FusionMAE: large-scale pretrained model to optimize and simplify diagnostic and control of fusion plasma

2025-09-16

AI Summary

In magnetic confinement fusion, the multiscale, nonlinear plasma dynamics lead to high diagnostic and control complexity, strong system coupling, and poor fault tolerance. To address these bottlenecks, we propose FusionMAE—a large-scale pretrained model based on a masked autoencoder architecture. Trained on 88-channel unlabeled diagnostic signals, FusionMAE achieves unified embedding representation and efficient compression of heterogeneous diagnostic data. It innovatively integrates joint compression-dimensionality reduction and missing-signal reconstruction, enabling virtual backup diagnostics. On missing-signal inference tasks, FusionMAE achieves 96.7% accuracy and exhibits three emergent capabilities: automated data analysis, a universal control-diagnostic interface, and multi-task cooperative control. These advances significantly reduce reliance on physical diagnostic hardware while enhancing operational robustness and control performance of fusion devices.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

In magnetically confined fusion device, the complex, multiscale, and nonlinear dynamics of plasmas necessitate the integration of extensive diagnostic systems to effectively monitor and control plasma behaviour. The complexity and uncertainty arising from these extensive systems and their tangled interrelations has long posed a significant obstacle to the acceleration of fusion energy development. In this work, a large-scale model, fusion masked auto-encoder (FusionMAE) is pre-trained to compress the information from 88 diagnostic signals into a concrete embedding, to provide a unified interface between diagnostic systems and control actuators. Two mechanisms are proposed to ensure a meaningful embedding: compression-reduction and missing-signal reconstruction. Upon completion of pre-training, the model acquires the capability for 'virtual backup diagnosis', enabling the inference of missing diagnostic data with 96.7% reliability. Furthermore, the model demonstrates three emergent capabilities: automatic data analysis, universal control-diagnosis interface, and enhancement of control performance on multiple tasks. This work pioneers large-scale AI model integration in fusion energy, demonstrating how pre-trained embeddings can simplify the system interface, reducing necessary diagnostic systems and optimize operation performance for future fusion reactors.

Problem

• Compress 88 diagnostic signals into unified embedding
• Provide reliable virtual backup for missing diagnostics
• Simplify system interface and enhance control performance

Innovation

• Large-scale pretrained model compresses diagnostic signals
• Virtual backup diagnosis infers missing data reliably
• Unified interface enhances control and simplifies systems