About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Energy Technology 2026: Advancement in Energy Materials - Theory, Simulation, Characterization, Application
|
| Presentation Title |
Explainable Inverse Design of Battery Materials via Multi-Model Learning and Conditional Filtering |
| Author(s) |
Tzu-Wei Wang, Gerfried Millner, Eason Yi-Sheng Chen |
| On-Site Speaker (Planned) |
Tzu-Wei Wang |
| Abstract Scope |
Lithium-ion batteries are critical in modern technologies, including electric vehicles, portable electronics, and grid storage. However, their long-term performance and safety are closely linked to intrinsic material features such as chemical composition, crystal structure, and bonding characteristics that govern key properties like capacity and voltage. While machine learning has demonstrated strong performance in predicting battery-related properties, most models still operate as black boxes: they achieve high accuracy but offer limited insight into the underlying physical mechanisms. To address this limitation, we combine explainable AI with multi-model learning to identify the key compositional and structural features that govern battery material performance. These high-impact features are then translated into quantifiable constraints that define a refined design space. Finally, a conditional generator proposes candidate compositions that meet these criteria, forming a closed-loop inverse framework that connects interpretable feature analysis with data-driven material generation for novel battery development. This project was supported by Nanyang Technological University under the URECA Undergraduate Research Programme. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Energy Conversion and Storage, Modeling and Simulation |