Abstract Details
Abstracts
Author: William F Messenger
Requested Type: Poster
Submitted: 2024-04-11 19:49:56
Co-authors: Brooks Howe, Luca Guazzotto, Dave Maurer, Evdokiya Kostadinova
Contact Info:
Auburn University
380 Duncan Dr
Auburn, Alabama 36849
United States of Ame
Abstract Text:
Disruptions of fusion plasma pose a major challenge to realizing nuclear fusion as a source of energy. To combat this challenge, it is crucial to understand the conditions and mechanisms causing disruptions for different magnetic field configurations. The goal of this research is to identify fundamental physical mechanisms leading to disruptions in fusion plasmas with 3D magnetic fields using data produced by Auburn University's Compact Toroidal Hybrid (CTH) plasma device. The CTH device can produce both stellarator-like and tokamak-like magnetic field configurations, which allows to investigate how three-dimensional magnetic field topology affects stability. Here we present initial results from the Machine Learning (ML) analysis of CTH discharges and discuss possible disruption mechanisms. The chosen algorithm to conduct this analysis is Sparse Identification of Nonlinear Dynamical systems (SINDy), a statistical learning model based on sparse regression. This model will allow us to identify dominant terms in a system that may impact the likelihood of disruptions. Specifically, we focus on electron density and temperature measures. Finally, we discuss implementation of CTH routines in disruption-py – a tool designed for efficient retrieval of large-scale tabular shot data from MDSplus, focusing on parameters used for analyzing disruptions.
Comments: