Finding Closure Models to Match the Time Evolution of Coarse Grained 2D Turbulence Flows Using Machine Learning

Round 1

Reviewer 1 Report

This manuscript used machine learning to obtain closure terms for coarse grained model of two-dimensional turbulent flow. These closure terms were related to the average flow employing a neural network based on simple structure and smoothed slightly. This manuscript can be improved considering the following comments:

1.-Introduction should include the main advantages of the proposed work in comparison with other works reported in the literature. In addition, introduction should add the main scientific contribution of the proposed work.

2.-Authors should improve the description of the different parameters used in all the proposed equations. The section of method should include more information of the assumptions used in the proposed equations.

3.-Authors should consider more discussion of the behavior of the main results.

4.-Which are the main limitations or challenges of the proposed method?

5.-What is the future research work?

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Please see the attached pdf file.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

1) The Abstract is too general and only descriptive. In the Abstract the Authors should add some of the most important results obtained in this research (its exact values), which cannot be found in other literature. Such addition will highlight the novelty of the presented paper already in the Abstract – at the moment from the Abstract cannot be seen any novel elements which are obtained in this paper. Therefore, the Abstract requires re-arrangement and addition of the most important obtained results.

2) Figures 1, 5 and 9 are missing markings mentioned in the title of each figure on each exact figure – (a), (b) and/or (c).

3) Call on each Figure in the paper text should be as close as possible to the exact Figure. At the moment, it is quite confusing to read this paper, because, for example – call on Figure 1 is placed after Figure 3 (Figure 1 is placed at the beginning of page 3, while call on Figure 1 is placed at the end of page 4). Many Figures are not inside proper sections or subsections, which additionally makes confusion (for example, I don’t see a reason why Figure 12 is placed after Conclusions section).

Such paper arrangement cause that the paper is extremely hard to read, it is essential to continuously turn back through the paper to find proper explanation, see a proper Figure, etc.

4) In the paper should be added a Nomenclature inside which will be listed and explained all abbreviations, symbols and markings used throughout the paper text. As many abbreviations, symbols and markings are not explained at all, adding a Nomenclature is required.

5) Line 114 – what it means: “rotating” the inputs? This “rotation” should be explained in detail, because from the presented explanation it is not clear what “rotation” exactly means.

6) There are missing any exact information about the used dataset. The used dataset should be presented and explained in detail, with all standard and relevant dataset parameters.

7) The cross-validation test results are completely missing – they should be added in the paper (only the statement that high correlation coefficient is achieved, without any evidences, is surely not sufficient).

8) The paper is only the theoretical one. In the paper is completely missing any validation or any connection with real turbulence flows. Therefore, a practical value of this paper is non-existing.

9) Scientific novelty and contribution of this paper to the specific research field are highly debatable and questionable – I don’t see any exact and important novelty or contribution.

10) As the Abstract, the Conclusions section should also be improved with the most important obtained results (its exact values). Also the Conclusions seem to be too descriptive and general, without any details obtained in the presented analysis.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The manuscript introduced an approach of using NN to provide a closure term for the coarse turbulence flow simulation. The manuscript is well organized, and the results are clear. Here are several minor concern on the paper.

1: Figure 5 shows the comparison of the vorticity for the DNS and LES with NN at t=11.36. Because the training data is from 2.37 to 27.27, the results shown in (t=11.36) Figure 5 still using the training data set. Is it possible to show the vorticity comparison for t=45.46?

2: It would be better to discuss how general are the trained NN. Is it possible used in other geometry? 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Authors have improved their manuscript considering the comments of reviewer.

Reviewer 2 Report

The authors have satisfactorily addressed  my concerns. This manuscript can be accepted in the present form.

Reviewer 3 Report

The Authors have performed the majority of the proposed corrections/additions. The paper can be published in a presented form.