An Efficient Siamese Network and Transfer Learning-Based Predictive Maintenance System for More Sustainable Manufacturing

Round 1

Reviewer 1 Report

The article proposes an FM for the generation of a PdM algorithm allowing the reuse of machines for greater sustainability. The authors aim to predict a failure maintenance time. Machine integrity was measured using a Siamese network thus obtaining reasonable results with the proposed FM provided by simulations and statistical analyses. The proposed FM allows for sustainable manufacturing models.

 

In general, the article is well structured and the introduction presents a clear outline of the problem.

 

The FM is well established and the simulations provided good statistical output to validate the conceptual model.

 

Results and discussion are in the same section, but it is acceptable for this case.

 

Issues for improvement:

 

- State a hypothesis or two during the introduction.

 

- Figure 1. Improve using arrows to describe FM flow. Finally, merge it with Figure 2.

 

- Provide, in the conclusions, statements in accordance with the previous hypotheses, in particular those of statistical production.

- Conlusion should be section 4.

 

Well done.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Explain the criteria used for window size selection used in the windowing approach.

Literature related to feature extraction for different data types needs to be included.

Line 247: The second is to cluster and label the reduced features via the k-means clustering method. Justify?  As k-means clustering is unsupervised algorithm. so explain the criteria for the selection of no clusters. 

 

Line 329: Why only one loss function (cross-entropy) is used? Justify its use.

Line 379: The classification performance of the proposed method has been verified on four different data sets. but the results section, the performance matrices related to classification are not evaluated.

How you have decided no of epochs for the training process? ref fig 7. justify?

Line 395: without training data thanks to the siamese network.  use technical language (Formal)

The data set description and relevance to the proposed work are not properly explained.

 

Title is too long. Try to make it short. 

Comments for author File: Comments.pdf

Moderate English grammar and spell check is required, along with sentence structure corrections 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The article contains statistical methods that have been well selected.

The structure of the article is acceptable. It would be worth considering whether, when describing the problem in question, you should add or modify the structure of the article in order to separate the subchapter "review of the literature". This is a note for the authors to consider. The more so, in many places the authors refer to numerous publications. Therefore, perhaps it is worth isolating a sub-chapter devoted to a review of the literature?

The final sub-Chapter "Conclusions" should be enriched with limitations and possibilities of further research. The authors do mention this. However, it is worth writing more about it in the context of the usefulness of further in-depth research and its limitations.

The subject matter of the article is interesting and from a practical point of view it can be an interesting contribution to the development of research.

The authors cite numerous publications.

The methods used have been synthetically characterized in terms of usefulness and weaknesses.

Correct language. Only in a few sentences should the linguistic correctness of the sentences be checked to make them more understandable for readers.

Author Response

Please see the attachment

Author Response File: Author Response.pdf