Uncovering flame physics with machine learning: application to the reaction rate in hydrogen flames

Abstract

Convolutional Neural Networks (CNNs) are used as analytical tools to investigate the relationship between the progress variable field and the local chemical source term in lean premixed hydrogen flames. Rather than employing machine learning for modelling, CNNs are leveraged to analyse the physical information contained in spatially resolved fields from direct numerical simulations. CNNs are particularly well suited for this task because they exploit spatial correlations and multi-scale structures, allowing an assessment of how spatial organisation influences the source term. The analysis demonstrates that the progress variable based on water, C_H₂O, contains all the information required to accurately parametrise the chemical source term whereas C_H₂ alone does not. The inclusion of the mixture fraction Z improves the accuracy of the latter but provides no significant additional information to the CNN when the spatial field of C_H2O is used as input. The same behaviour is observed in both a laminar thermodiffusively unstable flame and a turbulent slot-jet hydrogen flame, indicating that C_H₂O is a robust single variable for parametrisation. A complementary scale analysis in the laminar case shows that spatial features extending over at least two laminar flame thicknesses are required to reconstruct the source term accurately, thereby identifying the characteristic scale in the progress variable field that carries this information. These results demonstrate how machine learning can uncover physical dependencies that remain hidden to classical statistical analyses.

A physics-informed LLM agent for uncovering typhoon- and earthquake-induced turbine failures and offline yaw-pitch countermeasures
Linjun Fu, Zhipeng Xie, Shaohai Zhang +3
iScience29 Oct 2026Peer-reviewedCC BY 4.0General physicsCell PressCrossref
Physics-Informed Machine Learning for Response Tensors in Crystalline Magnets: Application to Kagome Metals
Yu-Hsuan Lin, Joel E. Moore
arXiv15 Oct 2026PreprintarXiv licenceFull textCondensed matterarXiv
Phase-field modeling of TiO2 nanocarving via reaction with hydrogen-bearing gas
Alireza Seifi, Sheikh Akbar, Yanzhou Ji
arXiv21 Aug 2026PreprintCC BY 4.0Full textCondensed matterarXiv
Ab-initio Investigation on h-Be3N2 Monolayer for Photocatalytic Hydrogen Evolution Reaction and Oxygen Evolution Reaction
Karan Patel, Hetvi Jadav, Himanshu Pandey
arXiv7 Oct 2026PreprintCC BY-NC-SA 4.0Full textCondensed matterarXiv
Transition Metal-Anchored Newly Synthesized Aza-Triphenylene Monolayer for Efficient Hydrogen Evolution Reaction: Insights from DFT
Soumendra Kumar Das, Jagannath Pradhan, Prasanjit Samal +2
arXiv25 Oct 2026PreprintCC BY 4.0Full textCondensed matterarXiv

Cite

BibTeX RIS

BibTeX

@misc{attili2026uncovering,
  title = {{Uncovering flame physics with machine learning: application to the reaction rate in hydrogen flames}},
  author = {Antonio Attili and Ludovico Nista and Tommaso Baffetti and Geveen Arumapperuma and Sofiane Al Kassar and Lukas Berger and Christoph D. K. Schumann and Temistocle Grenga and Heinz Pitsch},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.32552},
  eprint = {2609.32552},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.32552},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Uncovering flame physics with machine learning: application to the reaction rate in hydrogen flames
AU  - Antonio Attili
AU  - Ludovico Nista
AU  - Tommaso Baffetti
AU  - Geveen Arumapperuma
AU  - Sofiane Al Kassar
AU  - Lukas Berger
AU  - Christoph D. K. Schumann
AU  - Temistocle Grenga
AU  - Heinz Pitsch
PY  - 2026
DA  - 2026/09/26
PB  - arXiv
DO  - 10.48550/arxiv.2609.32552
UR  - https://arxiv.org/abs/2609.32552
AB  - Convolutional Neural Networks (CNNs) are used as analytical tools to investigate the relationship between the progress variable field and the local chemical source term in lean premixed hydrogen flames. Rather than employing machine learning for modelling, CNNs are leveraged to analyse the physical information contained in spatially resolved fields from direct numerical simulations. CNNs are particularly well suited for this task because they exploit spatial correlations and multi-scale structures, allowing an assessment of how spatial organisation influences the source term. The analysis demonstrates that the progress variable based on water, C_H_2O, contains all the information required to accurately parametrise the chemical source term whereas C_H_2 alone does not. The inclusion of the mixture fraction Z improves the accuracy of the latter but provides no significant additional information to the CNN when the spatial field of C_H2O is used as input. The same behaviour is observed in both a laminar thermodiffusively unstable flame and a turbulent slot-jet hydrogen flame, indicating that C_H_2O is a robust single variable for parametrisation. A complementary scale analysis in the laminar case shows that spatial features extending over at least two laminar flame thicknesses are required to reconstruct the source term accurately, thereby identifying the characteristic scale in the progress variable field that carries this information. These results demonstrate how machine learning can uncover physical dependencies that remain hidden to classical statistical analyses.
ER  -

Abstract

Related

Cite