Physics-Informed, Cost-Sensitive Fault Classification with Comparative Explainability for Predictive Maintenance: A SHAP–LIME Agreement Analysis

Physics-Informed, Cost-Sensitive Fault Classification with Comparative Explainability for Predictive Maintenance: A SHAP–LIME Agreement Analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Physics-Informed, Cost-Sensitive Fault Classification with Comparative Explainability for Predictive Maintenance: A SHAP–LIME Agreement Analysis Okoi Michael Obeten, Ahmed Jimoh This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9587760/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Predictive maintenance (PdM) is a cornerstone capability of Industry 4.0, enabling industrial systems to anticipate equipment failures before costly unplanned downtime occurs. Despite the widely use of machine learning approaches applied to the AI4I 2020 dataset, three research gaps persist. No prior study has exploited the dataset's causal structure to engineer physics-informed features encoding mechanical power, overstrain, and thermal dissipation failure mechanisms. Multi-class fault classification on AI4I has relied on accuracy-optimised objectives that treat all misclassification errors equally, ignoring the different operational severities across fault types. Furthermore, no study has quantitatively measured SHAP and LIME agreement using Spearman rank correlation, leaving practitioners without evidence-based guidance on which XAI method to trust per fault type. This paper addresses all three gaps. A physics-informed pipeline derives five causal interaction features from documented failure thresholds. A \((6\times6)\) fault-severity cost matrix provides class weights for cost-sensitive training of four classifiers: Logistic Regression, Support Vector Machine, Random Forest, and XGBoost, evaluated under stratified 5-fold cross-validation. Random Forest achieves a weighted F1-score of 0.988 and a total misclassification cost of 119, representing a 1.7% reduction over standard training. SHAP analysis reveals that Mechanical_Power_W and Overstrain_Index rank first and second globally, contributing over 60% of total feature importance. LIME explanations corroborate these rankings for thermally-driven faults but diverge for power and overstrain faults. Spearman rank correlation across four fault classes yields a mean \((\rho)\) of 0.641, indicating predominantly weak agreement reflecting the non-linear decision boundaries of tree-based models. SHAP should be preferred over LIME for non-linear fault classification in industrial PdM systems. Physical sciences/Engineering Physical sciences/Mathematics and computing Predictive Maintenance Explainable Artificial Intelligence SHAP LIME Fault Classification Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 13 May, 2026 Reviews received at journal 10 May, 2026 Reviewers agreed at journal 08 May, 2026 Reviewers invited by journal 08 May, 2026 Editor assigned by journal 08 May, 2026 Editor invited by journal 07 May, 2026 Submission checks completed at journal 06 May, 2026 First submitted to journal 06 May, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9587760","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":639907499,"identity":"44229bfb-79d1-42f4-bdfc-38161e7d340a","order_by":0,"name":"Okoi Michael Obeten","email":"data:image/png;base64,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","orcid":"","institution":"Auchi Polytechnic","correspondingAuthor":true,"prefix":"","firstName":"Okoi","middleName":"Michael","lastName":"Obeten","suffix":""},{"id":639907500,"identity":"97525317-9146-4f6c-82d2-b301a71bac27","order_by":1,"name":"Ahmed Jimoh","email":"","orcid":"","institution":"Auchi Polytechnic","correspondingAuthor":false,"prefix":"","firstName":"Ahmed","middleName":"","lastName":"Jimoh","suffix":""}],"badges":[],"createdAt":"2026-05-01 16:08:47","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9587760/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9587760/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109760505,"identity":"e145dfc1-5485-476f-8447-bf9ab494e5ea","added_by":"auto","created_at":"2026-05-22 07:28:46","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1317812,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript02.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9587760/v1_covered_104811f4-74c8-43f9-8422-14659ce7df30.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Physics-Informed, Cost-Sensitive Fault Classification with Comparative Explainability for Predictive Maintenance: A SHAP–LIME Agreement Analysis","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Predictive Maintenance, Explainable Artificial Intelligence, SHAP, LIME, Fault Classification","lastPublishedDoi":"10.21203/rs.3.rs-9587760/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9587760/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePredictive maintenance (PdM) is a cornerstone capability of Industry 4.0, enabling industrial systems to anticipate equipment failures before costly unplanned downtime occurs. Despite the widely use of machine learning approaches applied to the AI4I 2020 dataset, three research gaps persist. No prior study has exploited the dataset's causal structure to engineer physics-informed features encoding mechanical power, overstrain, and thermal dissipation failure mechanisms. Multi-class fault classification on AI4I has relied on accuracy-optimised objectives that treat all misclassification errors equally, ignoring the different operational severities across fault types. Furthermore, no study has quantitatively measured SHAP and LIME agreement using Spearman rank correlation, leaving practitioners without evidence-based guidance on which XAI method to trust per fault type. This paper addresses all three gaps. A physics-informed pipeline derives five causal interaction features from documented failure thresholds. A \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\((6\\times6)\\)\u003c/span\u003e\u003c/span\u003e fault-severity cost matrix provides class weights for cost-sensitive training of four classifiers: Logistic Regression, Support Vector Machine, Random Forest, and XGBoost, evaluated under stratified 5-fold cross-validation. Random Forest achieves a weighted F1-score of 0.988 and a total misclassification cost of 119, representing a 1.7% reduction over standard training. SHAP analysis reveals that Mechanical_Power_W and Overstrain_Index rank first and second globally, contributing over 60% of total feature importance. LIME explanations corroborate these rankings for thermally-driven faults but diverge for power and overstrain faults. Spearman rank correlation across four fault classes yields a mean \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\((\\rho)\\)\u003c/span\u003e\u003c/span\u003e of 0.641, indicating predominantly weak agreement reflecting the non-linear decision boundaries of tree-based models. SHAP should be preferred over LIME for non-linear fault classification in industrial PdM systems.\u003c/p\u003e","manuscriptTitle":"Physics-Informed, Cost-Sensitive Fault Classification with Comparative Explainability for Predictive Maintenance: A SHAP–LIME Agreement Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-18 22:56:30","doi":"10.21203/rs.3.rs-9587760/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"80861251767067136848842509033080796280","date":"2026-05-13T18:27:44+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-10T11:16:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"81555605953886710117636122294278470011","date":"2026-05-08T14:58:58+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-08T13:43:46+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-08T13:30:57+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-05-07T13:17:49+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-05-06T22:53:09+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-05-06T16:24:18+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"419908d2-eba8-4f3f-9d2e-b525e388aff2","owner":[],"postedDate":"May 18th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"80861251767067136848842509033080796280","date":"2026-05-13T18:27:44+00:00","index":35,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-10T11:16:15+00:00","index":32,"fulltext":""},{"type":"reviewerAgreed","content":"81555605953886710117636122294278470011","date":"2026-05-08T14:58:58+00:00","index":31,"fulltext":""},{"type":"reviewersInvited","content":"6","date":"2026-05-08T13:43:46+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-08T13:30:57+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-05-07T13:17:49+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-05-06T22:53:09+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-05-06T16:24:18+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":68110034,"name":"Physical sciences/Engineering"},{"id":68110035,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-05-18T22:56:30+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-18 22:56:30","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9587760","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9587760","identity":"rs-9587760","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}