Ashit Gupta, V. Masampally, Vishal Jadhav, A. Deodhar, V. Runkana
{"title":"Supervised Operational Change Point Detection using Ensemble Long-Short Term Memory in a Multicomponent Industrial System","authors":"Ashit Gupta, V. Masampally, Vishal Jadhav, A. Deodhar, V. Runkana","doi":"10.1109/SAMI50585.2021.9378683","DOIUrl":null,"url":null,"abstract":"Changes in operating conditions, environment, and deterioration of structural health of components over time leads to unplanned outages in industrial equipment. A multicomponent industrial system may fail when one or more of its components deteriorate beyond a certain limit. The deterioration is often a gradual and continuous process, culminating in sudden failure of an equipment. However, the components in a system may show some early signs of deterioration that might not be explicitly apparent even to domain experts. Therefore, advanced algorithms are required for early detection of these signatures of failure to enable corrective actions in time. A set of algorithms is presented here to detect signatures of failure from the continuous sensor data in a multicomponent system. Each system consists of four identical components, each with a different timing of failure. A set of Long Short-Term Memory (LSTM) based algorithms are employed to identify the onset of abnormal behavior. An ensemble framework, which minimizes the frequency of false and missed alarms is proposed and its performance is compared with other stand-alone algorithms. An ensemble approach on top of a set of LSTM-based models performed better than the individual algorithms.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"4","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/SAMI50585.2021.9378683","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 4
Abstract
Changes in operating conditions, environment, and deterioration of structural health of components over time leads to unplanned outages in industrial equipment. A multicomponent industrial system may fail when one or more of its components deteriorate beyond a certain limit. The deterioration is often a gradual and continuous process, culminating in sudden failure of an equipment. However, the components in a system may show some early signs of deterioration that might not be explicitly apparent even to domain experts. Therefore, advanced algorithms are required for early detection of these signatures of failure to enable corrective actions in time. A set of algorithms is presented here to detect signatures of failure from the continuous sensor data in a multicomponent system. Each system consists of four identical components, each with a different timing of failure. A set of Long Short-Term Memory (LSTM) based algorithms are employed to identify the onset of abnormal behavior. An ensemble framework, which minimizes the frequency of false and missed alarms is proposed and its performance is compared with other stand-alone algorithms. An ensemble approach on top of a set of LSTM-based models performed better than the individual algorithms.