Machine Learning-Driven Molecular Simulations for Alloy Microstructure Prediction and Control

• Applicant: Professorin Julija Zavadlav, Ph.D.
• Subject Area: Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 534045056

Alloy microstructure prediction and control are essential in advancing additive manufacturing technology. Of significant technological interest is, in particular, the laser powder bed fusion due to the high manufacturing flexibility, near-net-shape production, and efficient use of raw material. However, many materials, such as high-strength aluminum (Al) alloys, cannot be reliably printed as the involved large spatiotemporal thermal gradients can lead to solidification cracking. Recently, significant efforts have been focusing on developing novel alloy materials to control the solidification microstructure and, in turn, improve the quality of built products. Molecular dynamics (MD) simulations are ideally suited to elucidate the material/process-microstructure-property relations. They provide atomistic level insight into the solidification microstructural evolution. Nevertheless, the full potential of MD simulations is hindered by inaccurate molecular potentials leading to discrepancies between the predictions of simulations and experiments. Therefore, I aim to develop novel machine learning-based molecular models consistent with experimental data and employ them to predict the microstructure and mechanical properties of rapidly solidified Al alloys. This approach will enable me to reach the following objectives: (1) uncover the relations between the employed spatiotemporal gradients, solidification microstructure, and mechanical properties, (2) quantify the macrostructural changes due to thermal reheating cycles, and (3) determine the impact of different grain refiners and extract the optimal grain refiner chemical element and content. Overall, the project will provide guidance for the next-generation alloy material manufacturing, while the methodological advancements of data-driven molecular modeling will be of paramount importance in other areas of material engineering and beyond.

• DFG Programme: Research Grants