Details
| Original language | English |
|---|---|
| Article number | 107825 |
| Journal | NEURAL NETWORKS |
| Volume | 191 |
| Early online date | 11 Jul 2025 |
| Publication status | Published - Nov 2025 |
Abstract
This paper proposes a Curriculum-Transfer-Learning based physics-informed neural network (CTL-PINN) for long-term simulation of physical and mechanical behaviors. The main innovation of CTL-PINN lies in decomposing long-term problems into a sequence of short-term sub-problems. Initially, the standard PINN is employed to solve the first sub-problem. As the simulation progresses, subsequent time-domain problems are addressed using a curriculum learning approach that integrates information from previous steps. Furthermore, transfer learning techniques are incorporated, allowing the model to effectively utilize prior training data and solve sequential time-domain transfer problems. CTL-PINN combines the strengths of curriculum learning and transfer learning, overcoming the limitations of standard PINNs, such as local optimization issues, and addressing the inaccuracies over extended time domains encountered in CL-PINN and the low computational efficiency of TL-PINN. The efficacy and robustness of CTL-PINN are demonstrated through applications to nonlinear wave propagation, Kirchhoff plate dynamic response, and the hydrodynamic model of the Three Gorges Reservoir Area, showcasing its superior capability in addressing long-term computational challenges.
Keywords
- Curriculum Learning, Long-term Simulation, Physics-Informed Neural Network, Three Gorges Reservoir Area, Transfer Learning
ASJC Scopus subject areas
- Neuroscience(all)
- Cognitive Neuroscience
- Computer Science(all)
- Artificial Intelligence
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: NEURAL NETWORKS, Vol. 191, 107825, 11.2025.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Long-term simulation of physical and mechanical behaviors using curriculum-transfer-learning based physics-informed neural networks
AU - Guo, Yuan
AU - Fu, Zhuojia
AU - Min, Jian
AU - Lin, Shiyu
AU - Liu, Xiaoting
AU - Rashed, Youssef F.
AU - Zhuang, Xiaoying
N1 - Publisher Copyright: © 2025 Elsevier Ltd
PY - 2025/11
Y1 - 2025/11
N2 - This paper proposes a Curriculum-Transfer-Learning based physics-informed neural network (CTL-PINN) for long-term simulation of physical and mechanical behaviors. The main innovation of CTL-PINN lies in decomposing long-term problems into a sequence of short-term sub-problems. Initially, the standard PINN is employed to solve the first sub-problem. As the simulation progresses, subsequent time-domain problems are addressed using a curriculum learning approach that integrates information from previous steps. Furthermore, transfer learning techniques are incorporated, allowing the model to effectively utilize prior training data and solve sequential time-domain transfer problems. CTL-PINN combines the strengths of curriculum learning and transfer learning, overcoming the limitations of standard PINNs, such as local optimization issues, and addressing the inaccuracies over extended time domains encountered in CL-PINN and the low computational efficiency of TL-PINN. The efficacy and robustness of CTL-PINN are demonstrated through applications to nonlinear wave propagation, Kirchhoff plate dynamic response, and the hydrodynamic model of the Three Gorges Reservoir Area, showcasing its superior capability in addressing long-term computational challenges.
AB - This paper proposes a Curriculum-Transfer-Learning based physics-informed neural network (CTL-PINN) for long-term simulation of physical and mechanical behaviors. The main innovation of CTL-PINN lies in decomposing long-term problems into a sequence of short-term sub-problems. Initially, the standard PINN is employed to solve the first sub-problem. As the simulation progresses, subsequent time-domain problems are addressed using a curriculum learning approach that integrates information from previous steps. Furthermore, transfer learning techniques are incorporated, allowing the model to effectively utilize prior training data and solve sequential time-domain transfer problems. CTL-PINN combines the strengths of curriculum learning and transfer learning, overcoming the limitations of standard PINNs, such as local optimization issues, and addressing the inaccuracies over extended time domains encountered in CL-PINN and the low computational efficiency of TL-PINN. The efficacy and robustness of CTL-PINN are demonstrated through applications to nonlinear wave propagation, Kirchhoff plate dynamic response, and the hydrodynamic model of the Three Gorges Reservoir Area, showcasing its superior capability in addressing long-term computational challenges.
KW - Curriculum Learning
KW - Long-term Simulation
KW - Physics-Informed Neural Network
KW - Three Gorges Reservoir Area
KW - Transfer Learning
UR - http://www.scopus.com/inward/record.url?scp=105010846268&partnerID=8YFLogxK
U2 - 10.1016/j.neunet.2025.107825
DO - 10.1016/j.neunet.2025.107825
M3 - Article
AN - SCOPUS:105010846268
VL - 191
JO - NEURAL NETWORKS
JF - NEURAL NETWORKS
SN - 0893-6080
M1 - 107825
ER -