Independent Component Analysis Applied on Pulsed Thermographic Data for Carbon Fiber Reinforced Plastic Inspection: A Comparative Study

Round 1

Reviewer 1 Report

General comment:

Thanks to the authors for their effort. However, manuscript, as suggested by the title, is not a real review paper and does not reflect most of the valuable works from many research groups. If we suppose this is not a review paper, it doesn’t reflect significant experimental or theoretical achievements as well. Results presented, in my opinion, are not enough to conclude on a universal basis if the ICA methods used are really more efficient or less efficient. Therefore, there is the necessity to revise the paper in a major way and improve its quality.

Specific comments:

• The manuscript needs a thorough reading and some corrections. Some words especially in Introduction have been repeated several times.
• Please mind the use of acronyms, define for the first time and use the acronym throughout the manuscript.
• In the introduction, references are not located properly and sometimes makes it difficult to follow. Examples: page 2, first paragraph, references 9-13.
• Considering this manuscript as a review, it is preferred to cite the work of other researchers in the manuscript as well, and not just the work of your research group.
• Introduction, para 2, the definitions of ICA and PCA have been mixed, better to separate them or only define ICA (as the most relevant for this work).
• Section 2.1. para. 2: 120 Hz and 145 Hz in 30 sec, shall be 3600 and 4350 frames respectively. What is the correspondence between 2000 and 2200 frames? Why you have used these two specific frame rates?
• Section 2.2, page 3: why you have used 7 components and which components have been used to calculate SNR throughout the manuscript?
• Section 2.2, page 4, para. 2: This paragraph needs a serious review and rephrasing and a better explanation.
• Since you have used Image SNR as the merit, it would be better if you also depict the best images of each method for all or some defects. Moreover, Fig. 3, shows an image of the sample (not mentioned which method used) that shows a huge nonuniform heating pattern. Considering the method you have proposed as the merit, do you think this nonuniformity can affect your evaluation?
• Considering the data presented in Table 2, Can you explain why Infomax- and Quasi-Newton-ICA methods do not show any results (very low SNR values)? Why it only happens in the case of 145 Hz frame rate? Is the problem in the method and parameters or in the experiments?
• Looking at the data provided in Tables 1 and 2 and a simple graph of the date (of course not the way presented in Fig. 4-7), the SNR values do not show a monotonous change in depth and size, which in my opinion is related to the nonuniform heating pattern. Although PCA and ICA can somehow help removing this pattern, but still there’s no clear relationship between size, depth and SNR value. Can you better explain this?
• Referring to comment 9, the Discussion part instead of really discussing different aspects of the acquired results is more like an explanation of the data in words and not digging into the roots. This also applies to comment 8 regarding very low SNR obtained from two ICA methods.
• Figures 4 to 7 are stretched in the horizontal direction to fit into the page and are not presented properly, please change them and provide a more readable format with proper legends and scales in x and y axes.
• Section 4: there’s misleading information regarding the size of defects. In Section 2 it was mentioned that the lateral dimensions of each defect are 3, 5, 7, 10 and 15 mm, but in section 4, it is mentioned as 3, 5, 7, 10, 15 mm². Which one is correct?
• Section 5: Conclusion is very simple and weak, doesn’t literally make any conclusion except your investigation was not sufficient to evaluate and proper to compare these proposed methods. Moreover, you haven’t been able even to conclude that one of your proposed methods is better than the others.

Author Response

Dear Reviewer

We answered to your questions in the joint PDF file.

 

Best Regards

Julien FLEURET

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper is about the evaluation of the behavior of ICA applied to thermographic activity. The topic is quite specific, but interesting and the scientific approach is good.

The reviewer suggests to develop the state-of-art adding more reference and more critical comments.

Moreover, the authors should justify the choices of the techniques investigated in the paper.

Please comment figures 2 and 3 in the text.

The reviewer suggests to modify figures not clear and with deformed text (the graphs are not readable). Moreover, the text alignment of caption should be checked.

The reviewer recommends to check the paper, modify the phrases that are not clear, above all with respect to discussion and comments.

The reviewer strongly recommends to improve the readability of the figures and to add more specific comments on all of them.

Further comments are also recommended, above all about the applications of the results of the research.

Further recommendations are reported in the attached file.

Comments for author File: Comments.pdf

Author Response

Dear Reviewer

We answered to your questions in the joint PDF file.

 

Best Regards

Julien FLEURET

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper comes from the known research team, it represents a typical study where various data processing techniques are applied to experimental results obtained on a reference sample with plentiful defects with SNR being used as a figure of merit. The results are of interest for those dealing with data processing in thermal NDT. However, the paper needs major revision.

1. The main section of the paper is written better than Introduction and Conclusion, mainly because of English imperfections.
2. In the title, it is recommended to replace plural “analyses” with the singular “analysis”.
3. Perhaps, it would be appropriate to mention that any principal component technique suffers from a lack of a clear physical meaning of heat conduction in a sample under test being a sort of a mathematical exercise.
4. The considerable part of Discussion is occupied by statements on different behavior of SNR for defects of different depth and lateral size. In fact, classically, deeper defects produce lower SNR. It seems that the authors suppose that some implementations of the ICA techniques are able to stress defects at particular depths. This is disputable. I think that there might be explanation that real defects do not correspond to programmed ones. Many thermal NDT experts doubt that Teflon inserts, which are typical in ultrasonic inspection, are equally good in thermal NDT; boundary conditions between the composite and Teflon may vary thus leading to variations in temperature signals over defects. This issue should be explained clearer in the paper text. (By other words, deeper Teflon inserts may be surrounded with greater air gaps and thus produce higher signals than shallow defects).
5. The English needs editing (in the first two line of the manuscript, the word “field” is repeated 3 times; “very famous” is not good for a scientific paper, etc., see below)
6. It is wrong to say “PT consists of submitting a sample of interest to a pulse of energy” because this is a heat pulse is to be delivered to a sample under test.
7. Perhaps, “batch” can be replaced with “set”.
8. In the phrase “is our sound area”, the word “our” looks awkward.
9. Perhaps, “figure of merit” would be better than “metrics” and “values” - better than “scores”. But this is for authors’ decision.
10. Figure 4: it is better to use “defect lateral area” instead of “defect surface”.
11. There are mistakes and imperfections in the very first sentence in Conclusion: “We studied different implementation (??) of the ICA in order to evaluate their interest (??) for material inspection compaired (??) with state of the art methods sucha (??) as PCT and PPT”. All other phrases in Conclusion are either awkward or bear no useful information.

Author Response

Dear Reviewer

We answered to your questions in the joint PDF file.

 

Best Regards

Julien FLEURET

Author Response File: Author Response.pdf

Reviewer 4 Report

This manuscript conducted a comprehensive literature review on application of ICA on thermographic data for CFRP inspection. Overall, the topic of this research is interesting, and the manuscript was well organized and written. The detailed comments are provided as follows.

1. In title, please give the full name of CFRP.
2. In abstract, please provide full names of PCT and PPT when they appear for the first time.
3. Please illustrate the main innovation of this research. Why was ICA selected for processing thermographic data? Why not machine learning-based methods?
4. The main research challenges in this area should be included in this review.
5. Future research should be added in conclusion part.

Author Response

Dear Reviewer

We answered to your questions in the joint PDF file.

 

Best Regards

Julien FLEURET

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thanks to the authors for perfectly addressing all the comments and revising the manuscript extensively. This version is fine.

Reviewer 3 Report

The reviwer's comments are properly addressed. The manuscript is essentially improved. No objections for publishing.

Reviewer 4 Report

The authors well addressed the comments. I do not have further comments.