Szczegóły publikacji
Opis bibliograficzny
Application of physics-informed neural networks for coupled transport-reaction problem in solid oxide fuel cell anode / Shinichi Maeda, Masashi Kishimoto, Ren Matsukawa, Szymon BUCHANIEC, Yuting Guo, Hiroshi Iwai // International Journal of Heat and Mass Transfer ; ISSN 0017-9310 . — 2026 — vol. 264 art. no. 128723, s. 1–12. — Bibliogr. s. 11–12, Abstr. — Publikacja dostępna online od: 2026-04-02. — Sz. Buchaniec - dod. afiliacja: Kyoto University, Japan
Autorzy (6)
- Maeda Shinichi
- Kishimoto Masashi
- Matsukawa Ren
- AGHBuchaniec Szymon
- Guo Yuting
- Iwai Hiroshi
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 168172 |
|---|---|
| Data dodania do BaDAP | 2026-06-29 |
| Tekst źródłowy | URL |
| DOI | 10.1016/j.ijheatmasstransfer.2026.128723 |
| Rok publikacji | 2026 |
| Typ publikacji | artykuł w czasopiśmie |
| Otwarty dostęp |  |
| Czasopismo/seria | International Journal of Heat and Mass Transfer |
Abstract
Numerical prediction of the electrochemical performance of solid oxide fuel cells (SOFCs) requires solving complex partial differential equations (PDEs) that describe coupled multi-physics transport phenomena and electrochemical reactions. Conventional numerical methods are computationally intensive when comprehensively accounting for all factors influencing electrode performance, such as operating conditions and electrode microstructures. In this study, we apply physics-informed neural networks (PINNs) to simulate SOFC anodes, aiming to enable exhaustive performance prediction through a single training process. As a fundamental investigation into the applicability of PINNs for SOFC analysis, we focus on balancing the loss components that constitute the loss function and introduce two algorithms to improve the training process: the automatic satisfaction of boundary conditions (ASBC) and the dynamic weight strategy (DWS). The ASBC formulation is found to enhance prediction accuracy, while the DWS accelerates the training process. Additionally, a PINN model is constructed with anode overpotential, a typical operating parameter of SOFCs, as an input, enabling the prediction of anode performance under various overpotential conditions. After hyperparameter tuning using Bayesian optimization, the trained PINN successfully predicts overpotential characteristics. These results demonstrate the potential of PINNs for comprehensive performance prediction of SOFCs.
