A hybrid AI-CFD framework for optimizing heat transfer of a premixed methane-air flame jet on inclined surfaces
Issued Date
2025-05-01
Resource Type
eISSN
26662027
Scopus ID
2-s2.0-105002653896
Journal Title
International Journal of Thermofluids
Volume
27
Rights Holder(s)
SCOPUS
Bibliographic Citation
International Journal of Thermofluids Vol.27 (2025)
Suggested Citation
Kamma P., Loksupapaiboon K., Phromjan J., Suvanjumrat C. A hybrid AI-CFD framework for optimizing heat transfer of a premixed methane-air flame jet on inclined surfaces. International Journal of Thermofluids Vol.27 (2025). doi:10.1016/j.ijft.2025.101206 Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/109688
Title
A hybrid AI-CFD framework for optimizing heat transfer of a premixed methane-air flame jet on inclined surfaces
Author(s)
Corresponding Author(s)
Other Contributor(s)
Abstract
This study presents a novel integration of artificial intelligence (AI) and computational fluid dynamics (CFD) simulations to investigate and optimize the heat transfer characteristics of a premixed methane-air flame jet impinging on an inclined surface. Key parameters—including the mixture equivalence ratio (ϕ = 0.8–2.0), burner-to-plate distance (H/d = 2–6), Reynolds number (Re = 400–1200), and plate inclination angle (θ = 0°–90°)—were systematically analyzed to evaluate their effects on heat flux distribution and thermal efficiency. Using OpenFOAM, the laminar flame behavior was modeled under diverse conditions, revealing strong agreement with experimental data, with average errors of 6.23 % for flame height and 6.47 % for thermal efficiency. To reduce the computational expense of these simulations, a hybrid Artificial Neural Network-Genetic Algorithm (ANN-GA) model was developed. The ANN accurately predicted thermal efficiency based on operational parameters, while the GA optimized these inputs to achieve maximum thermal efficiency of 76.9955 %, closely matching the CFD-predicted value of 70.86 % (discrepancy:6.1355 %). The ANN-GA model demonstrated a low absolute error of 7.97 %, confirming its reliability and precision. This research is the first to establish a robust AI-driven framework for optimizing flame jet heat transfer performance on inclined surfaces, offering valuable insights for improving industrial heating processes and advancing the application of AI in thermal system design.