E-Thesis 56 views
Digital twinning of thermal problems in fusion energy systems / WIERA BIELAJEWA
Swansea University Author: WIERA BIELAJEWA
DOI (Published version): 10.23889/SUThesis.71619
Abstract
Digital Twinning (DT) technology is in the process of becoming an essential instrument for optimising efficiency, ensuring safety, and enhancing research productivity of the fusion energy experimental facilities. A digital twin represents a virtual counterpart of a physical system that maintains real-...
| Published: |
Swansea
2026
|
|---|---|
| Institution: | Swansea University |
| Degree level: | Doctoral |
| Degree name: | Ph.D |
| Supervisor: | Nithiarasu, P., and Baxter, M. |
| URI: | https://cronfa.swan.ac.uk/Record/cronfa71619 |
| first_indexed |
2026-03-12T14:15:00Z |
|---|---|
| last_indexed |
2026-03-13T05:25:17Z |
| id |
cronfa71619 |
| recordtype |
RisThesis |
| fullrecord |
<?xml version="1.0"?><rfc1807><datestamp>2026-03-12T14:24:29.9691039</datestamp><bib-version>v2</bib-version><id>71619</id><entry>2026-03-12</entry><title>Digital twinning of thermal problems in fusion energy systems</title><swanseaauthors><author><sid>1b4f8744a63a452048bfa72e7b14d92c</sid><firstname>WIERA</firstname><surname>BIELAJEWA</surname><name>WIERA BIELAJEWA</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2026-03-12</date><abstract>Digital Twinning (DT) technology is in the process of becoming an essential instrument for optimising efficiency, ensuring safety, and enhancing research productivity of the fusion energy experimental facilities. A digital twin represents a virtual counterpart of a physical system that maintains real-time synchronisation through the sensor-derived data. Within fusion energy experimental facilities, DT technology enables enhanced information extraction from the experimental data and facilitates optimal control. This work presents two distinct methodologies facilitating DT control of a sample.The first DT approach, Finite Element-based Digital Twinning (FE DT) is a dual-component control system comprising of a full-solution construction from limited data and a control mechanism. The solution reconstruction operates in near-real-time for small scale problems through a novel Finite Element (FE)-based data integration approach that transforms sparse measurements into comprehensive solutions for non-linear systems. This method integrates FE discretisation with a loss function minimisation to generate complete solutions from limited measurement data. Jacobian matrices are computed analytically rather than through Automatic Differentiation (AD). The control component consists of a digital (discrete) Proportional–Integral–Derivative (PID)controller that regulates the cooling water temperature of the test specimens. The modification of the solution construction approach involves its coupling with the thermal eigenvalue-based Reduced Order Modelling (ROM). This allows for the speed-up of the solution construction process whilst keeping the overall procedure the same.The second DT approach, Physics-Driven Machine Learning-based Digital Twinning (PD-ML DT), is rooted in Machine Learning (ML). It involves training two Neural Network (NN) using steady-state data. A two-part NN system is employed, forming a control loop. The first part, the heat flux NN, addresses the challenge of estimating thermal conditions. Its aim is to construct a steady-state equivalent of the heat flux based on the provided temperature measurements. The second part, the coolant NN, provides a solution for active thermal management. In this work, it determines the necessary coolant velocity to maintain the maximum sample temperature below a specified threshold. The additional adjustments are made to ensure the system’s effectiveness and robustness when dealing with dynamic system. These include extrapolating the constructed heat flux and selecting the maximum value from a range of past and extrapolated data points. These adjustments are critical as they account for the fact that both NNs are trained on steady-state data, allowing them to operate effectively in non-steady-state conditions. This combined approach enables real-time temperature monitoring and control addressing a challenge in thermal system management.The performance of two DT systems is demonstrated through the cooling control analysis of samples which were previously evaluated in a fusion energy experimental facility.Various system responses are generated under their control, and they are compared based on the accuracy, speed, and the resistance to measurement noise. The results show that they display advantages and disadvantages based on the information which could be provided during the experiment. Primarily, FE DT requires more measurements than PD-ML DT in the presence of noise; however, it has the ability to provide more information compared with PD-ML DT in the form of the full temperature field. As opposed with PD-ML DT, FE DT does not have instantaneous inference time; however, by employing ROM as part of the workflow it could be reduced.The reported development of DT process for thermal systems, powered by FE combined with PID controller or a dual NN control loop, holds significant implications for the future of fusion energy research and beyond. By enabling real-time monitoring and active thermal management, it provides a crucial tool for optimising the performance and ensuring the safety of experimental fusion facilities. This approach moves the progress closer to autonomous, self-correcting systems that can operate with greater efficiency and reliability. The methodology’s application extends beyond fusion, offering a template for managing thermal systems in a wide range of engineering fields where sparse measurements are involved.</abstract><type>E-Thesis</type><journal/><volume/><journalNumber/><paginationStart/><paginationEnd/><publisher/><placeOfPublication>Swansea</placeOfPublication><isbnPrint/><isbnElectronic/><issnPrint/><issnElectronic/><keywords>digital twinning, sparse data, finite element method, machine learning, nuclear fusion</keywords><publishedDay>17</publishedDay><publishedMonth>2</publishedMonth><publishedYear>2026</publishedYear><publishedDate>2026-02-17</publishedDate><doi>10.23889/SUThesis.71619</doi><url/><notes/><college>COLLEGE NANME</college><CollegeCode>COLLEGE CODE</CollegeCode><institution>Swansea University</institution><supervisor>Nithiarasu, P., and Baxter, M.</supervisor><degreelevel>Doctoral</degreelevel><degreename>Ph.D</degreename><degreesponsorsfunders>EPSRC Energy Programme</degreesponsorsfunders><apcterm/><funders>EPSRC Energy Programme</funders><projectreference/><lastEdited>2026-03-12T14:24:29.9691039</lastEdited><Created>2026-03-12T13:57:12.5230566</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Mechanical Engineering</level></path><authors><author><firstname>WIERA</firstname><surname>BIELAJEWA</surname><order>1</order></author></authors><documents><document><filename>Under embargo</filename><originalFilename>Under embargo</originalFilename><uploaded>2026-03-12T14:18:56.2313092</uploaded><type>Output</type><contentLength>23849826</contentLength><contentType>application/pdf</contentType><version>E-Thesis</version><cronfaStatus>true</cronfaStatus><embargoDate>2029-02-17T00:00:00.0000000</embargoDate><documentNotes>Copyright: the author, Wiera Bielajewa, 2026</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language></document></documents><OutputDurs/></rfc1807> |
| spelling |
2026-03-12T14:24:29.9691039 v2 71619 2026-03-12 Digital twinning of thermal problems in fusion energy systems 1b4f8744a63a452048bfa72e7b14d92c WIERA BIELAJEWA WIERA BIELAJEWA true false 2026-03-12 Digital Twinning (DT) technology is in the process of becoming an essential instrument for optimising efficiency, ensuring safety, and enhancing research productivity of the fusion energy experimental facilities. A digital twin represents a virtual counterpart of a physical system that maintains real-time synchronisation through the sensor-derived data. Within fusion energy experimental facilities, DT technology enables enhanced information extraction from the experimental data and facilitates optimal control. This work presents two distinct methodologies facilitating DT control of a sample.The first DT approach, Finite Element-based Digital Twinning (FE DT) is a dual-component control system comprising of a full-solution construction from limited data and a control mechanism. The solution reconstruction operates in near-real-time for small scale problems through a novel Finite Element (FE)-based data integration approach that transforms sparse measurements into comprehensive solutions for non-linear systems. This method integrates FE discretisation with a loss function minimisation to generate complete solutions from limited measurement data. Jacobian matrices are computed analytically rather than through Automatic Differentiation (AD). The control component consists of a digital (discrete) Proportional–Integral–Derivative (PID)controller that regulates the cooling water temperature of the test specimens. The modification of the solution construction approach involves its coupling with the thermal eigenvalue-based Reduced Order Modelling (ROM). This allows for the speed-up of the solution construction process whilst keeping the overall procedure the same.The second DT approach, Physics-Driven Machine Learning-based Digital Twinning (PD-ML DT), is rooted in Machine Learning (ML). It involves training two Neural Network (NN) using steady-state data. A two-part NN system is employed, forming a control loop. The first part, the heat flux NN, addresses the challenge of estimating thermal conditions. Its aim is to construct a steady-state equivalent of the heat flux based on the provided temperature measurements. The second part, the coolant NN, provides a solution for active thermal management. In this work, it determines the necessary coolant velocity to maintain the maximum sample temperature below a specified threshold. The additional adjustments are made to ensure the system’s effectiveness and robustness when dealing with dynamic system. These include extrapolating the constructed heat flux and selecting the maximum value from a range of past and extrapolated data points. These adjustments are critical as they account for the fact that both NNs are trained on steady-state data, allowing them to operate effectively in non-steady-state conditions. This combined approach enables real-time temperature monitoring and control addressing a challenge in thermal system management.The performance of two DT systems is demonstrated through the cooling control analysis of samples which were previously evaluated in a fusion energy experimental facility.Various system responses are generated under their control, and they are compared based on the accuracy, speed, and the resistance to measurement noise. The results show that they display advantages and disadvantages based on the information which could be provided during the experiment. Primarily, FE DT requires more measurements than PD-ML DT in the presence of noise; however, it has the ability to provide more information compared with PD-ML DT in the form of the full temperature field. As opposed with PD-ML DT, FE DT does not have instantaneous inference time; however, by employing ROM as part of the workflow it could be reduced.The reported development of DT process for thermal systems, powered by FE combined with PID controller or a dual NN control loop, holds significant implications for the future of fusion energy research and beyond. By enabling real-time monitoring and active thermal management, it provides a crucial tool for optimising the performance and ensuring the safety of experimental fusion facilities. This approach moves the progress closer to autonomous, self-correcting systems that can operate with greater efficiency and reliability. The methodology’s application extends beyond fusion, offering a template for managing thermal systems in a wide range of engineering fields where sparse measurements are involved. E-Thesis Swansea digital twinning, sparse data, finite element method, machine learning, nuclear fusion 17 2 2026 2026-02-17 10.23889/SUThesis.71619 COLLEGE NANME COLLEGE CODE Swansea University Nithiarasu, P., and Baxter, M. Doctoral Ph.D EPSRC Energy Programme EPSRC Energy Programme 2026-03-12T14:24:29.9691039 2026-03-12T13:57:12.5230566 Faculty of Science and Engineering School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Mechanical Engineering WIERA BIELAJEWA 1 Under embargo Under embargo 2026-03-12T14:18:56.2313092 Output 23849826 application/pdf E-Thesis true 2029-02-17T00:00:00.0000000 Copyright: the author, Wiera Bielajewa, 2026 true eng |
| title |
Digital twinning of thermal problems in fusion energy systems |
| spellingShingle |
Digital twinning of thermal problems in fusion energy systems WIERA BIELAJEWA |
| title_short |
Digital twinning of thermal problems in fusion energy systems |
| title_full |
Digital twinning of thermal problems in fusion energy systems |
| title_fullStr |
Digital twinning of thermal problems in fusion energy systems |
| title_full_unstemmed |
Digital twinning of thermal problems in fusion energy systems |
| title_sort |
Digital twinning of thermal problems in fusion energy systems |
| author_id_str_mv |
1b4f8744a63a452048bfa72e7b14d92c |
| author_id_fullname_str_mv |
1b4f8744a63a452048bfa72e7b14d92c_***_WIERA BIELAJEWA |
| author |
WIERA BIELAJEWA |
| author2 |
WIERA BIELAJEWA |
| format |
E-Thesis |
| publishDate |
2026 |
| institution |
Swansea University |
| doi_str_mv |
10.23889/SUThesis.71619 |
| college_str |
Faculty of Science and Engineering |
| hierarchytype |
|
| hierarchy_top_id |
facultyofscienceandengineering |
| hierarchy_top_title |
Faculty of Science and Engineering |
| hierarchy_parent_id |
facultyofscienceandengineering |
| hierarchy_parent_title |
Faculty of Science and Engineering |
| department_str |
School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Mechanical Engineering{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Aerospace, Civil, Electrical, General and Mechanical Engineering - Mechanical Engineering |
| document_store_str |
0 |
| active_str |
0 |
| description |
Digital Twinning (DT) technology is in the process of becoming an essential instrument for optimising efficiency, ensuring safety, and enhancing research productivity of the fusion energy experimental facilities. A digital twin represents a virtual counterpart of a physical system that maintains real-time synchronisation through the sensor-derived data. Within fusion energy experimental facilities, DT technology enables enhanced information extraction from the experimental data and facilitates optimal control. This work presents two distinct methodologies facilitating DT control of a sample.The first DT approach, Finite Element-based Digital Twinning (FE DT) is a dual-component control system comprising of a full-solution construction from limited data and a control mechanism. The solution reconstruction operates in near-real-time for small scale problems through a novel Finite Element (FE)-based data integration approach that transforms sparse measurements into comprehensive solutions for non-linear systems. This method integrates FE discretisation with a loss function minimisation to generate complete solutions from limited measurement data. Jacobian matrices are computed analytically rather than through Automatic Differentiation (AD). The control component consists of a digital (discrete) Proportional–Integral–Derivative (PID)controller that regulates the cooling water temperature of the test specimens. The modification of the solution construction approach involves its coupling with the thermal eigenvalue-based Reduced Order Modelling (ROM). This allows for the speed-up of the solution construction process whilst keeping the overall procedure the same.The second DT approach, Physics-Driven Machine Learning-based Digital Twinning (PD-ML DT), is rooted in Machine Learning (ML). It involves training two Neural Network (NN) using steady-state data. A two-part NN system is employed, forming a control loop. The first part, the heat flux NN, addresses the challenge of estimating thermal conditions. Its aim is to construct a steady-state equivalent of the heat flux based on the provided temperature measurements. The second part, the coolant NN, provides a solution for active thermal management. In this work, it determines the necessary coolant velocity to maintain the maximum sample temperature below a specified threshold. The additional adjustments are made to ensure the system’s effectiveness and robustness when dealing with dynamic system. These include extrapolating the constructed heat flux and selecting the maximum value from a range of past and extrapolated data points. These adjustments are critical as they account for the fact that both NNs are trained on steady-state data, allowing them to operate effectively in non-steady-state conditions. This combined approach enables real-time temperature monitoring and control addressing a challenge in thermal system management.The performance of two DT systems is demonstrated through the cooling control analysis of samples which were previously evaluated in a fusion energy experimental facility.Various system responses are generated under their control, and they are compared based on the accuracy, speed, and the resistance to measurement noise. The results show that they display advantages and disadvantages based on the information which could be provided during the experiment. Primarily, FE DT requires more measurements than PD-ML DT in the presence of noise; however, it has the ability to provide more information compared with PD-ML DT in the form of the full temperature field. As opposed with PD-ML DT, FE DT does not have instantaneous inference time; however, by employing ROM as part of the workflow it could be reduced.The reported development of DT process for thermal systems, powered by FE combined with PID controller or a dual NN control loop, holds significant implications for the future of fusion energy research and beyond. By enabling real-time monitoring and active thermal management, it provides a crucial tool for optimising the performance and ensuring the safety of experimental fusion facilities. This approach moves the progress closer to autonomous, self-correcting systems that can operate with greater efficiency and reliability. The methodology’s application extends beyond fusion, offering a template for managing thermal systems in a wide range of engineering fields where sparse measurements are involved. |
| published_date |
2026-02-17T05:39:02Z |
| _version_ |
1860430035323715584 |
| score |
11.099917 |