View source for LNF: Modelling of disruption types in thermonuclear plasmas and its recognition by means of machine learning techniques

== LNF - Nationally funded project ==

'''Title''': '''Modelling of disruption types in thermonuclear plasmas and its recognition by means of machine learning techniques'''

'''Reference''': PID2019-108377RB-C31

'''Programme and date''': Proyectos COORDINADOS DE I+D+i. MODALIDADES “GENERACIÓN DEL CONOCIMIENTO” Y “RETOS INVESTIGACIÓN” Año 2019

'''Programme type (Modalidad de proyecto)''': Tipo B  Coordinado

'''Area/subarea (Área temática / subárea)''': Energía y Transporte/Energía

'''Principal Investigator(s)''': [https://orcid.org/0000-0002-1622-3984 Jesús Vega] [https://orcid.org/0000-0002-5676-9631 Giuseppe Rattá]

'''Project type''': Proyecto coordinado

'''Start-end dates''': 01/06/2020 - 29/02/2024

'''Financing granted (direct costs)''': 84000 €

== Description of the project ==

Desde el punto de vista de la física, el proyecto ha conseguido su principal objetivo: determinar tipos de disrupciones en JET e identificar su origen físico. Se han desarrollado dos técnicas de aprendizaje automático (ambas de naturaleza muy diferente) y las dos reconocen tres tipos de disrupciones. Un primer grupo engloba a las debidas a la alta emisión de radiación. Un segundo conjunto agrupa a las resultantes de inestabilidades magneto-hidrodinámicas. Un tercer grupo, menos numeroso, abarca el resto de disrupciones. Es importante resaltar que ha sido la primera vez que por medios completamente diferentes y de manera no supervisada se clasifican bajo el mismo tipo un 80% de las disrupciones. Para conseguir estos logros, se han desarrollado técnicas no supervisadas de detección de precursores disruptivos partiendo de un conjunto de 22 señales de plasma. Se ha creado una plataforma web programable que implementa un entorno computacional distribuido sobre el que desplegar las técnicas no supervisadas y que son costosas desde el punto de vista de tiempo de cálculo. En este entorno se han integrado plataformas avanzadas de adquisición de datos con alto grado de heterogeneidad. Son plataformas con capacidades de tiempo real en las que conviven elementos de cálculo tales como CPUs, GPUs y FPGAs. En particular, se han usado tecnologías hardware (controladores rápidos) y software (Nominal Device Support) previstas tanto para los sistemas de instrumentación y control de ITER como para sus diagnósticos. Se han probado aplicaciones de tiempo real sobre ellas tales como predictores de disrupción, detección de anomalías, reconocimiento de precursores disruptivos y clasificación de disrupciones según los tipos mencionados anteriormente. Estas pruebas de concepto han demostrado, por un lado, la viabilidad bajo condiciones de tiempo real de los métodos de predicción de disrupciones y de los clasificadores desarrollados por los equipos del proyecto. Por otro lado, se ha verificado la idoneidad de las plataformas para predicción y clasificación de disrupciones bajo situaciones deterministas. Es muy importante resaltar que el grado de consecución global del proyecto “Modelado de tipos de disrupciones en plasmas termonucleares y su reconocimiento mediante técnicas de aprendizaje automático (DISRUPTION_ID)” puede estimarse en el 100%. Sin duda, esto solamente ha sido posible con la aportación coordinada de las respectivas especializaciones de los tres grupos involucrados. En otras palabras, el éxito del proyecto no se ha conseguido con trabajos inconexos de los participantes, sino que se ha conjugado en todo momento las necesidades específicas de I+D+i con la experiencia y el saber de cada equipo. Tampoco puede dejar de enfatizarse que el proyecto ha utilizado la base de datos del Tokamak más importante del mundo en la actualidad, el dispositivo JET, y que es el de prestaciones termonucleares más cercanas al futuro dispositivo ITER.

<!-- If applicable: references -->
== References ==

Publicaciones en revistas con “peer review” directamente relacionadas con los resultados del proyecto:

A. Murari, R. Rossi, T. Craciunescu, J. Vega, JET Contributors and M. Gelfusa. “A control oriented strategy of disruption prediction to avoid the configuration collapse of tokamak reactors”. Nature Communications 15, 2424 (2024). DOI: 10.1038/s41467-024-46242-7.

J. Vega, S. Dormido-Canto, R. Castro, D. Fernández, A. Murari on behalf of JET Contributors. “Real-time disruption prediction in multi-dimensional spaces leveraging diagnostic information not available at execution time”. Nuclear Fusion 64 (2024) 046010 (12pp). DOI: 10.1088/1741-4326/ad288a.

R. Correa, G. Farias, E. Fabregas, S. Dormido-Canto, I. Pastor and J. Vega. “Deep Learning Models to Reduce Stray Light in TJ-II Thomson Scattering Diagnostic”. Sensors 2024, 24, 2764 (15pp). DOI: 10.3390/s24092764.

R. Rossi, M. Gelfusa, T. Craciunescu, I. Wyss, J. Vega, A. Murari on behalf of JET Contributors. “A hybrid physics/data-driven logic to detect, classify, and predict anomalies and disruptions in tokamak plasmas”. Nuclear Fusion 64 (2024) 046017 (25pp). (https://doi.org/10.1088/1741-4326/ad2723).

G. A. Rattá, J. Vega, A. Murari, D. Gadariya, C. Stuart, G. Farías and JET Contributors. “Characterization of physics events in JET preceding disruptions”. Fusion Engineering and Design 189 (2023) 113468. DOI: 10.1016/j.fusengdes.2023.113468.

R. Castro, Y. Makushok, L. Abadie, B. Bauvir, A. Neto, J. Vega. “Storing EPICS process variables in HDF5 files for ITER”. Fusion Engineering and Design 194 (2023) 113697. DOI: 10.1016/j.fusengdes.2023.113697.

F. Esquembre, J. Chacón, J. Saenz, J. Vega, S. Dormido-Canto. “A programmable web platform for distributed access, analysis, and visualization of data”. Fusion Engineering and Design 197 (2023) 114049. DOI: 10.1016/j.fusengdes.2023.114049.

R. Rossi, M. Gelfusa, T. Craciunescu, L. Spolladore, I. Wyss, E. Peluso, J. Vega, C. F. Maggi, J. Mailloux, M. Maslov, A. Murari on behalf of JET Contributors. “A systematic investigation of radiation collapse for disruption avoidance and prevention on JET tokamak”. Matter and Radiation at Extremes 8, 046903 (2023) 21pp. DOI: 10.1063/5.0143193.

S.M. Gonzalez de Vicente, D. Mazon, M. Xu, S. Pinches, M. Churchill, A. Dinklage, R. Fischer, A. Murari, P. Rodriguez-Fernandez, J. Stillerman, J. Vega and G. Verdoolaege. “Summary report of the 4th IAEA Technical Meeting on Fusion Data Processing, Validation and Analysis (FDPVA)”. Nuclear Fusion 63 (2023) 047001. DOI: 10.1088/1741-4326/acbfce.

J. Vega, A. Murari, S. Dormido-Canto, G. A. Rattá, M. Gelfusa and JET Contributors. “Disruption prediction with artificial intelligence techniques in tokamak plasmas”. Nature Physics. 18, 741–750 (2022). https://doi.org/10.1038/s41567-022-01602-2.

D. Gadariya, J. Vega, C. Stuart, G. Rattá, P. Card, A. Murari, S. Dormido-Canto, JET Contributors. “Performance analysis of the centroid method predictor implemented in the JET real time network”. Plasma Physics and Controlled Fusion 64 (2022), 114003 (8pp). (https://doi.org/10.1088/1361-6587/ac963f).

M. Ruiz, J. Nieto, V. Costa, T. Craciunescu, E. Peluso, J. Vega, A. Murari, JET Contributors. “Acceleration of an Algorithm Based on the Maximum Likelihood Bolometric Tomography for the Determination of Uncertainties in the Radiation Emission on JET Using Heterogeneous Platforms”. Applied Sciences 12, 13 (2022), 6798. DOI: 10.3390/app12136798.

A. Murari, E. Peluso, T. Craciunescu, S. Dormido-Canto, M. Lungaroni, R. Rossi, L. Spolladore, J. Vega, M. Gelfusa and JET Contributors. “Frontiers in data analysis methods: from causality detection to data driven experimental design”. Plasma Phys. Control. Fusion 64 (2022) 024002 (12pp). (https://doi.org/10.1088/1361-6587/ac3ded). FI: 2.458, Q2

A Murari, E Peluso, L Spolladore, J Vega, M Gelfusa. “Considerations on Stellarator’s Optimization from the Perspective of the Energy Confinement Time Scaling Laws”. Applied Sciences 12, 6 (2022), 2862. DOI: 10.3390/app12062862.

A. Murari, E. Peluso, J. Vega, J. M. García-Regaña, J. L. Velasco, G. Fuchert and M. Gelfusa. “Scaling laws of the energy confinement time in stellarators without renormalization factors”. Nuclear Fusion 61 (2021) 096036 (12pp). https://doi.org/10.1088/1741-4326/ac0cbb. FI: 3,179, Q1

G. A. Rattá, J. Vega, A. Murari, D. Gadariya and JET Contributors. “PHAD: a phase-oriented disruption prediction strategy for avoidance, prevention and mitigation in JET”. Nuclear Fusion 61 (2021) 116055 (16pp). https://doi.org/10.1088/1741-4326/ac2637. FI: 3,179, Q1

G. Farias, E. Fabregas, I. Martínez, J. Vega, S. Dormido-Canto, H. Vargas, “Nuclear fusion pattern recognition by ensemble learning”, Complexity, ISSN: 1076-2787, vol. 2021, Article ID 1207167, pp: 9 pages, 2021. FI: 2.833, Q2

A. Murari, E. Peluso, M. Lungaroni, P. Gaudio, J. Vega, M. Gelfusa. “Data driven theory for knowledge discovery in the exact sciences with applications to thermonuclear fusion”. Scientific Reports 10 (2020) 19858. DOI: 10.1038/s41598-020-76826-4.

J. Vega, R. Castro, S. Dormido-Canto, G. A. Rattá, M. Ruiz. “Automatic recognition of plasma relevant events: Implications for ITER”. Fusion Engineering and Design 155 (2020) 111638. DOI: 10.1016/j.fusengdes.2020.111638.

R. Castro, J. Vega. “Smart decimation method for fusion research data”. Fusion Engineering and Design 159 (2020) 111814. DOI: 10.1016/j.fusengdes.2020.111814.

A. Murari, R. Rossi, E. Peluso, M. Lungaroni, P. Gaudio, M. Gelfusa, G. Ratta, J. Vega, and JET Contributors and ASDEX Upgrade Team. “On the transfer of adaptive predictors between different devices for both mitigation and prevention 
of disruptions”. Nuclear Fusion 60, 5 (2020) 056003. DOI: 10.1088/1741-4326/ab77a6.

G. Farias, E. Fabregas, S. Dormido-Canto, J. Vega, S. Vergara. “Automatic recognition of anomalous patterns in discharges by recurrent neural networks”. Fusion Engineering and Design 154 (2020) 111495. DOI: 10.1016/j.fusengdes.2020.111495

D. Mazon, S. M. González de Vicente, M. Churchill, A. Dinklage, R. Fischer, M. Jakubowski, A. Murari, M. Romanelli, J. Vega, G. Verdoolaege and M. Xu. “Summary report of the 3rd IAEA technical meeting on fusion data processing validation and analysis (FDPVA)”. Nuclear Fusion 60 (2020) 097002 (10pp). https://doi.org/10.1088/1741-4326/aba8dd. FI: 3,179, Q1

G. Farias, E. Fabregas, S. Dormido-Canto, J. Vega, S. Vergara, “Automatic recognition of anomalous patterns in discharges by Applying Deep Learning”, Fusion Science and Technology, ISSN: 1536-1055, vol. 76, pp: 925-932, 2020, FI: 1.1, Q4

M. Astrain, M. Ruiz, A. Carpeño, S. Esquembri, E. Barrera, J. Vega, A methodology to standardize the development of FPGA-based high-performance DAQ and processing systems using OpenCL, Fusion Engineering and Design, Volume 155, 2020, 111561, ISSN 0920-3796, https://doi.org/10.1016/j.fusengdes.2020.111561. FI: 1.453, Q3


Asistencia a congresos, conferencias o workshops relacionados con el proyecto:

J. Vega, S. Dormido-Canto, R. Ramírez, G. Farias, A. Murari, D. Gadariya and JET Contributors. “Real-time disruption prediction in multi-dimensional spaces with privileged information not available at execution time”. Fifth IAEA TM on Fusion Data Processing, Validation and Analysis JUN 12 - 15, 2023. Ghent University, Ghent, Belgium.

R. Castro, L. Abadie, O. Hoenen, S. Pinches, S. Simrock, P. Sawantdesai, P. Abreu, M. Osokawa, J. Vega, P. Martín, A. Luengo, J. Steeven. “MINT, ITER Interactive Data Visualization Tool”. Fifth IAEA TM on Fusion Data Processing, Validation and Analysis JUN 12 - 15, 2023. Ghent University, Ghent, Belgium.

J. Vega. “Machine learning for real-time disruption prediction: from massive training sets to data scarcity and privileged information”. IRFM CEA Cadarache (11th September 2023).

J. Vega, S. Dormido-Canto, A. Murari, J. D. Fernández. “Real-time disruption prediction in multi-dimensional spaces with privileged information not available at execution time”. 3rd International Conference on Computations for Science and Engineering. 20-23 September 2023, Naples, Italy.

J. Vega, G. A. Rattá, D. Gadariya, A. Murari, C. Stuart, S. Dormido-Canto and JET Contributors. “Review of a data-driven adaptive disruption predictor for mitigation based on a nearest centroid approach”. 2nd IAEA technical meeting on plasma disruptions and their mitigation (2022), Saint Paul lez Durance, France.

J. Vega, A. Murari, S. Dormido-Canto, G. A. Rattá, M. Gelfusa. “Predicting and Understanding Catastrophic Events: Tokamak Disruptions, an Issue for Thermonuclear Fusion, an Opportunity for Society”. 6th International Conference Frontiers in Diagnostic Technologies. (2022). Frascati, Italy.

A. Carpeño, M. Ruiz, V. Costa, D. Rivilla, J. Vega. “Analysis of the portability of a testing exchangeability using a randomized power martingale algorithm in FPGA-based devices”. 23th IEEE-NPSS Real Time Conference 1-5 August 2022.

G. A. Rattá, J. Vega, A. Murari, D. Gadariya, C. Stuart and JET Contributors. “Characterization of physics events in JET preceding disruptions”. 32nd Symposium on Fusion Technology. 18-23 September 2022. Dubrovnik, Croatia.

R. Castro, Y. Makushok, L. Abadi, J. Vega. “New EPICS PVAccess archiving system for ITER”. 32nd Symposium on Fusion Technology. 18-23 September 2022. Dubrovnik, Croatia.

J. Vega, G. A. Rattá, D. Gadariya. “Building a parsimonious disruption mitigation trigger”. 2022 WPSA General Meeting.

J. Vega. “Disruption prediction evolution: from traditional methods to estimation of the disruption time”. 1st JT-60SA Topical Group Meeting on “MHD Stability and Control”. October 2022.

J. Vega,  D. Gadariya, G. Rattá, A. Murari. “Anomaly Detection and Unsupervised Classification of Plasma Events”. 14th Chaotic Modeling and Simulation International Conference. 8 – 11 June 2021. Athens, Greece – Turned into Virtual. (http://www.cmsim.org/images/CHAOS2021_program-final.pdf).

G. A. Rattá,  J. Vega, A. Murari and JET Contributors. “Tiding up the chaos with Genetic Algorithms: examples in Magnetically Confined Nuclear Fusion”. 14th Chaotic Modeling and Simulation International Conference. 8 – 11 June 2021. Athens, Greece – Turned into Virtual. (http://www.cmsim.org/images/CHAOS2021_program-final.pdf).

J. Vega, S. Dormido-Canto, “Fusión nuclear y acceso seguro a entornos experimentales: el caso del estellarator TJ‐IIs”, II Congreso de Seguridad Digital y Ciberinteligencia-C1b3rWall, Escuela Nacional de Policía, Ávila, España, 21-23 de Junio de 2022.

J. Vega,  R. Dormido, S. Dormido-Canto, G. A. Rattá, D. Gadariya, A. Murari. “Prediction of Disruptive Events on the Route to Nuclear Fusion Reactors”. ISC High Performance 2021 Conference (Session: HPC for the Energy Transition). June 24th –July 2nd 2021. ISC-HPC.com (https://app.swapcard.com/widget/event/isc-high-performance-2021-digital/planning/UGxhbm5pbmdfNDQ0Nzg0).

J. Vega,  A. Murari, G. A. Rattá, S. Dormido-Canto, D. Gadariya and JET Contributors. “Disruption predictors in nuclear fusion by using machine learning methods: an overview”. 1st Workshop on Artificial Intelligence in Plasma Science. 20th – 22nd September 2021. Aix-en-Provence, France. (http://www.camt.eng.osaka-u.ac.jp/hamaguchi/WAIPS1/invitedSpeakerList.html).

G. A. Rattá,  J. Vega, A. Murari  and JET Contributors. “Disruption prediction strategy for mitigation, prevention and avoidance at JET using machine learning techniques”. 1st Workshop on Artificial Intelligence in Plasma Science. 20th – 22nd September 2021. Aix-en-Provence, France. (http://www.camt.eng.osaka-u.ac.jp/hamaguchi/WAIPS1/invitedSpeakerList.html).

J. Vega,  R. Dormido, S. Dormido-Canto, G. Rattá, D. Gadariya, A. Murari and JET Contributors. “Comparison of unsupervised methods to determine common patterns in the termination phase of disruptive discharges in JET”. 4th IAEA Technical Meeting on Fusion Data Processing, Validation and Analysis. 29th November – 6th December 2021. Virtual Event. (https://conferences.iaea.org/event/251/timetable/#20211129).

D. Gadariya,  J. Vega, C. Stuart, P. Card, A. Murari, G. A. Rattá, S. Dormido-Canto and JET Contributors. “Performance analysis of the centroid method predictor in the JET RT network”. 4th IAEA Technical Meeting on Fusion Data Processing, Validation and Analysis. 29th November – 6th December 2021. Virtual Event. (https://conferences.iaea.org/event/251/timetable/#20211129).

M. Gelfusa,  A. Murari, M. Lungaroni, R. Rossi, L. Spolladore, J. Vega and JET Contributors. “Open world learning: a new paradigm for disruption prediction”. 4th IAEA Technical Meeting on Fusion Data Processing, Validation and Analysis. 29th November – 6th December 2021. Virtual Event. (https://conferences.iaea.org/event/251/timetable/#20211129).

R. Rossi,  M. Gelfusa, J. Vega, A. Murari, JET Contributors, ASDEX-Upgrade Team, and MST1 Team. “Adaptive and Transfer Learning for Disruption Classification and Prevention on ASDEX-Upgrade and JET”. 4th IAEA Technical Meeting on Fusion Data Processing, Validation and Analysis. 29th November – 6th December 2021. Virtual Event. (https://conferences.iaea.org/event/251/timetable/#20211129).

F. Esquembre, J. Chacón, J. Sáenz, E. Fábregas, G. Farias, J. Vega, S. Dormido-Canto, “A programmable web platform for distributed data access, analysis and visualization”, 13th IAEA Technical Meeting on Control, Data Acquisition and Remote Participation for Fusion Research. July 5-8, 2021, Culham, United Kingdom.

G. Farias, D. Hidalgo, S. Cuellar, E. Fabregas, F. Esquembre, S. Dormido-Canto, J. Vega, “Reinforcement learning for building nuclear fusion classifiers from scratch”, 13th IAEA Technical Meeting on Control, Data Acquisition and Remote Participation for Fusion Research. July 5-8, 2021, Culham, United Kingdom.

G. Farias, R. Correa, H. Ramirez, E. Fabregas, F. Esquembre, S. Dormido-Canto, J. Vega, “Deep learning models to generate realistic new data in nuclear fusion”, 13th IAEA Technical Meeting on Control, Data Acquisition and Remote Participation for Fusion Research. July 5-8, 2021, Culham, United Kingdom.

F. Hernández-del-Olmo, N. Duro, E. Gaudioso, R. Dormido, J. Vega, “Correlation based method for sorting and filtering relevant features for unsupervised machine learning”, 13th IAEA Technical Meeting on Control, Data Acquisition and Remote Participation for Fusion Research. July 5-8, 2021, Culham, United Kingdom.

M. Ruiz,  J. Nieto, V. Costa, S. Esquembri, T. Craciunescu, E. Peluso, J. Vega, A. Murari  and JET contributors. “Acceleration of an algorithm based on the maximum likelihood bolometric tomography for the determination of uncertainties in the radiation emission on JET using heterogeneous platforms”. 13th Technical Meeting on Plasma Control Systems, Data Management and Remote Experiments in Fusion Research. 5-8 July 2021. Virtual (Culham Center for Fusion Energy).

J. Vega. “Comparison of unsupervised methods to determine common patterns in the termination phase of disruptive discharges in JET”. JET Task Force Meeting. Thursday 25th November 2021, Culham, UK.

G. A. Rattá. “A new phase-oriented disruption prediction strategy for mitigation, prevention and avoidance in JET”. JET Task Force Meeting. Thursday 11th March 2021, Culham, UK.

J. Vega, A. Murari and JET Contributors. “Predicting the dynamics on nonlinear instabilities: disruptions in tokamaks”. 13th CHAOS 2020 International Conference. 9-12 June 2020. Florence, Italy (http://www.cmsim.org/images/Book_of_Abstracts_CHAOS_2020--.pdf).

A. Murari, J. Vega. “Predicting collapse: adaptive and transfer learning”. 2020 EIROforum Workshop: Big data – from acquisition to data mining. 26 – 20 October 2020. Zurich. (https://indico.cern.ch/event/881752/contributions/4006555/attachments/2133448/3593073/Vega_Murari_EIROForum_2020.pdf).

M. Gelfusa, A. Murari, R. Rossi, M. Lungaroni, E. Peluso, G. Rattá, J. Vega. “On the Potential of Adaptive Predictors and their Transfer between Different Devices for both Mitigation and Prevention of Disruptions”. Report of Abstracts 112. (Virtual) Technical Meeting on Plasma Disruptions and their Mitigation. 20 July 2020-23 July 2020 WebExMeeting.

A. Murari, M. Gelfusa, M. Lungaroni, E. Peluso, J. Vega, P. Gaudio. “Investigating the Physics of the Tokamak Operational Boundaries using Machine Learning Tools”. Report of Abstracts 113. (Virtual) Technical Meeting on Plasma Disruptions and their Mitigation. 20 July 2020-23 July 2020 WebExMeeting.


Tesis doctorales finalizadas relacionadas con el proyecto:

Título: “Modelos de clasificación con medidas de confianza en predictores conformales aplicados a imágenes de fusión nuclear”
     Doctorando: Álvaro Antonio Olmedo Rodríguez.
     Directores: Jesús Antonio Vega Sánchez y Sebastián Dormido Canto.
     Fecha de Lectura: 30 de Septiembre de 2020.
     Organismo: UNED (Universidad Nacional de Educación a Distancia), Madrid
     Calificación: Sobresaliente cum laude.


Título: “Optimización de predictores de disrupciones en espacios bidimensionales”
     Doctorando: Francisco Javier Hernández Martín.
     Directores: Jesús Antonio Vega Sánchez y Sebastián Dormido Canto.
     Fecha de Lectura: 19 de Octubre de 2020.
     Organismo: UNED (Universidad Nacional de Educación a Distancia), Madrid


Título: “Real Time Performance Analysis of an Optimized Linear Disruption Predictor in JET”
     Doctorando: Dhaval Gadariya
     Directores:  Jesús Antonio Vega Sánchez y Sebastián Dormido Canto
     Fecha de Lectura: 29 de Junio de 2023
     Organismo: UNED (Universidad Nacional de Educación a Distancia), Madrid
     Calificación: Sobresaliente


<references /> 

<!-- DO NOT REMOVE THE FOLLOWING LINES -->
<hr>
[[LNF:Nationally Funded Projects|Back to list of nationally funded projects]]

[[Category:LNF Nationally Funded Projects]]