ST40 electromagnetic predictive studies supported by machine learning applied to experimental database
Abstract Nuclear fusion is entering the era of power plant-scale devices, which are now undergoing extensive studies to support the design phase. Plasma disruptions pose a high risk to these classes of devices because of the large stored thermal and magnetic energy which might jeopardize machine int...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2024-11-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-024-75798-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850061885627432960 |
|---|---|
| author | M. Scarpari S. Minucci G. Sias R. Lombroni P. F. Buxton M. Romanelli G. Calabrò |
| author_facet | M. Scarpari S. Minucci G. Sias R. Lombroni P. F. Buxton M. Romanelli G. Calabrò |
| author_sort | M. Scarpari |
| collection | DOAJ |
| description | Abstract Nuclear fusion is entering the era of power plant-scale devices, which are now undergoing extensive studies to support the design phase. Plasma disruptions pose a high risk to these classes of devices because of the large stored thermal and magnetic energy which might jeopardize machine integrity and availability. Therefore, disruptions within these devices must be virtually eliminated, and any disruptions which do happen must be highly mitigated. However, the characterisation, prediction and technology used to mitigate disruptions is still an area of active development. In this paper, the authors investigate the disruptions within ST40, with particular attention at the identification of causes and effects associated with disruptions, both from a physics basis and an engineering standpoint. This paper aims at presenting preliminary predictive analyses of ST40 plasma scenarios by exploiting Machine Learning techniques applied to an experimental database populated by plasma pulses executed during the ST40 2021–2022 experimental campaign. The database contains both disrupted and non-disrupted pulses. Using Machine Learning, common features within disruptions are automatically classified and identified, mapping the controllable operational space in terms of plasma displacement and variation of specific plasma internal parameters. The classification was validated by benchmarking the numerical reconstruction of the plasma dynamics with experimental data recovered from the plasma diagnostics. Subsequent Machine Learning analyses allowed the extrapolation of new disrupted plasma configurations for preliminary predictive simulations of the plasma column displacement. Thanks to the numerical simulations performed in MAXFEA environment, it is possible to investigate the plasma vertical displacement both during disrupted and regularly terminated plasma scenarios and to provide lessons to be learnt for the next ST40 experimental campaign and for the design of future ST devices. |
| format | Article |
| id | doaj-art-6b1d8097e87d49df8a37aa86d43f4996 |
| institution | DOAJ |
| issn | 2045-2322 |
| language | English |
| publishDate | 2024-11-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-6b1d8097e87d49df8a37aa86d43f49962025-08-20T02:50:04ZengNature PortfolioScientific Reports2045-23222024-11-0114111510.1038/s41598-024-75798-zST40 electromagnetic predictive studies supported by machine learning applied to experimental databaseM. Scarpari0S. Minucci1G. Sias2R. Lombroni3P. F. Buxton4M. Romanelli5G. Calabrò6Department of Economy, Engineering, Society and Business Organization (DEIM), University of TusciaDepartment of Engineering and Sciences, Faculty of Technological and Innovation Sciences, Universitas MercatorumElectrical and Electronic Engineering Department, Faculty of Engineering, University of CagliariDepartment of Economy, Engineering, Society and Business Organization (DEIM), University of TusciaTokamak Energy LtdTokamak Energy LtdDepartment of Economy, Engineering, Society and Business Organization (DEIM), University of TusciaAbstract Nuclear fusion is entering the era of power plant-scale devices, which are now undergoing extensive studies to support the design phase. Plasma disruptions pose a high risk to these classes of devices because of the large stored thermal and magnetic energy which might jeopardize machine integrity and availability. Therefore, disruptions within these devices must be virtually eliminated, and any disruptions which do happen must be highly mitigated. However, the characterisation, prediction and technology used to mitigate disruptions is still an area of active development. In this paper, the authors investigate the disruptions within ST40, with particular attention at the identification of causes and effects associated with disruptions, both from a physics basis and an engineering standpoint. This paper aims at presenting preliminary predictive analyses of ST40 plasma scenarios by exploiting Machine Learning techniques applied to an experimental database populated by plasma pulses executed during the ST40 2021–2022 experimental campaign. The database contains both disrupted and non-disrupted pulses. Using Machine Learning, common features within disruptions are automatically classified and identified, mapping the controllable operational space in terms of plasma displacement and variation of specific plasma internal parameters. The classification was validated by benchmarking the numerical reconstruction of the plasma dynamics with experimental data recovered from the plasma diagnostics. Subsequent Machine Learning analyses allowed the extrapolation of new disrupted plasma configurations for preliminary predictive simulations of the plasma column displacement. Thanks to the numerical simulations performed in MAXFEA environment, it is possible to investigate the plasma vertical displacement both during disrupted and regularly terminated plasma scenarios and to provide lessons to be learnt for the next ST40 experimental campaign and for the design of future ST devices.https://doi.org/10.1038/s41598-024-75798-zPlasma disruptionsST40Experimental databaseNumerical electromagnetic predictive simulationMachine learningSOM |
| spellingShingle | M. Scarpari S. Minucci G. Sias R. Lombroni P. F. Buxton M. Romanelli G. Calabrò ST40 electromagnetic predictive studies supported by machine learning applied to experimental database Scientific Reports Plasma disruptions ST40 Experimental database Numerical electromagnetic predictive simulation Machine learning SOM |
| title | ST40 electromagnetic predictive studies supported by machine learning applied to experimental database |
| title_full | ST40 electromagnetic predictive studies supported by machine learning applied to experimental database |
| title_fullStr | ST40 electromagnetic predictive studies supported by machine learning applied to experimental database |
| title_full_unstemmed | ST40 electromagnetic predictive studies supported by machine learning applied to experimental database |
| title_short | ST40 electromagnetic predictive studies supported by machine learning applied to experimental database |
| title_sort | st40 electromagnetic predictive studies supported by machine learning applied to experimental database |
| topic | Plasma disruptions ST40 Experimental database Numerical electromagnetic predictive simulation Machine learning SOM |
| url | https://doi.org/10.1038/s41598-024-75798-z |
| work_keys_str_mv | AT mscarpari st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase AT sminucci st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase AT gsias st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase AT rlombroni st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase AT pfbuxton st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase AT mromanelli st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase AT gcalabro st40electromagneticpredictivestudiessupportedbymachinelearningappliedtoexperimentaldatabase |