Self-Attention and Multi-Task Based Model for Remaining Useful Life Prediction with Missing Values
Round 1
Reviewer 1 Report
A very well-structured manuscript focused on Remaining useful life (RUL) prediction missing values. Congrats to the authors.
State of the art is relating the nowadays researches from the interest area, also the references are the majority of the past 10 years which reveals the up-to-date research.
Is not clear how was obtained the characteristics from fig 3…, must do a reinterpretation of the characteristic.
Table 2 and Table 3 do not clearly explain how was determinate the Improvement?!, just by reducing the influence of the spatiotemporal feature?
Row 336, listed some other methods used for comparison, this is a good aspect of the paper which provides more credibility to the result, but the acronyms, ” SVR, MLP, DLSTM, DCNN, and AGCNN”, where not even fully present what meaning they have, those were just linked to some references, this cannot be handy on reviewers side. Please add at least a short explanation of each acronym in completing the reference you link. The acronym Missing Rates ( MR) is first fully listed on page 10, in the description of table 4, but this appears in the article very often before. Please add the meaning of the acronym the first time it is listed.
As much as I concern, the simulation is not the same as the experiment…. So please reformulate chapter 3. Maybe can be considered more as a testing of the proposed method by inputting some simulated data.
The conclusions must be considering some effective improvements brought by the method proposed….
Overall, the work is based on a very well-done expert interpretation of the data resulted.
The proposed method can be useful for future researchers in order to improve the self-attention and multi-task-based model for remaining useful life predictions for missing values.
The paper is interesting and can be published in the journal with some major changes, regarding article presentation and structure form according to the journal template.
Regards to authors,
Reviewer
Author Response
Thanks for the constructive comments. Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
1. A revision of the English language (grammar and spelling) is suggested.
2. It is suggested to observe the self-citations following with the recommendations of COPE (Committee on Publication Ethics).
3. Citations are missing from essential parts of this article, e.g., equations, etc.
4. It is recommended to use benchmark publications in the bibliography whenever possible, e.g., CNN, ANN, LSTM, deep learning, etc. “principal component analysis”, in recognition of their respective inventors and, also, as a demonstration of erudition in relation to the subjects covered in this article.
5. Checking for correct citation in the text, e.g., the Paris-Erdogan Model is covered in reference [6]?, etc.
6. It is suggested to use soft colors in the illustrations.
7. What are the advantages and disadvantages that should be highlighted?
8. Thus, I suggest to the Authors highlight (objectively) the innovation of this proposal about literature.
Author Response
Thanks for the constructive comments. Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
1) The paper is written well, however, the originality and novelty are weak as all the formulas are well-known in the literature. Please clarify.
2) The texts in Figure 2 are not consistent with the conventional reading habit, which should rotate 180 degree.
3) In the formula (4), how does the constant alpha affect your prediction performance?
Author Response
Thanks for the constructive comments. Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have replied to all the queries and concerns with proper references. However, to improve the connectivity for readers, authors can develop their research in future articles by use of laboratory experiments.
Good luck with future research!
Reviewer 2 Report
---
