Machine-Learning enabled analysis of ELM filament dynamics in KSTAR

TL;DR

This paper introduces a deep learning model that automatically detects and analyzes ELM filament dynamics in tokamak plasmas, providing insights into their behavior during ELM events.

Contribution

A novel deep convolutional neural network trained on extensive ECEI data for automatic identification and analysis of ELM filaments in tokamak plasmas.

Findings

Model achieves 93.7% precision in filament detection

Identifies quasi-periodic oscillations in filament properties

Reveals correlations between filament dynamics and ELM phases

Abstract

The emergence and dynamics of filamentary structures associated with edge-localized modes (ELMs) inside tokamak plasmas during high-confinement mode is regularly studied using Electron Cyclotron Emission Imaging (ECEI) diagnostic systems. ECEI allows inference of electron temperature variations, often across a poloidal cross-section. Previously, detailed analyses of filamentary dynamics and classification of the precursors to ELM crashes have been done manually. We present a machine-learning-based model, capable of automatically identifying the position, spatial extent, and amplitude of ELM filaments. The model is a deep convolutional neural network that has been trained and optimized on an extensive set of manually labeled ECEI data from the KSTAR tokamak. Once trained, the model achieves a 93.7% precision and allows to robustly identify plasma filaments in unseen ECEI data. The trained model is used to characterize ELM filament dynamics in a single H-mode plasma shot. We identify quasi-periodic oscillations of the filaments' size, total heat content, and radial velocity. The detailed dynamics of these quantities appear strongly correlated with each other and appear qualitatively different during the pre-crash and ELM crash phases.