Leveraging machine learning to predict oxidation kinetics in interconnect alloys for solid oxide cell applications

In energy conversion technologies, solid oxide cells (SOCs) play a vital role, operating as both fuel and electrolysis cells. In SOC stacks, metallic interconnects are crucial, as they enable electrical connectivity, oxidant/fuel gas distribution, and structural stability. Oxidation during service is a significant challenge, compromising the long-term durability of interconnect alloys. Machine learning (ML) models are employed to predict the oxidation kinetics of SOC interconnects, addressing the complex interactions between alloy design and operating conditions. A comprehensive dataset is gathered from the literature, incorporating key inputs including alloy chemistry, service temperature, and atmosphere composition. The natural logarithm of the parabolic oxidation rate constant, ln( k p ), is regarded as the prediction target. Seven models are developed and tuned, achieving R 2 values of 0.90–0.95. Experimental oxidation data from SS441 interconnects, oxidized in air at 750–850 °C for up to 1500 h, are used as unseen validation data to confirm the predictive capability of the models. The integration of a data-driven approach with oxidation kinetics analysis provides deeper insight into the interplay between alloy design, service conditions, and oxidation behavior. The results demonstrate ML as a powerful tool for optimizing interconnect materials in SOC applications.