Performance Comparison of Machine Learning Disruption Predictors at JET

Aymerich, E.; Cannas, B.; Pisano, F.; Sias, Giuliana; Sozzi, C.; Stuart, Chris; Carvalho, P.; Fanni, Alessandra

doi:10.3390/app13032006

Cited by 7 publications

(4 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The cluster labels for B found with DBSCAN and those predicted with the linear classifier are then compared to verify whether a similar result is obtained. The classification performance is calculated using two different F1 scores [30,31], using the Scikit-learn functions: F1 macro , which calculate the average of the F1 score for every label and F1 weighted , in which the label F1 scores are weighted, to account for the different size of the groups.…”

Section: Data Analysis Toolsmentioning

confidence: 99%

Automated W7-X sawtooth crashes detection and characterization

Zanini,

Aymerich,

Böckenhoff

et al. 2024

Nucl. Fusion

View full text Add to dashboard Cite

Sawtooth crashes are observed during ECCD experiments at the superconducting optimized stellarator Wendelstein 7-X. The study and the characterization are necessary in order to understand under which condition ECCD can be driven without posing a risk to experimental operations. The development of automatic tools is crucial to speed up the analysis and use extensive datasets. In this work we report on the first attempt of using a data-driven approach to automatically characterise the sawtooth crashes. Cluster algorithms are applied to the dataset, confirming the existence of two distinct types of crashes. This approach allows to study the two groups separately and underlines the different plasma parameters that influence the sawtooth crash parameters, for instance crash amplitude and period.

show abstract

Section: Data Analysis Toolsmentioning

confidence: 99%

Automated W7-X sawtooth crashes detection and characterization

Zanini,

Aymerich,

Böckenhoff

et al. 2024

Nucl. Fusion

View full text Add to dashboard Cite

show abstract

“…If disruptive discharges trigger the alarm, it is true positive (TP); otherwise, it is false negative (FN). If non-disruptive discharges do not trigger an alarm, it is true negative; otherwise, it is false positive [28]. It is worth noting recall rate is 95.3%; among the 801 non-disruptive shots, 737 do not trigger any warning, and the false alarm rate is only 8%.…”

Section: Performance Of the Predictormentioning

confidence: 99%

Disruption prediction on EAST with different wall conditions based on a multi-scale deep hybrid neural network

et al. 2023

View full text Add to dashboard Cite

Plasma disruption is a very dangerous event for future tokamaks and fusion reactors. Therefore, predicting disruption is crucial for ensuring the safety and performance of reactors. In this study, the features of two deep learning algorithms are integrated to establish a multi-scale hybrid network disruption predictor. Firstly, 43 diagnostic signals are extracted by a convolutional neural network (CNN), and the time information is learned by a long short-term memory network. The predictor is trained and tested on a database containing ∼ tens of thousands non-all-metal wall discharges. Its area under the receiver operator characteristic curve (AUC), which is a common performanc metric for deep learning algorithms, reaches 0.97, and the true positive rate is ∼ 95.3% , while the false positive rate is ∼ 8%. Since EAST was upgraded in 2020, the wall condition has been upgraded from non-all-metal to all-metal. To examine the robustness of the predictor, the EAST disruption predictor is migrated to the all-metal wall experiment for the first time. It is again tested under the all-metal wall experimental data, and its warning performance decreases significantly, with an AUC of only 0.79. To improve the robustness of the predictor against disruptions, the convolutional attention mechanism is introduced into the CNN. After training and testing with the same data set, the warning performance for the all-metal wall data is improved, with the AUC value increasing to 0.84. To further improve the robustness of the predictor, the t-SNE algorithm is employed to explore the difference of the sample distribution before and after EAST upgradation, and the transfer learning algorithm is used to reduce this difference. By applying transfer learning to a small amount of discharges from the all-metal wall data, the warning performance in the all-metal wall experiment is further improved, and the AUC value increases to 0.93.

show abstract

“…The best results achieved on disruptions have been obtained by following the nowadays well established and quite popular approach based on machine learning (ML) techniques [13]. In Aymerich et al [15] a systematic comparison of the most widespread and used algorithms is performed. A Multilayer Perceptron Neural Network (MLP-NN), a Generative Topographic Mapping (GTM) and a Convolutional Neural Networks (CNN) algorithm have been considered.…”

Section: Overview Of the Issuementioning

confidence: 99%

New Challenges in Nuclear Fusion Reactors: From Data Analysis to Materials and Manufacturing

2023

View full text Add to dashboard Cite

The construction and operation of the first generation of magnetically controlled nuclear fusion power plants require the development of proper physics and the engineering bases. The analysis of data, recently collected by the actual largest and most important tokamak in the world JET, that has successfully completed his second deuterium and tritium campaign in 2021 (DTE2) with a full ITER like wall main chamber, has provided an important consolidation of the ITER physics basis. Thermonuclear plasmas are highly nonlinear systems characterized by the need of numerous diagnostics to measure physical quantities to guide, through proper control schemes, external actuators. Both modelling and machine learning approaches are required to maximize the physical understanding of plasma dynamics and at the same time, engineering challenges have to be faced. Fusion experiments are indeed extremely hostile environments for plasma facing materials (PFM) and plasma-facing components (PFC), both in terms of neutron, thermal loads and mechanical stresses that the components have to face during either steady operation or off-normal events. Efforts are therefore spent by the community to reach the ultimate goal ahead: turning on the first nuclear fusion power plant, DEMO, by 2050. This editorial is dedicated at reviewing some aspects touched in recent studies developed in this dynamic, challenging project, collected by the special issue titled “New Challenges in Nuclear Fusion Reactors: From Data Analysis to Materials and Manufacturing”.

show abstract

Performance Comparison of Machine Learning Disruption Predictors at JET

Cited by 7 publications

References 46 publications

Automated W7-X sawtooth crashes detection and characterization

Automated W7-X sawtooth crashes detection and characterization

Disruption prediction on EAST with different wall conditions based on a multi-scale deep hybrid neural network

New Challenges in Nuclear Fusion Reactors: From Data Analysis to Materials and Manufacturing

Contact Info

Product

Resources

About