Machine-learned accelerated discovery of oxidation-resistant NiCoCrAl high-entropy alloys

Designing oxidation-resistant bond coats from high-entropy alloys (HEAs) has been a serious concern primarily due to the limited understanding of how multiprincipal element interactions govern scale formation at extreme temperatures. Moreover, the use of conventional trial-and-error approaches is inefficient due to the wide possibilities of element combinations, environmental conditions, and other processing variables. In this study, a machine learning framework is trained on high-fidelity experimental data to predict the parabolic oxidation constant ( k p ) as a measure of oxidation resistance. The model not only achieves state-of-the-art precision ( R 2 = 0 . 91 ) but also reveals new oxidation trends in NiCoCrAl HEAs. The predictions reveal a clear, temperature-dependent transition in oxidation control, with Al-rich compositions exhibiting the lowest oxidation rates above 1100 °C. Compositions containing mixed Al-Cr ratios perform best at intermediate temperatures, and Cr-rich, low-Al alloys exhibit superior resistance at 850 °C. The incorporation of Hf and Y led to composition-dependent improvements, with Hf-added alloys exhibiting the lowest k p . The effectiveness of Y was observed in NiCo-lean alloys, while Y-Hf co-doped alloys showed saturated improvements. These results provide quantitative guidance for narrowing the compositional space of NiCoCrAl and identifying oxidation-resistant bond coats across service-relevant temperatures.