| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Microstructural Evolution and Material Properties Due to Manufacturing Processes: A Symposium in Honor of Anthony Rollett
|
| Presentation Title |
Accelerating Solidification Pathway Control in Steel Additive Manufacturing by Leveraging Active Learning and CALPHAD Modeling |
| Author(s) |
Joseph W. Aroh, Brian DeCost, Fan Zhang |
| On-Site Speaker (Planned) |
Joseph W. Aroh |
| Abstract Scope |
Additive manufacturing (AM) introduces a novel process regime within the rapid solidification of metal alloys. Typical AM process conditions straddle the boundaries between dendritic and planar solidification, and for many steels, the boundaries between primary ferrite and primary austenite solidification. In this work, we developed a CALPHAD-coupled interface response function (IRF) pipeline to predict phase (ferrite or austenite) and morphology (dendrite or planar) of Fe-based alloys. By leveraging active learning, we employ this IRF model for target steel compositions that result in changes in interface response classification at specified thermal gradients (G) and solidification rates (R). Using Gaussian process regression, we predict the interface temperatures of the respective phases/morphologies in a multi-dimensional composition space as a function of process variables G and R, effectively constructing a multi-compositional microstructure selection map. Selected alloy compositions were examined using in-situ synchrotron diffraction under various AM process conditions to probe the intricate multi-dimensional boundary in solidification pathway classified by the Gaussian process predictions. This work is directly inspired by Tony Rollett’s prolific contributions in computational modeling, machine learning, and synchrotron characterization of microstructural evolution in AM processes. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Solidification, Machine Learning |