skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Deep convolutional neural networks for multi-scale time-series classification and application to tokamak disruption prediction using raw, high temporal resolution diagnostic data

Abstract

In this paper we discuss recent advances in deep convolutional neural networks (CNN) for sequence learning, which allow identifying long-range, multi-scale phenomena in long sequences, such as those found in fusion plasmas. We point out several benefits of these deep CNN architectures, such as not requiring experts such as physicists to hand-craft input data features, the ability to capture longer range dependencies compared to the more common sequence neural networks (recurrent neural networks like long short-term memory (LSTM) networks), and the comparative computational efficiency. We apply this neural network architecture to the popular problem of disruption prediction in fusion energy tokamaks, utilizing raw data from a single diagnostic, the Electron Cyclotron Emission imaging (ECEi) diagnostic from the DIII-D tokamak. Initial results trained on a large ECEi dataset show promise, achieving an F1-score of ~91% on individual time-slices using only the ECEi data. This indicates the ECEi diagnostic by itself can be sensitive to a number of pre-disruption markers useful for predicting disruptions on timescales not only for mitigation but also avoidance. Future opportunities for utilizing these deep CNN architectures with fusion data are outlined, including impact of recent upgrades to the ECEi diagnostic.

Authors:
ORCiD logo [1];  [2]; ORCiD logo [3]
  1. Princeton Plasma Physics Lab. (PPPL), Princeton, NJ (United States)
  2. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
  3. Univ. of California, Davis, CA (United States)
Publication Date:
Research Org.:
General Atomics, San Diego, CA (United States); Princeton Plasma Physics Lab. (PPPL), Princeton, NJ (United States); Univ. of California, Davis, CA (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Fusion Energy Sciences (FES)
Contributing Org.:
DIII-D team
OSTI Identifier:
1708849
Alternate Identifier(s):
OSTI ID: 1633375; OSTI ID: 1642913; OSTI ID: 1777989
Grant/Contract Number:  
FC02-04ER54698; AC02–09CH11466; FG02-99ER54531; AC02-09CH11466
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
Physics of Plasmas
Additional Journal Information:
Journal Volume: 27; Journal Issue: 6; Journal ID: ISSN 1070-664X
Publisher:
American Institute of Physics (AIP)
Country of Publication:
United States
Language:
English
Subject:
70 PLASMA PHYSICS AND FUSION TECHNOLOGY; Optimization algorithms; Artificial intelligence; Artificial neural networks; Plasma confinement; Fusion energy; Machine learning; Tokamaks

Citation Formats

Churchill, R. M., Tobias, B., and Zhu, Y. Deep convolutional neural networks for multi-scale time-series classification and application to tokamak disruption prediction using raw, high temporal resolution diagnostic data. United States: N. p., 2020. Web. doi:10.1063/1.5144458.
Churchill, R. M., Tobias, B., & Zhu, Y. Deep convolutional neural networks for multi-scale time-series classification and application to tokamak disruption prediction using raw, high temporal resolution diagnostic data. United States. https://doi.org/10.1063/1.5144458
Churchill, R. M., Tobias, B., and Zhu, Y. 2020. "Deep convolutional neural networks for multi-scale time-series classification and application to tokamak disruption prediction using raw, high temporal resolution diagnostic data". United States. https://doi.org/10.1063/1.5144458. https://www.osti.gov/servlets/purl/1708849.
@article{osti_1708849,
title = {Deep convolutional neural networks for multi-scale time-series classification and application to tokamak disruption prediction using raw, high temporal resolution diagnostic data},
author = {Churchill, R. M. and Tobias, B. and Zhu, Y.},
abstractNote = {In this paper we discuss recent advances in deep convolutional neural networks (CNN) for sequence learning, which allow identifying long-range, multi-scale phenomena in long sequences, such as those found in fusion plasmas. We point out several benefits of these deep CNN architectures, such as not requiring experts such as physicists to hand-craft input data features, the ability to capture longer range dependencies compared to the more common sequence neural networks (recurrent neural networks like long short-term memory (LSTM) networks), and the comparative computational efficiency. We apply this neural network architecture to the popular problem of disruption prediction in fusion energy tokamaks, utilizing raw data from a single diagnostic, the Electron Cyclotron Emission imaging (ECEi) diagnostic from the DIII-D tokamak. Initial results trained on a large ECEi dataset show promise, achieving an F1-score of ~91% on individual time-slices using only the ECEi data. This indicates the ECEi diagnostic by itself can be sensitive to a number of pre-disruption markers useful for predicting disruptions on timescales not only for mitigation but also avoidance. Future opportunities for utilizing these deep CNN architectures with fusion data are outlined, including impact of recent upgrades to the ECEi diagnostic.},
doi = {10.1063/1.5144458},
url = {https://www.osti.gov/biblio/1708849}, journal = {Physics of Plasmas},
issn = {1070-664X},
number = 6,
volume = 27,
place = {United States},
year = {2020},
month = {6}
}

Works referenced in this record:

Wavelet Transforms and their Applications to Turbulence
journal, January 1992


Integrated modeling applications for tokamak experiments with OMFIT
journal, July 2015


Reconstruction of current profile parameters and plasma shapes in tokamaks
journal, November 1985


Commissioning of electron cyclotron emission imaging instrument on the DIII-D tokamak and first data
journal, October 2010


Deep learning
journal, May 2015


Machine learning control for disruption and tearing mode avoidance
journal, February 2020


Estimation of edge electron temperature profiles via forward modelling of the electron cyclotron radiation transport at ASDEX Upgrade
journal, December 2012


A systematic study of the class imbalance problem in convolutional neural networks
journal, October 2018


Relationship between locked modes and thermal quenches in DIII-D
journal, March 2018


A Survey on Transfer Learning
journal, October 2010


Exploratory Machine Learning Studies for Disruption Prediction Using Large Databases on DIII-D
journal, February 2018


Results of the JET real-time disruption predictor in the ITER-like wall campaigns
journal, October 2013


Predicting disruptive instabilities in controlled fusion plasmas through deep learning
journal, April 2019


Long Short-Term Memory
journal, November 1997


Requirements for Triggering the ITER Disruption Mitigation System
journal, April 2016


Imaging Techniques for Microwave Diagnostics
journal, March 2011


Development of KSTAR ECE imaging system for measurement of temperature fluctuations and edge density fluctuations
journal, October 2010


Theory of tokamak disruptions
journal, May 2012


Machine learning for disruption warnings on Alcator C-Mod, DIII-D, and EAST
journal, July 2019


Feature-wise transformations
journal, July 2018


New Trends in Microwave Imaging Diagnostics and Application to Burning Plasma
journal, May 2019


Chapter 3: MHD stability, operational limits and disruptions
journal, June 2007


Mitigation of sawtooth crash as a manifestation of MHD mode coupling prior to disruption of KSTAR plasma
journal, March 2019


Liquid crystal polymer receiver modules for electron cyclotron emission imaging on the DIII-D tokamak
journal, October 2018