Pith. sign in

REVIEW 4 major objections 5 minor 53 references

Early Prediction of Current Quench Events in the ADITYA Tokamak using Transformer based Data Driven Models

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A transformer model predicts tokamak current quenches 8–10 ms ahead with recall above 0.9, outperforming LSTM on ADITYA data.

desk verdict The transformer-vs-LSTM comparison is plausible on the filtered subset, but the headline 8–10 ms recall claim doesn't survive contact with the excluded half of the data. read the letter →

arxiv 2507.12797 v2 pith:34BHQ6UN submitted 2025-07-17 physics.plasm-ph

Jyoti Agarwal , Bhaskar Chaudhury , Jaykumar Navadiya , Shrichand Jakhar , Manika Sharma This is my paper · ORCID
classification physics.plasm-ph
keywords tokamakdisruptionpredictioncurrentquenchtransformerencoderLSTMbaselineADITYAmultivariatetimeseriesrecallearly
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that a transformer encoder trained on six ADITYA tokamak diagnostic signals can warn of an upcoming current quench early enough to act. On the filtered dataset, the model keeps recall above 0.9 even when the prediction threshold is 8–10 ms, while the LSTM baseline's recall declines. The authors argue that such lead time is practical for disruption mitigation on short-pulse machines like ADITYA, where the whole discharge lasts about 120 ms. They also report that the transformer stays more stable than LSTM as the warning window widens and across datasets with different disruption-to-nondisruption ratios. This is the first application of transformer architectures to ADITYA data for early current-quench prediction.

What carries the argument

The load-bearing object is the transformer encoder with self-attention, which lets each time step in the 600-step, six-channel diagnostic sequence weigh its dependence on every other step. The model is trained with binary cross-entropy to output a disruption probability for a time 80 steps (16 ms) ahead, and an alarm triggers when the probability exceeds 0.5. The comparison is made against an LSTM with the same input structure, alarm rule, and training procedure.

What would settle it

Apply the same trained models to the original 1407 shots without excluding slow-quench shots and compare recall at an 8–10 ms threshold; a drop below the reported level or below LSTM's would show the result depends on the filtering, while matching performance would support the generalization claim.

Watch

Extended reading notes

Core claim

The central claim is that transformer-based models are better than LSTM models at early current-quench prediction on ADITYA tokamak data. Using six diagnostic signals (plasma current, soft and hard X-rays, bolometer, H-alpha, and C-III), the transformer achieves recall above 0.9 up to a prediction threshold of 8–10 ms, whereas LSTM recall degrades faster with threshold. The transformer also maintains higher accuracy and more balanced precision–recall performance across naturally balanced and bias-adjusted training datasets, which the authors take as evidence of reliable generalization.

Load-bearing premise

The results rely on treating shots whose current quench times fall outside the median by more than 25% as outliers and removing them, but real ADITYA operation includes slow decays and controlled shutdowns, so performance on the filtered set may not carry over to unfiltered operation.

Editorial extensions

If this is right

  • A practical real-time current-quench warning on ADITYA is feasible with a lead time of about 8–10 ms, which the paper argues is enough to trigger mitigation on a 120 ms discharge.
  • Because the transformer keeps recall above 0.9 while LSTM recall declines, fewer actual disruptions would be missed at longer warning windows.
  • The model's stability across biased training sets suggests it can be deployed when operational conditions make disruptive and non-disruptive shots imbalanced.
  • The same architecture could be a starting point for real-time disruption-avoidance systems on other short-pulse tokamaks, though the paper flags that confirmation on long-duration tokamaks is still needed.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If attention weights are examined, the six selected diagnostics may reveal which signals carry the earliest precursor information, but the paper leaves that interpretability step to future work.
  • The 8–10 ms recall result was obtained after removing shots with quench times outside ±25% of the median; on an unfiltered operational dataset that includes slow decays and controlled shutdowns, recall and precision may differ.
  • The paper's suggested 5–10 ms mitigation window is based on an electromagnetic particle injector study; if faster actuators become available, the model's useful threshold could extend further.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper develops a transformer-encoder model for early prediction of current-quench events in the ADITYA tokamak, using six diagnostic time series from 725 selected shots, and compares its performance against an LSTM baseline. The authors report that the transformer maintains recall above 0.9 for prediction thresholds up to 8–10 ms, with greater stability than LSTM as the threshold increases, and they also study the effect of biasing the training set toward disruptive shots. The claimed contribution is the first transformer-based disruption predictor for ADITYA, with practical feasibility for short-pulse tokamaks.

Significance. If the central claims hold, the paper is a useful incremental contribution to tokamak disruption prediction: it applies a modern transformer architecture to ADITYA data, makes a direct quantitative comparison with LSTM, and explicitly addresses dataset diversity and class-imbalance sensitivity. The use of a larger ADITYA dataset than prior work and the attention to evaluation at multiple prediction thresholds are strengths, as is the clear statement that recall, not just accuracy, is the key operational metric. The main significance depends on whether the filtered dataset and the evaluation protocol support the strong practical-feasibility claim for real ADITYA operations.

major comments (4)
  1. [§2.1.3, §3.1] The data filtering step removes all shots with current-quench times outside ±25% of the median (~54–90 ms), reducing the dataset from 1407 to 725 shots. The paper itself notes in §2.1.2 that shots with quench times below 65 ms are often disruptive, and §2.1.3 states that high-quench-time events are often non-disruptive terminations or controlled shutdowns. Thus the evaluation in §3.1, which uses only the filtered 725-shot set, excludes both the early disruptive shots and the slow shutdowns that a deployed predictor on ADITYA would encounter. The abstract's claim of recall above 0.9 up to 8–10 ms and the claim of 'practical feasibility' are therefore established only for an artificially separated two-cluster distribution, not for the full distribution of ADITYA discharges. The authors should either report performance on the excluded shots as a separate class or distribution, or substantially qualify the practical-feasibility claim.
  2. [§2.1.3, §2.2.3] The exact label-encoding rule is never specified. Section 2.2.1 states that the target value 'encodes the proximity to a disruption event,' with values near 0 for normal operation and near 1 for imminent disruption, but no formula or algorithm is given for assigning y_t ∈ {0,1} to each time step, nor how the 80-step alignment and the 0.5 alarm threshold interact across shots of different lengths. Without a precise definition of the labeling rule, the reported precision/recall/accuracy numbers are not reproducible, and the TP/TN/FP/FN definitions in §2.2.4 are ambiguous for shots where quench occurs at intermediate times. The authors should provide the exact labeling procedure, including how the prediction horizon and the alarm threshold are applied in time.
  3. [§3.1] The test set for the threshold study consists of a single 30% random split (123 disruptive and 93 non-disruptive shots), and no confidence intervals, standard deviations, or repeated-seed results are reported. The observed differences between transformer and LSTM, particularly the claimed recall advantage of 8–10 ms, could be sensitive to the particular split and to the small number of non-disruptive test shots. The authors should report error bars or multiple-seed statistics, or at minimum state the number of runs over which the reported curves are averaged.
  4. [Abstract and §3.1] The abstract states that the transformer maintains recall 'above 0.9 even up to a prediction threshold of 8-10 ms,' but the results section does not report the actual numerical recall values; Figure 10 is the only evidence, and its values are hard to read from the figure as described. The authors should state the exact recall, precision, and accuracy values at thresholds of, say, 5, 8, 10, and 16 ms, so that the quantitative claim in the abstract is directly verifiable.
minor comments (5)
  1. [§2.1.3] The text says 'This removal is purely mathematical' but also argues that it 'improves the reliability of identifying truly non-disruptive cases'; these two statements sit in tension and should be clarified, since the removal is not neutral with respect to the physical interpretation.
  2. [§2.2.1] The notation in Equation (3) uses y_{t+80} both as the hidden state and as part of the input-output alignment; this is confusing and should be redefined, for example as h_{t+80} for the hidden state.
  3. [§2.1.4] The number 600 time steps is derived from 120 ms at 5 kHz, but the text says 'negative plasma current values are replaced with zero'; if this is a physical correction, it should be justified, and if it is an artifact removal, the fraction of affected shots/time steps should be reported.
  4. [§1, §2.1.1] There are several typographical errors and inconsistencies: 'Covolution' in the introduction, 'ADITY A' with a space throughout, and the diagnostic name 'C-111' versus 'C-III' in different places. These should be corrected.
  5. [§2.1.5] The diversity analysis computes Euclidean distance only on plasma current signals, but the models use six features; the claim that this measures diversity for the full feature space should be softened or supported with a comparison using the full feature set.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the transformer-vs-LSTM comparison is an internal benchmark on held-out shots, and no prediction reduces to its training labels by construction.

full rationale

The central claim—transformer recall above 0.9 at 8–10 ms lead time on the filtered ADITYA dataset—is supported by a held-out test split (30% of 725 shots) that is disjoint from the training data. Labels are defined by measured current-quench times, not by model outputs, and the LSTM baseline is evaluated under the same input/output alignment and alarm threshold of 0.5, so the reported superiority is not forced by construction. The ±25% median trim in Section 2.1.3 removes slow terminations and very early quenches, narrowing the evaluation distribution; this is a data-representativeness and generalization limitation, not a circular derivation, because the model must still learn to emit an alarm before the quench from pre-quench diagnostic signals. The selection of the 8 ms threshold after inspecting Figure 10 is a post hoc choice that inflates the statistical strength of the claim, but it does not make the prediction equal to the label by definition. The overlapping-author citation [52] is used for input-feature correlation analysis and is ancillary to the transformer-versus-LSTM comparison; it is not a load-bearing self-citation chain. Overall, the derivation is self-contained with respect to the empirical benchmark performed.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim depends on several hand-chosen parameters, especially the outlier trim range and alarm threshold, which affect the reported recall and accuracy. No new physical entities are introduced. The main domain assumptions are that current quench time is a sufficient label and that the filtered dataset is representative.

free parameters (7)
  • Alarm threshold = 0.5
    The sigmoid output is thresholded at 0.5 to fire an alarm (Section 2.2.1). No sensitivity analysis is provided, so the reported recall and precision depend on this hand-chosen value.
  • Outlier trim range = ±25% of median quench time (~72 ms)
    Shots with quench time outside 54-90 ms are discarded, reducing the dataset from 1407 to 725 shots (Section 2.1.3). This choice shapes the class distribution and directly affects all reported metrics.
  • Prediction horizon = 80 time steps (16 ms)
    Inputs X_t are aligned with labels y_{t+80} (Section 2.2.1), setting the forecasting horizon used in training and limiting the maximum usable lead time to 16 ms.
  • Diagnostic feature subset = plasma current, SXR, HXR, Bolo, H-alpha, C-III
    Six signals are used; they were selected by a correlation analysis reported in the authors' earlier paper [52] (Section 2.1.1), not re-derived here.
  • Operating warning threshold = 8 ms
    The bias study and conclusions use an 8 ms threshold, justified as suitable for ADITYA's 120 ms pulse (Section 3.2). Results at other thresholds are shown only qualitatively.
  • Transformer hyperparameters = not reported
    Number of encoder layers, attention heads, embedding dimension, and dropout are not given, so the comparison with LSTM cannot be independently calibrated.
  • Soft-label encoding rule = not specified
    The target y_t 'encodes the proximity to a disruption event' (Section 2.2.1) but the exact mapping from quench time to label values is not defined, making the training objective ambiguous.
assumptions (5)
  • domain assumption Current quench time is a valid label for disruption versus non-disruption.
    Section 2.1.3 uses fast current decay as the sole criterion for labeling; other disruption signatures (e.g., thermal quench, MHD activity) are not incorporated into the labels.
  • domain assumption The six selected diagnostics contain sufficient information to predict current quench.
    Feature selection rests on the correlation analysis in the authors' prior work [52] (Section 2.1.1); no ablation study or physics argument is given for sufficiency.
  • domain assumption The post-outlier-removal dataset (725 shots) is representative of ADITYA operations.
    Section 2.1.3 removes shots with quench times outside plus or minus 25% of the median, including slow terminations the authors associate with controlled shutdowns; this filtering may remove legitimate non-disruptive scenarios.
  • domain assumption Transformer and LSTM are trained under comparable conditions.
    Section 2.2.3 states both use binary cross-entropy, Adam, and early stopping, but no training curves or hyperparameters confirm equal convergence effort.
  • standard math Standard transformer attention mathematics and stochastic training are valid.
    The architecture follows Vaswani et al. [53]; the paper provides no formal verification but relies on established results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Early Prediction of Current Quench Events in the ADITYA Tokamak using Transformer based Data Driven Models." pith.science (2026). https://pith.science/paper/34BHQ6UN

@misc{pith2026250712797,
  author       = {Pith},
  title        = {Pith review of: Early Prediction of Current Quench Events in the ADITYA Tokamak using Transformer based Data Driven Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/34BHQ6UN}},
  note         = {Machine review of arXiv:2507.12797}
}
read the original abstract

Disruptions in tokamak plasmas, marked by sudden thermal and current quenches, pose serious threats to plasma-facing components and system integrity. Accurate early prediction, with sufficient lead time before disruption onset, is vital to enable effective mitigation strategies. This study presents a novel data-driven approach for predicting early current quench, a key precursor to disruptions, using transformer-based deep learning models, applied to ADITYA tokamak diagnostic data. Using multivariate time series data, the transformer model outperforms LSTM baselines across various data distributions and prediction thresholds. The transformer model achieves better recall, maintaining values above 0.9 even up to a prediction threshold of 8-10 ms, significantly outperforming LSTM in this critical metric. The proposed approach remains robust up to an 8 ms lead time, offering practical feasibility for disruption mitigation in ADITYA tokamak. In addition, a comprehensive data diversity analysis and bias sensitivity study underscore the generalization of the model. This work marks the first application of transformer architectures to ADITYA tokamak data for early current-quench prediction, establishing a promising foundation for real time disruption avoidance in short-pulse tokamaks.

Figures

Figures reproduced from arXiv: 2507.12797 by the authors.

Figure 1
Figure 1. Complete workflow to develop and testing of transformer model for current quench prediction in ADITYA tokamak [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Time evolution of plasma current (a), Soft x-ray signal of central [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. (a) Distribution of current quench time before outlier removal (b) [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: Disruption studies in ADITYA Tokamak:(a) Data diversity in earlier [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Euclidean distances between plasma current signals in the experimental dataset from ADITYA used for this study, shown for di [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: LSTM model architecture for predicting disruptions with 2 layers of [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Performance evaluation metrics based on the time window between the prediction time and the prediction threshold, illustrating (a) True negatives, (b) [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Variation of precision, recall, and accuracy as a function of predic [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: Performance comparison between Transformer and LSTM models across di [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 52 canonical work pages

  1. [1]

    L. A. Artsimovich, Tokamak devices, Nuclear Fusion 12 (2) (1972) 215

    work page 1972

  2. [2]

    S. T. Paruchuri, I. Ward, N. Rist, V . Graber, H. Al Khawaldeh, Z. Wang, E. Schuster, A. Pajares, J.- W. Juhn, Density regulation with disruption avoidance in next-generation tokamaks using a safe reinforcement learning-based controller, Fusion Engineering and Design 216 (2025) 115064

    work page 2025

  3. [3]

    L. E. Zakharov, S. A. Galkin, S. N. Gerasimov, et al., Un- derstanding disruptions in tokamaks, Physics of Plasmas 19 (5) (2012)

    work page 2012

  4. [4]

    A. H. Boozer, Theory of tokamak disruptions, Physics of Plasmas 19 (5) (2012) 058101

    work page 2012

  5. [5]

    Hassanein, T

    A. Hassanein, T. Sizyuk, V . Sizyuk, G. Miloshevsky, Im- pact of various plasma instabilities on reliability and per- formance of tokamak fusion devices, Fusion engineering and design 85 (7-9) (2010) 1331–1335

    work page 2010

  6. [6]

    Yoshino, D

    R. Yoshino, D. J. Campbell, E. Frederickson, Characteri- zation of disruption phenomenology in ITER, Tech. Rep. IAEA-CSP–8/C, IAEA (2001)

    work page 2001

  7. [7]

    F. C. Schuller, Disruptions in tokamaks, Plasma Physics and Controlled Fusion 37 (11A) (1995) A135

    work page 1995

  8. [8]

    Buzio, S

    M. Buzio, S. Walker, Determination of magnetic forces in tokamaks during axisymmetric plasma disruptions, Fu- sion engineering and design 81 (19) (2006) 2187–2203

    work page 2006

Show all 53 references
  1. [9]

    A. V . Nedospasov, Thermal quench in tokamaks, Nuclear Fusion 48 (3) (2008) 032002

    2008

  2. [10]

    Raman, W

    R. Raman, W. S. Lay, T. R. Jarboe, J. E. Menard, M. Ono, Electromagnetic particle injector for fast time response disruption mitigation in tokamaks, Nuclear Fusion 59 (1) (2018) 016021

    2018

  3. [11]

    K. H. Finken, M. Lehnen, S. A. Bozhenkov, Gas flow analysis of a disruption mitigation valve (DMV), Nuclear Fusion 48 (11) (2008) 115001

    2008

  4. [12]

    Dhyani, et al., A novel approach for mitigating disrup- tions using biased electrode in ADITY A tokamak, Nuclear Fusion 54 (8) (2014) 083023

    P. Dhyani, et al., A novel approach for mitigating disrup- tions using biased electrode in ADITY A tokamak, Nuclear Fusion 54 (8) (2014) 083023

    2014

  5. [13]

    A. H. Boozer, Plasma steering to avoid disruptions in ITER and tokamak power plants, Nuclear Fusion 61 (5) (2021) 054004

    2021

  6. [14]

    V . D. Pustovitov, Models and scalings for the disruption forces in tokamaks, Nuclear Fusion 62 (2) (2022) 026036

    2022

  7. [15]

    E. J. Strait, J. L. Barr, M. Baruzzo, J. W. Berkery, R. J. Buttery, P. C. de Vries, et al., Progress in disruption pre- vention for ITER, Nuclear Fusion 59 (11) (2019) 112012

    2019

  8. [16]

    Lehnen, K

    M. Lehnen, K. Aleynikova, P. B. Aleynikov, D. J. Camp- bell, P. Drewelow, N. W. Eidietis, et al., Disruptions in ITER and strategies for their control and mitigation, Jour- nal of Nuclear Materials 463 (2015) 39–48

    2015

  9. [17]

    Rea, et al., A real-time machine learning-based dis- ruption predictor in DIII-D, Nuclear Fusion 59 (9) (2019) 096016

    C. Rea, et al., A real-time machine learning-based dis- ruption predictor in DIII-D, Nuclear Fusion 59 (9) (2019) 096016

    2019

  10. [18]

    B. H. Guo, et al., Disruption prediction using a full con- volutional neural network on EAST, Plasma Physics and Controlled Fusion 63 (2) (2021) 025010

    2021

  11. [19]

    Zheng, et al., Disruption predictor based on neural net- work and anomaly detection on J-TEXT, Plasma Physics and Controlled Fusion 62 (4) (2020) 045008

    W. Zheng, et al., Disruption predictor based on neural net- work and anomaly detection on J-TEXT, Plasma Physics and Controlled Fusion 62 (4) (2020) 045008. 11

    2020

  12. [20]

    B. Dave, S. Patel, R. Shivani, S. Purohit, B. Chaudhury, Synthetic data generation using generative adversarial net- work for tokamak plasma current quench experiments, Contributions to Plasma Physics 63 (12 2022)

    2022

  13. [21]

    Vega, et al., Disruption prediction with artificial intel- ligence techniques in tokamak plasmas, Nuclear Fusion 60 (6) (2020) 066013

    J. Vega, et al., Disruption prediction with artificial intel- ligence techniques in tokamak plasmas, Nuclear Fusion 60 (6) (2020) 066013

    2020

  14. [22]

    Aymerich, et al., Disruption prediction at JET through deep convolutional neural networks using spatiotemporal information from plasma profiles, Nuclear Fusion 62 (7) (2022) 076027

    E. Aymerich, et al., Disruption prediction at JET through deep convolutional neural networks using spatiotemporal information from plasma profiles, Nuclear Fusion 62 (7) (2022) 076027

    2022

  15. [23]

    Aymerich, et al., Performance comparison of machine learning disruption predictors at JET, Applied Sciences 13 (3) (2023) 1749

    E. Aymerich, et al., Performance comparison of machine learning disruption predictors at JET, Applied Sciences 13 (3) (2023) 1749

    2023

  16. [24]

    Kates-Harbeck, et al., Predicting disruptive instabilities in controlled fusion plasmas through deep learning, Na- ture 568 (7753) (2019) 526–531

    J. Kates-Harbeck, et al., Predicting disruptive instabilities in controlled fusion plasmas through deep learning, Na- ture 568 (7753) (2019) 526–531

    2019

  17. [25]

    R. M. Churchill, et al., Deep convolutional neural net- works for multi-scale time-series classification and appli- cation to tokamak disruption prediction using raw, high temporal resolution diagnostic data, Physics of Plasmas 27 (6) (2020) 062510

    2020

  18. [26]

    J. X. Zhu, et al., Scenario adaptive disruption prediction study for next generation burning-plasma tokamaks, Nu- clear Fusion 61 (6) (2021) 064006

    2021

  19. [27]

    J. X. Zhu, et al., Hybrid deep-learning architecture for general disruption prediction across multiple tokamaks, Nuclear Fusion 61 (10) (2021) 106022

    2021

  20. [28]

    Croonen, et al., Investigation of machine learning tech- niques for disruption prediction using JET data, Plasma 6 (2) (2023) 351–366

    J. Croonen, et al., Investigation of machine learning tech- niques for disruption prediction using JET data, Plasma 6 (2) (2023) 351–366

    2023

  21. [29]

    Priyanka, J

    M. Priyanka, J. Sangeetha, C. Jayakumar, A review of tra- ditional and data-driven approaches for disruption predic- tion in different tokamaks, E3S Web of Conferences 477 (2024) 00039

    2024

  22. [30]

    Wang, et al., Robust control of q-profile and beta us- ing data-driven models on EAST, Fusion Engineering and Design 162 (2021) 112097

    S. Wang, et al., Robust control of q-profile and beta us- ing data-driven models on EAST, Fusion Engineering and Design 162 (2021) 112097

    2021

  23. [31]

    Dalsania, Z

    N. Dalsania, Z. Patel, S. Purohit, B. Chaudhury, An ap- plication of machine learning for plasma current quench studies via synthetic data generation, Fusion Engineering and Design 171 (2021) 112578

    2021

  24. [32]

    S. A. Sabbagh, J. W. Berkery, Y . S. Park, J. H. Ahn, Y . Jiang, J. D. Riquezes, et al., Disruption event charac- terization ex/p6-26 and forecasting in tokamaks, in: IAEA Fusion Energy Conference, IAEA, 2018

    2018

  25. [33]

    S. A. Sabbagh, J. W. Berkery, Y . S. Park, J. H. Ahn, Y . Jiang, J. D. Riquezes, et al., Tokamak disruption event characterization and forecasting research and expansion to real-time application, in: Proc. 28th Int. Conf. on Fusion Energy, IAEA, 2021

    2021

  26. [34]

    R. L. Tanna, et al., Overview of recent experimental re- sults from the ADITY A tokamak, Nuclear Fusion 57 (10) (2017) 102008

    2017

  27. [35]

    Dhyani, Biased electrode experiments in ADITY A toka- mak, Ph.D

    P. Dhyani, Biased electrode experiments in ADITY A toka- mak, Ph.D. thesis, Institute for Plasma Research (2014)

    2014

  28. [36]

    Purohit, et al., Characterization of the plasma current quench during disruptions in ADITY A tokamak, Nuclear Fusion 60 (7) (2020) 076018

    S. Purohit, et al., Characterization of the plasma current quench during disruptions in ADITY A tokamak, Nuclear Fusion 60 (7) (2020) 076018

    2020

  29. [37]

    Chattopadhyay, et al., Analysis of disruptive instabil- ities in ADITY A tokamak discharge, Indian Journal of Pure and Applied Physics 44 (12) (2006) 897–904

    A. Chattopadhyay, et al., Analysis of disruptive instabil- ities in ADITY A tokamak discharge, Indian Journal of Pure and Applied Physics 44 (12) (2006) 897–904

    2006

  30. [38]

    A. K. Chattopadhyay, et al., Instability analysis in ADITY A tokamak discharges with the help of soft x-ray, Journal of Plasma and Fusion Research Series 8 (2009) 685–690

    2009

  31. [39]

    R. L. Tanna, et al., Novel approaches for mitigating run- away electrons and plasma disruptions in ADITY A toka- mak, Nuclear Fusion 55 (6) (2015) 063010

    2015

  32. [40]

    Ghosh, et al., Disruption mitigation using ion cyclotron wave in ADITY A tokamak, Physics of Plasmas 28 (2) (2021) 022505

    J. Ghosh, et al., Disruption mitigation using ion cyclotron wave in ADITY A tokamak, Physics of Plasmas 28 (2) (2021) 022505

    2021

  33. [41]

    Sengupta, et al., Forecasting disruptions in the ADITY A tokamak using neural networks, Nuclear Fusion 40 (12) (2000) 1993

    A. Sengupta, et al., Forecasting disruptions in the ADITY A tokamak using neural networks, Nuclear Fusion 40 (12) (2000) 1993

    2000

  34. [42]

    Agarwal, et al., Deep sequence to sequence learning- based prediction of major disruptions in ADITY A toka- mak, Plasma Physics and Controlled Fusion 63 (7) (2021) 075006

    A. Agarwal, et al., Deep sequence to sequence learning- based prediction of major disruptions in ADITY A toka- mak, Plasma Physics and Controlled Fusion 63 (7) (2021) 075006

    2021

  35. [43]

    Joshi, J

    R. Joshi, J. Ghosh, N. Kalani, R. L. Tanna, Assessment of stacked LSTM, bidirectional LSTM, ConvLSTM2D, and auto encoders LSTM time series regression analysis at ADITY A-U tokamak, IEEE Transactions on Plasma Sci- ence 52 (9) (2024) 2776–2784

    2024

  36. [44]

    Joshi, J

    R. Joshi, J. Ghosh, N. Kalani, R. L. Tanna, Conceptual design and preliminary data analysis for classification of plasma disruption event at ADITY A-U tokamak, Tech. rep. (2022)

    2022

  37. [45]

    Muruganandham, S

    P. Muruganandham, S. Jayaraman, K. Tahiliani, R. Tanna, J. Ghosh, S. K. Pathak, N. Ramaiya, An advanced double-phase stacking ensemble technique with active learning classifier: Toward reliable disruption prediction in ADITY A tokamak, Review of Scientific Instruments 95 (9) (...

    2024

  38. [46]

    Zeyer, P

    A. Zeyer, P. Bahar, K. Irie, R. Schlüter, H. Ney, A com- parison of transformer and LSTM encoder decoder mod- els for ASR, in: IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), IEEE, 2019, pp. 8–15. doi:10.1109/ASRU46091.2019.9004025

    2019

  39. [47]

    Spangher, M

    L. Spangher, M. Bonotto, W. Arnold, D. Chayapathy, T. Gallingani, A. Spangher, F. Cannarile, D. Bigoni, E. de Marchi, C. Rea, Disruptionbench and Complimen- tary new models: Two Advancements in Machine Learn- ing Driven Disruption Prediction, Journal of Fusion En- ergy 44 (1) ...

    2025

  40. [48]

    S. B. Bhatt, D. Bora, B. N. Buch, ADITY A: The first in- dian tokamak, Indian Journal of Pure and Applied Physics 27 (9-10) (1989) 710–742

    1989

  41. [49]

    D. Raju, R. Jha, P. Kaw, S. Mattoo, Y . Saxena, A. Team, Mirnov coil data analysis for tokamak ADITY A, Pramana 55 (2000) 727–732

    2000

  42. [50]

    Atrey, D

    P. Atrey, D. Pujara, S. Mukherjee, Mhd mode bispectral analysis from density fluctuations in ADITY A discharges, Fusion Engineering and Design 130 (2018) 89–93

    2018

  43. [51]

    Bora, SST and ADITY A tokamak research in india, Brazilian journal of physics 32 (2002) 193–216

    D. Bora, SST and ADITY A tokamak research in india, Brazilian journal of physics 32 (2002) 193–216

    2002

  44. [52]

    Agarwal, B

    J. Agarwal, B. Chaudhury, S. Jakhar, N. Shah, S. Arora, D. Katrodia, M. Sharma, Automated labelling and corre- lation analysis of diagnostic signals from ADITY A toka- mak for developing ai-based disruption mitigation sys- tems, Radiation Effects and Defects in Solids 179 (7-8...

    2024

  45. [53]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017). 13

    2017

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.