About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Advancing the Frontier of Powder Materials Processing and Sintering: A MPMD/EPD Symposium in Honor of Eugene Olevsky
|
| Presentation Title |
Data-driven machine learning approach for materials characterization in powder-based additive manufacturing |
| Author(s) |
John S. Kang, Nicholas Satterlee, Runjian Jiang, Elisa Torresani, Eugene Olevsky |
| On-Site Speaker (Planned) |
John S. Kang |
| Abstract Scope |
Reliable microstructural characterization is essential for understanding and optimizing the performance of metal components produced via powder-based additive manufacturing (PBAM). Although 2D cross-sectional imaging is widely used for this purpose, conventional image processing techniques remain labor-intensive and insufficient for capturing complex morphological features at scale. This study introduces an integrated machine learning (ML) framework for automated, data-driven characterization of microstructures in PBAM parts. The framework processes cross-sectional images alongside material and process parameters, including powder distribution and sintering temperature, to automatically detect and quantify critical features such as pores, grain boundaries, and particle contact areas within 50 milliseconds per image. Leveraging advanced ML algorithms alongside novel data augmentation techniques, the method delivers accurate morphological analysis even with limited training data. This approach enables deeper insight into structure–property relationships and supports the development of scalable, high-throughput tools for process monitoring, quality assurance, and material qualification in PBAM systems. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Characterization, Machine Learning |