SSA-SL Transformer for Bearing Fault Diagnosis under Noisy Factory Environments
Round 1
Reviewer 1 Report
The SSA-SLTransformer framework is proposed to solve the bearing fault diagnostics. SSA is used to remove noise in the process of decomposing and reconstructing signals and SLTransformer is a new transformer model. It was confirmed that the SSA-SLTransformer has a higher accuracy performance than other methods through experiments. This paper is well structured. However, the English is poor:
- Some sentences are redundant, some repeated words should be deleted:
In the line 156、157、219-228、269、307、309、313-314、349
- In Figure 1, the frame size of the structure diagram should be drawn the same.
- In the line 138, “(3)” should be rewritten as “In formula (3)”.
- In Figure 2, the blocks in structure diagram are not aligned, the font size is different, the arrows are messed up.
- In the line 164, “two-way” and “bi-directive” are repeated in meaning.
- Some sentences’ grammar is wrong:
In the line 167、232、243、254、277、294、310
- The font size of Figure 3 is too small. The text in the figure 3 and 4 should be centered.
- In the line 203, this sentence is repeated.
- More experimental details such as parameter settings should be added.
- The advantage of Swish activation function has not been confirmed by a controlled trial.
Author Response
Dear Reviewer 1,
Special Issue "Advances in Fault Detection/Diagnosis of Electrical Power Devices" of Multidisciplinary Digital Publishing Institute (MDPI) Electronics
Thank you very much for your letter of April 14, 2022 regarding my paper ID electronics-1668990 entitled “SSA-SLTransformer for Bearing Fault Diagnosis under Noisy Factory Environments”.
Based on the reviewer’s comments, I have incorporated them in the revised version. The comments regarding our manuscript were extremely helpful to us in preparing a clearer version. We have rewritten many paragraphs according to the recommendations of the referees. In addition, the revised paper has been proofread by a good technical writer whose first language is English. Thank you very much for your advice. Attached is a copy of the revised version of the manuscript and a list of the revisions. Your acknowledgment will be greatly appreciated. Thanks once again for your significant help with this paper.
Thanks once again for your significant help with this paper.
Sincerely yours,
Seoyeong Lee, Graduate Student.
Department of Smart Factory Convergence (Advisor: Jongpil Jeong)
College of Software
Sungkyunkwan University (SKKU)
2066 Seobu-ro Jangan-Gu
Suwon 440-746, KOREA
Tel: +82-31-299-4260(Office)
Email: [email protected]
Reviewer 2 Report
The authors proposed the swish-LSTM transformer framework by redesigning the internal structure of the transformer using the swish activation function and long short-term memory. Then, the experiments of bearing failures were performed to demonstrate its effectiveness. Some comments and suggestions are given as below.
(1) As we all know, the feature extraction is vital to train intelligent fault diagnosis models. Therefore, it is suggested that the authors could review some advanced signal processing method for extracting weak fault features, such as [1-3].
[1] A second-order stochastic resonance method enhanced by fractional-order derivative for mechanical fault detection, Nonlinear Dynamics, 2021: 1-17.
[2] Blind deconvolution assisted with periodicity detection techniques and its application to bearing fault feature enhancement, Measurement, 2020, 159: 107804.
[3] Enhancement of rolling bearing fault diagnosis based on improvement of empirical mode decomposition denoising method, The International Journal of Advanced Manufacturing Technology, 2018, 97(5): 3099-3117.
(2) Section 1.1 shows the SSA Algorithm. In SSA, the singular value decomposition is a key ingredient, such as Hankel matrix. Even, the singular value decomposition is referred as [1].
[1] SVD principle analysis and fault diagnosis for bearings based on the correlation coefficient, Measurement Science and Technology, 2015, 26(8): 085014.
(3) Lots of articles on intelligent fault diagnosis of roller bearings using deep learning are able to directly learn useful signatures from the raw signal or its spectrum of roller bearings, and then use softmax to classify faults. In this manuscript, the authors use the singular spectrum analysis to preprocess the noisy vibration signals of roller bearings. Why the authors do not use deep learning to learn or extract fault signatures from the raw signals instead of the singular spectrum analysis?
(4) Fig. 10 and Fig. 11 show the accuracy graph of Non-applying SSA algorithm for noise data and the accuracy graph of applying SSA algorithm for noise data, respectively. It shows that preprocessing the noisy signals using SSA would improve the performance of deep learning models. Whether noise can be used to enhanced artificial intelligent models to design the noise robust model of AI?
(5) Fig. 6 and Fig. 7 show the raw signals and noisy versions of roller bearings. Please add the corresponding x-label and y-label.
Author Response
Dear Reviewer 2,
Special Issue "Advances in Fault Detection/Diagnosis of Electrical Power Devices" of Multidisciplinary Digital Publishing Institute (MDPI) Electronics
Thank you very much for your letter of April 14, 2022 regarding my paper ID electronics-1668990 entitled “SSA-SLTransformer for Bearing Fault Diagnosis under Noisy Factory Environments”.
Based on the reviewer’s comments, I have incorporated them in the revised version. The comments regarding our manuscript were extremely helpful to us in preparing a clearer version. We have rewritten many paragraphs according to the recommendations of the referees. In addition, the revised paper has been proofread by a good technical writer whose first language is English. Thank you very much for your advice. Attached is a copy of the revised version of the manuscript and a list of the revisions. Your acknowledgment will be greatly appreciated. Thanks once again for your significant help with this paper.
Thanks once again for your significant help with this paper.
Sincerely yours,
Seoyeong Lee, Graduate Student.
Department of Smart Factory Convergence (Advisor: Jongpil Jeong)
College of Software
Sungkyunkwan University (SKKU)
2066 Seobu-ro Jangan-Gu
Suwon 440-746, KOREA
Tel: +82-31-299-4260(Office)
Email: [email protected]
Reviewer 3 Report
The paper under review considers the issue of "Fault Diagnosis of Wind Energy Conversion Systems Using Gaussian Process Regression-based Multi-Class Random Forest "
In the reviewer’s opinion, in general, the paper is quite interesting.
However, there is important aspects that require authors comments:
1) In bearing diagnostics, the use of SVD and Hankel matrices is used
to filter the signal to enhance the vibration signal of the bearing.
The main problem is determining how many singular values to choose for signal reconstruction.
How do the authors choose the number of singular values for signal reconstruction???
Author Response
Dear Reviewer 3,
Special Issue "Advances in Fault Detection/Diagnosis of Electrical Power Devices" of Multidisciplinary Digital Publishing Institute (MDPI) Electronics
Thank you very much for your letter of April 14th 2022 regarding my paper ID electronics-1668990 entitled “SSA-SLTransformer for Bearing Fault Diagnosis under Noisy Factory Environments”.
Based on reviewer’s comments, I have incorporated them in the revised version. The comments of the regarding our manuscript were extremely helpful to us in preparing a clearer version. We have rewritten many paragraphs according to the recommendations of the referees. In addition, the revised paper has been proofread by a good technical writer whose first language is English. Thank you very much for your advisement. Attached are a copy of the revised version of the manuscript and a list of the revisions. Your acknowledgement will be greatly appreciated. Thanks once again for your significant help with this paper.
Thanks once again for your significant help with this paper.
Sincerely yours,
Seoyeong Lee, Graduate Student.
Department of Smart Factory Convergence (Advisor: Jongpil Jeong)
College of Software
Sungkyunkwan University (SKKU)
2066 Seobu-ro Jangan-Gu
Suwon 440-746, KOREA
Tel: +82-31-299-4260(Office)
Email: [email protected]
Round 2
Reviewer 2 Report
The revised version could be accepted now.
Author Response
Dear reviewer2,
Thanks to your advice, we think we made a good thesis because of your comments. thank you.
Sincerely yours,
Seoyeong Lee, Graduate Student.
Department of Smart Factory Convergence (Advisor: Jongpil Jeong)
College of Software
Sungkyunkwan University (SKKU)
2066 Seobu-ro Jangan-Gu
Suwon 440-746, KOREA
Tel: +82-31-299-4260(Office)
Email: [email protected]