Home Publications Research Data Research Grants Equipment Inventions & Patents Awards Additional Research Activities Faculties & Institutions Research Areas

RF-PINN reconstruction of species, flow fields and wall temperatures in a H2/air flame based on Raman thermometry

Frohberg F, Bräuer P, Casel M, Kandel P, Will S, Bauer F, Ghani A (2026)

---

Publication Type: Journal article

Publication year: 2026

Journal

Book Volume: 42

Article Number: 106447

DOI: 10.1016/j.proci.2026.106447

Abstract

We demonstrate that Reacting Flow Physics-Informed Neural Networks (RF-PINNs) successfully reconstruct the flow fields and species fields together with wall temperatures of a laminar premixed H2/air flame from experimentally measured temperatures. Solely based on spatially resolved temperatures from pure rotational Raman (PRR) spectroscopy, the RF-PINN infers velocities, density, pressure, species mass/mole fractions, reaction rates, and heat release rates and thus enables comprehensive insights into the underlying flame structure and chemistry. To accomplish this, the RF-PINN solves the governing equations for reacting flows in cylindrical coordinates using a detailed H2 mechanism with 11 species and 23 reactions, and the heat equation in the burner nozzle. Reconstructed molar fractions from the RF-PINN show excellent agreement with experimentally derived molar fractions, demonstrating that temperature measurements provide a suitable basis for quantitative field reconstructions. Validation against conjugate heat transfer Direct Numerical Simulation (DNS) confirms the accuracy of the reconstructed flame structure, including minor species and predicted wall temperatures. Novelty and significance statement: To our knowledge, this study presents the first detailed-chemistry Reactive Flow Physics-informed Neural Network that reconstructs flow and species fields and burner nozzle temperatures using Raman temperature measurements as the sole input. Previous PINN studies relied on simplified chemistry or diffusion models, multiple diagnostic inputs, or sooting conditions. In contrast, the RF-PINN applied to a H2/air flame incorporates detailed chemical kinetics, differential diffusion, and the full governing equations, including heat conduction in the burner nozzle. Using this combination of detailed-chemistry RF-PINNs and Raman thermometry, we infer complete reactive-flow and species fields, enabling reconstruction of experimentally inaccessible quantities while reducing the need for additional diagnostics, time, and cost. Moreover, boundary conditions reconstructed by the RF-PINN framework can serve as inputs for computational fluid dynamics simulations. The established framework can provide the foundation for optimization of reaction mechanisms or transport models, be extended to alternative fuels or applied to turbulent flames.

Authors with CRIS profile

Involved external institutions

How to cite

APA:

Frohberg, F., Bräuer, P., Casel, M., Kandel, P., Will, S., Bauer, F., & Ghani, A. (2026). RF-PINN reconstruction of species, flow fields and wall temperatures in a H2/air flame based on Raman thermometry. Proceedings of the Combustion Institute, 42. https://doi.org/10.1016/j.proci.2026.106447

MLA:

Frohberg, Fabio, et al. "RF-PINN reconstruction of species, flow fields and wall temperatures in a H2/air flame based on Raman thermometry." Proceedings of the Combustion Institute 42 (2026).

BibTeX: Download