About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Thermodynamics and Kinetics of Alloys
|
Presentation Title |
Symbolic Regression Search for a General Free Energy Function for BCC Alloys |
Author(s) |
Luke Wirth, Snigdhansu Chatterjee, Alexander Urban, Dallas Trinkle |
On-Site Speaker (Planned) |
Luke Wirth |
Abstract Scope |
Prediction of phase diagrams at new compositions enables discovery of novel materials for engineering applications. Identification of general thermodynamic functions for the free energies of crystal structures at finite temperatures would provide powerful tools for making these predictions. We search for such a function to describe BCC alloys based on physical features of their constituent elements. To prepare our model, we collect functions from thermodynamic database (TDB) files that calculate the free energy of the BCC phase, at given conditions, relative to stable reference phases at standard conditions for CALPHAD models. Data science tools, such as the sure independence screening and sparsifying operator (SISSO) compressed-sensing method, then connect coefficients that act on functions of temperature and composition to elemental features. The resultant models offer a physically meaningful way to make predictions about the stability of alloy phases in regions of composition space that have not been studied experimentally. |
Proceedings Inclusion? |
Planned: |
Keywords |
Modeling and Simulation, Machine Learning, |