| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Artificial Intelligence Applications in Integrated Computational Materials Engineering
|
| Presentation Title |
Data Assimilation of Multi-Phase-Field Model based on Physically Informed Neural Network |
| Author(s) |
Chang Liu, Meng Zhang, Junya Inoue, Ryoya Tsuruoka, Satoshi Noguchi |
| On-Site Speaker (Planned) |
Chang Liu |
| Abstract Scope |
The Phase-Field (PF) model is a powerful tool for simulating physical phenomena such as grain growth and phase transformation. However, the accuracy of simulations heavily relies on precise model parameters like interfacial energy and boundary mobility. Data assimilation techniques have been explored to efficiently identify these internal parameters of PF models. However, traditional finite difference (FD) methods encounter limitations in stability and computational cost, particularly for high-resolution and long-duration simulations. This work presents a novel approach to estimating the internal parameters of the Multi-Phase Field (MPF) model using Physics-Informed Neural Networks (PINNs). PINNs integrate physical laws into the neural network's training process, leveraging automatic differentiation to bypass the stability constraints of FD methods. This approach demonstrably achieves stable simulations with larger scales and significantly reduced computational times. Additionally, the model facilitates accurate analysis of internal parameters, paving the way for a deeper understanding of microstructural dynamics. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Modeling and Simulation, Machine Learning, |