About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Theory and Design of Metallic Glasses
|
| Presentation Title |
Simulation-Informed Machine Learning for Metallic Glass Characterization and Mechanical Response Prediction |
| Author(s) |
Michael L. Falk, Bin Xu, Zhao Wu, Jiayin Lu, Ahmed Elgailani, Wenjiang Huang, Rahul Meena, Michael D Shields, Franz Bamer, Chris Rycroft |
| On-Site Speaker (Planned) |
Michael L. Falk |
| Abstract Scope |
To enable design of additively manufactured amorphous metal parts, we develop two simulation-informed machine learning frameworks. The first utilizes a variational autoencoder to link experimental nanodiffraction data to atomistic data harvested through simulated quenching of a CuZr alloy using molecular dynamics. The autoencoder discerns structural signatures of glass thermal histories. We interpret the relevant latent dimensions and apply the autoencoder to experimental data. The second framework develops stochastic formulation of the plastic constitutive response. Through the interrogation of a 3D atomistic representative volume element, we quantify stress drops and structural changes. This informs a stochastic finite state automata that reproduces the material evolution to serve as a lower-scale constitutive model for a continuum implementation capable of achieving predictions on significantly larger length scales. Validation is undertaken in comparison with large scale molecular dyanamics simulations. This work is supported by the US NSF Grant Nos. DMR-2323718/DMR-2323719/DMR-2323720. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Characterization |