Pith. sign in

REVIEW 3 major objections 5 minor 40 references

Generating Bearing Vibration Signals at User-Specified Fault Probabilities Using PR-GAN and Counterfactual Methods

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Direct optimization of each signal against a frozen ensemble classifier can produce bearing vibration windows at user-chosen fault probabilities far more reliably than a trained probability-regularized GAN.

desk verdict A clean, honest application of known templates; the CF results are real but partly by design, and the realism question is left open. read the letter →

arxiv 2607.19455 v2 pith:BVZNIKT6 submitted 2026-07-21 cs.LG

Seyed Mohammadreza Alavi , Ardeshir Shojaeinasab , Reza Jalayer , Masoud Jalayer , Behnam Bahrak This is my paper
classification cs.LG
keywords bearingfaultdiagnosiscounterfactualgenerationprobability-targetedsynthesisvibrationsignalgenerativeadversarialnetworkspredictivemaintenanceclassifiercalibrationgray-zonesamples
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 solve a practical scarcity: in bearing vibration datasets, almost every window is confidently scored as healthy or faulty, so intermediate "gray-zone" fault probabilities are rare. It proposes two ways to synthesize windows at a user-chosen fault probability (0.25, 0.50, or 0.75) using a fixed, gradient-accessible ensemble classifier as a probability oracle: a trained residual GAN (PR-GAN) and a training-free per-sample counterfactual optimizer (CF). The paper's central empirical claim is that CF reaches the target with mean absolute probability error of 0.005–0.008 and a within-tolerance success rate of 1.000 on retained samples, whereas PR-GAN's error is 0.046–0.059 with success between 0.501 and 0.680. A sympathetic reader cares because controlled boundary-region samples could support maintenance-decision studies—but the authors explicitly caution that the uncalibrated oracle and source-closeness realism proxies leave the physical meaningfulness of the generated "borderline" samples open.

What carries the argument

The load-bearing object is the "probability oracle": a fixed, gradient-accessible heterogeneous ensemble classifier whose averaged output is treated as a continuous function of the input. Both generation methods steer this output to a target probability: PR-GAN trains a residual generator x̂ = x + Δ(x, p*) with a Wasserstein-GAN adversarial loss, a binary-cross-entropy alignment term between the classifier output and p*, and an L1 residual penalty; CF runs per-sample gradient-based optimization on BCE(p*, f(x_cf)) plus an L2 proximity term to the source, with amplitude clipping and an early-stop tolerance of 0.05. The paper also gives an intermediate-value existence argument (continuity, pat

What would settle it

Compute expected calibration error and reliability diagrams for the guidance ensemble, then run envelope-spectrum analysis at the characteristic bearing-fault frequencies on the CF-generated p* = 0.5 samples (or pass them through a second, independently trained classifier). If the samples are detected as out-of-distribution or the 0.5 outputs do not coincide with genuinely ambiguous vibration signatures, the central practical claim that these are useful borderline samples collapses.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that user-specified fault probabilities can be imposed on bearing vibration windows by directly optimizing each window against a frozen ensemble classifier: a per-sample counterfactual procedure using binary cross-entropy to a soft target plus an L2 proximity penalty achieved mean absolute error 0.005–0.008 and 100% within-tolerance success across all retained samples and both benchmark datasets, with smaller average L1 edits than the residual GAN. The trained PR-GAN, which augments a Wasserstein GAN with gradient penalty by adding a probability-alignment term and a residual edit, was systematically less reliable (MAE 0.046–0.059, success 0.501–0.680),

Load-bearing premise

The approach assumes the ensemble classifier's predicted probability is a trustworthy guide to real fault likelihood; if that output is uncalibrated—or if a p* = 0.5 edit is just an adversarial perturbation of the source rather than a plausible vibration condition—the generated "borderline" samples do not deliver the maintenance value the paper motivates.

Editorial extensions

If this is right

  • For a fixed classifier with accessible gradients, per-sample counterfactual optimization can hit a user-chosen fault probability almost exactly (mean absolute error ≤ 0.008) on every retained window, while keeping time-domain and spectral edits smaller than the trained GAN's.
  • The trained residual GAN is not a reliable way to control probability: on roughly a third to half of its retained samples it misses the ±0.05 tolerance, so amortized inference speed does not compensate for probability-steering failure on individual outputs.
  • Generated p* = 0.5 samples should be read as "the classifier is locally uncertain on this perturbation," not as calibrated 50% fault likelihood; the paper states this explicitly as a limitation.
  • The CF-versus-PR-GAN comparison is a system-level comparison, not a controlled ablation, so the reliability gap should not be attributed to any single design axis.
  • Probability-targeted samples provide a practical way to populate the boundary region of a fixed classifier with controlled probabilities that are otherwise scarce in standard bearing datasets, subject to the realism caveat.

Reading between the lines

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

  • A strong reading of the CF result is that it is a targeted adversarial perturbation under an L2 budget; whether that is useful for maintenance hinges on external validation (e.g., a second held-out classifier or envelope-spectrum features agreeing with the oracle), which the paper leaves as future work.
  • The 100% success rate may partly reflect the generous 300-step optimization budget and the oracle's smoothness; a fairer apples-to-apples comparison would give PR-GAN per-sample verification or a comparable per-sample budget, which the paper does not do.
  • If the guidance ensemble were replaced by a calibrated one, the per-sample targeting mechanism would likely still work—binary-cross-entropy targeting depends on relative ordering—but the physical meaning of the resulting probability would become stronger; this is a testable extension.
  • A direct downstream test: use the generated gray-zone samples as training augmentation and measure whether boundary robustness or uncertainty estimates improve; the paper lists this as future work but does not run it.
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

3 major / 5 minor

Summary. The manuscript addresses the scarcity of samples with intermediate predicted fault probabilities in bearing vibration classifiers. It proposes two generation methods: PR-GAN, a WGAN-GP extension with a residual generator and a BCE probability-alignment term against a frozen heterogeneous ensemble CNN, and CF, a Wachter-style per-sample counterfactual that optimizes a BCE plus L2 proximity objective with early stopping at tolerance epsilon=0.05. Experiments on CWRU and Paderborn report that CF achieves MAE_p 0.005--0.008 and SR_tau=1.000 on retained samples, whereas PR-GAN achieves MAE_p 0.046--0.059 and SR_tau 0.501--0.680 on retained samples; CF also has lower Delta-L1 and dPSD in most settings, while PR-GAN has lower reported runtime in most settings. The authors position both methods as instantiations of existing templates applied to vibration windows, and they include a GitHub repository for reproducibility.

Significance. If the results are taken at face value, the paper provides a reproducible comparison of training-based and training-free probability-targeted generation on two standard bearing datasets. The manuscript is honest about several confounders and limitations, and it ships code. However, the headline CF success rate is largely a consequence of the method's own stopping rule, and the realism evaluation is only a source-proximity proxy, with the paper itself conceding that p*=0.5 samples may be adversarial rather than physically meaningful. The central practical motivation---generating plausible borderline samples for maintenance decisions---therefore remains unsupported. These issues are addressable with additional reporting and experiments, so the paper has a useful core if revised substantially.

major comments (3)
  1. [§3.8, Algorithm 2; §3.9, Eq. (18)] CF's reported SR_tau=1.000 is a tautology. Algorithm 2 breaks when |p_f - p*| <= epsilon, with epsilon=0.05, and Eq. (18) defines success as |p_f - p*| <= tau with tau=0.05. Since N_failed=0, every returned sample satisfies the success criterion by construction; the 1.000 'success rate' is not an independent empirical outcome. The MAE_p values are not directly forced by the bound, but the comparison is still not like-for-like because PR-GAN has no early-exit oracle. Please report CF's error distribution under a fixed step budget (or at least how often the early-exit condition binds) and use the same evaluation procedure for both methods.
  2. [§3.9, Eqs. (17)--(18); Tables 4--5] All probability metrics are computed only on samples retained after the 0.25 filtering threshold. For PR-GAN this excludes 77, 74, and 3 samples (CWRU) and 20, 15, and 0 (Paderborn) out of 500. Thus SR_tau is a conditional proportion among retained samples, not the success rate over all attempted generations. For example, CWRU p*=0.25 PR-GAN's full-attempt success rate is roughly (500-77)/500 * 0.501 = 0.424, not 0.501. The paper does disclose the conditioning, but the abstract's presentation of 'success rates between 0.501 and 0.680' is misleading without full-attempt rates. Please report full-attempt MAE/SR or a combined metric such as retention rate times SR_tau.
  3. [§3.9, §4.5.3, §5] The central motivation is to generate practically useful borderline samples, but the realism metrics Delta-L1, Delta-TV, and dPSD measure proximity to the source, not physical plausibility. Section 4.5.3 explicitly concedes that p*=0.5 samples 'may lie closer to adversarial perturbations of the source than to physically meaningful borderline conditions.' Because no direct realism check (envelope-spectrum analysis, a realism classifier, or a downstream task) is provided, the practical-value claim is unsupported. This is a load-bearing limitation; either add at least one direct realism evaluation or substantially narrow the claims to 'edits that steer a fixed classifier's probability' rather than 'realistic bearing vibration signals.'
minor comments (5)
  1. [§4.3] The sentence 'The results are reported in Table 4.' appears twice consecutively. Please remove the duplicate.
  2. [Tables 4--5 and §4.5.1] The 'Time (minute)' column mixes PR-GAN training time with CF generation time. The text explains this, but the table should annotate the distinction more clearly (e.g., separate rows for training and per-sample generation) to avoid an apples-to-oranges runtime comparison.
  3. [§3.8, Algorithm 2] The values of alpha_ce and beta_l2 in the CF objective are not reported. Since they control the trade-off between probability alignment and source proximity, please include the values used in the experiments.
  4. [Abstract and §3.9] The abstract says CF 'requires smaller average L1 changes,' but CF's optimization objective uses an L2 proximity term; Delta-L1 is only an evaluation metric. Clarify this distinction to avoid confusion.
  5. [§3.2] The paper states that classifier calibration is not measured. This is a legitimate scoping choice, but it should be tied more prominently to the interpretation of p* in the abstract and conclusion, since 'fault probability' could be read as a physical probability rather than a model output.

Circularity Check

1 steps flagged · score 6.0 of 10

CF's reported success rate of 1.000 is the Algorithm 2 stopping rule restated as Eq. (18); the rest of the comparison is transparent but the headline probability-steering claim is partly forced.

  1. self definitional [§3.8 (Algorithm 2, lines 8–9) and §3.9 (Eq. 18)]
    "We set the stopping tolerance to ε = 0.05. ... Algorithm 2: 8: if |p_f − p∗| ≤ ε then 9: break ... SRτ = 1/M ∑_{i=1}^M 1(||pfault(xi)−p∗|| ≤ τ), where 1(·) is the indicator function, τ corresponds to the tolerance parameter which is 0.05."

    For CF, every sample that exits through Algorithm 2's early break satisfies |pf−p∗|≤0.05 by construction, and Eq. (18) defines success as exactly |pf−p∗|≤τ with τ=0.05. Thus SRτ=1.000 on retained samples is the stopping rule renamed as an evaluation result, not an empirical finding about steering accuracy. The MAE values are also measured on the same fixed, gradient-accessible oracle whose BCE drives the optimization, so they report fit to the generation objective. The comparison is therefore partly forced: PR-GAN has no per-sample early-stop guarantee, so the headline CF-vs-PR-GAN success-rate gap is not a fully independent test of the two methods.

full rationale

The one concrete circular step is the CF success-rate metric: the generation stopping tolerance ε=0.05 and the evaluation tolerance τ=0.05 are the same quantity, making CF's SRτ=1.000 definitional rather than predictive. The paper itself is transparent about the related confounds: §4.5.1 notes CF is given 300 Adam steps against a direct probability objective, §3.9 warns that the similarity metrics measure edit aggressiveness rather than absolute realism, §4.5.3 cautions that p∗=0.5 samples may be closer to adversarial perturbations than to physical borderline conditions, and the conclusion explicitly lists calibration and downstream validation as open. These are honest limitations, not circularity. I found no load-bearing self-citation chain, no imported uniqueness theorem, and no ansatz smuggled in by citation; PR-GAN is openly presented as a combination of existing components. The central derivation—that direct optimization of a real window against a fixed classifier can match a target probability—is self-contained and true by construction, but the headline 'success rate 1.000' is forced, so the circularity score is 6 rather than 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced; PR-GAN is a new model architecture but not a postulated entity that needs independent physical evidence. The main latent assumptions are about the meaningfulness of classifier probabilities, the realism proxies, and the evaluation filter.

free parameters (5)
  • alpha_ce / beta_l2 (CF objective weights) = not reported
    Balance the BCE probability term and the L2 proximity term in Equation 12; the values are not given and materially affect edit size and convergence.
  • CF max steps and stopping tolerance = S=300, epsilon=0.05
    Hand-chosen stopping rule; because the evaluation tolerance is the same 0.05, the reported SR_tau=1.000 is partly by construction.
  • PR-GAN loss weights (alpha_adv, lambda_bce, lambda_res) = not reported
    Control adversarial, probability-alignment, and residual terms in Equation 9; no values are reported.
  • PR-GAN WGAN-GP hyperparameters (lambda_gp, critic iterations) = not reported
    Gradient penalty weight and number of critic updates per generator update are not specified.
  • Retention threshold for evaluation = 0.25
    Samples with |p_fault-p*|>0.25 are discarded before computing MAE_p and SR_tau; this post-hoc filter affects all reported metrics.
assumptions (4)
  • domain assumption The admissible signal space is path-connected and the classifier probability f is continuous, with signals on both sides of p*.
    Used in the Section 3.4 intermediate-value theorem existence argument; CNN continuity is plausible, but path-connectivity of realistic vibration windows is unproven.
  • domain assumption The ensemble classifier's averaged probabilities are informative enough to guide generation.
    Stated in Section 3.2; calibration is explicitly not measured, so the semantic meaning of the target probability is left open.
  • ad hoc to paper Source-similarity metrics (Delta-L1, Delta-TV, dPSD) are valid proxies for signal realism.
    Introduced in Section 3.9; the paper notes that a method returning the source unchanged scores zero and that absolute realism is not directly measured.
  • domain assumption All fault types can be collapsed into a single binary fault class.
    Dataset preprocessing in Section 3.5; generated gray-zone samples may mix different fault types, which the evaluation does not distinguish.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generating Bearing Vibration Signals at User-Specified Fault Probabilities Using PR-GAN and Counterfactual Methods." pith.science (2026). https://pith.science/paper/BVZNIKT6

@misc{pith2026260719455,
  author       = {Pith},
  title        = {Pith review of: Generating Bearing Vibration Signals at User-Specified Fault Probabilities Using PR-GAN and Counterfactual Methods},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BVZNIKT6}},
  note         = {Machine review of arXiv:2607.19455}
}
read the original abstract

In bearing vibration datasets, most samples receive predicted fault probabilities close to 0 or 1, while samples with intermediate (gray-zone) probabilities are rare. Such borderline samples are important because they reflect conditions in which maintenance decisions may require additional inspection or a conservative response and are useful for studying decision boundaries. To address this scarcity, this paper proposes and compares two approaches that generate vibration signals whose predicted fault probability matches a target probability of 0.25, 0.50, or 0.75. We use the average output of a heterogeneous ensemble classifier with different architectures and random initializations as a fixed, gradient-accessible probability oracle. The first, training-based approach, Probability-Regularized Generative Adversarial Network (PR-GAN), extends Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) and edits a real signal through a residual generator while pushing the classifier output toward the target probability. The second is a training-free, per-sample Wachter-style counterfactual (CF) procedure that directly optimizes each input signal to reach the target probability while remaining close to the source signal. We evaluate both methods on the Case Western Reserve University (CWRU) and Paderborn bearing datasets using mean absolute target-probability error, time-domain total variation, and frequency-domain log power spectral density (log-PSD) differences. Across all settings, CF reaches the target with a mean absolute probability error of 0.005-0.008 and a within-tolerance success rate of 1.000 on retained samples, whereas PR-GAN's mean error is 0.046-0.059 with success rates between 0.501 and 0.680. CF therefore steers the probability more reliably and requires smaller average L1 changes, whereas PR-GAN has a lower reported runtime in most settings.

Figures

Figures reproduced from arXiv: 2607.19455 by the authors.

Figure 1
Figure 1. Schematic illustration of the proposed search process in the latent space. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. A fixed, gradient-accessible guidance model di [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Sliding-window cropping with length=256 and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Experimental setup of the CWRU bearing test rig [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Experimental setup of the Paderborn bearing test [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: PR-GAN training schematic. A generator produces [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Counterfactual (CF) method. Starting from a real signal [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Distribution of predicted fault probabilities for the CWRU and Paderborn datasets. [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: PCA projection of time-domain features, colored by the predicted fault probability. [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Generated vs. original samples for the inner race fault class of CWRU dataset. (The real label refers to the label [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Generated vs. original samples for the real inner ring fault class of Paderborn dataset. (The real label refers to the [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 7 linked inside Pith

  1. [1]

    Synthesizing rolling bearing fault samples in new conditions: A framework based on a modified CGAN

    Ahang, M., Jalayer, M., Shojaeinasab, A., Ogunfowora, O., Charter, T., Najjaran, H., 2022. Synthesizing rolling bearing fault samples in new conditions: A framework based on a modified CGAN. Sensors

    2022

  2. [2]

    What regularized auto-encoders learn from the data-generating distribution

    Alain, G., Bengio, Y., 2014. What regularized auto-encoders learn from the data-generating distribution. Journal of Machine Learning Research

    2014

  3. [3]

    Generalized denoising auto-encoders as generative modelsarXiv:1305.6663

    Bengio, Y., Yao, L., Alain, G., Vincent, P., 2013. Generalized denoising auto-encoders as generative modelsarXiv:1305.6663

    arXiv 2013

  4. [4]

    Basic vibration theory

    Blake, R.E., . Basic vibration theory. Chapter 2

  5. [5]

    Bearing data center.https://engineering.case.edu/bearingdatacent er

    Case Western Reserve University, Case School of Engineering, n.d. Bearing data center.https://engineering.case.edu/bearingdatacent er. Accessed: 2026-01-07

    2026

  6. [6]

    A new safe-level enabled borderline-smote for condition recognition of imbalanced dataset

    Chen, C., Shen, W., Yang, C., Fan, W., Liu, X., Li, Y., 2023. A new safe-level enabled borderline-smote for condition recognition of imbalanced dataset. IEEE Transactions on Instrumentation and Measurement

    2023

  7. [7]

    Abearingfaultdiagnosismethodinscenariosofimbalancedsamples and insufficient labeled samples

    Cheng, X., Lu, Y., Liang, Z., Zhao, L., Gong, Y., Wang, M., 2024. Abearingfaultdiagnosismethodinscenariosofimbalancedsamples and insufficient labeled samples. Applied Sciences

    2024

  8. [8]

    Generative alignment of posterior probabilities for source-free domain adaptation

    Chhabra, S., Venkateswara, H., Li, B., . Generative alignment of posterior probabilities for source-free domain adaptation

Show all 40 references
  1. [9]

    Instance-based counter- factual explanations for time series classificationarXiv:2009.13211

    Delaney, E., Greene, D., Keane, M.T., 2021. Instance-based counter- factual explanations for time series classificationarXiv:2009.13211

    2021 arXiv

  2. [10]

    Data Sets and Download — Bearing Data Center

    DesignandDriveTechnology(KAt),PaderbornUniversity,n.d. Data Sets and Download — Bearing Data Center. Website. URL:https: //mb.uni-paderborn.de/en/kat/research/bearing-datacenter/data-s ets-and-download. accessed January 7, 2026

    2026

  3. [11]

    Dropoutasabayesianapproximation: Representing model uncertainty in deep learning, in: Proceedings of the 33rd International Conference on Machine Learning, pp

    Gal,Y.,Ghahramani,Z.,2016. Dropoutasabayesianapproximation: Representing model uncertainty in deep learning, in: Proceedings of the 33rd International Conference on Machine Learning, pp. 1050– 1059

    2016

  4. [12]

    Rudin–osher–fatemi total variation denoising using split bregman

    Getreuer, P., 2012. Rudin–osher–fatemi total variation denoising using split bregman. Image Processing On Line

    2012

  5. [13]

    Reg- ularisation of neural networks by enforcing lipschitz continuity

    Gouk, H., Frank, E., Pfahringer, B., Cree, M.J., 2020. Reg- ularisation of neural networks by enforcing lipschitz continuity. arXiv:1804.04368

    2020 arXiv

  6. [14]

    Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., Courville, A.,

  7. [15]

    On calibration of modern neural networks, in: Proceedings of the 34th International Conference on Machine Learning

    Guo, C., Pleiss, G., Sun, Y., Weinberger, K.Q., 2017. On calibration of modern neural networks, in: Proceedings of the 34th International Conference on Machine Learning

    2017

  8. [16]

    Adaptive sv-borderline smote-svmalgorithmforimbalanceddataclassification

    Guo, J., Wu, H., Chen, X., Lin, W., 2024. Adaptive sv-borderline smote-svmalgorithmforimbalanceddataclassification. AppliedSoft Computing

    2024

  9. [17]

    Evaluating deep learning models for fault diagnosis of a rotating machinery with epistemic and aleatoric uncertainty

    Jalayer, R., Jalayer, M., Mor, A., Orsenigo, C., Vercellis, C., 2024. Evaluating deep learning models for fault diagnosis of a rotating machinery with epistemic and aleatoric uncertainty. arXiv preprint arXiv:2412.18980

    2024 arXiv

  10. [18]

    Obgan:Minorityoversamplingnearborderline with generative adversarial networks

    Jo,W.,Kim,D.,2022. Obgan:Minorityoversamplingnearborderline with generative adversarial networks. Expert Systems with Applica- tions

    2022

  11. [19]

    Therelativeperformance of ensemble methods with deep convolutional neural networks for image classification.arXiv:1704.01664

    Ju,C.,Bibaut,A.,vanderLaan,M.J.,2017. Therelativeperformance of ensemble methods with deep convolutional neural networks for image classification.arXiv:1704.01664

    2017 arXiv

  12. [20]

    Lessmeier, C., Kimotho, J.K., Zimmer, D., Sextro, W., 2016. Con- dition monitoring of bearing damage in electromechanical drive sys- tems by using motor current signals of electric motors: A benchmark datasetfordata-drivenclassification,in:EuropeanConferenceofthe Prognostics a...

    2016

  13. [21]

    Representations aligned counterfactual domain learning for open-set fault diagnosis under speed transient conditions

    Liu, S., Chen, J., Shi, Z., Song, L., He, S., 2025. Representations aligned counterfactual domain learning for open-set fault diagnosis under speed transient conditions. Knowledge-Based Systems

    2025

  14. [22]

    Counterfactual-augmented few-shotcontrastivelearningformachineryintelligentfaultdiagnosis with limited samples

    Liu,Y.,Jiang,H.,Yao,R.,Zeng,T.,2024. Counterfactual-augmented few-shotcontrastivelearningformachineryintelligentfaultdiagnosis with limited samples. Mechanical Systems and Signal Processing

    2024

  15. [23]

    Fault diagnosis based on counterfactual inference for the batch fermentation process

    Liu, Z., Lou, X., 2024. Fault diagnosis based on counterfactual inference for the batch fermentation process. ISA Transactions

    2024

  16. [24]

    A case study of conditional deep convolutional generative adversarial networks in machine fault diagnosis

    Luo, J., Huang, J., Li, H., 2021. A case study of conditional deep convolutional generative adversarial networks in machine fault diagnosis. Journal of Intelligent Manufacturing 32, 407–425

    2021

  17. [25]

    Imbalanced fault diagnosis of rolling bearing based on generative adversarial network: A com- parative study

    Mao, W., Liu, Y., Ding, L., Li, Y., 2019. Imbalanced fault diagnosis of rolling bearing based on generative adversarial network: A com- parative study. IEEE Access 7, 9515–9530

    2019

  18. [26]

    Bearing fault detection and diagnosis using case western reserve university dataset with deep learning approaches: A review

    Neupane, D., Seok, J., 2020. Bearing fault detection and diagnosis using case western reserve university dataset with deep learning approaches: A review. IEEE Access

    2020

  19. [27]

    Plug & play generative networks: Conditional iterative generation of images in latent space

    Nguyen,A.,Clune,J.,Bengio,Y.,Dosovitskiy,A.,Yosinski,J.,2017. Plug & play generative networks: Conditional iterative generation of images in latent space

    2017

  20. [28]

    A novel class imbalance-robust network for bearing fault diagnosis utilizing raw vibration signals

    Qian, W., Li, S., 2020. A novel class imbalance-robust network for bearing fault diagnosis utilizing raw vibration signals. Measurement 156, 107567

    2020

  21. [29]

    Improvement of Generative Adversarial Network and its application in bearing fault diagnosis: A review

    Ruan, D., Chen, X., Gühmann, C., Yan, J., 2023. Improvement of Generative Adversarial Network and its application in bearing fault diagnosis: A review. Lubricants

    2023

  22. [30]

    Improved techniques for training GANs arXiv:1606.03498

    Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., Chen, X., 2016. Improved techniques for training GANs arXiv:1606.03498

    2016 arXiv

  23. [31]

    A novel deep autoen- coder feature learning method for rotating machinery fault diagnosis

    Shao, H., Jiang, H., Zhao, H., Wang, F., 2017. A novel deep autoen- coder feature learning method for rotating machinery fault diagnosis. Mechanical Systems and Signal Processing 95, 187–204

    2017

  24. [32]

    A novel intelligent fault diagnosis method for rolling bearings based on wasserstein generative adversarial network and convolutional neural network under unbalanced dataset

    Tang, H., Gao, S., Wang, L., Li, X., Li, B., Pang, S., 2021. A novel intelligent fault diagnosis method for rolling bearings based on wasserstein generative adversarial network and convolutional neural network under unbalanced dataset. Sensors 21, 6754

    2021

  25. [33]

    Counterfactual expla- nations without opening the black box: Automated decisions and the gdpr

    Wachter, S., Mittelstadt, B., Russell, C., 2018. Counterfactual expla- nations without opening the black box: Automated decisions and the gdpr

    2018

  26. [34]

    Coun- terfactual data generation method for fault diagnosis of complex electromechanical systems

    Wang, C., Liu, J., Cao, J., Chen, X., Chen, L., Ji, Y., 2025. Coun- terfactual data generation method for fault diagnosis of complex electromechanical systems. IEEE Transactions on Instrumentation and Measurement

    2025

  27. [35]

    Wang, Z., Samsten, I., Miliou, I., Mochaourab, R., Papapetrou, P.,

  28. [36]

    Anewconvolutionalneural network-baseddata-drivenfaultdiagnosismethod.IEEETransactions on Industrial Electronics 65, 5990–5998

    Wen,L.,Li,X.,Gao,L.,Zhang,Y.,2018. Anewconvolutionalneural network-baseddata-drivenfaultdiagnosismethod.IEEETransactions on Industrial Electronics 65, 5990–5998

    2018

  29. [37]

    Counterfactual inference for generalized zero-shot compound-fault diagnosis

    Xu, J., Kong, H., Ding, X., Yuan, X., 2025. Counterfactual inference for generalized zero-shot compound-fault diagnosis. IEEE Transac- tions on Instrumentation and Measurement

    2025

  30. [38]

    Deep learning algorithms for bearing fault diagnostics—a comprehensive review

    Zhang, S., Zhang, S., Wang, B., Habetler, T.G., 2020. Deep learning algorithms for bearing fault diagnostics—a comprehensive review. IEEE Access 8, 29857–29881. M.Alavi et al. Page 15 of 15

    2020

  31. [2017]

    Improved training of wasserstein GANsarXiv:1704.00028

    arXiv

  32. [2024]

    Machine Learning

    Glacier:guidedlocallyconstrainedcounterfactualexplanations for time series classification. Machine Learning

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.