About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
Machine Learning–Supported High Throughput Crystal Plasticity Simulations of Stress Concentrations in Void-Containing Molybdenum Microstructures |
| Author(s) |
Talukder Musfika Tasnim Oishi, Jason Mayeur, Patxi Fernandez-Zelaia, Christopher Ledford, Michael Kirka, Marko Knezevic |
| On-Site Speaker (Planned) |
Talukder Musfika Tasnim Oishi |
| Abstract Scope |
Stress concentrations at microstructural features like voids cause failure of metallic alloys. In this paper, we develop a high-throughput computational framework to explore how local morphology and crystallographic orientations surroundings govern local stress amplifications in voided microstructures of polycrystalline molybdenum. Using DREAM.3D and DAMASK, we generated and simulated 1,200 columnar-grained 3D microstructures with [100], [111], and mixed orientations containing a centered void under uniaxial tension. Voxel-level stress data were processed to compute stress concentration factors (SCFs), building a rich dataset that links crystallography and void geometry to local mechanical response. This workflow not only enabled scalable screening of defect-sensitive regions but also lays the foundation for training machine learning models that predict SCF directly from microstructure. By bridging physical modeling with data-driven insights, this approach supports the design of more robust, damage-tolerant alloys and opens new pathways for microstructure-informed material optimization. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, ICME |