| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Artificial Intelligence Applications in Integrated Computational Materials Engineering
|
| Presentation Title |
Harnessing Graph Neural Networks for Classification of Unique Glassy Structures in CuZr Metallic Glasses |
| Author(s) |
Emily Gurniak, Suyue Yuan, Xuezhen Ren, Paulo S. Branicio |
| On-Site Speaker (Planned) |
Emily Gurniak |
| Abstract Scope |
Machine Learning techniques have emerged as important tools for studying chemistry and materials science. Here, we harness Graph Neural Networks (GNNs) to characterize the structures of unique states of CuZr metallic glasses (MGs). We use molecular dynamics to simulate the vitrification of CuZr from the liquid, employing quenching rates from 10^9 to 10^15 K/s to produce six unique MG states. We create a dataset containing 10,800 samples of 686 atoms, equally divided among the six states. We then train and evaluate the classification performance of GNNs, including Graph Attention Network (GAT), Graph Sample and AggreGatE (GraphSAGE), Graph Isomorphism Network (GIN), and Relational Graph Convolutional Network (RGCN). The GAT and GraphSAGE achieve the best performance, with an overall accuracy of 81%. These results demonstrate that GNNs can detect subtle differences in the structure of MGs, highlighting their potential application to other ordered and disordered materials. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Modeling and Simulation |