| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Advances in Multiphysics Modeling and Multi-modal Imaging of Functional Materials
|
| Presentation Title |
Deep Operator Learning for Battery Characterization: From Materials to Systems |
| Author(s) |
Wei Li, Ruqing Fang, Junning Jiao, Georgios N. Vassilakis, Juner Zhu |
| On-Site Speaker (Planned) |
Juner Zhu |
| Abstract Scope |
Recent advances in scientific machine learning (SciML) have shed light on modeling complex engineered systems, including batteries. Battery materials are functional through a combination of multiphysics processes, including pattern formation. Simulations of real patterns in battery materials incur significant computational costs, which could be alleviated by leveraging large image datasets. We developed Phase-Field DeepONet, a physics-informed operator neural network framework that predicts the dynamic responses of systems governed by gradient flows of free-energy functionals. The approach is demonstrated in solving the Allen-Cahn and Cahn-Hilliard equations. We also explored a technical roadmap to scale up material-level simulations to the system level by replacing governing equations of the sub-models with pre-trained neural networks (DeepONets). A successful preliminary example will be demonstrated on Li-metal solid-state batteries. The SciML-based framework shows significantly improved computational efficiency compared to conventional approaches. |