Multidisciplinary Optimization and Analysis of Stratospheric Airships Powered by Solar Arrays

Round 1

Reviewer 1 Report

Review of manuscript aerospace-1942371, titled "Multidisciplinary Optimization and Analysis of Stratospheric Airships Powered by Solar Arrays".

 

Overall Comments:

• Very interesting concept and thorough presentation of the theory and solution implementation. 
• We'd like to see more context on what's the current state of the art in this optimization exercise, and what it is the new proposition with this paper?
• In section 3, the authors present their proposed approach for the optimization exercise. How/why did they formulate this method? Where is the literature search to benchmark against and identify gaps in current practices, which improvements upon them will help justify the need for the proposed approach? 
• Just like your provided a flow chart of the optimization approach, an overall process schematic on the authors' research plan is needed. Within the context of the optimization, what research questions are being identified and how are these linked to the sensitivity analysis in the Results section? Please help the reader navigate and understand why your findings are key
• Fix broken reference/citations fields in the text

Author Response

Thank you for your valuable suggestions, according to your suggestions, we have revised the paper.

1、We'd like to see more context on what's the current state of the art in this optimization exercise, and what it is the new proposition with this paper?

Response:

According to the reviewer’s suggestion, we add the content “Ceruti et al. [22] demonstrated a shape optimization framework of hybrid airship based on added mass. Meng [23] proposed a multidisciplinary design optimization method of a lift-type hybrid airship. On the basis of building subsystem model, a Concurrent Subsystem Optimization algorithm based on Response Surface ( CSSO-RS) with the self-adaptive ability is put forward．To achieve a continuous flight and reduce the total mass and energy cost, Zhang [24] proposed a multi-phase strategy including the climb, daytime cruise, glide and nighttime cruise. However, they only considered the case in which solar array is symmetrically located on the upper surface of airship envelope.” in the third paragraph of section 1. And we also add the content “This simplification makes it difficult to obtain the solar cell layout with high power generation efficiency.” and “The proposed methodology can optimize the airships configurations by considering the envelope shape, the solar array layouts, etc., using detailed subsystem models. “ in paragraph 6 and 7 in section 1.

 

2、In section 3, the authors present their proposed approach for the optimization exercise. How/why did they formulate this method? Where is the literature search to benchmark against and identify gaps in current practices, which improvements upon them will help justify the need for the proposed approach?

Response: The proposed approach is formulated to improve the model more accurately, considering the factors such as the circumferential location of solar array, as discussed in section 1, based on the subsystem model and the airship design flowchart. For the choice of optimization method, as discussed in paragraph 2 in section 3, the hybrid method is used because it combines non-gradient based algorithm with gradient based algorithm, which can obtain the optimum solution efficiently and reliably.

 

3、Just like your provided a flow chart of the optimization approach, an overall process schematic on the authors' research plan is needed. Within the context of the optimization, what research questions are being identified and how are these linked to the sensitivity analysis in the Results section? Please help the reader navigate and understand why your findings are key

Response: According to the reviewer’s suggestion, a schematic research plan has been added in section 1 (Fig. 1). As shown in section 3.3, the research question is to minimize the total weight of airship powered by the solar array layouts subject to the power balance constraint and to the buoyancy constraint. For sensitivity analysis，as introduced in the Results section, to investigate the effects of these parameters on the optimal results, optimizations are conduct to obtain the corresponding optimal configurations of airships, including the solar array layout.

 

4、Fix broken reference/citations fields in the text

Response: The reference/citations fields in the text has been fixed.

Reviewer 2 Report

Comments for author File: Comments.pdf

Author Response

Dear reviewer, 

   Thanks for your valuable suggestions marked in the original manuscript, according to your notes, we have checked and revised the paper, which are marked in color in the revised manuscript.

1、References without [].

Response: the errors have been revised.

2、The English grammar errors.

Response: the errors have been revised, which are marked in red in the revised manuscript.

3、Some sentences are not clearly described

Response: the sentences have been rewritten more clearly.

4、some figures are marked in error.

Response: all figures have been checked and revised.

5、In conclusion section, the word “improvement” is confusing.

Response: the word “improvement” has been revised as “reduction”

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper is generally well written. The topic is interested and the performed study is indeed of interest. Specifically, it is interesting how you have included the effect of the operative conditions, such as latitude, or heading in the design, and you have documented it.

The MDO aspects and the optimization, since they are named in the title, deserve more space. E.g. how many iterations have been allowed to the algorithm? when it started to converge? was it difficult to satisfy the constrain? and so on

On the editing side, many of the references are either missing or not reported correctly. Also some of the images are missing (like the 10). A lot of editing and crosschecking of the figures is required.

The title of the paper includes MDO. In the paper there itself there are little further reference to that.

How the disciplines have been linked is not described. The optimization architecture is not reported either.

The only overview of the process is figure 5. If the objective is also to showcase the value of the MDO approach, at least that chart should be enlarged and enriched, along with a better description.

The introduction properly set up a framework for their case reporting the available literature and the limitations in those studies.

Introduction:

It clearly states the objective of the work.

reference 7 without []

Same for 13

The sentence: "However .. may increase" is very convoluted.

Missing references (a lot)

Other references with bracklets (will not report more of those)

Theory

This section presents a lot of formulas. Many of them are relatively simple but do their job in explaining clearly what has been done in the following sections.

Considering this, although they increase the total length of the paper, it is acceptable to state them explicitly instead of just reference the source.

However, formula's font and general readability could be improved (possibly related to the paper being in draft state).

This also affects the symbols used in the text with lines' separation behaving irregularly.

The impact of the temperature has not been included in the solar panel output evaluation. 

Are they supposed to operate at a nominal temperature? 

Does the day-night/shadowing of the panels due to relative orientation of the Sun have any effect?

Please clarify.

Also the impact, if any, of the stabilizers is not mentioned. If their design is such that they never cast a shadow over the solar panel, please state that.

 

Optimization Method

A quick mention to the concept of global and local minima could support the explanation about the different behavior of gradient and non-gradient.

A sentence to the effect of Hybrid algorithms try to combine the overall domain exploration properties of the non-gradient with the convergence of the gradient might help the non-expert reader. 

Is there a reason for the MIGA-NLPQL combination to be the algorithm of choice?

Figure 5 has "global optimal" or "stop criterion" as condition to exit the loop. The first is typically unknown so I don't see its value in this contest ,the latter would be the primary exit condition as it can be used both as limit on the number of experiments or the verify the convergence.

 

Table 1 set up can be improved. do variables have also a nominal value? like those that are later referenced to as "baseline" and the unit of measurements are not visible stated as they are.

Table 2, same comment. Since we are talking about a long list of symbols, please add a column with the full name of the variables.

Results and discussion

As the paper is about optimization, aside from the final results would be interesting to visualize the evolution of the solution over the length of the optimization runs.

A parallel coordinate plot would support the reader in this.

Also a note about how the results have evolved when the GA was leading compared to the NPQL would support the clain about the need for an hybrid algorithm.

Some figures that reports the results are missing.

 

4.3 for which latitude has been this study performed? section 4.4 has a sentece about the operative conditions, could be ised here as well.

Figure number 1314?

Several figures have wrong identification.

Author Response

Dear reviewer, 

   Thanks for your valuable suggestions，according to your notes, we have checked and revised the paper, which are marked in color in the revised manuscript.

1、The MDO aspects and the optimization, since they are named in the title, deserve more space. E.g. how many iterations have been allowed to the algorithm? when it started to converge? was it difficult to satisfy the constrain? and so on

Response: In the MIGA optimization, the sub-population size, number of islands, number of generations are all set to be 10, respectively, while in the NLPQL optimization, the number of max iterations is set to be 150. The above content has been added in the first paragraph of section 4.1. Besides, a figure which shows the comparison of optimization convergence history using hybrid algorithm, MIGA and NLPQL has been added.

2、On the editing side, many of the references are either missing or not reported correctly. Also some of the images are missing (like the 10). A lot of editing and crosschecking of the figures is required.

Response: According to the reviewer’s suggestion, the references and figures are checked and edited.

3、The title of the paper includes MDO. In the paper there itself there are little further reference to that. How the disciplines have been linked is not described. The optimization architecture is not reported either. The only overview of the process is figure 5. If the objective is also to showcase the value of the MDO approach, at least that chart should be enlarged and enriched, along with a better description.

Response: In paragraph 3 in section 1, reference about the MDO has been introduced, and some other reference has been added. The figure 6 in revised manuscript has been enlarged and an illustration of the optimization steps has been added in section 3.1.

Introduction:

4、reference 7 without [], Same for 13

Response: The errors have been revised.

5、The sentence: "However .. may increase" is very convoluted.

Response: The sentence has been revised to be ‘However, a design using solar arrays with higher area and mass might lead to a considerable decrease in the entire airship system mass, even though the mass fraction of energy system mass to total airship mass may increase.’

6、Missing references (a lot), Other references with bracklets (will not report more of those)

Response: We checked the whole paper and revised all these errors.

Theory

7、Formula's font and general readability could be improved (possibly related to the paper being in draft state). This also affects the symbols used in the text with lines' separation behaving irregularly.

Response: The formula's font has been revised and the general readability has been improved, and the symbols used in the text with lines' separation have also been revised.

8、The impact of the temperature has not been included in the solar panel output evaluation. Are they supposed to operate at a nominal temperature? 

Response: The solar panel are supposed to operate at a nominal temperature; hence the impact of the temperature has not been included in this paper. We add this illustration in section 2.2.3.

9、Does the day-night/shadowing of the panels due to relative orientation of the Sun have any effect?

Please clarify.

Response: The day-night/shadowing of the panels due to relative orientation of the Sun have effects on the power received by element, it can be seen from Equations 25, 24,12,11,9, and Fig.4

10、Also the impact, if any, of the stabilizers is not mentioned. If their design is such that they never cast a shadow over the solar panel, please state that.

Response: “The stabilizers are located at the tail of airship, and their design is such that they never cast a shadow over the solar panel.”  Above is added in the first paragraph in section 2.2.3.

Optimization Method

11、A quick mention to the concept of global and local minima could support the explanation about the different behavior of gradient and non-gradient.

Response: Global optimum is the optimum in the whole design space, while the local optimum is the optimum in certain local space. global optimum must be the local optimum, while local optimum is not necessary to be the global optimum. In the origin manuscript, we are very sorry to misrepresentation. In the revised manuscript, we revised the corresponding content as follows:  

“The non-gradient based algorithms can obtain the global optimum (optimum in the whole design space) without calculating the local gradient and strict requirement of the start point. However, they are generally inefficient. Gradient based algorithms are efficient but easy to obtain the local optimum (optimum in certain local space), and highly relies on the start searching points to obtain the global optimum. “

12、A sentence to the effect of Hybrid algorithms try to combine the overall domain exploration properties of the non-gradient with the convergence of the gradient might help the non-expert reader. 

Response: We revised the sentence in paragraph 1 of section 3.1 as “The hybrid algorithms, which combine the good overall domain exploration properties of the non-gradient based algorithm with the fast convergence properties of the gradient based algorithm, can generally obtain the optimum solution efficiently and reliably. “

13、Is there a reason for the MIGA-NLPQL combination to be the algorithm of choice?

Response: As illustrated in the paper, MIGA is developed based on the GA, it divides the individuals into a number of “islands”, and thus more easily to obtain the global optimum than GA. They can obtain the global optimum without calculating the local gradient and strict requirement of the start point. However, they are generally inefficient. NLPQL is a Gradient based algorithms, which is efficient but easy to obtain the local optimum, and highly relies on the start searching points to obtain the global optimum. Therefor, we use MIGA to obtain the initial starting point, and then use NLPQL to obtain the optimal solution.

14、Figure 5 has "global optimal" or "stop criterion" as condition to exit the loop. The first is typically unknown so I don't see its value in this contest ,the latter would be the primary exit condition as it can be used both as limit on the number of experiments or the verify the convergence.

Response: According to reviewer’s comment, we revised it in figure 6 of revised paper, only use “stop criterion”.

15、Table 1 set up can be improved. do variables have also a nominal value? like those that are later referenced to as "baseline" and the unit of measurements are not visible stated as they are.

Response: Table 1 has been improved, the variables have been described like those that are later referenced to as “baseline”.

16、Table 2, same comment. Since we are talking about a long list of symbols, please add a column with the full name of the variables.

Response: Table 2 has been improved, a column with the full name of the variables has been added.

Results and discussion

17、As the paper is about optimization, aside from the final results would be interesting to visualize the evolution of the solution over the length of the optimization runs. A parallel coordinate plot would support the reader in this. Also a note about how the results have evolved when the GA was leading compared to the NPQL would support the clain about the need for an hybrid algorithm.

Response: Figure 6 which shows the comparison of convergence history for hybrid algorithm, MIGA and NLPQL has been added in section 4.1, to clearly show the need for an hybrid algorithm.

18、Some figures that reports the results are missing.

Response: Error have been revised.

 19、4.3 for which latitude has been this study performed? section 4.4 has a sentence about the operative conditions, could be missed here as well.

Response: The latitude is 18°N. It has been added in section 4.3 and section 4.4.

20、Figure number 1314?

Response: The error has been revised.

21、Several figures have wrong identification.

Response: Errors have been revised.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Yes, it's adequately revised. 

Author Response

Dear reviewer,

        Thank you for your comments.