PNNL

Abstract

We identify compositionally complex alloys (CCAs) that offer exceptional mechanical properties for elevated temperature applications by employing machine learning to accelerate design, due to its scalability, rapidity, and reasonably accurate predictions. Machine-learning (ML) tools were implemented to predict the Young’s modulus of refractory-based CCAs by employing Gradient Boost, Ada Boost and XGBoost models. Our results, in conjunction with experimental validation, suggest that average valence electron count and difference in atomic radius are the key features dominating the Young’s modulus of CCAs and refractory-based CCAs. Moreover, XGBoost provided the best predictive capabilities among the three models. We present an approach that integrates high-quality validation data from experiments, literature data for training machine-learning models, and feature selection based on physical insights that opens a new avenue to optimize the desired materials property for different engineering applications.