A Systematic Mapping of the Advancing Use of Machine Learning Techniques for Predictive Maintenance in the Manufacturing Sector

Round 1

Reviewer 1 Report

I recommend this paper for publication.

Author Response

Dear Reviewer,

thanks You so much for Your positive response to Our proposal. I hope the paper  could be of support to the state of art analysis. We prepared a revised version of Our proposal -  You can find (If you prefer)  the new part and modified part  in revised form

Reviewer 2 Report

This paper provided an extensive review of recent advances in predictive maintenance with a systematic mapping. A few comments need to be addressed and revised before dressed before accepted.

1. There are some typos in the manuscript, e.g., there is an extra ']' in line 121.
2. The size of some figures needs to be adjusted, e.g. part of Figure 5 is missing.
3. Please indicate how data extraction in Section 3 is conducted. Do the authors use some natural language processing techniques to extract and organize key information?
4. Figure 3 is a very good illustration to present the structure of data-driven models, but the lines connecting block seems a little bit confusing. Please explain what are these lines used for, and reorganize them to make the relationship clear. Besides, it would be best if the authors could also compare the advantages and disadvantages of different approaches and their application in predictive maintenance.
5. According to Figure 15, major single domain machine learning/deep learning approaches have been utilized for predictive maintenance. I am wondering does domain adaptation/transfer learning techniques been widely used in predictive maintenance to mitigate the difference in the training set and test set? For instance, these techniques have been widely used in condition-based maintenance (e.g., A new intelligent fault identification method based on transfer locality preserving projection for actual diagnosis scenario of rotating machinery; Unsupervised Cross-domain Fault Diagnosis Using Feature Representation Alignment Networks for Rotating Machinery). I believe transfer learning could help to further improve the performance of predictive maintenance. And if applicable, I think it would be best if this topic could be discussed in a survey paper.
6. It would be best if the authors could present an overview of the mentioned papers with discussions of their strengths and weaknesses and classify them into different categories. For example, like Table 5 in "A survey on deep learning in medical image analysis". 

Author Response

Dear Reviewer,

thanks You so much for Your positive response to Our proposal. I hope the paper  could be of support to the state of art analysis. We prepared a revised version of Our proposal -  You can find (If you prefer)  the new part and modified part  in revised form. In the following text there are your question/suggestion and our reply 

 

This paper provided an extensive review of recent advances in predictive maintenance with a systematic mapping. A few comments need to be addressed and revised before dressed before accepted.

1. There are some typos in the manuscript, e.g., there is an extra ']' in line 121.

Thank You so much for Your careful note. We revised all the proposal in order to check and delete any typos error.

2. The size of some figures needs to be adjusted, e.g. part of Figure 5 is missing.

Sorry , It was related to the AS template and To a not careful revision of the margins. We provided correct

3. Please indicate how data extraction in Section 3 is conducted. Do the authors use some natural language processing techniques to extract and organize key information?

We included notes inside the paper. The  information about paper were extracted from  a self  designed  MatLab code ( for those information related to paper headings), form data mining approach using Latent Dirichlet allocation and from  proposal reading

4. Figure 3 is a very good illustration to present the structure of data-driven models, but the lines connecting block seems a little bit confusing. Please explain what are these lines used for, and reorganize them to make the relationship clear. Besides, it would be best if the authors could also compare the advantages and disadvantages of different approaches and their application in predictive maintenance.

Thank You so much. We included some notes about color and approach description in the revised copy . In the literature review ( as  per the references papers)  until the class of learning approach we can identify from supervised to semi- supervised approach ( grey color).  These proposals can implement NNS based methodology   with some  probability based  test ( for features extraction).  You should generally implement  a mixture of  data-driven models (blue connections). We Included some notes  caption of  Figures 3. We included notes in text

5. According to Figure 15, major single domain machine learning/deep learning approaches have been utilized for predictive maintenance. I am wondering does domain adaptation/transfer learning techniques been widely used in predictive maintenance to mitigate the difference in the training set and test set? For instance, these techniques have been widely used in condition-based maintenance (e.g., A new intelligent fault identification method based on transfer locality preserving projection for actual diagnosis scenario of rotating machinery; Unsupervised Cross-domain Fault Diagnosis Using Feature Representation Alignment Networks for Rotating Machinery). I believe transfer learning could help to further improve the performance of predictive maintenance. And if applicable, I think it would be best if this topic could be discussed in a survey paper.

 

Yes, thank you so much for your notes and for this interesting reading. Deep Learning is the evolution of machine learning. It technically  continue analyze data based on NNS while structuring algorithm in layers. We included  notes, in the proposal (chapter 1)  for condition based maintenance

6. It would be best if the authors could present an overview of the mentioned papers with discussions of their strengths and weaknesses and classify them into different categories. For example, like Table 5 in "A survey on deep learning in medical image analysis". 

 

We included a table with the selected references and their clusterisation. We specified geographic provenience, type of research and type of contribution. In the first version of the proposal it was not included  for the sake of length

Reviewer 3 Report

Please see attached.

Comments for author File: Comments.pdf

Author Response

Dear REVIEWER, thanks for your notes.. We included in the attached document the precise reply to Your suggestions . 

Author Response File: Author Response.pdf

Reviewer 4 Report

A systematic review model for machine learning for predictive maintenance in manufacturing system was studied in this manuscript. Though authors try to formulate error-free model, at their level best, however, still some major issues are there regarding writing, novelty, formatting etc. Thus, a major revision is required before considered further. My comments and suggestions are as follows:

1. The exact novelty, or necessity is missing in the abstract section, thus it is suggested to rewrite whole abstract with proper novelty of your research.
2. How sustainable manufacturing is performed? not mention anywhere. Thus, rewrite the introduction section with proper illustration of smart manufacturing and sustainable manufacturing.
3. Authors are suggested to add all the related research in the manuscript, as a review model was formulated by the authors.
4. Introduction and Literature review section are failed to provide exact novelty, finding. Thus, authors are advised to recreate those section very carefully.
5. Authors are suggested to provide more illustration about the difference of Predictive maintenance and Preventive maintenance.
6. Some picture is exceed the page limit and some citations are not in proper format. Correct all those typos. Moreover, some sentences are incomplete. Thus, authors are advised to revise their manuscript very carefully.
7. There are so many smart techniques (Autonomation, IoT, AI etc.) already used in those days in manufacturing system to make a smart manufacturing system. Thus, authors are advised to cover all those things and related literature in this study, which make the manuscript more reader friendly.
8. Conclusion should be rewritten with proper scientific way.
9. References should be proper journal format.
10. Overall English correction is needed.

Author Response

Dear Reviewer, 

In the attached file You can find the punctual  reply to Your comments. We did all the best in order to respect Your opinion and suggestion on the paper. you can track the revised version of the paper   in "revised" form

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

Thank you, authors done a good revision but still, some minor concerns are there. Thus, a minor revision still required.

1. Please remove all "I", "We", "Our" throughout the paper and rewrite the sentance.
2. Check the whole manuscript very carefully for typos.