A Machine Learning-Driven Framework for Predicting Strength and High Temperature Oxidation Properties in Refractory Complex, Concentrated Alloys

Refractory complex, concentrated alloys (RCCAs) offer new avenues for designing high strength and oxidation resistant materials at elevated temperature. However, RCCAs often exhibit multiple oxidation mechanisms including oxide volatilization, internal oxidation, external scale formation, and pesting. In this talk, we will present a new machine learning for accelerated materials discovery (ML-AMD) framework that utilizes multi-fidelity and multi-cost experiments with physics-based modeling. New semi-high-throughput methods for characterizing hardness and oxidation resistance will be presented, and methods for implementing high-throughput simulations into the ML-AMD framework will be expounded. Our recent efforts utilizing this framework have resulted in the discovery of ultra-high hardness single phase body-centered cubic RCCAs with up to 35at.% Al content. We have additionally created an open-source, automated framework housed in NanoHub for experimental data ingestion and storage of high temperature oxidation experimental data. Predictions of mass gain behavior of RCCAs utilizing machine learning with physics-based descriptors will be presented, and recent efforts to design new RCCAs with superior oxidation resistance will be shown.