Solar Cell Cracks and Finger Failure Detection Using Statistical Parameters of Electroluminescence Images and Machine Learning

Round 1

Reviewer 1 Report

This paper presented a methodology to use electroluminescence images for solar cell cracks and finger failure detection, where statistical parameters were extracted from the histogram of these images, and three standard machine learning techniques, SVM, RF and kNN, were investigated. In general, the paper can generate impact, and it is easy to read and follow. However, there are some issues, in terms of structure, literature review and formatting, which should be addressed. (1) The whole section of methodology description should be restructured, according to the procedure as described in Figure 3. (2) In lines 192-193, explain what are these our proposed statistical feature vector (V1) and the commonly used combined vector (V2). Table A1 should also be included in the main report around here. (3) In the Introduction, it was mentioned the state-of-the-art techniques, such as CNN, but it was not compared with the standard classifiers that were used in the paper (SVM, RF and kNN). This should be discussed in the comparison/discussion section, together with other state-of-the-art methods, such as RNN, autoencoder and GAN, etc., see ref, Deep recurrent entropy adaptive model for system reliability monitoring, IEEE TII; Visually interpretable profile extraction with an autoencoder for health monitoring of industrial systems, ICARM. (4) A lot of references regarding Python libraries and machine learning online studies are unnecessary, e.g. 34, 36, 38, 40, 41, 43, 44, 45. Instead, please use the original references, or papers who were using same techniques to solve engineering problems, e.g. Acoustic-based engine fault diagnosis using WPT, PCA and Bayesian optimization, Applied Sciences. (5) Fonts in Figure 1 could be made clearer. Check the fonts and format of Lines 125-144. Tables 1, 2 and 3 could be improved – using the actual table. Check line 178 – it should fit in main texts. Check format of Figure 6.

Author Response

Please see the attachment. We hope that we have adequately addressed all the comments. However, if more information is needed, we will be happy to provide it. We thank the reviewer again for the comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors present a method for the detection and classification of errors in photovoltaic cells based on statistical processing of electroluminescence images and machine learning. The work is interesting and well presented so I think it should be accepted for publication. However, the authors should undertake some corrections previously:

• Revise and unify the format of the text throughout the document (for example, the font size changes from some passages to others)
• Try to enlarge the size of the texts included in the figures so that they can be read correctly (especially in figures 1, 3, 4 and 7)
• Define all the variables included in Table 1 of Appendix A
• Review the citations and correctly apply the referencing standards, unifying the criteria. For example: i) in some references the name of the authors is placed before the surname and in others the opposite, ii) sometimes only the initial of the name of the authors is placed and in others the full name, iii) the sign = after "doi", "isbn", "vol", ... should be deleted or replaced by : ; iv) when quoting web pages, include date of last access, etc.
• The title of table 3 and the table should be consecutive on the same page
• The title of table 1 in Appendix A and the table must be consecutive on the same page
• The title of figure 6 should follow the figure, not preceding it
• Table 1 in Appendix A should be named differently so as not to be mistaken for table 1 on page 5

Author Response

Please see the attachment. We hope that we have adequately addressed all the comments. However, if more information is needed, we will be happy to provide it. We thank the reviewer again for the comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed most of the comments. However, the presentation/formatting of the revised version has become worse, e.g. Table A8 should be an actual table; Fig. 1B could also be improved. Furthermore, other sate-of-the-art neural networks should be mentioned and discussed however briefly nonetheless in the literature review.

Author Response

Please see the attachment.

We hope that we have adequately addressed all the comments. However, if more information is needed, we will be happy to provide it.

We thank you again for the comments.

Author Response File: Author Response.docx