A high-throughput physics- and data-driven framework for High-Entropy Alloy development

Linking the relationships between composition, thermo-fluid properties, solidification dynamics, and the resulting microstructure is of great importance in increasing the adoption of High Entropy Alloys (HEAs) into practical applications. However, predicting these characteristics presents significant challenges due to the complex, multi-component nature of HEAs resulting in a large design space with a vast range of thermo-fluid properties. To -address this challenge, a multi-physics simulation framework and a data-driven surrogate modeling approach are implemented, to provide a method to rapidly assess how changes in HEA composition will lead to changes in material properties and as-solidified, cast microstructure. The physics-based framework sequentially couples CALPHAD models for material properties, computational fluid dynamics (CFD) for thermal conditions during solidification, and Cellular Automata (CA) for microstructure. A data-driven surrogate model, which can reconstruct microstructure statistics from the thermal conditions, is trained via a reduced-dimensional form of the microstructure output of the physics-based framework. Dimensionality reduction is achieved through a statistical representation of microstructure via Angularly Resolved Chord Length Distributions (ARCLDs), and identification of the primary ARCLD modes via Principal Component Analysis (PCA). This surrogate modeling approach provides a significant speedup for predicting as-cast microstructure features from HEA composition and solidification conditions, and is highly adaptable to other solidification processes.