Boundary Reconstruction of Tokamak Plasma Based on Deep Neural Networks

Abstract

[Introduction] In order to realize the real-time reconstruction of plasma shape and position in tokamak, a visible light edge reconstruction algorithm based on fully connected neural network is proposed based on the analysis of camera calibration algorithm. [Method] The function of the algorithm was to establish the corresponding relationship between the pixel coordinate system and the tokamak coordinate system, and then realize the plasma visible light edge reconstruction. [Result] On the basis of the algorithm, few-shot learning is added to further improve the reconstruction algorithm of fully connected neural network. [Conclusion] Experimental results show that the algorithm can accurately reconstruct the plasma visible light edge, and also meet the real-time requirements of the system.

Keywords

• tokamak
• camera calibration
• visible light edge reconstruction algorithm
• fully connected neural network
• few-shot learning