Hierarchical Multi-to-Single-Modal Knowledge Distillation for Disruption Prediction in EAST

2026-07-05

AI Summary

This work addresses the high computational overhead of multimodal fusion in tokamak plasma disruption prediction, despite its improved discriminative performance. To this end, the authors propose a hierarchical many-to-one modality knowledge distillation framework: during training, a multimodal teacher model leverages both visible-light images and time-series data, while at inference, only a lightweight time-series student model is deployed. Knowledge transfer is achieved through a three-level distillation process—graph structure, representation, and decision—augmented by a prototype-guided spatiotemporal hypergraph module to effectively distill multimodal information into the unimodal student. Experiments on the EAST dataset with 640 discharges demonstrate that the proposed method maintains high prediction accuracy while substantially reducing computational cost, offering an efficient and practical solution for real-time disruption warning.

Abstract

Plasma disruption is a critical threat to tokamak safety. Existing data-driven predictors mainly rely on time-series diagnostic signals, while visible images provide complementary spatial cues including plasma deformation, local brightening, and radiation-structure evolution. Although the image modality improves the model's discriminative capability, it also substantially increases the computational cost during inference. To address this issue, we propose a hierarchical multi-to-single-modal knowledge distillation framework for disruption prediction on a synchronized EAST multimodal dataset. During training, visible images and time-series signals are used to train a multimodal teacher, which learns disruption precursor representations through Transformer-based encoders and a prototype-guided spatiotemporal hypergraph module. During inference, only the time-series student is retained, with multimodal knowledge transferred through graph-structure-level, representation-level, and decision-level distillation. On the 640-discharge EAST dataset, the results demonstrate that the proposed framework can preserve the discriminative advantages of multimodal learning while substantially reducing inference cost, and providing an effective route for efficient disruption prediction in EAST. The source code of this paper will be released on https://github.com/Event-AHU/OpenFusion.

Problem

• plasma disruption prediction
• multimodal learning
• knowledge distillation
• computational cost
• tokamak safety

Innovation

• knowledge distillation
• multimodal learning
• plasma disruption prediction
• spatiotemporal hypergraph
• tokamak