Prediction of growth velocity of undercooled multicomponent metallic alloys using a machine learning approach

Establishing growth kinetics of undercooled metallic alloys is essential to understand microstructure evolution during solidification processing. The complexity of the physical processes have hitherto limited growth models to simple assumptions that do not lead to satisfactory predictive capability. Experimental measurements of growth velocity as a function of undercooling are also tedious. The current study uses experimental and literature data to train the machine learning algorithms. Five algorithms are trained: Random Forest, Bagging Regressor, Gradient Boosting Regressor, XGBoost, and Artificial Neural Network. A labeled data set of 910 was used, with 70% data for training and 30 % data for testing. An R2 cross-validation score of more than 0.89 was obtained for ANN. The trained algorithms are used to predict the growth velocity of medium and high entropy alloys show good compatibility with the experimental data.