A Bearing Fault Detection Method Based on EMDWS-CNT-BO

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

machines-3860210

A Bearing Fault Detection Method Based on EMDWS-CNT-BO

 

This paper introduces a bearing fault detetion method called EMDWS-CNT-BO.Generally, the novelty of the present combination is normal. However, both the conetens and discussions are enough to support it for publication. Some issue as follow：

1.Why use EMDWS? Please give comparisons with other signal processing methods?

2. Too many abbrevations are not well explained, such as ‘MDWS’, etc. Moreover, ‘deep learning architecture (CNN-Transformer)’, etc. Please thoroughly revise the paper.
3. The present model is just suitbable for close-set detection, however, for open-set detection, how to active AI model is a big bone problem must be carefully discussed. Please see: the FEM simulation-driven AI models, DOI10.3390/app6120414 published in 2016 might be the early research metioned this point. Thereafter, many researchers focused on the numerial simulation-aided/driven AI models and its improvement versions (different names). Please add discussion about this point.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

In this manuscript, the authors propose A Bearing Fault Detection Method Based on EMDWS-CNT-BO. Overall, the paper is well-structured, the literature review is thorough, and the experimental results demonstrate the effectiveness of the proposed method. However, the approach is essentially a straightforward concatenation of existing techniques and lacks genuine theoretical novelty. My specific comments are as follows:

1. The proposed “EMDWS-CNT-BO” method is merely a simple combination of several off-the-shelf components, offering no substantial theoretical or structural innovation. Although the authors claim to have “innovatively integrated data pre-processing with hyper-parameter optimization,” such practices are already common in fault diagnosis and do not constitute a significant academic breakthrough.
2. In the methodology section, the authors fail to justify why the “EMD + wavelet denoising” combination was chosen, nor do they compare its advantages or disadvantages with other mainstream denoising strategies.
3. The description of the Bayesian optimization module is vague; crucial details such as specific parameter configurations are missing.
4. In the experiments, the authors only compare the proposed method with self-constructed baselines such as EMD-CNN, EMD-Transformer, and EMD-CNN-Transformer, but do not benchmark against recent state-of-the-art (SOTA) methods, making it difficult to demonstrate any genuine superiority.
5. Although the model achieves over 99 % accuracy on multiple datasets, the results are presented without confidence intervals, standard deviations, or any hypothesis testing, preventing the reader from assessing the statistical significance of the reported improvements.
6. I recommend that the authors provide an in-depth discussion of the strengths and weaknesses of the proposed diagnostic framework and outline clear directions for future research.
7. The manuscript contains numerous typographical errors; please revise carefully.
8. The captions of Figures 6 and 7 appear to be incorrect.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The paper proposes a fault diagnosis strategy for bearing faults. Overall, the paper is organized, but too dense and lengthy. The reviewer comments are as follows:

• Could the authors explain in detail how the proposed approach differs from the literature? Other works also employ the same tools for diagnosing bearing faults:

[1] Y. Han et al., "MT-ConvFormer: A Multitask Bearing Fault Diagnosis Method Using a Combination of CNN and Transformer," in IEEE Transactions on Instrumentation and Measurement, vol. 74, pp. 1-16, 2025

[2] Zhenya Wang, Tao Liu, Xing Wu, Chang Liu, “A diagnosis method for imbalanced bearing data based on improved SMOTE model combined with CNN-AM”, Journal of Computational Design and Engineering, Volume 10, Issue 5, October 2023, Pages 1930–1940.

[3] Lu, Y., Wang, Z., Xie, R. et al. Bayesian optimized deep convolutional network for bearing diagnosis. Int J Adv Manuf Technol 108, 313–322 (2020).

• Page 15, lines 533-535: According to the reviewer understanding, no-load condition was used for training purposes, whereas loading conditions were considered to verify the method effectiveness. Why is such approach considered?
• Information about the vibration sensor positioning, loading conditions and other test conditions is missing.
• The caption of Fig. 6 is not consistent with the figure content. Moreover, the labels of the figure are fairly small. The font size should be increased in this figure and in the subsequent ones.
• When comparing the models performance for the three datasets (including the one obtained experimentally by the authors), it is stated that the accuracy differs the most when considering the self-built dataset. What explains such behaviour?
• The evaluation of the models performance is missing an assessment of the implementation effort. Is the improvement of accuracy compensated by the (likely) increase in computational effort?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

No further comments.

Reviewer 2 Report

Comments and Suggestions for Authors

accept

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have properly addressed the reviewer comments. There are no further observations to add.