FusionPapers
図版検索トレンドwiki日本の研究
© 2026 FUSIONPAPERS
About法務情報
トップに戻る

Interpretation of machine-learning-based disruption models for plasma control

Matthew S Parsons2017年Plasma Physics and Controlled FusionIF 2.2出版社

While machine learning techniques have been applied within the context of fusion for predicting plasma disruptions in tokamaks, they are typically interpreted with a simple 'yes/no' prediction or perhaps a probability forecast. These techniques take input signals, which could be real-time signals from machine diagnostics, to make a prediction of whether a transient event will occur. A major criticism of these methods is that, due to the nature of machine learning, there is no clear correlation between the input signals and the output prediction result. Here is proposed a simple method that could be applied to any existing prediction model to determine how sensitive the state of a plasma is at any given time with respect to the input signals. This is accomplished by computing the gradient of the decision function, which effectively identifies the quickest path away from a disruption as a function of the input signals and therefore could be used in a plasma control setting to avoid them. A numerical example is provided for illustration based on a support vector machine model, and the application to real data is left as an open opportunity.

日本語訳

機械学習技術はトカマクにおけるプラズマ破壊の予測のために核融合の文脈で応用されてきたが、それらは典型的には単純な「はい/いいえ」の予測または確率予報として解釈される。これらの技術は、機械からのリアルタイム診断信号となり得る入力信号を取り込み、過渡事象が発生するかどうかの予測を行う。これらの手法に対する主な批判は、機械学習の性質上、入力信号と出力予測結果の間に明確な相関関係が存在しないことである。ここでは、任意の既存の予測モデルに適用可能な簡潔な手法を提案し、プラズマの状態が各時点で入力信号に対してどの程度敏感であるかを決定する。これは、決定関数の勾配を計算することによって達成され、入力信号の関数として破壊から遠ざかる最速の経路を効果的に特定する。したがって、この手法はプラズマ制御において破壊を回避するために使用できる。数値例としてサポートベクターマシンモデルに基づく実例を示し、実データへの適用は今後の課題として残されている。

wiki

Plasma disruptionPlasma control
この論文にはまだAI要約がありません。

関連論文

Determining the prediction limits of models and classifiers with applications for disruption prediction in JET

2017Nuclear Fusion

Tokamak divertor plasma emulation with machine learning

2024Nuclear Fusion

Forecasting disruptions in the ADITYA tokamak using neural networks

2000Nuclear Fusion

Disruption prediction and model analysis using LightGBM on J-TEXT and HL-2A

2021Plasma Physics and Controlled Fusion

Vi-DP: low-latency video-based disruption prediction with multi-FOV fusion in EAST

2026Nuclear Fusion

Disruption predictor based on neural network and anomaly detection on J-TEXT

2020Plasma Physics and Controlled Fusion

High-fidelity and explainable machine learning method for neoclassical toroidal viscosity torque prediction

2026Nuclear Fusion

Design of HL-2A plasma position predictive model based on deep learning

2020Plasma Physics and Controlled Fusion

Machine learning based disruption prediction using long short-term memory in KSTAR

2025Nuclear Fusion

Performance prediction of radio frequency based negative ion source using fusion neural network model

2025Nuclear Fusion