Learning-Based Super-Resolution Imaging of Turbulent Flames in Both Time and 3D Space Using Double GAN Architectures

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper is based on a double-layer GAN network architecture, achieving high-resolution 3D reconstruction of swirling flames and jet flames, and enhancing the spatiotemporal resolution of the reconstructed images. Using 3D turbulent swirling flames and jet flames data provided by LES as training data, the first layer of the paper architecture achieves spatial up-sampling of the flames. The output from the first layer serves as the input for the temporal up-sampling GAN network in the second layer, ultimately producing high-resolution, high-frame-rate 3D flame images. Below are the suggested modifications:

Major comments:

a)     One previous publication that shows very similar methodology and scenario, authors should announce their novelty. If necessary, comparison should also be added.

b)     As the authors mentioned, this architecture improves the spatiotemporal resolution of flame images. However, the results presented by the authors only demonstrate an improvement in spatial resolution, with no evidence of enhanced temporal resolution.

c)     The double-layer GAN architecture designed by the authors takes the LR images of swirling flames and jet flames as input and outputs SR images. The authors describe both types of flames in a very common manner. For the design of a single-layer GAN network, it might be better to discuss the two types of flames separately or to highlight any special preprocessing applied to the data or the network.

Minor comments:

a)     In your abstract, GAN stands for Generative Adversarial Network, not Generative Discriminative Network.

b)     In line 190, you mentioned that your HR dataset is your ground truth later in the paper. Why do you refer to it as the output here?

c)     In equation (2), your mean square error loss function is totally wrong, it should be divided by the number of voxels you have. The authors should read more textbooks.

d)     For example, in line 241, there are many instances of oddly sized symbols throughout your paper, which affect its aesthetic quality. The authors need to identify and correct these inconsistencies.

Also, GAN has been a powerful and widely used method that promotes combustion diagnostics. A more detailed review should be given. For example, 

(1)Zhang, W., Dong, X., Liu, C. et al. Generating planar distributions of soot particles from luminosity images in turbulent flames using deep learning. Appl. Phys. B 127, 18 (2021). https://doi.org/10.1007/s00340-020-07571-9   

(2) Xiaogang Cheng, Fei Ren, Zhan Gao, Luoxi Wang, Lei Zhu, Zhen Huang,

Predicting 3D distribution of soot particle from luminosity of turbulent flame based on conditional-generative adversarial networks, Combustion and Flame, Volume 247, 2023, 112489, ISSN 0010-2180, https://doi.org/10.1016/j.combustflame.2022.112489.   

(3)Anthony Carreon, Shivam Barwey, Venkat Raman, A generative adversarial network (GAN) approach to creating synthetic flame images from experimental data, Energy and AI, Volume 13, 2023, 100238, ISSN 2666-5468, https://doi.org/10.1016/j.egyai.2023.100238.

Comments on the Quality of English Language

minor editing is required

Author Response

Please refer to the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

What types of flames does the spatiotemporal super-resolution (SR) reconstruction model address?

What is the main objective of using double generative adversarial network (GAN) architectures in this study?

What type of training data is used to update the GAN model parameters?

The literature survey has to be updated with recent statistical data (see below - 

Veeraraghavan, S. M., Kaliyaperumal, G., Dillikannan, D., & De Poures, M. V. (2024). Influence of Hydrogen induction on performance and emission characteristics of an agricultural diesel engine fuelled with cultured Scenedesmus obliquus from industrial waste. Process Safety and Environmental Protection, 187, 1576–1585. https://doi.org/10.1016/j.psep.2024.05.042

 

De Poures, M. V., Dillikannan, D., Kaliyaperumal, G., Thanikodi, S., Ağbulut, Ü., Hoang, A. T., Mahmoud, Z., Shaik, S., Saleel, C. A., & Afzal, A. (2023). Collective influence and optimization of 1-hexanol, fuel injection timing, and EGR to control toxic emissions from a light-duty agricultural diesel engine fueled with diesel/waste cooking oil methyl ester blends. Process Safety and Environmental Protection, 172, 738–752. https://doi.org/10.1016/j.psep.2023.02.054

Ü. Ağbulut, M. Ubaidullah, R. Saravanan, J. Giri, and S. F. Shaikh, “Waste to fuel: A detailed combustion, performance, and emission characteristics of a CI engine fuelled with sustainable fish waste management augmentation with alcohols and nanoparticles,” Energy, vol. 299, p. 131412, Jul. 2024, doi: 10.1016/j.energy.2024.131412.

Author Response

Please refer to the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Provide a more detailed explanation of the GAN architecture, including the roles of the generator and discriminator networks. Include diagrams to illustrate the model structure and the training process. Explain the choice of training data and how it was generated.

Provide a more detailed explanation of these metrics and why they are suitable for evaluating the performance of your model. Discuss the significance of the obtained values in the context of your study.

Suggestion: Include a more detailed comparison with traditional methods, highlighting the specific advantages of your approach.

Discuss the potential generalization of your approach to other types of flames or turbulent structures.

Include a more detailed discussion on the noise immunity of your model. Provide examples or experimental results to support this claim and discuss potential applications where noise immunity is crucial.

Expand the conclusion to include a detailed discussion of potential future work. Outline specific areas for improvement and possible extensions of your research.

Author Response

请参阅附件。

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Suggested Corrections are incorporated. recommended for publication.

Reviewer 3 Report

Comments and Suggestions for Authors

All the revisions are fine