Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging turbulence data from fusion experiments

Minjun J. Choi

TL;DR

The paper addresses extracting quantitative turbulence transport information from 2D fusion-plasma fluctuation measurements by organizing methods into spectral, statistical, and PINN-based approaches. It surveys noise-robust spectral diagnostics (frequency, wavenumber, and multi-wave couplings), statistical characterizations (Gaussianity, self-similarity, and chaos), and physics-informed neural networks for missing-field prediction and model validation. Key contributions include detailed practical examples from devices like KSTAR and DIII-D, demonstrations of energy transfer and multi-wave couplings, and a publicly available Python toolkit (fluctana) to implement these techniques. The work advances the practical utilization of 2D turbulence data to diagnose transport processes and validate turbulence models in fusion experiments.

Abstract

Various methods for leveraging turbulent fluctuation measurements from fusion plasma experiments are introduced, along with selected application examples. These can be categorized into spectral methods, statistical methods, and physics informed neural network based methods, and they are most effective for two-dimensional turbulence measurements, which are now widely accessible. Extracting more information from turbulence data would pave the way for a better understanding of plasma turbulence transport in fusion experiments.

Leveraging turbulence data from fusion experiments

TL;DR

The paper addresses extracting quantitative turbulence transport information from 2D fusion-plasma fluctuation measurements by organizing methods into spectral, statistical, and PINN-based approaches. It surveys noise-robust spectral diagnostics (frequency, wavenumber, and multi-wave couplings), statistical characterizations (Gaussianity, self-similarity, and chaos), and physics-informed neural networks for missing-field prediction and model validation. Key contributions include detailed practical examples from devices like KSTAR and DIII-D, demonstrations of energy transfer and multi-wave couplings, and a publicly available Python toolkit (fluctana) to implement these techniques. The work advances the practical utilization of 2D turbulence data to diagnose transport processes and validate turbulence models in fusion experiments.

Abstract

Various methods for leveraging turbulent fluctuation measurements from fusion plasma experiments are introduced, along with selected application examples. These can be categorized into spectral methods, statistical methods, and physics informed neural network based methods, and they are most effective for two-dimensional turbulence measurements, which are now widely accessible. Extracting more information from turbulence data would pave the way for a better understanding of plasma turbulence transport in fusion experiments.
Paper Structure (27 sections, 28 equations, 12 figures)

This paper contains 27 sections, 28 equations, 12 figures.

Table of Contents

  1. Introduction
  2. Spectral methods
  3. Noise suppressed frequency spectrum
  4. Practical example
  5. Frequency spectrum with a long time correlated noise
  6. The local dispersion relation
  7. Practical example
  8. Noise suppressed local wavenumber-frequency spectrum
  9. Practical example
  10. Bicoherence to detect three-wave coupling
  11. Quadratic energy transfer estimation
  12. Practical example
  13. Tricoherence to detect four-wave coupling
  14. Practical example
  15. Statistical methods
  16. ...and 12 more sections

Figures (12)

  • Figure 1: The Fourier transform coefficient in the complex plane and the ensemble average
  • Figure 2: The power-law power spectrum indicates that the event size would exhibit the power-law distribution. Two adjacent channels (square boxes) on the same flux surface are used to reduce noise components in the power spectrum. Reprinted from ChoiNF2019. Copyright 2019 IOP
  • Figure 3: (Color online) (top) Two time series data: $x(t_j) = 2 a_j + \varepsilon_{1j}$ (black) and $y(t_j) = a_j \varepsilon_{2j}$ (blue) where $a_j$ is a sinusoidal signal and $\varepsilon_{1,2}$ are FARIMA(0, 0.3, 0) processes. (bottom) The squared ordinary coherence (black crosses) and squared Laplace coherence (red crosses).
  • Figure 4: (Color online) The local dispersion measurements at different locations to estimate the flow shearing rate outside a magnetic island. The estimated shearing rate is larger near (a) the O-point than (b) the X-point, consistent with the weaker fluctuation power near the O-point than the X-point Choi:2017ez. Reprinted from Choi:2021fs. Copyright 2021 Springer Nature
  • Figure 5: (Color online) The estimated local wavenumber $\overline{K}(f) = \sum_K K \langle S_L(K,f) \rangle_p$ and the spectral width $\sigma_K(f) = \sqrt{\sum_K (K - \overline{K}(f))^2 \langle S_L(K,f) \rangle_p}$ in two different edge localized mode suppression phases A and B by the resonant magnetic perturbation field. Reprinted from ChoiPoP2022. Copyright 2022 AIP
  • ...and 7 more figures