Physics-Informed Deep Learning Model for Line-integral Diagnostics Across Fusion Devices

2024-11-27

AI Summary

To address the slow reconstruction speed and poor generalizability of 2D plasma profile reconstruction in nuclear fusion devices, this paper proposes Onion, a physics-informed neural network. Onion introduces a novel multiplicative physics-constraint mechanism that explicitly embeds the line-integral forward model into the network architecture and employs a customized loss function grounded in soft X-ray absorption physics. The framework supports diverse backbone networks and is compatible with real diagnostic data from EAST and HL-2A, as well as scalable synthetic data generation. Experiments demonstrate that Onion reduces the relative error (E_1) by 71% on synthetic data and by 27% on experimental data. The embedded physical constraints significantly enhance back-projection fidelity and cross-device generalizability. Overall, Onion provides a high-accuracy, end-to-end differentiable inversion framework suitable for real-time plasma diagnostics in fusion experiments.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Non-convex Optimization

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Rapid reconstruction of 2D plasma profiles from line-integral measurements is important in nuclear fusion. This paper introduces a physics-informed model architecture called Onion, that can enhance the performance of models and be adapted to various backbone networks. The model under Onion incorporates physical information by a multiplication process and applies the physics-informed loss function according to the principle of line integration. Experimental results demonstrate that the additional input of physical information improves the model's ability, leading to a reduction in the average relative error E_1 between the reconstruction profiles and the target profiles by approximately 52% on synthetic datasets and about 15% on experimental datasets. Furthermore, the implementation of the Softplus activation function in the final two fully connected layers improves model performance. This enhancement results in a reduction in the E_1 by approximately 71% on synthetic datasets and about 27% on experimental datasets. The incorporation of the physics-informed loss function has been shown to correct the model's predictions, bringing the back-projections closer to the actual inputs and reducing the errors associated with inversion algorithms. Besides, we have developed a synthetic data model to generate customized line-integral diagnostic datasets and have also collected soft x-ray diagnostic datasets from EAST and HL-2A. This study achieves reductions in reconstruction errors, and accelerates the development of diagnostic surrogate models in fusion research.

Problem

• Reconstructs 2D plasma profiles efficiently
• Incorporates physical information in deep learning
• Reduces errors in fusion diagnostics models

Innovation

• Physics-informed deep learning model
• Onion architecture with multiplication process
• Softplus activation function enhancement