Bradley Ezard , Ling Li , Hong Hao , Ruhua Wang , Senjian An
{"title":"Robust structural damage detection with deep multiple instance learning for sensor fault tolerance","authors":"Bradley Ezard , Ling Li , Hong Hao , Ruhua Wang , Senjian An","doi":"10.1016/j.engstruct.2025.119957","DOIUrl":null,"url":null,"abstract":"<div><div>Many structural health monitoring systems rely on signals collected from sensors to localise and quantify damage on a structure. In the last decade, many machine learning models have been proposed to detect structural damage. These models in general are trained by data generated from finite element analyses and are used for structural damage detection based on the data measured at the same degrees of freedom of the structure as those used to train the model. Sensor failure – where one or more sensors does not produce a usable signal – is a common and significant problem, especially under extreme conditions such as severe impact or natural disasters like cyclones and earthquakes, leading to the trained model not applicable for damage detection because of unavailability of data at some degrees of freedom. Despite this, few methods have been developed to address such a challenge. This paper proposes a deep learning approach which views structural damage identification as a case of multiple instance learning to address sensor failure. The new method is trained and evaluated on numerical simulations, followed by validation on an experimental case. The results of the studies show strong performance in accurately predicting structural damage with data from less number of sensors compared to those used in initial training of the model, even when more than half of the original sensors fail.</div></div>","PeriodicalId":11763,"journal":{"name":"Engineering Structures","volume":"331 ","pages":"Article 119957"},"PeriodicalIF":5.6000,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Engineering Structures","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0141029625003487","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, CIVIL","Score":null,"Total":0}
引用次数: 0
Abstract
Many structural health monitoring systems rely on signals collected from sensors to localise and quantify damage on a structure. In the last decade, many machine learning models have been proposed to detect structural damage. These models in general are trained by data generated from finite element analyses and are used for structural damage detection based on the data measured at the same degrees of freedom of the structure as those used to train the model. Sensor failure – where one or more sensors does not produce a usable signal – is a common and significant problem, especially under extreme conditions such as severe impact or natural disasters like cyclones and earthquakes, leading to the trained model not applicable for damage detection because of unavailability of data at some degrees of freedom. Despite this, few methods have been developed to address such a challenge. This paper proposes a deep learning approach which views structural damage identification as a case of multiple instance learning to address sensor failure. The new method is trained and evaluated on numerical simulations, followed by validation on an experimental case. The results of the studies show strong performance in accurately predicting structural damage with data from less number of sensors compared to those used in initial training of the model, even when more than half of the original sensors fail.
期刊介绍:
Engineering Structures provides a forum for a broad blend of scientific and technical papers to reflect the evolving needs of the structural engineering and structural mechanics communities. Particularly welcome are contributions dealing with applications of structural engineering and mechanics principles in all areas of technology. The journal aspires to a broad and integrated coverage of the effects of dynamic loadings and of the modelling techniques whereby the structural response to these loadings may be computed.
The scope of Engineering Structures encompasses, but is not restricted to, the following areas: infrastructure engineering; earthquake engineering; structure-fluid-soil interaction; wind engineering; fire engineering; blast engineering; structural reliability/stability; life assessment/integrity; structural health monitoring; multi-hazard engineering; structural dynamics; optimization; expert systems; experimental modelling; performance-based design; multiscale analysis; value engineering.
Topics of interest include: tall buildings; innovative structures; environmentally responsive structures; bridges; stadiums; commercial and public buildings; transmission towers; television and telecommunication masts; foldable structures; cooling towers; plates and shells; suspension structures; protective structures; smart structures; nuclear reactors; dams; pressure vessels; pipelines; tunnels.
Engineering Structures also publishes review articles, short communications and discussions, book reviews, and a diary on international events related to any aspect of structural engineering.