Reference(id=1325495704264769859, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou Bin, Liu Shenghua, Hooi B, et al. BeatGAN: Anomalous rhythm detection using adversarially generated time series[C]//Proc of the 28th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2019: 4433−4439, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)]), AuthorCompany(id=1325495692478775527, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=3, ext=[AuthorCompanyExt(id=1325495692487164136, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692478775527, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3School of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310058), AuthorCompanyExt(id=1325495692495552745, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692478775527, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浙江大学航空航天学院 杭州 310058)])], figs=[ArticleFig(id=1325495696207511821, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Figure 1, caption=Performance of point and contextual anomalies in time and frequency domains, figureFileSmall=7i+/u6eYdPD7Mz1Ce13aeg==, figureFileBig=3ZRq4eGGuaLCGdy5pw3m/A==, tableContent=null), ArticleFig(id=1325495696291397902, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=图1, caption=点异常和上下文异常在时域和频域的表现, figureFileSmall=7i+/u6eYdPD7Mz1Ce13aeg==, figureFileBig=3ZRq4eGGuaLCGdy5pw3m/A==, tableContent=null), ArticleFig(id=1325495696522084623, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Figure 2, caption=Overall architecture of time-series anomaly detection method based on time-frequency enhanced contrastive learning, figureFileSmall=Q2v4VyDFUagw5Po9B3RhjA==, figureFileBig=Mhz9h/OUYib1SWDbuBYpaQ==, tableContent=null), ArticleFig(id=1325495696605970704, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=图2, caption=基于时频增强的对比学习时序异常检测方法的整体框架, figureFileSmall=Q2v4VyDFUagw5Po9B3RhjA==, figureFileBig=Mhz9h/OUYib1SWDbuBYpaQ==, tableContent=null), ArticleFig(id=1325495696735994129, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Figure 3, caption=Analysis results of time-frequency enhancement intensity, figureFileSmall=qwmgDHtLthnranl5vvUI8g==, figureFileBig=DlPaBdC2yGfQkzo5U2ZicA==, tableContent=null), ArticleFig(id=1325495696891183378, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=图3, caption=时频增强强度分析结果, figureFileSmall=qwmgDHtLthnranl5vvUI8g==, figureFileBig=DlPaBdC2yGfQkzo5U2ZicA==, tableContent=null), ArticleFig(id=1325495696979263763, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Figure 4, caption=Comprehensive efficiency analysis results, figureFileSmall=TTLNlAJ+uqYYB/In9Is0wQ==, figureFileBig=WOiKvGsPkcE0LcQTkqoQ7Q==, tableContent=null), ArticleFig(id=1325495697071538452, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=图4, caption=综合效能分析结果, figureFileSmall=TTLNlAJ+uqYYB/In9Is0wQ==, figureFileBig=WOiKvGsPkcE0LcQTkqoQ7Q==, tableContent=null), ArticleFig(id=1325495697134453013, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Table 1, caption=Statistical Information of the Datasets, figureFileSmall=null, figureFileBig=null, tableContent=
数据集维度样本数异常率/%
训练集验证集测试集
MSL5546653116647372910.5
SMAP251081462703742761712.8
SWAT513960009900044991912.1
PSM25105984264978784127.8
), ArticleFig(id=1325495697226727702, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=表1, caption=

数据集统计信息

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集维度样本数异常率/%
训练集验证集测试集
MSL5546653116647372910.5
SMAP251081462703742761712.8
SWAT513960009900044991912.1
PSM25105984264978784127.8
), ArticleFig(id=1325495697306419479, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Table 2, caption=Anomaly Detection Results Comparison on Different Datasets %, figureFileSmall=null, figureFileBig=null, tableContent=
方法MSLSMAP
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
注: 黑体数值为指标最优值,下划线数字为指标次优值。
DAGMM[34]90.0586.9388.4669.4266.8287.5856.3168.5553.4281.37
DSVDD[12]83.4092.6987.8078.1693.7587.2560.0372.5160.7486.71
BeatGAN[36]91.4485.4288.3374.1582.7993.9155.4169.7087.8872.12
OmniAnomaly[6]90.8286.4788.5975.6876.8492.3456.1869.8584.5863.53
TranAD[35]90.7294.7392.6880.6583.3893.1271.3380.7882.8284.58
TimesNet[8]87.3285.4286.3681.4192.0686.4765.4874.5389.0179.60
DCdetector[21]92.0894.4493.2485.6595.3693.1898.8495.9390.9097.20
STEN[9]84.0797.0890.1184.4995.3995.8496.6396.6694.1498.16
TFMAE[24]92.0096.2994.0985.1594.4694.9198.9296.8892.4597.75
MtsCID[23]92.2996.3994.5487.6989.6994.9899.4497.1692.4095.57
TF-ECL(本文)96.2596.4294.8187.8796.2196.6198.9797.5894.2998.25
方法SWATPSM
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
DAGMM[34]90.0586.9388.4661.1562.2292.0790.0991.0771.1472.21
DSVDD[12]90.1485.6890.8259.7974.7592.9890.2791.6083.5190.30
BeatGAN[36]91.4485.4288.3352.1768.7296.5890.1693.2678.4284.76
OmniAnomaly[6]90.8286.4788.5966.2564.4095.7190.0992.8278.7178.91
TranAD[35]90.7294.7392.6872.3387.6297.4498.1997.9281.9286.92
TimesNet[8]87.3285.4286.3678.7795.3397.6498.2297.9392.6795.24
DCdetector[21]92.0894.4493.2497.2798.7497.1498.7497.9492.1593.14
STEN[9]84.0796.0890.1191.7897.7498.6797.6998.1894.8896.03
TFMAE[24]92.0096.2994.0997.2699.0398.4398.1298.5391.9092.68
MtsCID[23]93.7998.9896.7989.1297.7399.5797.3298.4490.4190.24
TF-ECL(本文)96.4410098.1897.3599.2398.8198.3198.5592.8796.17
), ArticleFig(id=1325495697407082776, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=表2, caption=

不同数据集的异常检测结果对比

, figureFileSmall=null, figureFileBig=null, tableContent=
方法MSLSMAP
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
注: 黑体数值为指标最优值,下划线数字为指标次优值。
DAGMM[34]90.0586.9388.4669.4266.8287.5856.3168.5553.4281.37
DSVDD[12]83.4092.6987.8078.1693.7587.2560.0372.5160.7486.71
BeatGAN[36]91.4485.4288.3374.1582.7993.9155.4169.7087.8872.12
OmniAnomaly[6]90.8286.4788.5975.6876.8492.3456.1869.8584.5863.53
TranAD[35]90.7294.7392.6880.6583.3893.1271.3380.7882.8284.58
TimesNet[8]87.3285.4286.3681.4192.0686.4765.4874.5389.0179.60
DCdetector[21]92.0894.4493.2485.6595.3693.1898.8495.9390.9097.20
STEN[9]84.0797.0890.1184.4995.3995.8496.6396.6694.1498.16
TFMAE[24]92.0096.2994.0985.1594.4694.9198.9296.8892.4597.75
MtsCID[23]92.2996.3994.5487.6989.6994.9899.4497.1692.4095.57
TF-ECL(本文)96.2596.4294.8187.8796.2196.6198.9797.5894.2998.25
方法SWATPSM
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
DAGMM[34]90.0586.9388.4661.1562.2292.0790.0991.0771.1472.21
DSVDD[12]90.1485.6890.8259.7974.7592.9890.2791.6083.5190.30
BeatGAN[36]91.4485.4288.3352.1768.7296.5890.1693.2678.4284.76
OmniAnomaly[6]90.8286.4788.5966.2564.4095.7190.0992.8278.7178.91
TranAD[35]90.7294.7392.6872.3387.6297.4498.1997.9281.9286.92
TimesNet[8]87.3285.4286.3678.7795.3397.6498.2297.9392.6795.24
DCdetector[21]92.0894.4493.2497.2798.7497.1498.7497.9492.1593.14
STEN[9]84.0796.0890.1191.7897.7498.6797.6998.1894.8896.03
TFMAE[24]92.0096.2994.0997.2699.0398.4398.1298.5391.9092.68
MtsCID[23]93.7998.9896.7989.1297.7399.5797.3298.4490.4190.24
TF-ECL(本文)96.4410098.1897.3599.2398.8198.3198.5592.8796.17
), ArticleFig(id=1325495697541300505, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Table 3, caption=Average Experimental Results of Different Anomaly Detection Methods %, figureFileSmall=null, figureFileBig=null, tableContent=
指标平均值DAGMMDSVDDBeatGANOmniAnomalyTranADTimesNetDCdetectorSTENTFMAEMtsCIDTF-ECL(本文)
注: 黑体数值为指标最优值,下划线数字为指标次优值。
$ \overline P $89.9488.4493.3492.4293.0189.6993.6290.6694.3495.1697.03
$ \overline R $80.0682.1779.1179.8189.7483.6496.6296.8797.4198.0398.42
$ \overline F 1 $84.1485.6884.9184.9691.0286.3195.0993.7695.9196.7397.28
$ \overline {VUS {\text{-}}PR} $63.7870.5573.1676.3079.4385.4691.4991.3291.6989.9193.11
$ \overline {VUS {\text{-}} ROC} $70.6686.3877.1170.9285.6290.5696.1196.8395.9893.3197.46
), ArticleFig(id=1325495697641963802, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=表3, caption=

不同异常检测方法的平均实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
指标平均值DAGMMDSVDDBeatGANOmniAnomalyTranADTimesNetDCdetectorSTENTFMAEMtsCIDTF-ECL(本文)
注: 黑体数值为指标最优值,下划线数字为指标次优值。
$ \overline P $89.9488.4493.3492.4293.0189.6993.6290.6694.3495.1697.03
$ \overline R $80.0682.1779.1179.8189.7483.6496.6296.8797.4198.0398.42
$ \overline F 1 $84.1485.6884.9184.9691.0286.3195.0993.7695.9196.7397.28
$ \overline {VUS {\text{-}}PR} $63.7870.5573.1676.3079.4385.4691.4991.3291.6989.9193.11
$ \overline {VUS {\text{-}} ROC} $70.6686.3877.1170.9285.6290.5696.1196.8395.9893.3197.46
), ArticleFig(id=1325495697725849883, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=EN, label=Table 4, caption=Ablation Study on Different Datasets %, figureFileSmall=null, figureFileBig=null, tableContent=
方法MSLSMAP
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
注: 黑体数值为指标最优值,下划线数字为指标次优值。
TF-ECL(本文)96.2596.4294.8187.8796.2196.6198.9797.5894.2998.25
TF-ECLw/o NM87.1591.3789.2183.8695.6496.1496.8296.4884.0298.21
TF-ECLw/o FE90.3591.6591.0185.6395.7781.0198.5388.9187.9097.34
TF-ECLw/o TE91.5895.6893.5986.6095.9496.1596.5796.5692.6698.23
方法SWATPSM
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
TF-ECL(本文)96.4410098.1897.3599.2398.8198.3198.5592.8796.17
TF-ECLw/o NM93.8898.1895.9894.8499.0697.0796.6796.8792.8296.06
TF-ECLw/o FE93.6799.9996.7397.0999.2193.2795.5094.3789.0094.43
TF-ECLw/o TE94.0398.8996.4495.9999.1396.7694.6496.1390.4494.99
), ArticleFig(id=1325495697864261916, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, language=CN, label=表4, caption=

不同数据集上的消融实验

, figureFileSmall=null, figureFileBig=null, tableContent=
方法MSLSMAP
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
注: 黑体数值为指标最优值,下划线数字为指标次优值。
TF-ECL(本文)96.2596.4294.8187.8796.2196.6198.9797.5894.2998.25
TF-ECLw/o NM87.1591.3789.2183.8695.6496.1496.8296.4884.0298.21
TF-ECLw/o FE90.3591.6591.0185.6395.7781.0198.5388.9187.9097.34
TF-ECLw/o TE91.5895.6893.5986.6095.9496.1596.5796.5692.6698.23
方法SWATPSM
PRF1VUS-PRVUS-ROCPRF1VUS-PRVUS-ROC
TF-ECL(本文)96.4410098.1897.3599.2398.8198.3198.5592.8796.17
TF-ECLw/o NM93.8898.1895.9894.8499.0697.0796.6796.8792.8296.06
TF-ECLw/o FE93.6799.9996.7397.0999.2193.2795.5094.3789.0094.43
TF-ECLw/o TE94.0398.8996.4495.9999.1396.7694.6496.1390.4494.99
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550364, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202550364, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202550364, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202550364, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1772294400000, fullTextJson=null, articleText=null, reference=null)
收藏切换
基于时频增强的对比学习时序异常检测方法
葛伦 1, 2 , 王意洁 1, 2, * , 张莹 3 , 周小晖 1, 2
计算机研究与发展 | 面向新质生产力的数据存储与处理专题 2026,63(3): 615-626
收起
收藏切换
计算机研究与发展 |面向新质生产力的数据存储与处理专题 2026 , 63 (3) : 615 -626
基于时频增强的对比学习时序异常检测方法
全屏
[Author(id=1325495692592021739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=gelun8686@163.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495692713656558, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692592021739, language=EN, stringName=Lun Ge, firstName=Lun, middleName=null, lastName=Ge, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"

born in 1986. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

"}, bioImg=null, bioContent=

born in 1986. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495692789154031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692592021739, language=CN, stringName=葛伦, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
2国防科技大学计算机学院 长沙 410073, bio={"img":"L6f9pEzlo5eWmgA6mVB3Gg==","content":"

1986年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

"}, bioImg=L6f9pEzlo5eWmgA6mVB3Gg==, bioContent=

1986年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)])]), Author(id=1325495692877234417, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=wangyijie@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1325495692961120500, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692877234417, language=EN, stringName=Yijie Wang, firstName=Yijie, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"

born in 1971. PhD, professor, PhD supervisor. Her main research interests include big data analysis, cloud computing, and distributed storage

"}, bioImg=null, bioContent=

born in 1971. PhD, professor, PhD supervisor. Her main research interests include big data analysis, cloud computing, and distributed storage

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495693061783797, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495692877234417, language=CN, stringName=王意洁, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
2国防科技大学计算机学院 长沙 410073, bio={"img":"NI2rBT+0Q7FSe8ezGbG4Ag==","content":"

1971年生。博士,教授,博士生导师。主要研究方向为大数据分析、云计算、分布式存储

"}, bioImg=NI2rBT+0Q7FSe8ezGbG4Ag==, bioContent=

1971年生。博士,教授,博士生导师。主要研究方向为大数据分析、云计算、分布式存储

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)])]), Author(id=1325495693141475575, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495693229555961, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693141475575, language=EN, stringName=Ying Zhang, firstName=Ying, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3School of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310058, bio={"content":"

born in 1983. PhD candidate. Her main research interests include anomaly detection, data mining, and space trajectory intent recognition

"}, bioImg=null, bioContent=

born in 1983. PhD candidate. Her main research interests include anomaly detection, data mining, and space trajectory intent recognition

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495693309247738, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693141475575, language=CN, stringName=张莹, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3浙江大学航空航天学院 杭州 310058, bio={"img":"zlFnA2t9nNrng9NPgR7J+g==","content":"

1983年生。博士研究生。主要研究方向为异常检测、数据挖掘、空间轨道意图识别

"}, bioImg=zlFnA2t9nNrng9NPgR7J+g==, bioContent=

1983年生。博士研究生。主要研究方向为异常检测、数据挖掘、空间轨道意图识别

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692478775527, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=3, ext=[AuthorCompanyExt(id=1325495692487164136, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692478775527, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3School of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310058), AuthorCompanyExt(id=1325495692495552745, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692478775527, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浙江大学航空航天学院 杭州 310058)])]), Author(id=1325495693409911036, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495695112798464, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693409911036, language=EN, stringName=Xiaohui Zhou, firstName=Xiaohui, middleName=null, lastName=Zhou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"

born in 2000. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

"}, bioImg=null, bioContent=

born in 2000. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495695209267457, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, authorId=1325495693409911036, language=CN, stringName=周小晖, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
2国防科技大学计算机学院 长沙 410073, bio={"img":"VzRvj3Pt5kn3I0kiDxo7PA==","content":"

2000 年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

"}, bioImg=VzRvj3Pt5kn3I0kiDxo7PA==, bioContent=

2000 年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495692281643233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=1, ext=[AuthorCompanyExt(id=1325495692290031842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073), AuthorCompanyExt(id=1325495692302614755, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692281643233, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073)]), AuthorCompany(id=1325495692382306532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, xref=2, ext=[AuthorCompanyExt(id=1325495692390695141, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1325495692403278054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495688502575310, companyId=1325495692382306532, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2国防科技大学计算机学院 长沙 410073)])])]
葛伦1, 2 , 王意洁1, 2, * , 张莹3, 周小晖1, 2
作者信息
  • 1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
  • 2国防科技大学计算机学院 长沙 410073
  • 3浙江大学航空航天学院 杭州 310058
通讯作者:
王意洁()
作者简介:

1986年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

1971年生。博士,教授,博士生导师。主要研究方向为大数据分析、云计算、分布式存储

1983年生。博士研究生。主要研究方向为异常检测、数据挖掘、空间轨道意图识别

2000 年生。博士研究生。主要研究方向为智能运维、异常检测、深度学习

Time-Series Anomaly Detection Method Based on Time-Frequency Enhanced Contrastive Learning
Lun Ge1, 2 , Yijie Wang1, 2, * , Ying Zhang3, Xiaohui Zhou1, 2
Affiliations
  • 1National Key Laboratory of Parallel and Distributed Computing (National University of Defense Technology), Changsha 410073
  • 2College of Computer Science and Technology, National University of Defense Technology, Changsha 410073
  • 3School of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310058
  • born in 1986. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

    born in 1971. PhD, professor, PhD supervisor. Her main research interests include big data analysis, cloud computing, and distributed storage

    born in 1983. PhD candidate. Her main research interests include anomaly detection, data mining, and space trajectory intent recognition

    born in 2000. PhD candidate. His main research interests include AIOps, anomaly detection, and deep learning

出版时间: 2026-03-01 doi: 10.7544/issn1000-1239.202550364
文章导航
收藏切换

人工智能、大数据、云计算等新兴产业的飞速发展,催生了大规模云计算中心的广泛应用。这些云计算中心在运行过程中持续产生海量级运维数据,高效、准确的异常检测是保障智能运维和系统可靠性的关键环节。由于设备老化、环境突变、动态负载等因素,运维数据往往包含数据噪声污染、异常信号耦合、设备状态动态漂移等问题。当前主流的异常检测方法通常依赖数据纯净性假设,通过对正常数据的特征学习并评估重构误差来实现异常判别,而数据质量的降低会严重影响异常检测效果。针对上述问题,提出基于时频增强的对比学习时序异常检测方法(TF-ECL),通过构建时频双视图增强机制,在时域和频域实施可控增强,基于时频多视图对比学习的方法提升了模型抗噪能力。同时,利用异常信号在时频空间的表现差异性,构建双域协同检测机制,利用时频跨域对比误差进行异常定位。在多个真实时序数据集的实验结果表明,相较于之前的多维时间序列异常检测方法,TF-ECL的F1分数,VUS-PR,VUS-ROC等指标都优于或达到现有的SOTA方法水平,表现出稳健而优异的性能。

多维时间序列  /  时序异常检测  /  对比学习  /  时域增强  /  频域增强

The rapid development of emerging industries such as artificial intelligence, big data, and cloud computing has led to the widespread application of large-scale cloud computing centers, which continuously generate massive amounts of operational data during operation. Efficient and accurate anomaly detection is a key part in ensuring intelligent operation and maintenance as well as the system reliability. Due to factors such as equipment aging, sudden environmental changes, and dynamic loads, operational data often contain problems such as data noise pollution, abnormal signal coupling, and dynamic drift in equipment status. Current mainstream anomaly detection methods typically rely on assumptions of data purity, identifying anomalies by learning normal data features and evaluating reconstruction errors. However, industrial-grade intelligent systems face inherent challenges such as sensor noise contamination, dynamic state drift, and coupled anomaly signals due to their complex topological structures and heterogeneous operating environments, all of which degrade data quality and compromise detection efficacy. To address these issues, we propose a time-series anomaly detection method based on time-frequency enhanced contrastive learning(TF-ECL). It introduces a dual-view enhancement mechanism to implement controllable augmentations in temporal and frequency domains, thereby improving model noise immunity based on cross-view contrastive learning. Meanwhile, based on the differential representations of abnormal signals in time-frequency space, a dual-domain collaborative detection mechanism is developed to achieve abnormal localization using time-frequency contrastive errors. Experimental results from multiple real-time data sets demonstrate that compared with previous multidimensional time-series anomaly detection methods, TF-ECL achieves robust and excellent performance, with metrics such as F1 score, VUS-PR, and VUS-ROC surpassing or matching those of current state-of-the-art (SOTA) methods.

multivariate time series  /  time-series anomaly detection  /  contrastive learning  /  time-domain augmentation  /  frequency-domain augmentation
葛伦, 王意洁, 张莹, 周小晖. 基于时频增强的对比学习时序异常检测方法. 计算机研究与发展, 2026 , 63 (3) : 615 -626 . DOI: 10.7544/issn1000-1239.202550364
Lun Ge, Yijie Wang, Ying Zhang, Xiaohui Zhou. Time-Series Anomaly Detection Method Based on Time-Frequency Enhanced Contrastive Learning[J]. Journal of Computer Research and Development, 2026 , 63 (3) : 615 -626 . DOI: 10.7544/issn1000-1239.202550364
人工智能、大数据、云计算等新兴产业飞速发展,对计算、存储等资源的需求呈快速增长趋势,而云计算中心作为金融、交通等关键行业计算、存储资源的重要提供方,其稳定、可靠运行尤为重要。这些云计算中心在运行过程中持续产生海量级运维数据,对此类时序数据的高效、准确的异常检测是保障智能运维和系统可靠性的关键环节。由于设备老化、环境突变、动态负载等因素,运维数据往往包含数据噪声污染、异常信号耦合、设备状态动态漂移等问题。当前主流的异常检测方法通常依赖数据纯净性假设,通过对正常数据的特征学习并基于重构误差来判别异常,而数据质量的降低会严重影响异常检测效果。
我们注意到异常信号在不同表征空间的表现形式存在显著差异,如图1所示。以典型点异常和上下文异常为例:时域中的突发尖峰(点异常)在频域表现为特定低频分量的能量突增,而持续性偏离(上下文异常)图1中阴影区域所示,则对应频域中相位谱幅值的系统性偏移,这种时频域的表征分离特性为异常检测提供了新的视角。
近年来,基于深度学习的重构方法因其强大的非线性表征能力,在异常检测领域取得显著进展[1]。基于重构的方法是一种重要的异常检测方法,通过学习正常模式的数据特征,利用重构误差来识别异常。但在现实场景中,复杂的外部环境干扰会对时序数据质量造成较大影响,对基于重构的深度学习异常检测方法构成严峻挑战:
1)传感器噪声与设备漂移会扭曲原始信号特征,导致模型在噪声污染下难以准确捕捉真实的数据分布;
2)混杂在训练集的异常数据可能误导模型学习到错误的映射关系;
3)时序数据的高度动态变化性会导致模型难以稳定捕捉潜在特征表示,从而使得依赖单一视图的重构方法在复杂波动中易收敛至局部最优解。
针对上述挑战,提出了一种基于时频增强的对比学习时序异常检测方法(time-series anomaly detection method based on time-frequency enhanced contrastive learning,TF-ECL)。本文的主要贡献包括3个方面:
1) 提出了一种基于时频双视图增强的对比学习异常检测框架,通过引入频域视图的可控数据增强来提升数据特征提取能力,弥补了单视图在复杂时序数据中特征提取鲁棒性不足的缺陷。
2) 基于异常信号在时域和频域的表征差异,提出一种时频双视图KL散度对比的异常判别准则。该设计充分发挥频域重构对上下文异常的高敏感性与时域重构对点异常的敏感性优势,形成互补检测机制。同时借助KL散度的概率分布量化能力及区间归一化特性,显著放大微弱波形畸变特征,从而提升复杂时序异常的鉴别能力。
3) 在多个多维时间序列真实数据集上评估了TF-ECL方法。对比实验结果表明TF-ECL获得了平均最佳F1分数(0.9728)、平均最佳VUS-PR值(0.9311)和平均最佳VUS-ROC值(0.9746),相较于最先进的方法分别提升了0.55个百分点、1.42个百分点、0.63个百分点。进一步,通过消融实验、时频增强强度分析以及综合效能损耗分析证明了提出的多维时间序列异常检测结构设计的合理有效性。
时间序列异常检测技术经历了从统计学模型、传统机器学习到深度学习方法的演进,其中深度学习方法凭借其强大的非线性特征提取能力已成为主流范式[2]。基于重构的深度学习方法(如自动编码器[3]、变分自编码器[4]及Transformer[5]架构)通过学习正常数据的重建模式,以重构误差定位异常:如OmniAnomaly[6]通过变分推理建模高维时序数据的潜在分布以实现无监督检测;Anomaly Transformer[7]设计了一种双分支异常注意力机制,并利用最小最大化策略优化关联差异,从而增强检测能力;TimesNet[8]通过多周期张量分解捕捉复杂时序动态,基于重构误差实现异常判别;STEN[9]设计了基于预测的时间特征提取模块和基于距离的空间特征提取模块,能够更好地学习正常模式数据中的时空特征;周小晖等人[10]提出基于融合学习的无监督异常检测方法FLAD,通过多层融合编码器进行特征提取,利用重构误差进行异常检测;袁子淇等人[11]通过构建时频多模态生成对抗网络模型,对正常时序信息的特征分布进行联合学习,实现更有效的异常检测。
另外,基于端到端异常检测方法能够提升对未知异常的识别能力,如Deep SVDD[12](后文简写为“DSVDD”)通过压缩正常样本至超球中心直接构建决策边界,在高维数据异常检测中体现优势。近年来,图神经网络(graph neural network,GNN)因其多尺度关联建模的优势被广泛应用于异常检测,GAD-NR[13]基于邻域重构实现图异常检测;GeneralDyG[14]结合动态子图采样与时空Transformer实现动态图的高效异常定位;吴彦文等人[15]通过设计图神经网络预测模块,结合重构模块进行图结构学习,以端到端的方式实现时序数据异常检测。同时,大模型技术的兴起推动异常检测向通用化与自适应方向演进:AnomalyLLM[16]通过时序指令微调适配通用大语言模型,利用上下文感知突破小样本检测瓶颈;AnomalyBERT[17]设计多尺度时序掩码自编码架构,借助跨模态对比学习提升瞬态与渐变异常检测灵敏度。
然而实际场景中训练集的异常数据混杂与时序动态漂移性仍严重制约模型性能。值得注意的是,对比学习在图像分类任务中已展现出接近有监督方法的表征能力,这一特性为异常检测领域提供了新启示,通过构建解耦的鲁棒特征空间,可有效对抗异常污染与动态漂移干扰。对比学习的异常检测方法现有工作主要沿2个方向推进:一是通过注入合成异常来构造负样本对,并根据原始数据与负样本的差异来判断异常。如TimeAutoAD,CTAD,DACAD等方法[18-20],其依赖先验异常分布假设,在开放场景中易受未知异常干扰。二是免负样本构建,通过数据多视图对比实现特征学习。如DCdetector[21]与PatchAD[22]分别在时序维度构建多尺度视图,利用对比损失提取时序不变性特征,在性能上超越了传统重构方法。然而,这类方法主要聚焦于时域空间的表征对齐,未能有效利用频域空间特征信息。
关注到复杂变化的时序数据在频域空间会表现出频域特征稀疏性,更利于模型学习,研究人员也提出了一些关于频域学习的方法:如MtsCID[23]设计了时域多尺度注意力图和频域卷积模块,通过交互捕捉时域和频域特征,促进了模型异常检测能力提升;袁子淇等人[11]基于时域和频域的多模态对抗学习,实现无监督时间序列异常检测;而与本文技术路线更具有相似性的是TFMAE[24]方法,其通过对时频双域数据实施可学习掩码操作,并基于时域-频域视图互信息最大化来优化模型训练,再利用多视图重构差异实现异常检测。但该方法对相位信息在时序连续性中的关键作用缺乏显式约束,相位谱的随机掩蔽会破坏时序波形的局部连续性特征,导致异常检测的时空一致性下降。
针对这一缺陷,提出相位保持约束策略,在频域增强过程中固定相位谱核心分量不变,仅对功率谱非主频段施加可控扰动。该设计既完整保留原始时序的拓扑结构特征,又通过跨域对比学习强化异常敏感表征的判别能力,实现鲁棒性与检测精度的协同提升。
针对时间序列异常检测任务,问题定义如下:在无监督场景下,给定多元时间序列训练数据${{{\boldsymbol{X}}}_{{\text{train}}}} = ({{{\boldsymbol{x}}}_1},{{{\boldsymbol{x}}}_2}, … ,{{{\boldsymbol{x}}}_{{T}}})$,其中每个时间步$ {{{\boldsymbol{x}}}_t} \in {\mathbb{R}^d} $包含$d$维特征,目标是构建模型检测未知测试序列${{{\boldsymbol{X}}}_{{\text{test}}}} = ({{\boldsymbol{x}}}_{1}' , {{\boldsymbol{x}}}_2' , … , {{\boldsymbol{x}}}_{{T}}' )$中的异常点,并输出异常标签序列$ {{{\boldsymbol{Y}}}_{{\text{test}}}} = ({{{{y}}}_1},{{{{y}}}_2}, … , {{{{y}}}_{{T}}}) $。其中,${{{{y}}}_t} \in \{ 0,1\} $,“1”表示异常事件,“0”表示正常事件。
针对训练数据噪声污染与分布漂移引发的重构基准失准问题,本研究提出TF-ECL方法,该方法通过时域、频域的可控增强及双视图编码实现时序数据核心特征提取,利用异常数据在时域、频域表现的差异性进行异常检测。模型由5个核心组件构成: 1)时域增强模块;2)时域编码器;3)频域增强模块;4)频域编码器;5)时频对比模块。模型整体框架如图2所示。
训练数据首先经过实例规范化模块,实现数据分布规范化的数据预处理。时域增强模块主要使用随机抖动和掩码进行数据增强;频域增强模块主要分为功率谱和相位谱2部分,在保持相位不变约束下实施可控频域增强。时域编码器和频域编码器采用Transformer模型进行特征提取,编码器输出特征经过空间投影器后,分别将时域嵌入特征$ {{\tilde {\boldsymbol{Z}}}_{\text{T}}} $和频域嵌入特征$ {{\tilde {\boldsymbol{Z}}}_{\text{F}}} $统一映射至时域空间,实现频域表征向时域表征的强制对齐。构建跨域正负样本,通过最大化不同视图的表征差异,推动对比学习架构的稳定训练。该方法通过时频双视图的特征解耦与隐空间协同优化机制有效提升模型在噪声干扰下的模型鲁棒能力。
时间序列数据通常呈现非平稳性,其均值、方差等统计特性会随时间发生变化,导致模型在训练阶段学习到的数据分布与实际推理阶段的数据分布存在偏差,这种偏差会造成模型性能下降。尤其对于Transformer的模型,其注意力机制对输入分布的敏感性可能放大非平稳性影响,造成注意力图模糊化与预测结果过平滑化。为缓解上述问题,本文使用文献[25]提出的RevIN方法进行数据实例规范化。通过对单个变量的时序表示进行独立归一化,消除其非平稳性特征。给定输入序列${{{\boldsymbol{X}}}^{b \times l \times d}}$,实例规范化公式为
$ {{{\boldsymbol{X}}}_{{\text{norm}}}} = \frac{{{{{\boldsymbol{X}}}_d} - mean({{{\boldsymbol{X}}}_d})}}{{std({{{\boldsymbol{X}}}_d}) + \varepsilon }}\;。 $
经过实例规范化后,本文遵循原始Transformer的嵌入编码策略[26],将输入序列的数值特征与位置信息分别通过词嵌入与位置嵌入投影至隐空间,并通过残差叠加构建表征信息,嵌入向量${{{\boldsymbol{X}}}_{{\text{emb}}}}$表示为
$ {{{\boldsymbol{X}}}_{{\text{emb}}}} = dropout[{E_{{\text{value}}}}({{{\boldsymbol{X}}}_{{\text{norm}}}}) + {E_{pos}}(t)]\text{,} $
其中位置编码基于正弦函数将观测值的绝对顺序信息融入特征向量。位置编码公式为
$ E_{pos}(t)=\left\{\begin{gathered}\sin(pos/10\; 000^{2i/d_{\text{model}}}),\; i\in Even\ , \\ \cos(pos/10\; 000^{2i/d_{\text{model}}}),\; i\in Odd\; 。 \\ \end{gathered}\right. $
在现实世界中,传感器生成的时序数据通常包含大量噪声,因此模型在不同环境下的性能稳健性非常重要。针对时间序列的数据增强技术已得到广泛研究,该技术已被证明能有效提升模型对噪声的鲁棒性,常用的时序数据增强方法包括随机抖动、幅度缩放、窗口扭曲、时序翻转和随机掩码等[27]。本文采用了随机抖动加随机掩码的方法进行数据增强,主要目的是在保留时序数据时序关联性不变前提下,增加随机噪声,使得时间编码器聚焦学习时序数据核心特征。时域增强特征表示为
$ {{{\boldsymbol{X}}}_{{\text{augT}}}} = {G_{{\text{jitter}}}}(\alpha,{{{\boldsymbol{X}}}_{{\text{emb}}}}) + {G_{{\text{mask}}}}(\beta,{{{\boldsymbol{X}}}_{{\text{emb}}}})\text{,} $
其中抖动系数$\alpha = 0.1$,掩码系数$\beta = 0.1$。
本文使用基于多头自注意力机制的Transformer编码器作为时域特征提取器,其核心优势在于具有较强的全局依赖建模能力。时域特征编码过程表示为
$ {{\tilde {\boldsymbol{X}}}_{\text{T}}} = Encode{r_{\text{T}}}({{{\boldsymbol{X}}}_{{\text{augT}}}}) + {{{\boldsymbol{b}}}_{\text{T}}}\text{,} $
其中$ {{\tilde {\boldsymbol{X}}}_{\text{T}}} $为编码后特征,${{{\boldsymbol{X}}}_{{\text{augT}}}}$为时域增强特征,${{{\boldsymbol{b}}}_{\text{T}}}$为可学习的偏置参数。通过多头自注意力机制能够实现不同维度特征的有效提取,多头自注意力公式为
$ {{\boldsymbol{Head}}_h} = attention({{{\boldsymbol{Q}}}_h},{{{\boldsymbol{K}}}_h},{{{\boldsymbol{V}}}_h}) = softmax\left(\frac{{{{{\boldsymbol{Q}}}_h}{{\boldsymbol{K}}}_h^{\text{T}}}}{{\sqrt {d/h} }}\right){{{\boldsymbol{V}}}_h} \text{,} $
$ {{{\boldsymbol{Q}}}_h} = {{{\boldsymbol{X}}}_{{\text{augT}}}}{{\boldsymbol{W}}}_h^q,{{{\boldsymbol{K}}}_h} = {{{\boldsymbol{X}}}_{{\text{augT}}}}{{\boldsymbol{W}}}_h^k,{{{\boldsymbol{V}}}_h} = {{{\boldsymbol{X}}}_{{\text{augT}}}}{{\boldsymbol{W}}}_h^v \text{,} $
其中$h $为注意力头数,${{\boldsymbol{W}}}_h^q,{{\boldsymbol{W}}}_h^k,{{\boldsymbol{W}}}_h^v$为各注意力头的投影矩阵,多头注意力经过拼接后的输出特征为
$ {{\tilde {\boldsymbol{Z}}}_{\text{T}}} = concat({{\boldsymbol{Head}}_1},{{\boldsymbol{Head}}_2},\;…,{{\boldsymbol{Head}}_h}){{\boldsymbol{W}}}\;。$
本文提出了基于频域解耦的联合表征方法,通过解耦频域物理意义正交的分量,来抑制噪声对时序不变性特征的干扰。首先通过快速傅里叶变换(fast Fourier transform,FFT)将原始时序${{{\boldsymbol{X}}}_{\text{T}}}$映射至频域空间${{{\boldsymbol{X}}}_{\text{F}}}$,随后将其解耦为功率谱${{{\boldsymbol{P}}}_{\text{F}}}$与相位谱${{{\boldsymbol{\varphi }}}_{\text{F}}}$两个互补子空间,表示为
$ \left\{\begin{aligned}&{\boldsymbol{X}}_{\text{F}}=FFT({\boldsymbol{X}}_{\text{T}})\text{,}\\ &{\boldsymbol{P}}_{\text{F}}={\left|{\boldsymbol{X}}_{\text{F}}\right|}^{2}\text{,}\\ &{\boldsymbol{\varphi}}_{\text{F}}=\text{arctan}\left(\frac{\text{Im}({\boldsymbol{X}}_{\text{F}})}{\text{Re}({\boldsymbol{X}}_{\text{F}})}\right)\text{,}\end{aligned}\right.$
其中${{{\boldsymbol{P}}}_{\text{F}}}$表征频域能量分布,${{{\boldsymbol{\varphi}} }_{\text{F}}}$表征时序相位关系。需指出的是,频域增强过程中的过度扰动可能导致时域信号的结构性失真。为在增强模型鲁棒性与抑制表征塌缩之间实现均衡,本文采用可控频域扰动策略:仅针对功率谱幅值低于阈值的成分进行随机添加和移除操作,不对相位谱信息进行修改,以保持时序动态模式的关键不变性。通过约束频带能量调制范围,在保留时序关键模式前提下引入可控多样性。具体方法是[28]:在功率谱成分添加操作时,首先计算频谱最大振幅${A}$,然后随机选择$\theta $倍振幅小于$\omega {A}$的频率成分,将其振幅提升至$\omega {A}$。移除操作采用掩码技术随机丢弃$\theta $倍频率成分。频域增强公式为
$ {{{\boldsymbol{P}}}_{{\text{augF}}}} = {G_{{\text{add}}}}(\omega,{{{\boldsymbol{P}}}_{\text{F}}}) + {G_{{\text{drop}}}}(\theta,{{{\boldsymbol{P}}}_{\text{F}}})\text{,} $
其中${G_{{\text{add}}}}$为增强频率,${G_{{\text{drop}}}}$为删除频率,$\omega $为缩放因子,$\theta $为增删系数,本文设置$\omega = 0.1$,$\theta = 0.1$。
频域编码器由功率谱和相位谱2个独立的Transformer编码器组成。增强后的功率谱${{{\boldsymbol{P}}}_{{\text{augF}}}}$特征和未增强的相位谱${{{\boldsymbol{\varphi}} }_{\text{F}}}$特征通过双通道编码器进行频域特征提取,并生成对应特征${{{\boldsymbol{H}}}_{\text{amp}}},{{{\boldsymbol{H}}}_{\rm{phi }}}$,表示为
$ {{{\boldsymbol{H}}}_{\text{amp}}} = Encode{r_{\text{amp}}}({{{\boldsymbol{P}}}_{{\text{augF}}}}) + {{{\boldsymbol{b}}}_{\text{amp}}}\text{,} $
$ {{{\boldsymbol{H}}}_{{{\mathrm{phi}} }}} = Encode{r_{{{\mathrm{phi}} }}}({{{\boldsymbol{\varphi}} }_{\text{F}}}) + {{{\boldsymbol{b}}}_{{{\mathrm{phi}} }}} \text{,} $
其中${{{\boldsymbol{b}}}_{\text{amp}}},{{{\boldsymbol{b}}}_{{{\mathrm{phi}} }}}$为可学习的偏置参数。
在时序数据分析中,时域与频域表征的异构性常导致跨域语义割裂和模式冲突的问题,文献[29]直接使用时域和频域特征进行对比,忽视了傅里叶对偶理论中时频信号的能量守恒约束,致使模型难以捕捉跨域一致性特征。
因此,本文基于特征对比一致性考虑,将频域学习后特征${{{\boldsymbol{H}}}_{\text{amp}}},{{{\boldsymbol{H}}}_{{{\mathrm{phi}} }}}$,通过傅里叶逆变换到时域再进行对比,在信息论层面最小化时频跨域互信息差异。具体操作是:先基于欧拉公式,将编码后的幅度与相位映射为复数频域信号,频域信号表示为
$ {{{\boldsymbol{X}}}_{{\text{augF}}}} = {{{\boldsymbol{H}}}_{\text{amp}}}(\cos {{{\boldsymbol{H}}}_{{{\mathrm{phi}} }}} + {\text{j}}\sin {{{\boldsymbol{H}}}_{{{\mathrm{phi}} }}}) \text{,} $
再将复数频域信号通过正交归一化逆傅里叶变换恢复为时域信号,表示为
$ {\tilde{{\boldsymbol{Z}}}_{\text{F}}} = \frac{1}{{\sqrt N }}\sum\limits_{k = 0}^{N/2} {{{{\boldsymbol{X}}}_{{\text{augF}}}}\exp \left({\text{j}}\frac{{2{\text{π}}kn}}{N}\right)} \;。$
本节对训练阶段对比学习框架的关键部件进行阐述,主要包括时频域投影头、跨域正负样本对构建、对比损失函数3部分。
1)时频域投影头
在对比学习中,投影头可分离任务无关特征与任务相关特征,增强隐变量对噪声的鲁棒性,对于时域编码器输出$ {{\tilde {\boldsymbol{Z}}}_{\text{T}}} $,设计多层感知机(multilayer perceptron,MLP)作为时域投影头,将其映射至时域输出空间:
$ {{\tilde {\boldsymbol{X}}}_{\text{T}}} = MLP({{\tilde {\boldsymbol{Z}}}_{\text{T}}})\;。 $
频域投影头和时域特征投影头相同,使用多层感知机将频域隐变量特征$ {{\tilde {\boldsymbol{Z}}}_{\text{F}}} $也映射至时域输出空间:
$ {{\tilde {\boldsymbol{X}}}_{\text{F}}} = MLP({{\tilde {\boldsymbol{Z}}}_{\text{F}}})\;。$
2)跨域正负样本对构建
对于同一个样本${X_i}$,在时域、频域特征的嵌入向量相似,而不同样本之间时域和频域特征的嵌入向量相距较远。基于此,本文将时域隐变量$ {{\tilde {\boldsymbol{Z}}}_{\text{T}}} $和输出变量$ {{\tilde {\boldsymbol{X}}}_{\text{T}}} $、频域隐变量$ {{\tilde {\boldsymbol{Z}}}_{\text{F}}} $和输出变量$ {{\tilde {\boldsymbol{X}}}_{\text{F}}} $交叉组成跨域样本对。其中,同一样本之间记为正样本对$ ({\tilde{{\boldsymbol{X}}}}_{\text{T}}^{i},{\tilde{{\boldsymbol{Z}}}}_{\text{F}}^{i}) $,$ ({\tilde{{\boldsymbol{X}}}}_{\text{F}}^{i},{\tilde{{\boldsymbol{Z}}}}_{\text{T}}^{i}) $;不同样本之间记为负样本对$ ({\tilde{{\boldsymbol{X}}}}_{\text{T}}^{j},{\tilde{{\boldsymbol{Z}}}}_{\text{F}}^{j}) $,$ ({\tilde{{\boldsymbol{X}}}}_{\text{F}}^{j}, {\tilde{{\boldsymbol{Z}}}}_{\text{T}}^{j}) $。该对比样本对构建方式既增大了对比信息量,也扩大了时频不同视图之间的对比差异,有利于模型的对比学习。
3)对比损失函数
本文采用对比学习中广泛应用的归一化温度缩放交叉熵损失作为损失函数,结合梯度冻结策略[30]实现模型的稳定训练。为构建时频双域互监督机制,总损失函数${L_{\text{C}}}$由时域对比损失${L_{\text{T}}}$与频域对比损失${L_{\text{F}}}$加权组成,通过平衡系数$\lambda $进行权重调节,损失函数为
$ {L_{\text{C}}} = \lambda {L_{\text{T}}}({{\tilde {\boldsymbol{X}}}_{\text{T}}}DT({{\tilde {\boldsymbol{Z}}}_{\text{F}}})) + (1 - \lambda ){L_{\text{F}}}({{\tilde {\boldsymbol{X}}}_{\text{F}}}DT({{\tilde {\boldsymbol{Z}}}_{\text{T}}})) \text{,} $
$ {L_{{\text{T}},i}} = - \log \dfrac{{\exp (sim({\tilde {\boldsymbol{X}}}_{\text{T}}^i,{\tilde {\boldsymbol{Z}}}_{\text{F}}^j)/\tau )}}{{\displaystyle\sum\limits_{{x_j} \in D} {{\text{exp}}(sim({\tilde {\boldsymbol{X}}}_{\text{T}}^i,{\tilde {\boldsymbol{Z}}}_{\text{F}}^j)/\tau )} }} \text{,} $
$ {L_{{\text{F}},i}} = - \log \dfrac{{\exp (sim({\tilde {\boldsymbol{X}}}_{\text{F}}^i,{\tilde {\boldsymbol{Z}}}_{\text{T}}^j)/\tau )}}{{\displaystyle\sum\limits_{{x_j} \in D} {{\text{exp}}(sim({\tilde {\boldsymbol{X}}}_{\text{F}}^i,{\tilde {\boldsymbol{Z}}}_{\text{T}}^j)/\tau )} }}\text{,} $
其中$sim({{\boldsymbol{u}}},{{\boldsymbol{v}}}) = {{{\boldsymbol{u}}}^{\text{T}}}{{\boldsymbol{v}}}/(||{{\boldsymbol{u}}}||||{{\boldsymbol{v}}}||)$表示余弦相似度,$ DT( \cdot ) $表示梯度冻结,$\tau $为调节尺度温度参数。${x_j} \in D$表示其他增强样本集合,平衡系数$\lambda = 0.2$。需要注意的是:若同时更新$ {{\tilde {\boldsymbol{X}}}_{\text{T}}} $与$ {{\tilde {\boldsymbol{Z}}}_{\text{F}}} $,模型易陷入平凡解,导致对比学习失效。本文通过冻结隐变量参数$ {{\tilde {\boldsymbol{Z}}}_{\text{T}}} $与$ {{\tilde {\boldsymbol{Z}}}_{\text{F}}} $,维持特征空间的结构稳定性,避免模型训练过程中的表征塌陷。
本文的异常检测机制基于跨模态表征一致性假设:正常样本在时域与频域视图中保持特征一致性,而异常样本则表现出跨视图表征不一致性。鉴于时域重构对点异常敏感以及频域重构对上下文异常更具敏感性,我们利用时频双视图的信号差异特性构建互补检测机制。为强化上下文异常检测能力,采用时频双视图重构分布与原始输入分布的KL(Kullback-Leibler)散度[31]作为异常度量基准。相较于传统重构异常分数,经归一化至[0,1]区间的KL散度能更有效量化概率分布差异,并对持续波形畸变信息产生显著放大效应。
具体步骤如下:在推理阶段,测试数据不进行时域和频域增强,直接进入时域编码器和频域编码器进行重构,通过测试数据分布与不同视图输出数据分布的KL散度差异来计算异常分数,表示为
$ score(t) = {D_{{\text{KL}}}}({{\boldsymbol{X}}}_{{{\text{test}}}}||{\tilde {\boldsymbol{X}}}_{\text{T}}) - {D_{{\text{KL}}}}({{\boldsymbol{X}}}_{{{\text{test}}}}||{\tilde {\boldsymbol{X}}}_{\text{F}})\;。 $
最后根据异常得分和预先确定的阈值进行异常判别,即当得分超过阈值时检测为异常。异常标签计算结果表示为
$ {{\tilde {{y}}}_t} = \left\{ \begin{gathered}1,\;score(t) \geqslant \delta , \\0,\;score(t) \lt \delta \;。 \\ \end{gathered} \right.$
本文在4个经过广泛验证的多元时间序列数据集上开展系统性评估,数据集介绍如表1所示。
1)MSL是NASA发布的火星车传感器数据,包含航天器运行时的突发性机械故障异常,异常类型以瞬时尖峰和持续性漂移为主。
2)SMAP是NASA土壤湿度卫星监测数据,异常由传感器信号中断或环境干扰引发,呈现低频周期性扰动特征。
3)SWAT是工业水处理系统安全监测数据,涵盖51个传感器与执行器,主要包含网络攻击引发的异常。
4)PSM是eBay公开的服务器集群性能指标,覆盖CPU利用率、内存占用等多维度时序,异常反映服务器节点故障。
为了更全面地比较时间序列异常检测性能,本文采用了2类评价指标:
第1类指标为经过调整策略[32]的精确率(precision,P)、召回率(recall,R)和F1分数,上述指标和点调整策略广泛用于现有的异常检测方法,指标计算见式(22)~(24)。点调整策略具体实现是:如果检测到某一段中的单个观测值为异常,则识别该段都为异常值。P,R,F1分数指标计算如下
$ P = \frac{{TP}}{{TP + FP}} \text{,} $
$ R = \frac{{TP}}{{TP + FN}} \text{,} $
$ F1 = \frac{{2 \times P \times R}}{{P + R}} \text{,} $
其中TP表示正确检测为异常的异常,FP表示错误检测为异常的正常,TN表示正确识别为正常的正常,FN表示错误识别为正常的异常。
由于F1分数对检测阈值的设定较为敏感,且在数据分布极端不平衡时评价能力不足,本文选取和阈值无关的体积不确定性评分VUS(volume under surface)系列指标VUS-ROC和VUS-PR[33]作为第2类评价指标。该指标也被近期多篇文献使用[21-24]。VUS-ROC和VUS-PR是对传统ROC和PR曲线下面积指标的扩展,表示通过量化模型对异常事件持续时间的敏感性。其核心思想是在3维参数空间内计算积分体积,数学形式定义为
$ VUS = \frac{1}{K}\sum\limits_{k = 1}^K {\int_{{\tau _{\min }}}^{{\tau _{\max }}} {AUC(f(\tau ,{\theta _k})){\text{d}}\tau } } \;。$
我们将本文方法TF-ECL与10种基线方法进行综合对比评估,基线模型包括:基于密度估计方法DAGMM[34];基于聚类的方法DSVDD[12];基于重构的方法OmniAnomaly[6]、TranAD[35]、TimesNet[8]、STEN[9]、MtsCID[23];基于生成对抗的方法BeatGAN[36];基于对比学习的方法DCdetector[21]、TFMAE[24]。为避免设备差异对实验结果的影响,文中结果通过官方代码库复现,与TF-ECL保持一致的训练环境。
所有基线方法的实验均在配备英伟达GeForce RTX 5060 GPU显卡的机器上进行,使用PyTorch 2.7.0进行实现。在模型训练阶段,设置样本批次训练大小为64、滑动时间窗口大小为96、迭代轮次为5、学习率为0.0001,模型训练采用早停策略。实验对每个数据集都进行5次重复实验,最终实验结果取5次重复实验结果的平均值。所有方法的检测阈值均为点调整后获得最佳F1分数时对应的阈值。
11种方法在4个数据集的对比实验结果如表2所示,不同方法在所有数据集的指标平均值如表3所示。从表3中可以看出,TF-ECL在所有数据集上的各项指标平均值均优于所有对比方法:平均准确率为97.03%、平均召回率为98.42%、平均F1分数为97.28%、平均VUS-PR为93.11%、平均VUS-ROC为97.46%,较次优方法分别提高了1.87个百分点、0.39个百分点、0.55个百分点、1.42个百分点、0.63个百分点。
同时,如表2所示,TF-ECL在不同数据集上的单项评价指标均位列前2。其中,在所有数据集上的F1分数和VUS-ROC指标均优于其他对比方法。在MSL和SMAP数据集上,TF-ECL有最高的精确率(96.25%和96.61%),较次优方法分别提升了3.96个百分点和0.77个百分点;在SWAT数据集上,TF-ECL方法有最高的精确率(96.44%)和召回率(100%),较最优方法分别提升了2.65个百分点和1.02个百分点。
另外,与时域特征学习里表现最出色的STEN相比,TF-ECL方法的平均F1分数、VUS-PR、VUS-ROC指标分别提升了3.52个百分点、1.79个百分点、0.63个百分点。说明相较时域特征提取的方法,利用异常信号在频域视图的表征差异性可以提升时间序列异常检测的性能。
而相较引入频域信息后表现最出色的MtsCID方法,TF-ECL方法的平均F1分数、VUS-PR、VUS-ROC指标分别提升了0.55个百分点、3.19个百分点、4.15个百分点。说明本文使用的基于相位不变的频域可控增强策略,较其他方法更有利于保持数据间的潜在时序关系,实现时序特征提取和异常检测效果的共同提升。
综上,TF-ECL能够更加有效学习多维时间序列的时间依赖特征,引入的频域视图能够有效降低数据噪声和数据分布漂移对模型性能影响,提升异常检测效果,弥补了单视图在复杂时序数据中特征提取鲁棒性不足的缺陷。
为了进一步验证本文各个模块设计的合理性和有效性,在全部实验数据集对TF-ECL模型进行消融实验。4种模型结构分别为:
1) TF-ECL模型。本文提出的原始模型,使用实例规范化,并采用时域、频域增强进行多视图对比。
2) TF-ECLw/o(without)NM模型。在TF-ECL原始模型中去除了实例规范化模块。
3) TF-ECLw/o(without)FE模型。在TF-ECL原始模型中去除了频域增强机制,使用原始数据代替。
4) TF-ECLw/o(without)TE模型。在TF-ECL原始模型中去除了时域增强机制,使用原始数据代替。
以上4种方法在全部实验数据集上的各项指标实验结果如表4所示。消融实验结果表明,移除任一核心模块均显著降低模型性能:1)去除实例规范化后导致平均F1分数、VUS-PR、VUS-ROC分别下降2.65个百分点、4.15个百分点、0.21个百分点,主要源于测试集与训练集的数据分布漂移偏差导致模型性能下降。2)移除频域增强致使平均F1分数、VUS-PR、VUS-ROC分别下降4.53个百分点、2.89个百分点、0.77个百分点,主要原因是频域增强视图具有抑制噪声和异常数据对模型性能影响的能力,缺少该模块会导致模型泛化能力降低,进而影响异常判别能力。3)去除时域增强致使平均F1分数、VUS-PR、VUS-ROC分别下降1.61个百分点、1.67个百分点、0.45个百分点,主要表现在时域局部突变异常感知能力不足,缺少该模块后无法与频域增强视图构成跨域样本对,影响模型对数据表征的学习。
完整的TF-ECL模型通过实例规范化、频域增强与时域增强的协同作用,在噪声抑制、跨域特征对及动态模式捕捉3个维度实现优化,最终在所有数据集全部指标上都展现出最优或次优结果,验证了多视图增强机制与跨域对比框架设计的有效性。
为探究时域增强强度与频域增强强度对模型性能的影响,本文在时域抖动系数$\alpha \in [0.1,0.9]$、时域掩码率$\beta \in [0.1,0.7]$和频域增删系数$\theta \in [0.1,0.9]$范围内进行网格搜索实验。实验结果表明,过大的时域抖动或频域增删会破坏时序本身特征,导致模型检测能力下降。时域增强与频域增强的分析结果如图3所示。
1)时域抖动系数$\alpha $。随机噪声注入过强导致时序波形过度扭曲,破坏正常模式连续性,当$\alpha = 0.1$时,SWAT,PSM,SMAP数据集的F1分数为最优值,而$\alpha = 0.3$时,所有数据集的F1分数均大幅下降。
2)时域掩码系数$\beta $。与时域抖动系数分析结果类似,一定程度的掩码率有助于提升模型鲁棒性,掩码系数过大会导致时序特征过度缺失,降低模型的特征学习效果,当$\beta = 0.1$时,各数据集的F1分数为最优值。
3)频域增删系数$\theta $。当频率增删系数过大时,主频能量泄漏与相位谱畸变引发频域表征失真,各数据集的F1分数均出现不同程度下降(下降0.52%~4.66%)。实验表明,$\theta = 0.1$可有效增强频域判别性,同时避免波形拓扑结构破坏。
综上,时域与频域增强的协同参数优化是保障模型性能的关键,过度增强将破坏跨域表征一致性,而增强不足则削弱异常敏感特征提取能力。
为验证TF-ECL模型在工业场景下的实用性,本文对比分析了TF-ECL与TranAD,BeatGAN,TimesNet,DCdetector,TFMAE等基线模型在SWAT数据集上的单轮训练时间和显存占用情况下的F1分数,结果如图4所示。其中,横轴为单轮次训练时间,纵轴为F1分数,气泡大小代表模型训练显存占用情况。实验结果表明:TF-ECL在获得最高的F1分数同时,在GPU显存利用率与训练速度之间取得平衡。TF-ECL单轮训练耗时254 s,低于同样使用对比学习的基线方法DCdetector(1200 s)和TFMAE(268 s),以及其他基线方法DSVDD(633 s),BeatGAN(430 s),OmniAnomaly(330 s),TimesNet(270 s),TranAD(288 s),略高于STEN(222 s)和DAGMM(189 s)。
同时,相较DCdetector(2112 MB),TFMAE (1647 MB)等对比学习方法,TF-ECL(760 MB)的显存占用显著降低,其原因是TF-ECL使用时频多视图在时域对齐后的特征向量来直接构建正负样本对,而未使用DCdetector等在批次内逐样本间构建正负样本对的方法,从而大大降低了正负样本对的训练数量,进而大幅降低显存占用。实验证明,TF-ECL在异常检测精度、训练效率与计算消耗之间实现了有效平衡,满足工业监控系统对高精度、低消耗的核心需求,具备较强实际应用价值。
针对时序数据异常检测中单视图建模的鲁棒性不足问题,本文提出基于时频增强的多视图对比学习方法,通过构建时频双视图增强机制,利用时频重构分布差异来进行异常检测.TF-ECL在时域引入随机抖动与掩码增强局部时序鲁棒性,在频域固定相位主分量并扰动非主频段功率谱以维持波形连续性,结合跨域对比学习策略与梯度冻结机制,有效抑制噪声干扰并有效捕捉异常特征。实验表明,TF-ECL在多个真实数据集上的F1分数、VUS-PR、VUS-ROC指标优于基线模型。未来工作将进一步探究时域与频域数据增强的随机性对模型稳定性的影响,探索和解释时域和频域增强参数的选择依据和关联机理,优化增强策略的动态自适应机制,以进一步提升复杂动态场景下的检测鲁棒性。
作者贡献声明:葛伦提出主要研究思路,完成实验并撰写论文;王意洁提出实验方案和指导意见以及修改和审核论文;张莹协助完成部分实验,提出论文修改思路;周小晖指导论文写作,参与论文修订。

参考文献 引证文献
排序方式:
1
Pang Guansong, Shen Chunhua, Cao Longbing, et al. Deep learning for anomaly detection: A review[J]. ACM Computing Surveys, 2021, 54(2): 1−38
2
Kim D, Park S, Choo J, et al. When model meets new normals: Test-time adaptation for unsupervised time-series anomaly detection[C]//Proc of the 38th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2024: 406−413
3
Wang Zexin, Pei Changhua, Ma Minghua, et al. Revisiting VAE for unsupervised time series anomaly detection: A frequency perspective [C]//Proc of the 24th ACM Web Conf. New York: ACM, 2024: 3009−3017
4
Iglesias G, Talavera E. Data augmentation techniques in time series domain: A survey and taxonomy[J]. Neural Computing and Applications, 2023, 35(14): 1123−1145
5
Zhou Tian, Ma Ziqing, Wen Qingsong, et al. FEDformer: Frequency enhanced decomposed Transformer for long-term series forecasting[C]//Proc of the 39th Int Conf on Machine Learning. New York: ACM, 2022: 78−85
6
Su Ya, Zhao Youjian, Niu Chenhao, et al. Robust anomaly detection for multivariate time series through stochastic recurrent neural network[C]//Proc of the 25th ACM SIGKDD Int Conf on Knowledge Discovery and Data Mining. New York: ACM, 2019: 2828−2837
7
Xu Jiehui, Wu Haixu, Wang Jianmin, et al. Anomaly Transformer: Time series anomaly detection with association discrepancy[C/OL]//Proc of the 10th Int Conf on Learning Representations. 2022[2025-06-01]. https://openreview.net /pdf?id=LzQQ89U1qm_
8
Wu Haixu, Hu Tengge, Liu Yong, et al. TimesNet: Temporal 2D-variation modeling for general time series analysis[C/OL]//Proc of the 11th Int Conf on Learning Representations. 2023[2025-06-01]. https://openreview.net/pdf?id= ju_Uqw384Oq
9
Chen Yutong, Xu Hongzuo, Pang Guangsong, et al. Self-supervised spatial-temporal normality learning for time series anomaly detection[C]//Proc of the 2024 European Conf on Machine Learning and Principles and Practice of Knowledge Discovery in Databases. Berlin: Springer, 2024: 183−195
10
周小晖, 王意洁, 徐鸿祚, 等. 基于融合学习的无监督多维时间序列异常检测[J]. 计算机研究与发展, 2023, 60(3): 496−508
Zhou Xiaohui, Wang Yijie, Xu Hongzuo, et al. Unsupervised multivariate time series anomaly detection via fusion learning[J]. Journal of Computer Research and Development, 2023, 60(3): 496−508 (in Chinese)
11
袁子淇, 孙庆赟, 周号益, 等. MNDetector: 基于多层网络的异常访问检测方法[J]. 计算机研究与发展, 2025, 62(3): 765−778
Yuan Ziqi, Sun Qingyun, Zhou Haoyi, et al. MNDetector: Anomaly access detection method based on multiplex network[J]. Journal of Computer Research and Development, 2025, 62(3): 765−778 (in Chinese)
12
Ruff L, Vandermeulen R, Goernitz N, et al. Deep one class classification[C]//Proc of the 35th Int Conf on Machine Learning. New York: ACM, 2018: 4393−4402
13
Roy A, Sh Juan, Li Jia. GAD-NR: Graph anomaly detection via neighborhood reconstruction[C]//Proc of the 17th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2024: 576−585
14
Yang Xiao, Zhao Xuejiao, Shen Zhiqi. A generalizable anomaly detection method in dynamic graphs[C]//Proc of the 2025 AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025: 481−495
15
吴彦文, 谭溪晨, 葛迪, 等. 结合重构和图预测的多元时序异常检测框架[J]. 计算机工程与应用, 2024 , 60 (13): 301−310
Wu Yanwen, Tan Xichen, Ge Di, et al. Combining reconstruction and graph prediction for multi-dimensional time series anomaly detection framework[J]. Computer Engineering and Applications, 2024, 60 (13): 301−310 (in Chinese)
16
Liu Chen, He Shibo, Zhou Qihang, et al. Large language model guided knowledge distillation for time series anomaly detection[C]//Proc of the 33rd Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2024: 451−465
17
Jeong Y, Yang E, Ryu J H, et al. AnomalyBERT: Self-supervised transformer for time series anomaly detection using data degradation scheme[C/OL]//Proc of the 11th Int Conf on Learning Representations. 2023[2025-06-01]. https://arxiv.org/pdf/2305.04468v1.pdf
18
Jiao Yang, Yang Kai, Song Dongjing, et al. TimeAutoAD: Autonomous anomaly detection with self-supervised contrastive loss for multivariate time series[J]. IEEE Transactions on Network Science and Engineering, 2022, 9(3): 1604−1619
19
Kim H, Kim S, Min S, et al. Contrastive time-series anomaly detection[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(12): 1−14
20
Darban Z Z, Yang Yiyuan, Webb C I, et al. DACAD: Domain adaptation contrastive learning for anomaly detection in multivariate time series[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 37(8): 4485−4496
21
Yang Yiyuan, Zhang Chaoli, Zhou Tian, et al. DCdetector: Dual attention contrastive representation learning for time series anomaly detection[C]//Proc of the 29th ACM SIGKDD Conf on Knowledge Discovery and Data Mining. New York: ACM, 2023: 571−586
22
Zhong Zhijie, Yu Zhiwen, Yang Yiyuan, et al. PatchAD: A lightweight patch-based MLP mixer for time series anomaly detection[J]. IEEE Transactions on Big Data, 2025, 1(15): 234−245
23
Xie Yongzheng, Zhang Hongyu, Babar M A , et al. Multivariate time series anomaly detection by capturing coarse grained intra and inter variate dependencies[C]//Proc of the 2025 World Wide Web Conf. New York: ACM, 2025: 697−705
24
Fang Yuchen, Xie Jiadong, Chen Lu, et al. Temporal-frequency masked autoencoders for time series anomaly detection[C]//Proc of the 40th IEEE Int Conf on Data Engineering. Piscataway, NJ: IEEE, 2024: 158−172
25
Kim T, Kim J, Tae Y, et al. Reversible instance normalization for accurate time-series forecasting against distribution shift[C/OL]//Proc of the 10th Int Conf on Learning Representations. 2022[2025-06-01]. https://openreview.net/ forum?id =cGDAkQo1C0p
26
Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]//Proc of the 31st Int Conf on Neural Information Processing Systems. Cambridge, MA: MIT, 2017: 5998−6008
27
Xu Hongzuo, Wang Yijie, Pang Guangsong, et al. RoSAS: Deep semi-supervised anomaly detection with contamination-resilient continuous supervision[J]. Information Processing and Management, 2023, 60(5): 1059−1072
28
Zhang Xiang, Zhao Ziyuan, Tsiligkaridis T, et al. Self-supervised contrastive pre-training for time series via time-frequency consistency[C]//Proc of the 36th Conf on Neural Information Processing Systems. Cambridge, MA: MIT, 2022: 431−446
29
Peng Furong, Luo Jiachen, Lu Xuan, et al. Cross-domain contrastive learning for time series clustering[C]//Proc of the 38th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2024: 8921−8929
30
Chen Xinlei, He Kaiming. Exploring simple siamese representation learning[C]//Proc of the 2021 IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 15745−15757
31
Zhang Yufeng, Pan Jialu, Liu Wanwei, et al. Kullback-Leibler divergence-based out of distribution detection with flow based generative models[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 36(4): 1683−1697
32
Sarfraz M, Chen Meiyen, Layer L, et al. Position: Quo vadis, unsupervised time series anomaly detection?[C]//Proc of the 41st Int Conf on Machine Learning. New York: ACM, 2024: 187−196
33
John P, Paul B, Themis P, et al. Volume under the surface: A new accuracy evaluation measure for time-series anomaly detection[J]. Proceedings of the VLDB Endowment, 2022, 15(11): 2774−2787
34
Zong Bo, Song Qi, Min M R, et al. Deep autoencoding gaussian mixture model for unsupervised anomaly detection[C/OL]//Proc of the 6th Int Conf on Learning Representations. 2018[2025-06-01]. https://openreview.net/ forum? id=BJJLHbb0-
35
Tuli S, Casale G, Jennings N R, et al. TranAD: Deep transformer networks for anomaly detection in multivariate time series data[C]//Proc of the 48th Int Conf on Very Large Data Bases. New York: ACM, 2022: 338−351
36
Zhou Bin, Liu Shenghua, Hooi B, et al. BeatGAN: Anomalous rhythm detection using adversarially generated time series[C]//Proc of the 28th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2019: 4433−4439
2026年第63卷第3期
139
63
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202550364
  • 接收时间:2025-05-30
  • 首发时间:2026-11-06
  • 出版时间:2026-03-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-05-30
  • 修回日期:2025-10-21
  • 录用日期:2025-10-31
基金
作者信息
    1并行与分布计算全国重点实验室(国防科技大学) 长沙 410073
    2国防科技大学计算机学院 长沙 410073
    3浙江大学航空航天学院 杭州 310058

通讯作者:

王意洁()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550364
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏