{"id":{"repo_id":"tdl","oai_identifier":"oai:tdl-ir.tdl.org:2104/13778"},"canonical_url":"https://search.dev.ndltd.org/etd/tdl/oai:tdl-ir.tdl.org:2104/13778","repository":{"repo_id":"tdl","name":"Texas Digital Library","base_url":"https://tdl-ir.tdl.org/server/oai/request"},"display":{"title":"Fault detection in multivariate processes : handling autocorrelation, contamination, and small sample sizes in engineered systems.","abstract":"Multivariate statistical process monitoring is commonly used to detect faults, or unexpected deviations from normal operating behavior, in complex processes. Faults are typically identified by comparing process behavior in real-time to previously observed in-control process behavior. Many existing methods perform well when monitoring multivariate processes, but they typically require that certain assumptions are met, including multivariate normality and independence. Furthermore, the availability of a large set of in-control historical data is also usually assumed, but often unstated. However, in some cases, minimal historical data are available, and the assumptions of multivariate normality and/or independence may be violated. Furthermore, even when a large set of historical data is available, it may be contaminated with outliers, making it difficult to mathematically characterize in-control behavior. In this work, we explore methods to account for small sample sizes, violations of independence, and contaminated data. In particular, we propose and evaluate non-parametric methods to identify thresholds separating in-control from out-of-control behavior that account for nonnormality and dependence. An extensive literature review covering approaches for handling small sample sizes and contamination is presented, including examples illustrating the shortcomings and trade-offs of existing methods, and finally, a novel Bayesian approach using robust estimators is proposed to handle situations in which historical data are both limited and contaminated with outliers.","abstract_html":"Multivariate statistical process monitoring is commonly used to detect faults, or unexpected deviations from normal operating behavior, in complex processes. Faults are typically identified by comparing process behavior in real-time to previously observed in-control process behavior. Many existing methods perform well when monitoring multivariate processes, but they typically require that certain assumptions are met, including multivariate normality and independence. Furthermore, the availability of a large set of in-control historical data is also usually assumed, but often unstated. However, in some cases, minimal historical data are available, and the assumptions of multivariate normality and/or independence may be violated. Furthermore, even when a large set of historical data is available, it may be contaminated with outliers, making it difficult to mathematically characterize in-control behavior. In this work, we explore methods to account for small sample sizes, violations of independence, and contaminated data. In particular, we propose and evaluate non-parametric methods to identify thresholds separating in-control from out-of-control behavior that account for nonnormality and dependence. An extensive literature review covering approaches for handling small sample sizes and contamination is presented, including examples illustrating the shortcomings and trade-offs of existing methods, and finally, a novel Bayesian approach using robust estimators is proposed to handle situations in which historical data are both limited and contaminated with outliers.","abstract_has_math":false,"creators":["Grimm, Taylor R., 1998-"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Hering, Amanda S.","Newhart, Kathryn B."],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-05","date_published":"2025-05","updated_at":"2026-07-27T21:19:19Z","subjects":["Fault detection.","Statistical process monitoring.","Robust methods.","Multivariate statistics."],"languages":["en"],"rights":["No access - Contact librarywebmaster@baylor.edu","Baylor University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. Contact libraryquestions@baylor.edu for inquiries about permission."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/2104/13778","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Hering, Amanda S.","Newhart, Kathryn B."]},{"key":"dc:creator","label":"Author","values":["Grimm, Taylor R., 1998-"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-02-10T23:07:24Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-05"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Fault detection.","Statistical process monitoring.","Robust methods.","Multivariate statistics."]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["No access - Contact librarywebmaster@baylor.edu","Baylor University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. Contact libraryquestions@baylor.edu for inquiries about permission."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://hdl.handle.net/2104/13778"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/2104/13778"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Multivariate statistical process monitoring is commonly used to detect faults, or unexpected deviations from normal operating behavior, in complex processes. Faults are typically identified by comparing process behavior in real-time to previously observed in-control process behavior. Many existing methods perform well when monitoring multivariate processes, but they typically require that certain assumptions are met, including multivariate normality and independence. Furthermore, the availability of a large set of in-control historical data is also usually assumed, but often unstated. However, in some cases, minimal historical data are available, and the assumptions of multivariate normality and/or independence may be violated. Furthermore, even when a large set of historical data is available, it may be contaminated with outliers, making it difficult to mathematically characterize in-control behavior. In this work, we explore methods to account for small sample sizes, violations of independence, and contaminated data. In particular, we propose and evaluate non-parametric methods to identify thresholds separating in-control from out-of-control behavior that account for nonnormality and dependence. An extensive literature review covering approaches for handling small sample sizes and contamination is presented, including examples illustrating the shortcomings and trade-offs of existing methods, and finally, a novel Bayesian approach using robust estimators is proposed to handle situations in which historical data are both limited and contaminated with outliers."]},{"key":"dc:title","label":"Title","values":["Fault detection in multivariate processes : handling autocorrelation, contamination, and small sample sizes in engineered systems."]}]}],"canonical_facts":{"dc:contributor":["Hering, Amanda S.","Newhart, Kathryn B."],"dc:creator":["Grimm, Taylor R., 1998-"],"dc:date.accessioned":["2026-02-10T23:07:24Z"],"dc:date.issued":["2025-05"],"dc:description.abstract":["Multivariate statistical process monitoring is commonly used to detect faults, or unexpected deviations from normal operating behavior, in complex processes. Faults are typically identified by comparing process behavior in real-time to previously observed in-control process behavior. Many existing methods perform well when monitoring multivariate processes, but they typically require that certain assumptions are met, including multivariate normality and independence. Furthermore, the availability of a large set of in-control historical data is also usually assumed, but often unstated. However, in some cases, minimal historical data are available, and the assumptions of multivariate normality and/or independence may be violated. Furthermore, even when a large set of historical data is available, it may be contaminated with outliers, making it difficult to mathematically characterize in-control behavior. In this work, we explore methods to account for small sample sizes, violations of independence, and contaminated data. In particular, we propose and evaluate non-parametric methods to identify thresholds separating in-control from out-of-control behavior that account for nonnormality and dependence. An extensive literature review covering approaches for handling small sample sizes and contamination is presented, including examples illustrating the shortcomings and trade-offs of existing methods, and finally, a novel Bayesian approach using robust estimators is proposed to handle situations in which historical data are both limited and contaminated with outliers."],"dc:identifier":["https://hdl.handle.net/2104/13778"],"dc:identifier.uri":["https://hdl.handle.net/2104/13778"],"dc:language":["en"],"dc:rights":["No access - Contact librarywebmaster@baylor.edu","Baylor University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. Contact libraryquestions@baylor.edu for inquiries about permission."],"dc:subject":["Fault detection.","Statistical process monitoring.","Robust methods.","Multivariate statistics."],"dc:title":["Fault detection in multivariate processes : handling autocorrelation, contamination, and small sample sizes in engineered systems."],"dc:type":["Thesis"]},"updated_at":"2026-07-27T21:19:19Z"}