About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Chemistry and Physics of Interfaces
|
| Presentation Title |
A machine learning informed continuum damage model to simulate void nucleation and growth in metal microstructures during shock loading. |
| Author(s) |
Abhijith Thoopul Anantharanga, Brandon Runnels |
| On-Site Speaker (Planned) |
Abhijith Thoopul Anantharanga |
| Abstract Scope |
Structural materials in defense and aerospace applications experience high-impact dynamic loads that drive spallation through void nucleation, growth, and coalescence. Traditional damage models fail to capture the microstructural dependencies of nucleation, producing unphysical void nucleation under shock loading. This work presents a unified machine learning-continuum damage framework for predicting and simulating void nucleation and growth in polycrystalline metals. An attention-enhanced U-Net, trained on grain boundary energy and crystallographic orientation, predicts void nucleation probability fields along grain boundaries. Intermediate feature maps identify which grains and boundaries most influence nucleation, offering mechanistic insight into damage initiation. These probability fields inform a continuum damage model coupled with crystal plasticity, nucleating voids at physically motivated locations and evolving them through growth and coalescence mechanisms under shock loading. The coupled framework reproduces experimentally observed spall behavior, providing a novel approach to predicting void nucleation and simulating void growth under shock loading. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Other, Other |