Refractory high-entropy alloys (RHEAs) and complex concentrated alloys (RCCAs) are vital for high-temperature applications beyond the capabilities of Ni-based superalloys. Traditional methods for predicting oxidation resistance in these alloys are often inaccurate and resource-intensive. This study introduces a novel approach using Gradient Boosted Decision Trees (GBDT), an artificial intelligence technique, to predict specific mass gain due to oxidation. Utilizing a dataset synthesized from extensive literature and characterized by diverse alloy compositions and oxidation conditions, the model was trained using Iterated Nested k-fold Cross Validation with Shuffling (INKCVS). Our findings demonstrate that the GBDT model achieves a good balance between accuracy and generalization capacity in predicting oxidation resistance, as validated experimentally with selected alloys. This approach not only enhances prediction accuracy but also significantly reduces the need for extensive experimental testing, facilitating rapid development of new high-performance materials.

Refractory high-entropy alloys (RHEAs) [1], refractory complex concentrated alloys (RCCAs) and high entropy superalloys (HESAs) [2] have emerged as significant candidates for high-temperature applications, offering promising alternatives to conventional Ni-based superalloys, which are reaching their operational temperature limits. As future technologies demand materials that can endure higher temperatures, for applications such as aerospace engines, nuclear reactors, and thermal protection systems, the development of alloys capable of performing under such extreme conditions has become crucial. RHEAs and RCCAs, characterized by a mix of multiple principal refractory elements, often display superior mechanical properties and higher melting points compared to traditional materials [3,4]. These alloys, drawn from a palette of nine refractory metals including Zr, Hf, V, Nb, Ta, Cr, Mo, W, and Re, with minor additions of Al, Si, or Ti, are designed to withstand temperatures well beyond the 1000 • C threshold, competing with Ni-based superalloys.

Despite their high-temperature mechanical performance, a significant challenge in the development of RHEAs and RCCAs is their susceptibility to oxidation [5-9], a critical factor in many high-temperature environments, which can severely impair their mechanical properties. Such a high temperature oxidation process is governed by complex thermodynamic and kinetic factors that involve the formation, growth, dissolution, and spalling of oxide layers. Traditionally, predicting the oxidation behavior of alloys has relied on empirical observations and complex physical models, which are both resource-intensive and limited in their predictive accuracy, particularly for new alloy compositions. These methods often fail to effectively navigate the vast design space required for developing innovative materials. Given these significant limitations, there is a pressing need for more adaptive and scalable models.

Recognizing the limitations of traditional methods, we propose a novel AI-driven approach to predict the oxidation resistance of RHEAs/ RCCAs. Unlike conventional empirical or thermo-kinetic models, our AI techniques efficiently analyze extensive datasets to uncover patterns not