Generative Adversarial Networks in Combustion: Flame Image Generation for Clean and Predictive Combustion Modeling Using Real Data
Date
2025
Authors
Singh, C.
Chinnici, M.
Kor, A.L.
Evans, M.J.
Medwell, P.R.
Chinnici, A.
Editors
Zimmermann, A.
Schmidt, R.
Jain, L.C.
Howlett, R.J.
Schmidt, R.
Jain, L.C.
Howlett, R.J.
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Book chapter
Citation
Event/exhibition information: KES-HCIS 2024: International KES Conference on Human Centred Intelligent Systems 2024, Madeira, Portugal, 19/06/2024-21/06/2024
Source details - Title: Human Centred Intelligent Systems: Proceedings of KES-HCIS 2024 Conference, 2025 / Zimmermann, A., Schmidt, R., Jain, L.C., Howlett, R.J. (ed./s), vol.414, Ch.10, pp.99-110
Statement of Responsibility
Conference Name
Abstract
This study explores the application of Generative Adversarial Networks (GANs) in combustion science, utilizing a flame image dataset. By comparing the produced images with the original dataset, we qualitatively analyze the discriminator and generator loss to assess the performance of the GAN. The results show improvements in the discriminator’s ability to distinguish between real and generated images, as well as improvements in the generator’s ability to add missing details, leading to the generation of images that are more realistic. Fundamental properties of flames are well captured in the resulting images, despite the absence or distortion of minor details. The study advances the fields of AI, image processing, and combustion science by highlighting possible uses in the creation of synthetic images and data augmentation. Overall, our qualitative analysis enriches comprehension of combustion science.
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
Copyright 2025 The Author(s), under exclusive license to Springer Nature Singapore