Gaussian process tomography based on Bayesian data analysis for soft x-ray and AXUV diagnostics on EAST*

Abstract

This work presents the Gaussian process tomography (GPT) based on Bayesian data analysis and its applications in soft x-ray (SXR) and absolute extreme ultraviolet spectroscopy (AXUV) diagnostics on experimental advanced superconducting tokamak (EAST). This is the first application of the GPT method in the AXUV diagnostic system in fusion devices. It is found that even if only horizontal detector arrays are used to reconstruct the two-dimensional (2D) distribution of SXR and AXUV emissivity fields, the GPT method performs robustly and extremely fast, which enables the GPT method to provide real-time feedback on impurity transport and fast magnetohydrodynamics (MHD) events. By reconstructing SXR emissivity in the poloidal cross section on EAST, an m/n = 1/1 internal kink mode has been observed, and the plasma redistribution due to the kink mode is clearly visible in the reconstructions, where m is the poloidal mode number and n is the toroidal mode number. Sawtooth-like internal disruptions extended throughout the entire plasma core and mainly driven by the m/n = 2/1 mode have been acquired. During the sawtooth-like internal disruption crash phase, the conversion from an m = 2 mode to an m = 1 mode is observed. Using the reconstructed AXUV emissivity field we were able to observe the process of impurity accumulated in the plasma core and the mitigation of core impurity due to neon injection in the plasma edge. The data from all other diagnostics involved in the analysis shows that the reconstructions from AXUV measurements are reliable.

Investigation of the compatibility of pellet fueling with ELM-free H-mode plasmas in EAST tokamak

“…An extreme ultra-violet (EUV) spectrometer is used to monitor the highly ionized impurity content by measuring the EUV line emission in the wavelength range of 2-50 nm [27]. A 46channel soft x-ray (SXR) array is used to measure the SXR emission from the core plasma [28]. Moreover, some routine diagnostics, such as mirnov probes, EUV and filterscope are also used.…”

A Gaussian process guide for signal regression in magnetic fusion

“…GPR has been getting increased attention in recent years in fusion and plasma applications. In [11] GPs are used to infer soft x-ray emissivity distributions from noisy line integral measurements, while in [12,13] GPs are similarly used to develop a GPs tomography method to reconstruct local soft xray emissivity in WEST and EAST, and similar approaches in ASDEX [14] upgrade and Wendelstein 7-X [15]. In [16][17][18] GPRs are applied to simulation studies with an eye towards eventual application to full experimental regimes, while in [19] studies are performed on three-dimensional equilibrium reconstruction from both experimental data as well as simulation codes.…”

Application of deep learning to soft x-ray tomography at EAST

“…Many reconstruction techniques have been carried out, such as the analytical method [4], pixel method [5] or a mixture. In the EAST tokamak, the Fourier-Bessel analytical method [1,6] or Gaussian process tomography [7] are applied for SXT. Unfortunately, the traditional SXT methods cannot avoid a heavy time cost because of the multi-loop computation, which is the major difficulty of real-time SXT for tokamak operation.…”