Physics-informed neural networks for solving steady-state heat conduction inverse problems
Advances in Differential Equations and Control Processes · 05.03.2026
DOI: 10.59400/adecp3584 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advances in Differential Equations and Control Processes, online erschienen am 05.03.2026. Verfasst von Tangwei Liu und 5 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This paper investigates numerical methods for a class of three-dimensional steady-state heat conduction inverse problems. By employing Physics-Informed Neural Networks (PINNs), the three-dimensional heat conduction inverse problems are reformulated as optimization problems with respect to a properly defined loss function. Two cases with different additional conditions are considered: one incorporates an additional boundary temperature gradient condition, while the other involves an additional partial internal temperature measurement. Corresponding efficient algorithms are developed to solve the resulting optimization problems. To optimize the performance of the proposed numerical framework, systematic sensitivity analyses are performed to rigorously justify the selection of key hyperparameters (e.g., activation functions and network architecture). Additionally, to validate the mathematical effectiveness and noise robustness of the algorithm, this study primarily employs synthetic data with controlled noise levels for quantitative evaluation. Numerical results demonstrate that the proposed method can efficiently and accurately approximate the solutions to the three-dimensional (3D) heat conduction inverse problems for both polynomial and non-polynomial cases. In addition, a theoretical analysis is provided to interpret the method's stability against noise amplification. Future work will focus on applying the proposed framework to real-world field data to further validate its practical engineering value.
Zitieren
APA
Liu, T., Zou, Q., Ruan, X., Yan, D., Kafle, J., & Lan, Z. (2026). Physics-informed neural networks for solving steady-state heat conduction inverse problems. Advances in Differential Equations and Control Processes. https://doi.org/10.59400/adecp3584
Vancouver
Liu T, Zou Q, Ruan X, Yan D, Kafle J, Lan Z. Physics-informed neural networks for solving steady-state heat conduction inverse problems. Advances in Differential Equations and Control Processes. 2026. doi:10.59400/adecp3584
BibTeX
@article{Liu2026,
author = {Liu, Tangwei and Zou, Qiong and Ruan, Xiaoqing and Yan, Dingding and Kafle, Jeevan and Lan, Zhongzhou},
title = {Physics-informed neural networks for solving steady-state heat conduction inverse problems},
journal = {Advances in Differential Equations and Control Processes},
year = {2026},
doi = {10.59400/adecp3584}
}
EUVIMED