2025/07/17 by Jyoti Agarwal, Bhaskar Chaudhury, Agarwal, Jyoti +7
Computer Science · Materials Science · Physics and Astronomy · #FOS: Physical sciences #Fusion materials and technologies #Magnetic confinement fusion research #Plasma Physics (physics.plasm-ph) #Time Series Analysis and Forecasting
paper · pdf · doi:10.48550/arxiv.2507.12797
openalex publication_date 2025/07/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Disruptions in tokamak plasmas, marked by sudden thermal and current quenches, pose serious threats to plasma-facing components and system integrity. Accurate early prediction, with sufficient lead time before disruption onset, is vital to enable effective mitigation strategies. This study presents a novel data-driven approach for predicting early current quench, a key precursor to disruptions, using transformer-based deep learning models, applied to ADITYA tokamak diagnostic data. Using multivariate time series data, the transformer model outperforms LSTM baselines across various data distributions and prediction thresholds. The transformer model achieves better recall, maintaining values above 0.9 even up to a prediction threshold of 8-10 ms, significantly outperforming LSTM in this critical metric. The proposed approach remains robust up to an 8 ms lead time, offering practical feasibility for disruption mitigation in ADITYA tokamak. In addition, a comprehensive data diversity analysis and bias sensitivity study underscore the generalization of the model. This work marks the first application of transformer architectures to ADITYA tokamak data for early current-quench prediction, establishing a promising foundation for real time disruption avoidance in short-pulse tokamaks.