| Zugriffsnummer | 56562 |
| Dokumenttyp | Zeitschriftenartikel |
| Peer Review | mit Peer Review |
| Sprache | Englisch |
| Titel | DNN surrogate-assisted mechanism optimization of THF/isopropanol blends: Shock tube autoignition and CO time-histories for lean premixed combustion |
| Autor(in); Institution |
Wang, Guanyu; 3.3, Physikalische Chemie, PTB-Braunschweig
|
| Quelle/Jahr | Applications in Energy and Combustion Science: 27 (2026), 1 - 15 |
| Artikelnummer | 100531 |
| ISSN | 2666-352X (online) |
| DOI | |
| Verlag | Amsterstam: Elsevier Ltd |
| Freie Schlagworte | tetrahydrofuran ; isopropanol ; e-fuel blends ; shock tube ; surrogate-assisted optimization ; artificial neural networks |
| Zusammenfassung | The aviation sector’s transition to sustainable energy requires characterizing novel e-fuel candidates such as tetrahydrofuran (THF) and isopropanol. This study investigates the autoignition properties of THF/isopropanol blends under lean conditions (φ = 0.25, 0.5 and 0.9) relevant to Lean Premixed Prevaporized (LPP) combustion. Shock tube experiments were conducted at 0.89–1.41 atm and 1280–1788 K, measuring ignition delay times (IDT) via OH* chemiluminescence and time-resolved CO histories via Tunable Diode Laser Absorption Spectroscopy (TDLAS) across isopropanol mole fractions of 0–50%. A recently developed kinetic mechanism for THF/isopropanol blends, validated against Jet Stirred Reactor data from our group, was found to exhibit discrepancies under shock tube conditions. To resolve these, a Surrogate-Assisted Genetic Algorithm (SAGA) framework was developed, coupling Deep Neural Network (DNN) surrogates with a genetic optimizer to simultaneously adjust 26 key pre-exponential factors. The DNN surrogates achieved R2 > 0.999 on independent test data and reduced end-to-end optimization time by approximately three orders of magnitude. The optimizer converged consistently across 1000 independent trials, with 88% of function values within 5% of the global best. The optimized mechanism shows excellent agreement with measured IDTs and CO profiles and is cross-validated against independent laminar flame speed data not included in the optimization targets. Applied to LPP operability, the mechanism indicates a trade-off between flashback propensity and blowout resistance, with a favourable operating point near φ = 0.65 and 0–50% THF that is largely preserved across the pressures examined. |
| Kostenfreier Zugang | Open Access Gold |
| Rechteinformation | CC BY 4.0 ; Creative Commons Attribution 4.0 License |
| Themenbereich der Metrologie | Metrologie in der Chemie und Stoffeigenschaften |
| Förderinformationen (1) |
Förderername: Deutsche Forschungsgemeinschaft (DFG)
Förderer ID: 0000 0001 2096 9829 Förderer ID Typ: ISNI Titel der Förderung: EXC 2163 – Sustainable and Energy Efficient Aviation Förderungsnummer: 390881007 (EXC 2163) URI der Förderung: https://www.tu-braunschweig.de/en/se2a https://gepris.dfg.de/gepris/projekt/390881007 |
Zitierung
Wang, G., Nadiri, S., Li, M., Agarwal, S., Fernandes, R., & Shu, B. (2026). DNN surrogate-assisted mechanism optimization of THF/isopropanol blends: Shock tube autoignition and CO time-histories for lean premixed combustion. Applications in Energy and Combustion Science, 27, 1–15. https://doi.org/10.1016/j.jaecs.2026.100531