About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
AI-Enabled Materials Processing: Integrating Accelerated Experimental Workflows and Processing-Aware Machine Learning
|
| Presentation Title |
Closed-loop machine learning guided exploration of processing pathways reveals optimal microstructures |
| Author(s) |
Sreenivas Raguraman, Maitreyee Sharma Priyadarshini, Andrew Kim, Tunde Ayodeji, Veronica Ivanovskaya, Beril Ulugun, Camryn Byrum, Adam Griebel, Paulette Clancy, Timothy Weihs |
| On-Site Speaker (Planned) |
Sreenivas Raguraman |
| Abstract Scope |
AI-enabled materials design often emphasizes composition- or microstructure-based descriptors, but practical materials development requires processing pathways that can be experimentally implemented and can reliably produce targeted combinations of properties. Here, we demonstrate a closed-loop, processing-aware machine-learning workflow to optimize thermomechanical pathways in a fixed-composition ZX10 magnesium alloy. Processing descriptors spanning strain, heat-treatment temperature, duration, and sequence were used as model inputs, while hardness and corrosion rate served as rapid experimental feedback signals. Across three experimental rounds, the workflow evolved from a broad property-ratio search to a multi-objective, feasibility-informed strategy that identified implementable pathways with improved hardness and reduced corrosion. Post-optimization characterization shows that the best-performing pathways converged toward a common microstructural state consisting of ultrafine grains, comparable defect densities, similar texture, and finely distributed precipitates, even without microstructural input. These findings show that processing-aware machine learning can guide processing toward favorable microstructural states without requiring microstructural inputs during optimization. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Process Technology, Biomaterials |