Decoding weight-gain patterns in tungsten-containing refractory high-entropy alloys under high-temperature oxidation through machine learning

This work investigates the high-temperature (850 °C) oxidation behavior of tungsten-containing high-entropy alloys (HEAs) and develops a predictive framework for their oxidation behavior. Oxide scale evolution was characterized using XRD, SEM, and EDS, revealing that moderate W additions (0.05W and 0.1W) achieved the lowest parabolic oxidation rate constants (∼1.5 × 10 -10 and ∼1.4 × 10 -10 g 2 /cm 4 ·s, respectively), whereas excess W (0.5W) increased the rate constant to ∼8.6 × 10 -10 g 2 /cm 4 ·s. These results confirm that controlled W incorporation enhances oxidation resistance, while excessive W destabilizes protective scales. To complement experiments, machine learning models were trained to predict oxidation-induced mass gain. Among them, the random forest algorithm provided the best predictive performance, with a correlation coefficient (R) of 0.999 and minimal mean squared error. By integrating quantitative oxidation data with predictive modeling, this study delivers new insights into W’s role in scale stability and demonstrates machine learning as a powerful tool for guiding HEA design.