Berlin 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 53: Data Driven Material Science: Big Data and Workflows VI

MM 53.4: Vortrag

Donnerstag, 21. März 2024, 11:00–11:15, C 243

Datadriven thermodynamic modeling with CALPHAD — Tobias Spitaler and •Lorenz Romaner — Montanuniversität Leoben, Department Werkstoffwissenschaft, Leoben, Österreich

CALPHAD models and computational thermodynamics play an essential role in materials science and in the development of novel materials. In the CALPHAD method, thermodynamic quantities and phase diagrams are calculated from a parameterized model of the Gibbs free energy, which is stored in a thermodynamic database. The reliability of the calculated quantities relies on the correctness and quality of the thermodynamic database. With new computational tools and statistical methods, the database creation can be accelerated and uncertainty can be quantified, which is propagated from the input data to the quantities of interest.

We combine heterogeneous data from experiment and simulation in CALPHAD modeling and use statistical tools to propagate the uncertainty of the model parameters to quantities of interest. We demonstrate parameter optimization and uncertainty quantification in the phase diagram of selected systems (e.g. W-Ti, Fe-C). Uncertainty quantification of phase boundaries, invariant points and other quantities of interest are demonstrated. With the statistical methods regions with high uncertainty in the composition space can be identified and the potential experiments with the highest information proposed. With the data-driven and statistical approach to CALPHAD modeling, new thermodynamic databases can be obtained in a faster and more reproducible way.

Keywords: CALPHAD; Phase diagram; Thermodynamic modelling; Metals