{"id":{"repo_id":"exeter","oai_identifier":"oai:figshare.com:article/31015171"},"canonical_url":"https://search.dev.ndltd.org/etd/exeter/oai:figshare.com:article/31015171","repository":{"repo_id":"exeter","name":"University of Exeter","base_url":"https://api.figshare.com/v2/oai"},"display":{"title":"Digital Twin for Machine Tools and Manufacturing Systems","abstract":"This doctoral research develops an integrated Digital Twin (DT) and Cyber- Physical System (CPS) framework for machine tools and manufacturing systems, addressing the challenges of model fidelity, knowledge extraction and cost- effective deployment in industrial environments. The study makes three key contributions. Firstly, it proposes a lightweight hybrid modelling approach that fuses finite element analysis with neural networks to predict multi-physics behaviours of machine components. Secondly, it presents a semantic text- analysis pipeline that transforms unstructured maintenance logs into interpretable fault categories. Thirdly, it designs and validates an edge-deployable predictive maintenance system that achieves real-time performance under stringent resource constraints. The framework is evaluated through three experimental cases. The first one is a virtual–physical dynamic modelling case for rotating shafts using FEM and NARX networks. The second one is an industrial text-mining case involving over 2000 lines of historical fault records from an automotive connecting-rod production line. And the third one is an edge-intelligence prototype for gearbox monitoring built on low-cost micro-controllers. Across the three cases, the results demonstrate high-fidelity dynamic prediction (error <5%), robust unsupervised fault clustering (silhouette score up to 0.9), and fast real-time response for edge sensing and actuation (0.3 s latency with 62% improvement over a PID baseline). 5 The digital twin that is developed and analysed in this research is an outcome of and extensive cycles of process that includes design, simulation and feedback- based refinement, the virtual model collaborates with its real-world counterpart for increased fidelity and robustness. Furthermore, the proposed DT framework enables a scalable deployment of its components from sensor-level data acquisition to higher-level of platform services in an incremental, stepwise manner of theoretical integration and methodology over specific quantitative analysis. In summary, the dissertation provides a formal and conceptually grounded overview of how integrating digital twins within CPS architectures, paired with an iterative development methodology, can advance intelligent manufacturing. However, several limitations remain, including the lack of thermal–mechanical coupling in the DT models, the reliance on enterprise- specific datasets for text analysis and the limited generalisability of the current edge prototype. These constraints offer clear opportunities for future work in multi-physics fusion, cross-factory validation, and scalable cloud–edge intergrations.<p></p>","abstract_html":"This doctoral research develops an integrated Digital Twin (DT) and Cyber- Physical System (CPS) framework for machine tools and manufacturing systems, addressing the challenges of model fidelity, knowledge extraction and cost- effective deployment in industrial environments. The study makes three key contributions. Firstly, it proposes a lightweight hybrid modelling approach that fuses finite element analysis with neural networks to predict multi-physics behaviours of machine components. Secondly, it presents a semantic text- analysis pipeline that transforms unstructured maintenance logs into interpretable fault categories. Thirdly, it designs and validates an edge-deployable predictive maintenance system that achieves real-time performance under stringent resource constraints. The framework is evaluated through three experimental cases. The first one is a virtual–physical dynamic modelling case for rotating shafts using FEM and NARX networks. The second one is an industrial text-mining case involving over 2000 lines of historical fault records from an automotive connecting-rod production line. And the third one is an edge-intelligence prototype for gearbox monitoring built on low-cost micro-controllers. Across the three cases, the results demonstrate high-fidelity dynamic prediction (error &lt;5%), robust unsupervised fault clustering (silhouette score up to 0.9), and fast real-time response for edge sensing and actuation (0.3 s latency with 62% improvement over a PID baseline). 5 The digital twin that is developed and analysed in this research is an outcome of and extensive cycles of process that includes design, simulation and feedback- based refinement, the virtual model collaborates with its real-world counterpart for increased fidelity and robustness. Furthermore, the proposed DT framework enables a scalable deployment of its components from sensor-level data acquisition to higher-level of platform services in an incremental, stepwise manner of theoretical integration and methodology over specific quantitative analysis. In summary, the dissertation provides a formal and conceptually grounded overview of how integrating digital twins within CPS architectures, paired with an iterative development methodology, can advance intelligent manufacturing. However, several limitations remain, including the lack of thermal–mechanical coupling in the DT models, the reliance on enterprise- specific datasets for text analysis and the limited generalisability of the current edge prototype. These constraints offer clear opportunities for future work in multi-physics fusion, cross-factory validation, and scalable cloud–edge intergrations.&lt;p&gt;&lt;/p&gt;","abstract_has_math":false,"creators":["Albert Li (21041474)"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-01-05T00:00:00Z","date_published":"2026-01-05T00:00:00Z","updated_at":"2026-07-27T19:34:56Z","subjects":["Digital Twin","Machine Tools","MES"],"languages":[],"rights":["All rights reserved","Open Access after 2027-07-05"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["10779/exe.31015171.v1"],"render_values":[{"text":"10779/exe.31015171.v1","href":null,"code":true}]}]},"links":{"outbound_url":null,"outbound_label":null,"outbound_source":null},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Albert Li (21041474)"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026-01-05T00:00:00Z"]},{"key":"dc:relation","label":"Dc Relation","values":["https://figshare.com/articles/thesis/Digital_Twin_for_Machine_Tools_and_Manufacturing_Systems/31015171"]},{"key":"dc:type","label":"Dc Type","values":["Text","Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Digital Twin","Machine Tools","MES"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["All rights reserved","Open Access after 2027-07-05"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["10779/exe.31015171.v1"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This doctoral research develops an integrated Digital Twin (DT) and Cyber- Physical System (CPS) framework for machine tools and manufacturing systems, addressing the challenges of model fidelity, knowledge extraction and cost- effective deployment in industrial environments. The study makes three key contributions. Firstly, it proposes a lightweight hybrid modelling approach that fuses finite element analysis with neural networks to predict multi-physics behaviours of machine components. Secondly, it presents a semantic text- analysis pipeline that transforms unstructured maintenance logs into interpretable fault categories. Thirdly, it designs and validates an edge-deployable predictive maintenance system that achieves real-time performance under stringent resource constraints. The framework is evaluated through three experimental cases. The first one is a virtual–physical dynamic modelling case for rotating shafts using FEM and NARX networks. The second one is an industrial text-mining case involving over 2000 lines of historical fault records from an automotive connecting-rod production line. And the third one is an edge-intelligence prototype for gearbox monitoring built on low-cost micro-controllers. Across the three cases, the results demonstrate high-fidelity dynamic prediction (error <5%), robust unsupervised fault clustering (silhouette score up to 0.9), and fast real-time response for edge sensing and actuation (0.3 s latency with 62% improvement over a PID baseline). 5 The digital twin that is developed and analysed in this research is an outcome of and extensive cycles of process that includes design, simulation and feedback- based refinement, the virtual model collaborates with its real-world counterpart for increased fidelity and robustness. Furthermore, the proposed DT framework enables a scalable deployment of its components from sensor-level data acquisition to higher-level of platform services in an incremental, stepwise manner of theoretical integration and methodology over specific quantitative analysis. In summary, the dissertation provides a formal and conceptually grounded overview of how integrating digital twins within CPS architectures, paired with an iterative development methodology, can advance intelligent manufacturing. However, several limitations remain, including the lack of thermal–mechanical coupling in the DT models, the reliance on enterprise- specific datasets for text analysis and the limited generalisability of the current edge prototype. These constraints offer clear opportunities for future work in multi-physics fusion, cross-factory validation, and scalable cloud–edge intergrations.<p></p>"]},{"key":"dc:title","label":"Title","values":["Digital Twin for Machine Tools and Manufacturing Systems"]}]}],"canonical_facts":{"dc:creator":["Albert Li (21041474)"],"dc:date":["2026-01-05T00:00:00Z"],"dc:description":["This doctoral research develops an integrated Digital Twin (DT) and Cyber- Physical System (CPS) framework for machine tools and manufacturing systems, addressing the challenges of model fidelity, knowledge extraction and cost- effective deployment in industrial environments. The study makes three key contributions. Firstly, it proposes a lightweight hybrid modelling approach that fuses finite element analysis with neural networks to predict multi-physics behaviours of machine components. Secondly, it presents a semantic text- analysis pipeline that transforms unstructured maintenance logs into interpretable fault categories. Thirdly, it designs and validates an edge-deployable predictive maintenance system that achieves real-time performance under stringent resource constraints. The framework is evaluated through three experimental cases. The first one is a virtual–physical dynamic modelling case for rotating shafts using FEM and NARX networks. The second one is an industrial text-mining case involving over 2000 lines of historical fault records from an automotive connecting-rod production line. And the third one is an edge-intelligence prototype for gearbox monitoring built on low-cost micro-controllers. Across the three cases, the results demonstrate high-fidelity dynamic prediction (error <5%), robust unsupervised fault clustering (silhouette score up to 0.9), and fast real-time response for edge sensing and actuation (0.3 s latency with 62% improvement over a PID baseline). 5 The digital twin that is developed and analysed in this research is an outcome of and extensive cycles of process that includes design, simulation and feedback- based refinement, the virtual model collaborates with its real-world counterpart for increased fidelity and robustness. Furthermore, the proposed DT framework enables a scalable deployment of its components from sensor-level data acquisition to higher-level of platform services in an incremental, stepwise manner of theoretical integration and methodology over specific quantitative analysis. In summary, the dissertation provides a formal and conceptually grounded overview of how integrating digital twins within CPS architectures, paired with an iterative development methodology, can advance intelligent manufacturing. However, several limitations remain, including the lack of thermal–mechanical coupling in the DT models, the reliance on enterprise- specific datasets for text analysis and the limited generalisability of the current edge prototype. These constraints offer clear opportunities for future work in multi-physics fusion, cross-factory validation, and scalable cloud–edge intergrations.<p></p>"],"dc:identifier":["10779/exe.31015171.v1"],"dc:relation":["https://figshare.com/articles/thesis/Digital_Twin_for_Machine_Tools_and_Manufacturing_Systems/31015171"],"dc:rights":["All rights reserved","Open Access after 2027-07-05"],"dc:subject":["Digital Twin","Machine Tools","MES"],"dc:title":["Digital Twin for Machine Tools and Manufacturing Systems"],"dc:type":["Text","Thesis"]},"updated_at":"2026-07-27T19:34:56Z"}