High-Resistance Connection Fault Diagnosis in Ship Electric Propulsion System Using Res-CBDNN

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes a Res-Convolutional BiLSTM Deep Neural Network (Res-CBDNN) for fault detection and diagnosis of the ship's electric propulsion system. The following are my suggestions/comments.

 

• My major concern is the lack of comparisons with existing literature. While considering and comparing with CNN and CBDNN is a positive step, it is insufficient to demonstrate a significant contribution. To strengthen the work, I highly recommend incorporating baselines, comparisons with other relevant studies, and clearly highlighting the improvements achieved through your contributions. Without benchmarks or comparisons, it is difficult to assess the novelty and effectiveness of the proposed approach, and the results presented in isolation do not provide a clear picture of its value.
• The comparison with CNN and CBDNN needs improvement. Clearly indicate the reference numbers (in the tables) of the specific works. Include these references and explicitly compare the proposed Res-CBDNN with the existing CNN and CBDNN methods. Without proper citations and comparisons, it appears as if you are simply comparing three models without justification.
• Further, the metrics used in the research may not provide deeper insights. It's good to consider a few more metrics.
• It’s recommended to check for more suitable Transfer learning approaches and check their performances [19].
• In the abstract, it’s good to briefly discuss existing state-of-the-art methods used for fault detection and diagnosis in the literature and the major issues of those methods. This helps the reader to clearly understand the need for the proposed Res-CBDNN and its importance to the field.
• The introduction section does not have good connectivity and looks like a summary version of papers from the literature. I strongly recommend the authors to make changes to this. The reader should gain a good understanding of the research area, major breakthroughs, and the research gaps leading to a clear problem statement. Subsequently, the contributions of this paper's work should be highlighted.
• Some of the equations are not cited in the text, check for these instances.

Author Response

Dear Reviewer,

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “High-Resistance Connection Fault Diagnosis in Ship Electric Propulsion System Using Res-Convolutional Bidirectional Long Short Term Memory Deep Neural Network”(jmse-2921462). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researche. We have studied comments carefully and have made corrections which we hope meets with approval. Revised portions are marked in red in the paper. The main corrections in the paper and the respondses to the reviewer’s comments are as flowing

 

Point 1: To strengthen the work, I highly recommend incorporating baselines, comparisons with other relevant studies, and clearly highlighting the improvements achieved through your contributions.

Response 1: I am sorry that this part was not clear in the original manuscript. I should have explained that in this paper, based on the results of CNN diagnosis model, we evaluate the two improved models, namely CBDNN and res-CBDNN. In the field of fault diagnosis of marine electric propulsion system, the accuracy and robustness of diagnosis results are usually paid more attention, that is, the ability of maintaining high accuracy under the condition of noise pollution, comparing CBDNN with res-CBDNN, the diagnostic results show that the proposed method has some advantages over CNN and CBDNN in terms of diagnostic accuracy and anti-jamming ability, it can be used in HRC fault diagnosis of ship propulsion system.

 

Point 2: The comparison with CNN and CBDNN needs improvement. Clearly indicate the reference numbers (in the tables) of the specific works.

Response 2: As suggested by the reviewer, we have added Ref.32 and Ref.33. The CNN model and the CBDNN model are established in reference 32 and 33, respectively. Although they were originally used to diagnose other types of faults, we modified their structure to produce CNN and CBDNN models that can be used for HRC fault diagnosis. Based on CNN, the CBDNN and Res-CBDNN is improved by adding Bilstm network and ResNet, and finally compared with the Res-CBDNN, it is concluded that the proposed method can guarantee more than 85% diagnostic accuracy in the presence of noise.

 

Point 3: It’s recommended to check for more suitable Transfer learning approaches and check their performances [19].

Response 3: Incorrect references have been corrected. We were really sorry for our careless mistakes. Thank you for your reminder.

 

Point 4: In the abstract, it’s good to briefly discuss existing state-of-the-art methods used for fault detection and diagnosis in the literature and the major issues of those methods.

Response 4: We have re-written this part according to the Reviewer’s suggestion.

 

Point 5: The introduction section does not have good connectivity and looks like a summary version of papers from the literature.

Response 4: We sincerely appreciate the valuable comments. Based on the research and development of HRC fault diagnosis, this paper combs the relevant literature, analyzes the advantages and disadvantages of the methods mentioned in the literature, and the progressive relationship among the different methods, hoping to form a clear problem statement, then, we rewrite the last paragraph of the introduction, focusing on the contribution of this paper..

We tried our best to improve the manuscript and made some changes marked in red in revised paper which will not influence the content and framework of the paper. We appreciate for Editors/Reviewers’ warm work earnestly, and hope the correction will meet with approval. Once again, thank you very much for your comments and suggestions.

In all, I found the reviewer’s comments are quite helpful, and I revised my paper point-by-point. Special thanks to you for your good comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

My comments on your work are as follows:

The meaning of the abbreviations mentioned in line 81 is not clear.

The assertion (lines 91-92) is questionable, since gas turbines are used on ships to a much lesser extent than diesel engines. The reason for this is that gas turbines with an output of less than approx. 50 MW have a significantly lower efficiency than diesel engines. In addition, the efficiency of gas turbines is significantly reduced by the reduction in load. COGAS and COGES propulsion systems can achieve the same or higher efficiency than diesel engines, but only at nominal load.

The sentence (lines 91-95) is too long.

Figure 1 does not show the typical layout of a ship electric system with electric transmission. Normally, the ship has one main switchboards board, which is divided into two or more sections for reliability reasons. In propulsion systems with electrical transmission, the high-voltage and low-voltage switchboards also differ.

It is necessary to harmonize the corresponding text and Figure 3 with regard to the to the name of the figure "System performance in alarm and fault condition".

Should "Alter" or "Alert" be written in Figure 4?

It is recommended to separate the units of measurement from the names of the physical quantities in one of the two ways, e.g. current, A or current (A).

The symbols of the physical quantities should be written in italics.

In Table 2, the unit of measurement for torque (Nm) should be corrected.

Figure 11 – The units of measurement next to the axis designations are missing.

The method to which the data set refers must be specified in line 260.

Clarification of the designation Acc?

Correct spelling of the unit for the noise level dB in Figure 13.

Highlight the year of publication of reference 6 by using the "bold" option.

 

Best regards!Top of Form

Author Response

Dear Reviewer,

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “High-Resistance Connection Fault Diagnosis in Ship Electric Propulsion System Using Res-Convolutional Bidirectional Long Short Term Memory Deep Neural Network”(jmse-2921462). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researche. We have studied comments carefully and have made corrections which we hope meets with approval. Revised portions are marked in red in the paper. The main corrections in the paper and the respondses to the reviewer’s comments are as flowing

 

Point 1: The meaning of the abbreviations mentioned in line 81 is not clear.

Response 1: We will give the full name of the method in line 81. support vector machines(SVM), K-nearest neighbor(KNN);

 

Point 2:. The assertion (lines 91-92) is questionable.

Response 2: We agree with the Reviewer that gas turbines are not as efficient as diesel engines as the main engine to drive the propellers. But the gas turbines and diesel engines here shall be interpreted as prime movers of the generator set.That is, together with the generators, it constitutes a gas turbine generator set and a diesel generator set, and serving as the generating units of the ship, to provide high quality and stable power for ship electric propulsion system.

 

Point 3:. The sentence (lines 91-95) is too long.

Response 3: We have re-written this part (line 91-95) according to the Reviewer's suggestion..

 

Point 4:. Figure 1 does not show the typical layout of a ship

Response 4: We have modified Figure 1

 

Point 5: Harmonize the corresponding text and Figure 3

Response 5: We have changed the text for Figure 3 as "System performance of Alert and Degraded "

 

Point 6: Should "Alter" or "Alert" be written in Figure 4?

Response 6: The word should be "Alert" and has now been corrected. We were really sorry for our careless mistakes. Thank you for your reminder.

 

Point 7-10: Separate the units of measurement from the names of the physical quantities, The symbols of the physical quantities should be written in italics; In Table 2, the unit of measurement for torque (Nm) should be corrected. Figure 11 – The units of measurement next to the axis designations are missing.

Response 7-10: We have checked and filled in all the units of physical quantities in the diagrams. (Figure 4, Table 2, Figure 11)

 

Point 11:. The method to which the data set refers must be specified in line 260.

Response 11: We choose 10,400 simulated data as training data set in section 4.1 to train the three networks, and compare their training effects

 

Point 12: Clarification of the designation Acc?

Response 12: Acc is an abbreviation for Accurcy. For the sake of context unity, it is all changed to Accurcy

 

Point 13: Correct spelling of the unit for the noise level dB in Figure 13.

Response 13: The noise level in Figure 13 has been corrected to "dB"

 

Point 14: Highlight the year of publication of reference 6 by using the "bold" option.

Response 14: We have highlighted the year of publication of reference 6 by using the "bold" option.

We tried our best to improve the manuscript and made some changes marked in red in revised paper which will not influence the content and framework of the paper. We appreciate for Editors/Reviewers’ warm work earnestly, and hope the correction will meet with approval. Once again, thank you very much for your comments and suggestions.

In all, I found the reviewer’s comments are quite helpful, and I revised my paper point-by-point. Special thanks to you for your good comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The paper has been revised well. There are no other comments.