Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks

Rea, Christina; Montes, Kevin J.; Pau, A.; Granetz, R. S.; Sauter, O.

doi:10.1080/15361055.2020.1798589

Title: Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks

Abstract

Here in this paper we lay the groundwork for a robust cross-device comparison of data-driven disruption prediction algorithms on DIII-D and JET tokamaks. In order to consistently carry on a comparative analysis, we define physics-based indicators of disruption precursors based on temperature, density, and radiation profiles that are currently not used in many other machine learning predictors for DIII-D data. These profile-based indicators are shown to well-describe impurity accumulation events in both DIII-D and JET discharges that eventually disrupt. The univariate analysis of the features used as input signals in the data-driven algorithms applied on the data of both tokamaks statistically highlights the differences in the dominant disruption precursors. JET with its ITER-like wall is more prone to impurity accumulation events, while DIII-D is more subject to edge-cooling mechanisms that destabilize dangerous magnetohydrodynamic modes. Even though the analyzed data sets are characterized by such intrinsic differences, we show through a few examples that the inclusion of physics-based disruption markers in data-driven algorithms is a promising path toward the realization of a uniform framework to predict and interpret disruptive scenarios across different tokamaks. As long as the destabilizing precursors are diagnosed in a device-independent way, the knowledge that data-driven algorithms learnmore »« less

Authors:

^[1];

^[2];

^[1];

^[3]

Massachusetts Inst. of Technology (MIT), Cambridge, MA (United States). Plasma Science and Fusion Center
Ecole Polytechnique Federale Lausanne (Switzlerland). Swiss Plasma Center
Ecole Polytechnique Federale Lausanne (Switzlerland)

Publication Date:: Mon Sep 21 00:00:00 EDT 2020

Research Org.:: General Atomics, San Diego, CA (United States)

Sponsoring Org.:: USDOE; EURATOM; Swiss National Science Foundation (SNF)

OSTI Identifier:: 1661188

Grant/Contract Number:: FC02-04ER54698; SC0014264; 633053

Resource Type:: Accepted Manuscript

Journal Name:: Fusion Science and Technology

Additional Journal Information:: Journal Volume: 76; Journal Issue: 8; Journal ID: ISSN 1536-1055

Publisher:: American Nuclear Society

Country of Publication:: United States

Language:: English

Subject:: 70 PLASMA PHYSICS AND FUSION TECHNOLOGY; Disruption; machine learning; DIII-D, JET

Citation Formats


                    Rea, Christina, Montes, Kevin J., Pau, A., Granetz, R. S., and Sauter, O. Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks.  United States: N. p., 2020. 
Web.  doi:10.1080/15361055.2020.1798589.

Copy to clipboard


                    Rea, Christina, Montes, Kevin J., Pau, A., Granetz, R. S., & Sauter, O. Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks.  United States.  https://doi.org/10.1080/15361055.2020.1798589

Copy to clipboard


                    Rea, Christina, Montes, Kevin J., Pau, A., Granetz, R. S., and Sauter, O. Mon .  
"Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks".  United States.  https://doi.org/10.1080/15361055.2020.1798589.  https://www.osti.gov/servlets/purl/1661188.

Copy to clipboard


                    
@article{osti_1661188,

  title        = {Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks},

  author       = {Rea, Christina and Montes, Kevin J. and Pau, A. and Granetz, R. S. and Sauter, O.},

  abstractNote = {Here in this paper we lay the groundwork for a robust cross-device comparison of data-driven disruption prediction algorithms on DIII-D and JET tokamaks. In order to consistently carry on a comparative analysis, we define physics-based indicators of disruption precursors based on temperature, density, and radiation profiles that are currently not used in many other machine learning predictors for DIII-D data. These profile-based indicators are shown to well-describe impurity accumulation events in both DIII-D and JET discharges that eventually disrupt. The univariate analysis of the features used as input signals in the data-driven algorithms applied on the data of both tokamaks statistically highlights the differences in the dominant disruption precursors. JET with its ITER-like wall is more prone to impurity accumulation events, while DIII-D is more subject to edge-cooling mechanisms that destabilize dangerous magnetohydrodynamic modes. Even though the analyzed data sets are characterized by such intrinsic differences, we show through a few examples that the inclusion of physics-based disruption markers in data-driven algorithms is a promising path toward the realization of a uniform framework to predict and interpret disruptive scenarios across different tokamaks. As long as the destabilizing precursors are diagnosed in a device-independent way, the knowledge that data-driven algorithms learn on one device can be re-used to explain a disruptive behavior on another device.},

  doi          = {10.1080/15361055.2020.1798589},

  journal      = {Fusion Science and Technology},

  number       = 8,

  volume       = 76,

  place        = {United States},

  year         = {Mon Sep 21 00:00:00 EDT 2020},

  month        = {Mon Sep 21 00:00:00 EDT 2020}

}

Copy to clipboard

Journal Article:

Free Publicly Available Full Text

Accepted Manuscript (DOE)

Publisher's Version of Record

https://doi.org/10.1080/15361055.2020.1798589

Other availability

Search WorldCat to find libraries that may hold this journal

Save / Share:

Export Metadata

Save to My Library

Works referenced in this record:

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

Meneghini, O.; Smith, S. P.; Lao, L. L.
Nuclear Fusion, Vol. 55, Issue 8
DOI: 10.1088/0029-5515/55/8/083008

A machine learning approach based on generative topographic mapping for disruption prevention and avoidance at JET
journal, August 2019

Pau, A.; Fanni, A.; Carcangiu, S.
Nuclear Fusion, Vol. 59, Issue 10
DOI: 10.1088/1741-4326/ab2ea9

Deep Learning for Plasma Tomography and Disruption Prediction From Bolometer Data
journal, January 2020

Ferreira, Diogo R.; Carvalho, Pedro J.; Fernandes, Horacio
IEEE Transactions on Plasma Science, Vol. 48, Issue 1
DOI: 10.1109/TPS.2019.2947304

2D tomography with bolometry in DIII‐D ^a)
journal, February 1995

Leonard, A. W.; Meyer, W. H.; Geer, B.
Review of Scientific Instruments, Vol. 66, Issue 2
DOI: 10.1063/1.1146006

Statistical analysis of disruptions in JET
journal, April 2009

de Vries, P. C.; Johnson, M. F.; Segui, I.
Nuclear Fusion, Vol. 49, Issue 5
DOI: 10.1088/0029-5515/49/5/055011

Thomson scattering diagnostic upgrade on DIII-D
journal, October 2010

Ponce-Marquez, D. M.; Bray, B. D.; Deterly, T. M.
Review of Scientific Instruments, Vol. 81, Issue 10
DOI: 10.1063/1.3495759

A First Analysis of JET Plasma Profile-Based Indicators for Disruption Prediction and Avoidance
journal, July 2018

Pau, A.; Fanni, A.; Cannas, B.
IEEE Transactions on Plasma Science, Vol. 46, Issue 7
DOI: 10.1109/TPS.2018.2841394

Survey of disruption causes at JET
journal, April 2011

de Vries, P. C.; Johnson, M. F.; Alper, B.
Nuclear Fusion, Vol. 51, Issue 5
DOI: 10.1088/0029-5515/51/5/053018

Review of MARFE phenomena in tokamaks
journal, February 1987

Lipschultz, B.
Journal of Nuclear Materials, Vol. 145-147
DOI: 10.1016/0022-3115(87)90306-0

Statistical analysis of m / n = 2/1 locked and quasi-stationary modes with rotating precursors at DIII-D
journal, November 2016

Sweeney, R.; Choi, W.; La Haye, R. J.
Nuclear Fusion, Vol. 57, Issue 1
DOI: 10.1088/0029-5515/57/1/016019

Scaling of the MHD perturbation amplitude required to trigger a disruption and predictions for ITER
journal, December 2015

de Vries, P. C.; Pautasso, G.; Nardon, E.
Nuclear Fusion, Vol. 56, Issue 2
DOI: 10.1088/0029-5515/56/2/026007

An advanced disruption predictor for JET tested in a simulated real-time environment
journal, January 2010

Rattá, G. A.; Vega, J.; Murari, A.
Nuclear Fusion, Vol. 50, Issue 2
DOI: 10.1088/0029-5515/50/2/025005

A reduced resistive wall mode kinetic stability model for disruption forecasting
journal, May 2017

Berkery, J. W.; Sabbagh, S. A.; Bell, R. E.
Physics of Plasmas, Vol. 24, Issue 5
DOI: 10.1063/1.4977464

ITER on the road to fusion energy
journal, December 2009

Ikeda, Kaname
Nuclear Fusion, Vol. 50, Issue 1
DOI: 10.1088/0029-5515/50/1/014002

Methods for interpreting and understanding deep neural networks
journal, February 2018

Montavon, Grégoire; Samek, Wojciech; Müller, Klaus-Robert
Digital Signal Processing, Vol. 73
DOI: 10.1016/j.dsp.2017.10.011

Tokamak-agnostic actuator management for multi-task integrated control with application to TCV and ITER
journal, October 2019

Vu, N. M. Trang; Blanken, T. C.; Felici, F.
Fusion Engineering and Design, Vol. 147
DOI: 10.1016/j.fusengdes.2019.111260

A real-time machine learning-based disruption predictor in DIII-D
journal, July 2019

Rea, C.; Montes, K. J.; Erickson, K. G.
Nuclear Fusion, Vol. 59, Issue 9
DOI: 10.1088/1741-4326/ab28bf

Assessment of controllers and scenario control performance for ITER first plasma
journal, September 2019

Walker, Michael L.; Welander, Anders; Humphreys, David
Fusion Engineering and Design, Vol. 146
DOI: 10.1016/j.fusengdes.2019.03.050

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

Rea, Cristina; Granetz, Robert S.
Fusion Science and Technology, Vol. 74, Issue 1-2
DOI: 10.1080/15361055.2017.1407206

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

Vega, Jesús; Dormido-Canto, Sebastián; López, Juan M.
Fusion Engineering and Design, Vol. 88, Issue 6-8
DOI: 10.1016/j.fusengdes.2013.03.003

Overview of the SPARC tokamak
journal, September 2020

Creely, A. J.; Greenwald, M. J.; Ballinger, S. B.
Journal of Plasma Physics, Vol. 86, Issue 5
DOI: 10.1017/S0022377820001257

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

Kates-Harbeck, Julian; Svyatkovskiy, Alexey; Tang, William
Nature, Vol. 568, Issue 7753
DOI: 10.1038/s41586-019-1116-4

Advanced tokamak operation using the DIII–D plasma control system
journal, September 2003

Humphreys, D. A.; Ferron, J. R.; Garofalo, A. M.
Fusion Engineering and Design, Vol. 66-68
DOI: 10.1016/S0920-3796(03)00322-3

Direct measurements of internal structures of born-locked modes and the key role in triggering tokamak disruptions
journal, April 2019

Du, X. D.; Shafer, M. W.; Hu, Q. M.
Physics of Plasmas, Vol. 26, Issue 4
DOI: 10.1063/1.5085329

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

Montes, K. J.; Rea, C.; Granetz, R. S.
Nuclear Fusion, Vol. 59, Issue 9
DOI: 10.1088/1741-4326/ab1df4

Automatic disruption classification at JET: comparison of different pattern recognition techniques
journal, May 2006

Cannas, B.; Cau, F.; Fanni, A.
Nuclear Fusion, Vol. 46, Issue 7
DOI: 10.1088/0029-5515/46/7/002

The impact of the ITER-like wall at JET on disruptions
journal, November 2012

de Vries, P. C.; Arnoux, G.; Huber, A.
Plasma Physics and Controlled Fusion, Vol. 54, Issue 12
DOI: 10.1088/0741-3335/54/12/124032

Progress in disruption prevention for ITER
journal, June 2019

Strait, E. J.; Barr, J. L.; Baruzzo, M.
Nuclear Fusion, Vol. 59, Issue 11
DOI: 10.1088/1741-4326/ab15de

Similar Records in DOE PAGES and OSTI.GOV collections:

Progress Towards Interpretable Machine Learning-based Disruption Predictors Across Tokamaks

Dataset Rea, C. ; Montes, K. J. ; Pau, A. ; ...

In this paper we lay the groundwork for a robust cross-device comparison of data-driven disruption prediction algorithms on DIII-D and JET tokamaks. In order to consistently carry on a comparative analysis, we define physics-based indicators of disruption precursors based on temperature, density, and radiation profiles that are currently missing for DIII-D data. These profile-based indicators are shown to well-describe impurity accumulation events in both DIII-D and JET discharges that eventually disrupt. Thanks to the univariate analysis on the features used in such data-driven applications on both tokamaks, we are able to statistically highlight differences in the dominant disruption precursors: JETmore »« less
https://doi.org/10.7910/DVN/QRZOQ3

View Dataset
Integrated deep learning framework for unstable event identification and disruption prediction of tokamak plasmas

Journal Article Zhu, J. X. ; Rea, C. ; Granetz, R. S. ; ... - Nuclear Fusion

Abstract The ability to identify underlying disruption precursors is key to disruption avoidance. In this paper, we present an integrated deep learning (DL) based model that combines disruption prediction with the identification of several disruption precursors like rotating modes, locked modes, H-to-L back transitions and radiative collapses. The first part of our study demonstrates that the DL-based unstable event identifier trained on 160 manually labeled DIII-D shots can achieve, on average, 84% event identification rate of various frequent unstable events (like H-L back transition, locked mode, radiative collapse, rotating MHD mode, large sawtooth crash), and the trained identifier can bemore »« less
https://doi.org/10.1088/1741-4326/acb803
Disruption prediction investigations using Machine Learning tools on DIII-D and Alcator C-Mod

Journal Article Rea, C. ; Granetz, R. S. ; Montes, K. ; ... - Plasma Physics and Controlled Fusion

Using data-driven methodology, we exploit the time series of relevant plasma parameters for a large set of disrupted and non-disrupted discharges to develop a classification algorithm for detecting disruptive phases in shots that eventually disrupt. Comparing the same methodology on different devices is crucial in order to have information on the portability of the developed algorithm and the possible extrapolation to ITER. Therefore, we use data from two very different tokamaks, DIII-D and Alcator C-Mod. We then focus on a subset of disruption predictors, most of which are dimensionless and/or machine-independent parameters, coming from both plasma diagnostics and equilibrium reconstructions,more »« less
Cited by 47
https://doi.org/10.1088/1361-6587/aac7fe

Full Text Available
DIII-D Studies of Massive Gas Injection Fast Shutdowns for Disruption Mitigation

Conference Hollmann, E ; Jernigan, T ; Antar, G ; ...

Injection of massive quantities of gas is a promising technique for fast shutdown of ITER for the purpose of avoiding divertor and first wall damage from disruptions. Previous experiments using massive gas injection (MGI) to terminate discharges in the DIII-D tokamak have demonstrated rapid shutdown with reduced wall heating and halo currents (relative to natural disruptions) and with very small runaway electron (RE) generation [1]. Figure 1 shows time traces which give an overview of shutdown time scales. Typically, of order 5 x 10{sup 22} Ar neutrals are fired over a pulse of 25 ms duration into stationary (non-disrupting) discharges.more »« less
Full Text Available
Implementation of AI/DEEP learning disruption predictor into a plasma control system

Journal Article Tang, William ; Dong, Ge ; Barr, Jayson ; ... - Contributions to Plasma Physics

This paper reports on advances in the state-of-the-art deep learning disruption prediction models based on the Fusion Recurrent Neural Network (FRNN) originally introduced in a 2019 NATURE publication [https://doi.org/10.1038/s41586-019-1116-4]. In particular, the predictor now features not only the “disruption score,” as an indicator of the probability of an imminent disruption, but also a “sensitivity score” in real time to indicate the underlying reasons for the imminent disruption. This adds valuable physics interpretability for the deep learning model and can provide helpful guidance for control actuators now implemented into a modern plasma control system (PCS). The advance is a significant stepmore »« less
https://doi.org/10.1002/ctpp.202200095

Similar Records

Title: Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks

Abstract

Citation Formats

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

A machine learning approach based on generative topographic mapping for disruption prevention and avoidance at JET journal, August 2019

Deep Learning for Plasma Tomography and Disruption Prediction From Bolometer Data journal, January 2020

2D tomography with bolometry in DIII‐D a) journal, February 1995

Statistical analysis of disruptions in JET journal, April 2009

Thomson scattering diagnostic upgrade on DIII-D journal, October 2010

A First Analysis of JET Plasma Profile-Based Indicators for Disruption Prediction and Avoidance journal, July 2018

Survey of disruption causes at JET journal, April 2011

Review of MARFE phenomena in tokamaks journal, February 1987

Statistical analysis of m / n = 2/1 locked and quasi-stationary modes with rotating precursors at DIII-D journal, November 2016

Scaling of the MHD perturbation amplitude required to trigger a disruption and predictions for ITER journal, December 2015

An advanced disruption predictor for JET tested in a simulated real-time environment journal, January 2010

A reduced resistive wall mode kinetic stability model for disruption forecasting journal, May 2017

ITER on the road to fusion energy journal, December 2009

Methods for interpreting and understanding deep neural networks journal, February 2018

Tokamak-agnostic actuator management for multi-task integrated control with application to TCV and ITER journal, October 2019

A real-time machine learning-based disruption predictor in DIII-D journal, July 2019

Assessment of controllers and scenario control performance for ITER first plasma journal, September 2019

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

Overview of the SPARC tokamak journal, September 2020

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

Advanced tokamak operation using the DIII–D plasma control system journal, September 2003

Direct measurements of internal structures of born-locked modes and the key role in triggering tokamak disruptions journal, April 2019

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

Automatic disruption classification at JET: comparison of different pattern recognition techniques journal, May 2006

The impact of the ITER-like wall at JET on disruptions journal, November 2012

Progress in disruption prevention for ITER journal, June 2019

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

A machine learning approach based on generative topographic mapping for disruption prevention and avoidance at JET
journal, August 2019

Deep Learning for Plasma Tomography and Disruption Prediction From Bolometer Data
journal, January 2020

2D tomography with bolometry in DIII‐D ^a)
journal, February 1995

Statistical analysis of disruptions in JET
journal, April 2009

Thomson scattering diagnostic upgrade on DIII-D
journal, October 2010

A First Analysis of JET Plasma Profile-Based Indicators for Disruption Prediction and Avoidance
journal, July 2018

Survey of disruption causes at JET
journal, April 2011

Review of MARFE phenomena in tokamaks
journal, February 1987

Statistical analysis of m / n = 2/1 locked and quasi-stationary modes with rotating precursors at DIII-D
journal, November 2016

Scaling of the MHD perturbation amplitude required to trigger a disruption and predictions for ITER
journal, December 2015

An advanced disruption predictor for JET tested in a simulated real-time environment
journal, January 2010

A reduced resistive wall mode kinetic stability model for disruption forecasting
journal, May 2017

ITER on the road to fusion energy
journal, December 2009

Methods for interpreting and understanding deep neural networks
journal, February 2018

Tokamak-agnostic actuator management for multi-task integrated control with application to TCV and ITER
journal, October 2019

A real-time machine learning-based disruption predictor in DIII-D
journal, July 2019

Assessment of controllers and scenario control performance for ITER first plasma
journal, September 2019

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

Overview of the SPARC tokamak
journal, September 2020

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

Advanced tokamak operation using the DIII–D plasma control system
journal, September 2003

Direct measurements of internal structures of born-locked modes and the key role in triggering tokamak disruptions
journal, April 2019

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

Automatic disruption classification at JET: comparison of different pattern recognition techniques
journal, May 2006

The impact of the ITER-like wall at JET on disruptions
journal, November 2012

Progress in disruption prevention for ITER
journal, June 2019