About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Mechanics at the Extremes: Bridging Length-Scales From Nanoscale to Bulk
|
| Presentation Title |
Scaling ductility from microscale to bulk using deep learning and uncertainty quantification |
| Author(s) |
Laura Z. Vietz, Rebecca Divine, Ashley Spear |
| On-Site Speaker (Planned) |
Laura Z. Vietz |
| Abstract Scope |
Experimentally characterizing bulk materials can be challenging and cost-prohibitive, especially for materials used in harsh operating environments (e.g., materials used in extreme temperatures or under irradiation). Some experiments may be limited to small-scale testing, which includes sample sizes well below the representative volume element (RVE). Sub-RVE sizes, volumes smaller than the RVE, exhibit scattered mechanical properties due to size effects, making it challenging to characterize bulk properties. There is a need to establish a scaling relationship between sub-RVE and RVE mechanical properties. This proof-of-concept work uses a 3D convolutional neural network (CNN) to predict bulk properties from microscale test specimens. The dataset used to train the 3D CNN consists of different-sized sub-RVE synthetic microstructures and their corresponding tensile response for four FCC materials. Uncertainty quantification methods are investigated to evaluate model confidence. This work can improve small-scale testing by providing researchers with adequate material information and by reducing material waste. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Computational Materials Science & Engineering, Modeling and Simulation |