Shape Optimization with a Flattening-Based Morphing Method
Round 1
Reviewer 1 Report
This manuscript presents a flattening-based morphing method and its application in shape optimization problems. The presented method seems interesting. However, it is demonstrated in a single, rather simple, geometry. Since the authors claim that their method overcomes the disadvantages of other methods, to my opinion, they should demonstrate the capabilities of the method in a set of different cases/examples where other methods fail. For this reason, I recommend a major revision of the article before being accepted for publication. My comments are summarized below:
1) Introduction (page 1, lines 32-33). The authors state that a CAD-based parameterization is so simple. This conclusion is extracted for 2-3 references where only a few variables of the CAD are used for the design. To my opinion, this is not correct; it is not a general conclusion since a CAD-based parameterization can be very complicated as well.
2) Introduction (pages 1-2, lines 40-41). Regarding NURBS-based parameterization the authors state that “… generating the global and radical shape changes from the baseline design is difficult to accomplish.” Why is that? Please clarify.
3) Section 2, page 3. The definition of anchor points is not well explained. Is it an automated or, at least, semi-automated procedure or a manual one? What about a more complex, for instance with winglets, geometry?
4) How the proposed method handles surfaces that are not allowed to vary (such as the pylon and/or the nacelle) in the optimization?
5) It is not clear, at least from figure 3, whether the mapping to the parametric space includes the whole geometry or different mappings for the lower and upper side of the BWB are required.
6) Page 6, lines 156-160. How the one-to-one correspondence is ensured after the remeshing? Please clarify.
7) The presented method requires the existence of a few (at least two of three) baseline models. Do you think that in more complicated geometry one may possess, or generate, more than one baseline models? Usually, we have a single, quite complicated, baseline model. Could you comment on that?
8) For the shape optimization (with only two design variables) a DoE with 19 samples was carried out. Is there a specific reason for selecting this number of samples? I assume that the distance between the design points is a user-defined value. Any other value could have been selected. Is this selection related to the computational cost of the simulation?
9) How the constraint 0<= ω_1 + ω_2 <=1 is handled? Does the algorithm select ω_1 and then adapts the bounds of ω_2? Please explain
10) What is the cost of selecting the ‘optimal’ ANN vis-à-vis to the simulation cost? Is it negligible or not?
Minor, typo errors:
Page 3, lines 93-94: the terms vortex and vortices appear. They should be changed to vertex and vertices, respectively.
Page 3, line 119: e_{ij} of M or of U?
· Page 4, line 125: the boundary of M or of U?
· Page 4, line 136: “To explain about this”. Please rephrase.
· Page 5, caption of Figure 3: topologicall -> topologically
· Page 10, line 284: “constraints…is satisfied” -> are satisfied.
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Review report:
Shape optimization with a flattening-based morphing method
This work investigates a new design method, called flattening-based morphing, which can easily generate various shaped designs for shape optimization problems.
Moreover, the work provide a interesting example of a complete optimization loop including parametrization, investigation of design space, surrogate generation and optimization. In this regards, the work is interesting and worth publishing.
However, I recommend to include some clarifications and further work to improve the quality of the publication. Also, I recommend a detailed review of the English, some sentences are difficult to understand.
1) I don’t understand Lines 2-3. It appears to be a copy&paste from the introduction with no clear relation with previous sentences in the abstract.
2) More or less the same for lines 3-4
3) Lines 69. For not expert, can you explain what is a star-shaped configurations and why they are used in multi-resolution morphing?
4) The flattening-based morphing seems very well suited for simple geometries. How do you plan to apply this method to more complex geometries?
5) Can this method be applied to quadrilateral meshes?
6) How do you plan to manage the junctions where the parametrizations are different and the relation between points is not bijective?
7) Can you explain in more detail the role of the anchor points? How are they selected? Which is the sensitivity of the method to those points?
Regarding the ANN:
8) Provide additional details of the accuracy of the surrogate model. No data of the test cases are provided. (line 249).
9) I assume full factorial design has been used for the distribution of the total of 19 design points of experiment step. But this is not mentioned.
10) Can you provide computational cost of the simulation and surrogate model generation?
11) It will be interesting to show the performance of the methods in a simple wing fuselage configuration, where to different bodies with connection parts are present.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Review of a manuscript titled “Shape optimization with a flattening-based morphing method” by Kim and Oh
The authors have presented a new morphing method that would effectively give morphing for non-star-shaped objects. The work is very interesting and presented well. Certain aspects in relation to CFD analysis must be presented in the manuscript. The following aspects must explain in detail for a manuscript to be published.
-
1. How are baseline models chosen? It would be beneficial to the research community to share the CAD of baseline models as part of appendix data or details of geometry as given in Ref 43. Without this, the work becomes secluded without being able to work on a similar line in future.
-
2. The authors have cited work related to blended-wing-body configuration in general like
Peifeng, L. I., et al. "Aerodynamic design methodology for blended wing body transport." Chinese Journal of Aeronautics 25.4 (2012): 508-516.
Sun, C., Song, B., & Wang, P. (2015). Parametric geometric model and shape optimization of an underwater glider with blended-wing-body. International Journal of Naval Architecture and Ocean Engineering, 7(6), 995-1006.
Panagiotou, P., & Yakinthos, K. (2017). Parametric aerodynamic study of Blended-Wing-Body platforms at low subsonic speeds for UAV applications. In 35th AIAA Applied Aerodynamics Conference (p. 3737).
-
3. One of the major aspects missing in the numerical section is validation with experimental data. It is essential to take care of this aspect to have confidence on the used turbulence model
-
4. In table 1, it would be good to specify the first grid length used for layers.
-
5. Which software is used for CFD simulations?
-
6. What are computational settings like convergence criterion, numerical schemes used etc?
-
7. 19 design points in the design of experiments should be stated as part of the appendix.
-
8. Fit in figure 9 is actually scatter a lot. The representation of the regression model is not that good.
-
9. What were polynomials generated in hidden layers? What model is used in ANN feedback?
-
10. It would essential for the authors to compare their optimum BWB with existing literature and comment.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The authors have correctly addressed the recommendations and suggestions of my review and certainly the current version of the paper is improved. I recommend the publication in its present form.
Reviewer 3 Report
The authors have responded to all queries well.
