Abstract: Oxidation resistance is a critical metric for assessing the high-temperature property of superalloys. Traditional models are often constrained by the parabolic rate law, limiting their ability to simulate complex oxidation behavior. This study introduces a hybrid machine learning model that combines a onedimensional convolutional neural network with a long short-term memory network to predict oxidation behavior with high accuracy (R2 = 0.981) and smoothness. The model demonstrates improved predictive performance across various stages of oxidation, successfully fitting a wide range of oxidation kinetics and accurately estimating the activation energy for the Co-9W-9Al-0.12B alloy. It also identifies the critical Cr content range for the transition from internal to external oxidation in Co-based superalloys, which aligns well with experimental results and theoretical calculations. Although this study focuses on Co-based superalloys, the versatility extends its applicability to other superalloy systems, paving the way for future research in materials science.

Key words: Superalloys, Oxidation, Machine learning, Kinetics, Oxidation mechanism