Research on Fault Diagnosis Algorithm of Ship Electric Propulsion Motor

Round 1

Reviewer 1 Report

1- Some parts of the paper are so lengthy (e.g., part 2) and it can be shortened or removed to an appendix.

2- Please clarify the novelty of the article. What are your contributions to the field?

3- Please add a part to describe simulations executed to get the results to increase the clarity and readability of the article.

4- The quality of some of the figures should be improved (figures 7, 8, and 9).

5- The conclusion section is not well organized. Please reorganize it and enrich it with some data. 

Author Response

Responses to Reviewer 1 Comments

Dear Editors and Reviewers:

We thank you very much for giving us an opportunity to revise our manuscript. Thanks a lot for the helpful comments and recommends and thank you for your time spent. According to the reviewers' comments, we have revised the manuscript.

The main corrections in the paper and the responds to the comments are as follows:

1. Some parts of the paper are so lengthy (e.g., part 2) and it can be shortened or removed to an appendix.
   Response: Thanks very much for your affirmation to my manuscript. I am sorry that this problem has troubled you. The second part of the article involves the basic principle of the method applied in this paper. If the content of this part is deleted, it may lead to the reader 's inability to understand the content of the improved algorithm in the third part, so I retained the content of the second part. Hope to get your understanding.
   2. Please clarify the novelty of the article. What are your contributions to the field?

Response: According to your suggestion, the fifth part of this article has been modified. The fifth part summarizes the main work and contributions of this paper.
3.    Please add a part to describe simulations executed to get the results to increase the clarity and readability of the article.
Response: To improve the readability of the article, the fourth part of the article was modified. The feasibility of the algorithm in this paper is illustrated by ablation experiments, and the results of each experiment are analyzed and compared.

4. The quality of some of the figures should be improved (figures 7, 8, and 9).

Response: I am sorry for the trouble that the figure clarity brings to your reading. In the revised manuscript, text descriptions are added to Figure 7, Figure 8, and Figure 9 to facilitate you to understand the content of each picture. However, due to the limitations of manuscript layout and image content, it is difficult to further improve the clarity of the image. Hope to get your understanding.

5. The conclusion section is not well organized. Please reorganize it and enrich it with some data.

Response: The revised manuscript has revised the conclusion part. In the fifth part of the article, the main contributions of this paper are summarized, and the future work is prospected.

Author Response File: Author Response.docx

Reviewer 2 Report

1.     Introduction section must be written on more quality way, i.e. more up-to-date references addressed. Research gap should be delivered on more clear way with directed necessity for the conducted research work

2.     The literature review is very short. The authors should add more recently published papers in the same area of research and make a detailed literature review.

3.     Add the Nomenclatures and Abbreviations in order to improve the quality and readability of the manuscript.

4. The diagrams are not clear to understand

5. simulation results are not good and clarity

6.Compare your proposed works to existing works 

 

Author Response

Responses to Reviewer 2 Comments

Dear Editors and Reviewers:

We thank you very much for giving us an opportunity to revise our manuscript. Thanks a lot for the helpful comments and recommends and thank you for your time spent. According to the reviewers' comments, we have revised the manuscript.

The main corrections in the paper and the responds to the comments are as follows:

1. Introduction section must be written on more quality way, i.e., more up-to-date references addressed. Research gap should be delivered on more clear way with directed necessity for the conducted research work.

Response: Thanks very much for your affirmation to my manuscript. According to your suggestion, the introduction of the article was modified. In the introduction part of the article, firstly, the research background and significance of the article are introduced, and the current results of permanent magnet synchronous motor fault diagnosis are introduced, and many research results are listed. Secondly, the application of deep learning method in fault diagnosis of permanent magnet synchronous motor is introduced, and the current research results are analyzed, and the limitations of the research methods are pointed out. Then, the current achievements on the diagnosis of inter-turn short circuit fault of permanent magnet synchronous motor are introduced, and the shortcomings of the results are pointed out. Finally, the main work of this paper is introduced, and the chapter of the paper is briefly introduced.
2.    The literature review is very short. The authors should add more recently published papers in the same area of research and make a detailed literature review.

Response: According to your suggestion, the introduction of the article was modified. In the introduction part of the article, the recent papers published in fault diagnosis of permanent magnet synchronous motor, fault diagnosis of permanent magnet synchronous motor based on deep learning and fault diagnosis of inter-turn short circuit of permanent magnet synchronous motor are listed, and the papers are expounded and analyzed.

3. Add the Nomenclatures and Abbreviations in order to improve the quality and readability of the manuscript.

Response: According to your suggestions, terms and abbreviations are added to the article to improve the quality of the article.

4. The diagrams are not clear to understand.

Response: I am sorry that this problem has troubled you. In order to improve the readability of the article, each simulation result is illustrated in the fourth part of the article to help you better understand the meaning of the chart.

5. Simulation results are not good and clarity.

Response: I am sorry for the trouble that the figure clarity brings to your reading. In the revised manuscript, each picture is given a text description, and in order to increase the clarity of the picture, the simulation results of the original Figure 13 and Figure 14 are displayed and explained one by one. However, due to the limitations of manuscript layout and picture content, it is difficult to further improve the clarity of the picture for Figure 7, Figure 8, and Figure 9. Hope to get your understanding.

6. Compare your proposed works to existing works.

Response: According to your suggestion, in the introduction part of the article, the shortcomings of the existing works are pointed out through literature review, and the improvement content of this article is described in detail.

Author Response File: Author Response.docx

Reviewer 3 Report

Major comment:

1.       The authors used an LSTM-based model for fault diagnosis and optimized its parameter with the whale algorithm. Unfortunately, throughout the manuscript, any reasons or explanations of why LSTM was used among various deep learning models were not found. Other models (e.g., RNN and 1D-CNN) can also be used for time-series data[R1-R3]. It is required to prove the superiority of LSTM for fault diagnosis of ship electric propulsion motors.

[R1] Park, Chan Hee, et al. "A feature inherited hierarchical convolutional neural network (FI-HCNN) for motor fault severity estimation using stator current signals." International Journal of Precision Engineering and Manufacturing-Green Technology 8 (2021): 1253-1266.

[R2] Hoang, Duy Tang, and Hee Jun Kang. "A motor current signal-based bearing fault diagnosis using deep learning and information fusion." IEEE Transactions on Instrumentation and Measurement 69.6 (2019): 3325-3333.

[R3] Seera, Manjeevan, et al. "Fault detection and diagnosis of induction motors using motor current signature analysis and a hybrid FMM–CART model." IEEE transactions on neural networks and learning systems 23.1 (2011): 97-108.

2.       The authors decompose signals with a kurtosis value greater than three and reconstruct the signal. It’s a reasonable approach that uses statistical features as a criterion; however, there is no explanation of why kurtosis is selected among various statistical features and how the kurtosis value of three was determined. The reviewer thinks it is essential to fully explain all aspects of the proposed method. Therefore, the reviewer recommends that the authors include a specific explanation and analysis that can verify why and how kurtosis is used.

3.     The authors performed a comparison study through figures 10 to 13 to verify the performance of the proposed method. In this study, the authors compared the unimproved and improved components of the proposed method. The reviewer thinks that to verify the performance of the proposed method properly, an ablation study should be performed by eliminating each component of the proposed method one by one. In this perspective, the reviewer cannot agree with the contribution of the proposed method. As the proposed method consists of three modules (Improved VMD, Improved multi-scale, Improved Bi LSTM), it is recommended to conduct an ablation study based on the three modules. For example, the proposed method can be compared with the improved Bi LSTM method. Also, it is advisable to provide a clear explanation of the comparison model in the ablation study.

4.     Finally, the reviewer has a severe concern about the proposed method validation. In section 4, the author uses one dataset obtained from Simulink in MATLAB. Also, the author claims the proposed method achieves 100% fault diagnostic accuracy. However, the reviewer thinks the proposed method overfits the dataset. It is recommended to use an additional dataset to verify the generalization ability of the proposed method. For example, the open dataset containing inter-turn short-circuit data can be used.

Minor comment:

1.       It is highly recommended to improve the resolution of all figures.

2.       Typos and missing letters are found manuscript. For example, many words do not have a '-' in Bi-LSTM words. The authors should thoroughly review and proofread the manuscript.

3.       Most figures are unclear. A part of the box of figure 1 is cut off, and the target class handwriting of Figures 13 and 15 is awkward. Also, Chinese are found in figure 9.

 

 

Author Response

Responses to Reviewer 3 Comments

Dear Editors and Reviewers:

We thank you very much for giving us an opportunity to revise our manuscript. Thanks a lot for the helpful comments and recommends and thank you for your time spent. According to the reviewers' comments, we have revised the manuscript.

The main corrections in the paper and the responds to the comments are as follows:

1. The authors used an LSTM-based model for fault diagnosis and optimized its parameter with the whale algorithm. Unfortunately, throughout the manuscript, any reasons or explanations of why LSTM was used among various deep learning models were not found. Other models (e.g., RNN and 1D-CNN) can also be used for time-series data[R1-R3]. It is required to prove the superiority of LSTM for fault diagnosis of ship electric propulsion motors.

Response: Thanks very much for your affirmation to my manuscript. I am sorry that this problem has troubled you. In the introduction of the revised article, the reasons why LSTM neural networks are used in deep learning models are explained. In the process of fault diagnosis of permanent magnet synchronous motor, CNN is usually used as the feature extraction model of time series. LSTM is usually used as a fault diagnosis model and as a variant of RNN neural network, which can solve the problem of disappearance of traditional RNN gradient.

2. The authors decompose signals with a kurtosis value greater than three and reconstruct the signal. It’s a reasonable approach that uses statistical features as a criterion; however, there is no explanation of why kurtosis is selected among various statistical features and how the kurtosis value of three was determined. The reviewer thinks it is essential to fully explain all aspects of the proposed method. Therefore, the reviewer recommends that the authors include a specific explanation and analysis that can verify why and how kurtosis is used.

Response: I am sorry that this problem has troubled you. Using kurtosis value three as the standard is a common method in the process of time domain signal processing. In the fourth part of the modified article, specific references are given. Hope to get your understanding.

3. The authors performed a comparison study through figures 10 to 13 to verify the performance of the proposed method. In this study, the authors compared the unimproved and improved components of the proposed method. The reviewer thinks that to verify the performance of the proposed method properly, an ablation study should be performed by eliminating each component of the proposed method one by one. In this perspective, the reviewer cannot agree with the contribution of the proposed method. As the proposed method consists of three modules (Improved VMD, Improved multi-scale, Improved Bi LSTM), it is recommended to conduct an ablation study based on the three modules. For example, the proposed method can be compared with the improved Bi LSTM method. Also, it is advisable to provide a clear explanation of the comparison model in the ablation study.

Response: According to your suggestion, the fourth part of the article has been significantly modified. The ablation experiment of the improved VMD, improved multi-scale and improved Bi LSTM modules was studied respectively, and in order to improve the readability of the article, a text description was added to each simulation result.

4. Finally, the reviewer has a severe concern about the proposed method validation. In section 4, the author uses one dataset obtained from Simulink in MATLAB. Also, the author claims the proposed method achieves 100% fault diagnostic accuracy. However, the reviewer thinks the proposed method overfits the dataset. It is recommended to use an additional dataset to verify the generalization ability of the proposed method. For example, the open dataset containing inter-turn short-circuit data can be used.

Response: I am sorry that this problem has troubled you. At present, most of the research on inter-turn short circuit fault of permanent magnet synchronous motor is based on the data set obtained by Simulink simulation program. Due to the limitation of conditions, it is difficult to find an open data set containing inter-turn short circuit data on the network. In the future, we will expand the data set and conduct more extensive experimental research on other fault types of PMSM. Hope to get your understanding.

Minor comment:

1. It is highly recommended to improve the resolution of all figures.

Response: I am sorry for the trouble that the figure clarity brings to your reading. In the revised manuscript, each picture is given a text description, and in order to increase the clarity of the picture, the simulation results of the original Figure 13 and Figure 14 are displayed and explained one by one. However, due to the limitations of manuscript layout and picture content, it is difficult to further improve the clarity of the picture for Figure 7, Figure 8, and Figure 9. Hope to get your understanding.

2. Typos and missing letters are found manuscript. For example, many words do not have a '-' in Bi-LSTM words. The authors should thoroughly review and proofread the manuscript.

Response: According to your suggestion, the manuscript has been thoroughly reviewed and proofread.

3. Most figures are unclear. A part of the box of figure 1 is cut off, and the target class handwriting of Figures 13 and 15 is awkward. Also, Chinese are found in figure 9.

Response: I am sorry for the trouble caused by unclear figures to you reading the article. In the revised manuscript, a text description is added to each simulation result. Because the purpose of Figure 9 is only to show the waveform diagram decomposed by VMD, it has no practical effect on the reading of the article, and because the image content is too compact, it is difficult to improve the clarity of the image, so we choose to delete Figure 9. Hope to get your understanding.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

No further comments.

Author Response

Dear Editors and Reviewers:

We thank you very much for giving us an opportunity to revise our manuscript. Thanks a lot for the helpful comments and recommends and thank you for your time spent. Thank you again for your recognition of our article.

Reviewer 2 Report

 Improve the quality of figure 8 (a),(b) and (c). The X and Y axis scales are not visible

 

Author Response

Dear Editors and Reviewers:

We thank you very much for giving us an opportunity to revise our manuscript. Thanks a lot for the helpful comments and recommends and thank you for your time spent. According to the reviewers' comments, we have revised the manuscript.

The main corrections in the paper and the responds to the comments are as follows:

1. Improve the quality of figure 8 (a), (b) and (c). The X and Y axis scales are not visible.

Response: Thanks very much for your affirmation to my manuscript. I am sorry for the trouble that the figure clarity brings to your reading. In the revised manuscript, I improve the quality of figure 8 (a), (b) and (c) to make the coordinate scale clearer.

Author Response File: Author Response.docx