About this Abstract |
| Meeting |
2023 TMS Annual Meeting & Exhibition
|
| Symposium
|
Frontiers in Solidification: An MPMD Symposium Honoring Jonathan A. Dantzig
|
| Presentation Title |
Machine Learning Enhanced Operando Study of the Nucleation and Evolution of Complex Intermetallic Phases in Solidification |
| Author(s) |
Kang Xiang, Jiawei Mi |
| On-Site Speaker (Planned) |
Kang Xiang |
| Abstract Scope |
Microstructural control via the solidification process is the most widely used technology in processing of metallic alloys. Recent advances in synchrotron X-ray based radiography, tomography and scattering techniques have made it possible to observe in real-time the nucleation and evolution of complex metallic phases in the solidification processes. These real-time and in situ research have provided much more insight in understanding the underlying physics in phase transformation at multilength and multi-time scale. For a typical X-ray high-speed radiography and/or tomography experiment with a few tens of TB data collected. It is a challenging task to achieve both efficient and high-fidelity phase segmentation, especially difficult for the convoluted complex multiphases in 3D space.
Here, we used a machine learning-based data analysis algorithm to segment the complex and convoluted Fe-based intermetallic phases of an Al-Cu-Fe-Si recycled alloy obtained in operando condition during solidification.
Using the machine learning approach, the initial Fe-phase nucleus and the subsequent gradual branching of the phases along different crystallographic directions/planes are revealed clearly. The Chinese-script type Fe phase nucleation and growth dynamics are fully characterized and presented here for the first time. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Solidification, Aluminum, Characterization |