Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET

A. Murari, M. Lungaroni, E. Peluso, P. Gaudio, J. Vega, S. Dormido-Canto, M. Baruzzo, M. Gelfusa

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

Detecting disruptions with sufficient anticipation time is essential to undertake any form of remedial strategy, mitigation or avoidance. Traditional predictors based on machine learning techniques can be very performing, if properly optimised, but do not provide a natural estimate of the quality of their outputs and they typically age very quickly. In this paper a new set of tools, based on probabilistic extensions of support vector machines (SVM), are introduced and applied for the first time to JET data. The probabilistic output constitutes a natural qualification of the prediction quality and provides additional flexibility. An adaptive training strategy ‘from scratch’ has also been devised, which allows preserving the performance even when the experimental conditions change significantly. Large JET databases of disruptions, covering entire campaigns and thousands of discharges, have been analysed, both for the case of the graphite and the ITER Like Wall. Performance significantly better than any previous predictor using adaptive training has been achieved, satisfying even the requirements of the next generation of devices. The adaptive approach to the training has also provided unique information about the evolution of the operational space. The fact that the developed tools give the probability of disruption improves the interpretability of the results, provides an estimate of the predictor quality and gives new insights into the physics. Moreover, the probabilistic treatment permits to insert more easily these classifiers into general decision support and control systems.
Original languageEnglish
Article number056002
Pages (from-to)-
JournalNuclear Fusion
Volume58
Issue number5
DOIs
Publication statusPublished - 2 Mar 2018
Externally publishedYes

Fingerprint

education
predictions
decision support systems
machine learning
avoidance
output
qualifications
estimates
inserts
classifiers
preserving
flexibility
coverings
graphite
requirements
physics

All Science Journal Classification (ASJC) codes

  • Nuclear and High Energy Physics
  • Condensed Matter Physics

Cite this

Murari, A., Lungaroni, M., Peluso, E., Gaudio, P., Vega, J., Dormido-Canto, S., ... Gelfusa, M. (2018). Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET. Nuclear Fusion, 58(5), -. [056002]. https://doi.org/10.1088/1741-4326/aaaf9c
Murari, A. ; Lungaroni, M. ; Peluso, E. ; Gaudio, P. ; Vega, J. ; Dormido-Canto, S. ; Baruzzo, M. ; Gelfusa, M. / Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET. In: Nuclear Fusion. 2018 ; Vol. 58, No. 5. pp. -.
@article{c3791a66d9874da9839c12e3e303efc4,
title = "Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET",
abstract = "Detecting disruptions with sufficient anticipation time is essential to undertake any form of remedial strategy, mitigation or avoidance. Traditional predictors based on machine learning techniques can be very performing, if properly optimised, but do not provide a natural estimate of the quality of their outputs and they typically age very quickly. In this paper a new set of tools, based on probabilistic extensions of support vector machines (SVM), are introduced and applied for the first time to JET data. The probabilistic output constitutes a natural qualification of the prediction quality and provides additional flexibility. An adaptive training strategy ‘from scratch’ has also been devised, which allows preserving the performance even when the experimental conditions change significantly. Large JET databases of disruptions, covering entire campaigns and thousands of discharges, have been analysed, both for the case of the graphite and the ITER Like Wall. Performance significantly better than any previous predictor using adaptive training has been achieved, satisfying even the requirements of the next generation of devices. The adaptive approach to the training has also provided unique information about the evolution of the operational space. The fact that the developed tools give the probability of disruption improves the interpretability of the results, provides an estimate of the predictor quality and gives new insights into the physics. Moreover, the probabilistic treatment permits to insert more easily these classifiers into general decision support and control systems.",
author = "A. Murari and M. Lungaroni and E. Peluso and P. Gaudio and J. Vega and S. Dormido-Canto and M. Baruzzo and M. Gelfusa",
year = "2018",
month = "3",
day = "2",
doi = "10.1088/1741-4326/aaaf9c",
language = "English",
volume = "58",
pages = "--",
journal = "Nuclear Fusion",
issn = "0029-5515",
publisher = "IOP Publishing Ltd.",
number = "5",

}

Murari, A, Lungaroni, M, Peluso, E, Gaudio, P, Vega, J, Dormido-Canto, S, Baruzzo, M & Gelfusa, M 2018, 'Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET', Nuclear Fusion, vol. 58, no. 5, 056002, pp. -. https://doi.org/10.1088/1741-4326/aaaf9c

Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET. / Murari, A.; Lungaroni, M.; Peluso, E.; Gaudio, P.; Vega, J.; Dormido-Canto, S.; Baruzzo, M.; Gelfusa, M.

In: Nuclear Fusion, Vol. 58, No. 5, 056002, 02.03.2018, p. -.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET

AU - Murari, A.

AU - Lungaroni, M.

AU - Peluso, E.

AU - Gaudio, P.

AU - Vega, J.

AU - Dormido-Canto, S.

AU - Baruzzo, M.

AU - Gelfusa, M.

PY - 2018/3/2

Y1 - 2018/3/2

N2 - Detecting disruptions with sufficient anticipation time is essential to undertake any form of remedial strategy, mitigation or avoidance. Traditional predictors based on machine learning techniques can be very performing, if properly optimised, but do not provide a natural estimate of the quality of their outputs and they typically age very quickly. In this paper a new set of tools, based on probabilistic extensions of support vector machines (SVM), are introduced and applied for the first time to JET data. The probabilistic output constitutes a natural qualification of the prediction quality and provides additional flexibility. An adaptive training strategy ‘from scratch’ has also been devised, which allows preserving the performance even when the experimental conditions change significantly. Large JET databases of disruptions, covering entire campaigns and thousands of discharges, have been analysed, both for the case of the graphite and the ITER Like Wall. Performance significantly better than any previous predictor using adaptive training has been achieved, satisfying even the requirements of the next generation of devices. The adaptive approach to the training has also provided unique information about the evolution of the operational space. The fact that the developed tools give the probability of disruption improves the interpretability of the results, provides an estimate of the predictor quality and gives new insights into the physics. Moreover, the probabilistic treatment permits to insert more easily these classifiers into general decision support and control systems.

AB - Detecting disruptions with sufficient anticipation time is essential to undertake any form of remedial strategy, mitigation or avoidance. Traditional predictors based on machine learning techniques can be very performing, if properly optimised, but do not provide a natural estimate of the quality of their outputs and they typically age very quickly. In this paper a new set of tools, based on probabilistic extensions of support vector machines (SVM), are introduced and applied for the first time to JET data. The probabilistic output constitutes a natural qualification of the prediction quality and provides additional flexibility. An adaptive training strategy ‘from scratch’ has also been devised, which allows preserving the performance even when the experimental conditions change significantly. Large JET databases of disruptions, covering entire campaigns and thousands of discharges, have been analysed, both for the case of the graphite and the ITER Like Wall. Performance significantly better than any previous predictor using adaptive training has been achieved, satisfying even the requirements of the next generation of devices. The adaptive approach to the training has also provided unique information about the evolution of the operational space. The fact that the developed tools give the probability of disruption improves the interpretability of the results, provides an estimate of the predictor quality and gives new insights into the physics. Moreover, the probabilistic treatment permits to insert more easily these classifiers into general decision support and control systems.

UR - http://www.scopus.com/inward/record.url?scp=85043604001&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85043604001&partnerID=8YFLogxK

U2 - 10.1088/1741-4326/aaaf9c

DO - 10.1088/1741-4326/aaaf9c

M3 - Article

VL - 58

SP - -

JO - Nuclear Fusion

JF - Nuclear Fusion

SN - 0029-5515

IS - 5

M1 - 056002

ER -

Murari A, Lungaroni M, Peluso E, Gaudio P, Vega J, Dormido-Canto S et al. Adaptive predictors based on probabilistic SVM for real time disruption mitigation on JET. Nuclear Fusion. 2018 Mar 2;58(5):-. 056002. https://doi.org/10.1088/1741-4326/aaaf9c