Detection and Monitoring of Pitting Progression on Gear Tooth Flank Using Deep Learning

Round 1

Reviewer 1 Report

This paper proposed a machine vision method for detection and monitoring of pitting at the gear tooth. The idea of applying faster-RCNN is good. However, the proposed method is lack of innovation, which is more like an application of an existing method.

 

Some detailed comments:

• In the introduction, you could mention the problem or deficiency of existing researches, so the significance and innovation of your method can be highlighted.

 

• The second part “Gear failure” is also a kind of introduction. I suggest putting it together with the first part, and adjust the structure of the introduction.

 

• It is not necessary to spend one chapter describing the faster-RCNN. Instead, present the innovation of your method in this part.

 

• The data collection part could be more specific. For example, the angle of camera and Photographic light condition should be mentioned.

 

• The title of the fourth part should be “4. Data collection”, instead of “3. Data Collection”.

 

• The second paragraph of the conclusion is the result discussion, which makes your conclusion lengthy. You could make it shorter or put it in the previous chapter.

 

The overall readability of this paper is good. Most figures and tables can clearly present useful information. However, the innovation of this paper is insufficient. I strongly recommend that the author add some innovative ideas and provide some scientific guidance to readers.

Author Response

Dear Madam/Sir,

The authors would like to thank you for your effort and valuable suggestions for improving the paper.

The responses to your review are given with red text according to comments and with appropriate line numbering.

------------

Review with replays:

 

This paper proposed a machine vision method for detection and monitoring of pitting at the gear tooth. The idea of applying faster-RCNN is good. However, the proposed method is lack of innovation, which is more like an application of an existing method.

 

Some detailed comments:

In the introduction, you could mention the problem or deficiency of existing researches, so the significance and innovation of your method can be highlighted.

We emphasized the deficiency of existing researches as well as innovativeness of our paper in the separate part of Introduction. (lines 83-87)

 

The second part “Gear failure” is also a kind of introduction. I suggest putting it together with the first part, and adjust the structure of the introduction.

Authors had the same idea/dilemma during paper writing. We integrate “Gear failure” in Introduction. (lines 88-110)

 

It is not necessary to spend one chapter describing the faster-RCNN. Instead, present the innovation of your method in this part.

The authors think that we need to explain at least the most important parts of used neural network. In principle, reviewer is right, we did not change anything about this neural network. The innovation is practical application of an existing neural network which corresponds to the name of the journal - “Applied science”. We gave some more data about our approach especially in “3. Data collection” (lines 194-212, 217-223). Also lines 83-87 are giving more closer innovativeness and contribution.

 

The data collection part could be more specific. For example, the angle of camera and Photographic light condition should be mentioned.

We expand the part of the paper that is dealing with dataset according to reviewer’s comment. (lines 178-186)

 

The title of the fourth part should be “4. Data collection”, instead of “3. Data Collection”.

We made a change according to reviewer’s comment. (line 159)

 

The second paragraph of the conclusion is the result discussion, which makes your conclusion lengthy. You could make it shorter or put it in the previous chapter.

The comment make sense since this is duplication of mentioning the results. We decrease this chapter by showing just most important results. (lines 290-292)

 

The overall readability of this paper is good. Most figures and tables can clearly present useful information. However, the innovation of this paper is insufficient. I strongly recommend that the author add some innovative ideas and provide some scientific guidance to readers.

There is no innovativeness in the neural network since this is the practical use of existing neural network and we were using the neural network as a tool for detection of pitting on an image. On the other hand, we think that the main idea of the paper is testing innovative idea that is giving answer to the question how to monitor pitting progression using machine vision with quantification. Fact that this is novelty and innovative is that this approach is not publish up to now. Up to now, only few papers were dealing with just detection of pitting, which could be enough for publishing a scientific paper. But in this case we went two steps further: 1) we achieved monitoring; 2) measurement based on machine vision in mm.

Author Response File: Author Response.docx

Reviewer 2 Report

This contribution presents original ideas in the study and advances the previous research in this area. The paper deserves publication. The level of the originality of contribution to the existing knowledge with an emphasis on the paper’s innovativeness in both theory development and methodology used in the study is very high.

This work makes a significant practical contribution and it makes impact on the research work on the research community.

The quality of arguments, the critical analysis of concepts, theories and findings, and consistency and coherency of debate are well addressed in this paper. 

The paper has a good writing style in term of accuracy, clarity, readability, organization, and formatting. 

In figure 1, the proposed architecture of Faster R-CNN deep learning network for defect detection is proposed. Please address the problem of ist Training more in depth and discuss more in depth the optimization Problem which is defined for this Task including ist non convexity.

The paper is a good one and deserves publication

 

Author Response

Dear Madam/Sir,

The authors would like to thank you for your effort and valuable suggestion for improving the paper.

The responses to your review are given with red text according to comments and with appropriate line numbering.

-----------

Review with replay:

 

This contribution presents original ideas in the study and advances the previous research in this area. The paper deserves publication. The level of the originality of contribution to the existing knowledge with an emphasis on the paper’s innovativeness in both theory development and methodology used in the study is very high.

This work makes a significant practical contribution and it makes impact on the research work on the research community.

The quality of arguments, the critical analysis of concepts, theories and findings, and consistency and coherency of debate are well addressed in this paper. 

The paper has a good writing style in term of accuracy, clarity, readability, organization, and formatting. 

1. In figure 1, the proposed architecture of Faster R-CNN deep learning network for defect detection is proposed. Please address the problem of ist Training more in depth and discuss more in depth the optimization Problem which is defined for this Task including ist non convexity.

We expand the “Data collection” in which we describe the most important part with appropriate explanations and references. Here, we gave more data about dataset; how we made; why we made such a dataset; what was our intention…; How we come to parameters of used neural network that are presented in the paper, also we put parameters for different classes that are recognized etc;  (3. Data collection). Important parts are given with appropriate references. (lines 178-223)

The paper is a good one and deserves publication

Author Response File: Author Response.docx

Reviewer 3 Report

The article is devoted to developing a machine vision system for the process of checking the detection of pitting and monitoring its progression. The study's relevance is confirmed by the fact that gears are essential elements of machines subjected to heavy loads. In some cases, gearboxes are critical to driving machines that need to run almost every day for more extended periods. Tooth surface damage is typical in working gears, and pitting is the most common damage. For the regular operation of gears, it is necessary to regularly determine the occurrence and extent of the damaged tooth surface caused by pitting. The authors propose a machine vision system to check the detection of pitting and monitor its progression. The implemented monitoring system uses a faster R-CNN network to identify and position pitting on a specific tooth, allowing for monitoring.

Despite the satisfactory quality of the article, some shortcomings need to be corrected.

1. The abstract should include obtained results.
2. The aim of the research should be defined.
3. The argumentation of the proposed neural network architecture should be included.
4. The known approaches should be separated from the model proposed by the authors.
5. It should be described why the authors used a learning rate of 0,005 and trained the neural network for 12000 iterations.
6. It should be justified why the authors use 90% of the dataset for training and 10% for testing.
7. The scientific and practical novelty of the study should be highlighted.
8. English should be proofread.

In summarizing my comments, I recommend that the manuscript is accepted after major revision. 

Author Response

Dear Madam/Sir,

The authors would like to thank you for your effort and valuable suggestions for improving the paper.

The responses to your review are given with red text according to comments and with appropriate line numbering. All the changes are in track mode.

-----------

Review with replays:

 

The article is devoted to developing a machine vision system for the process of checking the detection of pitting and monitoring its progression. The study's relevance is confirmed by the fact that gears are essential elements of machines subjected to heavy loads. In some cases, gearboxes are critical to driving machines that need to run almost every day for more extended periods. Tooth surface damage is typical in working gears, and pitting is the most common damage. For the regular operation of gears, it is necessary to regularly determine the occurrence and extent of the damaged tooth surface caused by pitting. The authors propose a machine vision system to check the detection of pitting and monitor its progression. The implemented monitoring system uses a faster R-CNN network to identify and position pitting on a specific tooth, allowing for monitoring.

Despite the satisfactory quality of the article, some shortcomings need to be corrected.

1. The abstract should include obtained results.

We add the essence of obtained results according to reviewers comment. (lines 20-22)

 

2. The aim of the research should be defined.

Thank you for this comment. We add separate part in Introduction that is dealing with aim and contribution of the paper and research idea. (lines 111-122)

 

3. The argumentation of the proposed neural network architecture should be included.

We add explanation for this in the first chapter of „Description of the method”: (lines 128-131)

 

4. The known approaches should be separated from the model proposed by the authors.

We wrote about this in the introduction but since it is not clear we will expand this part. For detection of pitting, there are three known approaches: 1) analysis of vibrations (lines 39-61); 2) acoustic analysis (lines 39-61); 3) machine vision (lines 62-82). We wrote about all this in Introduction and gave the references that we could find.

 

5. It should be described why the authors used a learning rate of 0,005 and trained the neural network for 12000 iterations.

We add explanations with new appropriate references. (lines 187-212)

 

6. It should be justified why the authors use 90% of the dataset for training and 10% for testing.

We add explanation with new appropriate references. (lines 213-223)

 

7. The scientific and practical novelty of the study should be highlighted.

We will add appropriate text in Introduction according to reviewer’s comment. (lines 111-121)

 

8. English should be proofread.

We gave the paper after we add all changes according to reviewers comments to professional proofreading. All the changes are in track mode.

In summarizing my comments, I recommend that the manuscript is accepted after major revision. 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Compared with the previous edition, it has been improved a lot. The idea and overall readability of this paper is good.

Some statements need to be more concise, for instance, the section 3 “Data collection” is a bit lengthy. The author needs to carefully check the grammar and format before publishing.

Reviewer 3 Report

Thanks for the authors for considering my comments and recommendations. In my opinion, now the paper can be accepted.