A Novel Fault Diagnosis Method for High-Temperature Superconducting Field Coil of Superconducting Rotating Machine

Round 1

Reviewer 1 Report

I cannot recommend this manuscript for publication. The authors should improve the manuscript and, at least, carry out the following comments:

1. Acronyms should be defined in the manuscript.

2. The introduction should be improved and the authors should avoid reference overkill, making more sentences that show the main authors' contributions.

3. The methodology section and results and discussion section should be presented more clearly. They are mixed in the manuscript. The authors should divide section 2 into one section for methodology and another one for results and discussion.

4. Additional experiments should be carried out.

Author Response

Thank you for your kind comments on our paper. The manuscript was modified to reflect comments. The answers for each comments are as follows.

1. Acronyms should be defined in the manuscript.

It was reflected in the attached paper.

2. The introduction should be improved and the authors should avoid reference overkill, making more sentences that show the main authors' contributions.

It was reflected in the attached paper.

3. The methodology section and results and discussion section should be presented more clearly. They are mixed in the manuscript. The authors should divide section 2 into one section for methodology and another one for results and discussion.

It was also reflected in the attached paper. Please check attached paper.

4.  Additional experiments should be carried out.

- As you commented, we understand that additional experiments are required to help readers underatand. However, the purpose of paper is propose a new quech detection method that does not exist using resonance filter. Therefore, we presented the limitations of detecting quench in time domain, and compared the advantages of our method based on the experimental results in section 3. We also duscussed the limitations of our method in section 3. So, please check the revised paper.

Reviewer 2 Report

The paper presents a method for quench detection in high-temperature 12 superconductor (HTS) rotating machinery.

Despite I am not an expert of this type of industrial equipment, I found the following issues:

Issue 1

The introduction is too short. A proper literature review of existing method in literature for quench detection is totally missing. The Authors should properly discuss pros and cons of existing methods. Further, the Authors should explain which is the main contribution of their work in the field.

Issue 2

A Section “Problem statement”  which describes Authors’ assumptions on the available data and states the problem is totally missing.

Issue 3

The methodology to perform quench detection is limited to FFT analysis and results are just shown pictorially through Figure 12. I expected the Authors to develop a proper fault detection module. For example, a simple logistic regression classifier with input the selected amplitude(s) can be developed for the fault detection task.

Issue 4

There are many typos and English style should improved.

Author Response

Thank you for your kind comments on our paper. The manuscript was modified to reflect comments. The answers for each comments are as follows.

 

1. The introduction is too short. A proper literature review of existing method in literature for quench detection is totally missing. The Authors should properly discuss pros and cons of existing methods. Further, the Authors should explain which is the main contribution of their work in the field.

It was reflected in the attached paper.

2. A Section “Problem statement”  which describes Authors’ assumptions on the available data and states the problem is totally missing.

It was reflected in the attached paper.

3. The methodology to perform quench detection is limited to FFT analysis and results are just shown pictorially through Figure 12. I expected the Authors to develop a proper fault detection module. For example, a simple logistic regression classifier with input the selected amplitude(s) can be developed for the fault detection task.

As your comment, regression classifier for quench protection has already been developed ("Quench analysis and protection circuit design of a superconducting magnet system for RISP 28GHz ECR ion source", S. Song, T. K. Ko, S. Choi, and M. C. Ahn.). However, the purpose of this paper is to preesnt a new quench detection method. Therefore, the regression classifier for protection was not mentioned in this paper.

4. There are many typos and English style should improved.

- The paper was revised by a native speaker. So, please check attached paper.

Round 2

Reviewer 1 Report

The manuscript has been improved, but the authors should carry out the following minor changes:

1. The references used in the text should be idicated properly. For instance, “N. Nanato et al.” should be “Nanato et al. [6] and Wei et al. [7]” in line 44; “N. Nanato et al. [8]” should be “Nanato and Nakamura” in line 48; etc.

2. Moreover, I do not understand why several references used in the previous manuscript version have been deleted.

Author Response

Thank you for your kind comments on our paper. The manuscript was modified to reflect comments. The answers for each comments are as follows.

 

1. The reference has been modified to indicate properly. Please check revised version.

2. We removed references with duplicate meanings that were going to be presented throughout the paper and added a new sentence indicating the contribution of the main author.

 

 

Reviewer 2 Report

Despite some improvements, some issues are not properly addressed. The Authors to answer to some issues have attached a paper. This is not sufficient. Modifications should be properly addressed in the manuscript itself. The methodology to detect the quench anomaly is still limited to visual inspection, no automatic method has been developed for this task. Further there is no comparison with other methods of literature. No metric is used to assess the quality of the detection (e.g., percentage of missed alarms, percentage of false alarms, mean delay detection time, etc.).

Summarising, the revised version does not address some important issues.  So, the paper is not ready for publication in the current form. 

Author Response

Thank you for your kind comments on our paper. The manuscript was modified to reflect comments. The answers for each comments are as follows.

 

1. When applying the 1 uV/cm criterion, as shown in figure 11 (a), even the visual inspection is impossible with the conventional method. However, as can be seen in figure 12 (a), quench detection is possible using the method presented in this paper. There are various methods for protection after quench detection, and there is a method of automatically shutting off power when a value equal to or greater than the threshold is measured using a relay. However, since this is related to protection rather than quench detection, it is not mentioned in this paper. It is newly mentioned in the paper of IEC criterion, which is one of the methods applied for quench detection in normal time domain. In addition, a percentage of false alarm was calculated and added in relation to figure12 (a). In the FFT signal processing, the change of quench detection time through window size adjustment is mentioned. A reference about detection time for safe protection has been added.

 

Round 3

Reviewer 2 Report

The Authors state at page 9:

"The quench voltage of the INS coil used in the experiment is 1.3 mV
 based on 1μV/cm [15]. In Figure 12 (a), when the INS coil is in the superconducting state, the standard
 deviation is 0.8 mV, which is 0.5 mV less than the quench voltage. Also the sampling rate for data
 acquisition was 1 kHz, so quench detection was performed every 0.1 s [16]. In this case, the probability
 that a false alarm will occur due to the measured value being higher than the quench voltage is 4.5
%. However, when the quench voltage is set to 3 times, the probability that a false alarm occur
converges to 0 % [17]. "

 

Despite these results are interesting I suggest the following changes:

1) The Authors should deeply discuss how they have computed this performance.

2) I suggest to report the results in a Table.

3) Probability is not a percentage. The Authors should talk about "false alarm rate", etc.

4) The AUthors should report the missed alarm rate as well.

 

Finally, before the conclusion session, I suggest the Authors to add a section

"Discussion and outlooks", where the AUthors discuss the limits of their work and improvements that can be addressed in future works.

Author Response

Thank you for your kind comments on our paper. The manuscript was modified to reflect comments. Please check attached paper.

Author Response File: Author Response.docx