Physics-informed neural network for fatigue life prediction of irradiated austenitic and ferritic/martensitic steels

This study proposes a Physics-Informed Neural Network (PINN) framework to predict the low-cycle fatigue (LCF) life of irradiated austenitic and ferritic/martensitic (F/M) steels used in nuclear reactors. During operation, these materials undergo cyclic loading and irradiation at elevated temperatures, resulting in complex degradation mechanisms that traditional empirical or purely data-driven models often fail to capture accurately. The developed PINN model incorporates physical fatigue life constraints into its loss function, improving prediction accuracy, reliability, and generalizability. Trained on 495 data points, including both irradiated and unirradiated conditions, the model is more robust than traditional machine learning models, such as Random Forest, Gradient Boosting, eXtreme Gradient Boosting, and the conventional Neural Network. Model interpretability assessed via SHapley Additive exPlanations analysis revealed that strain amplitude, irradiation dose, and test temperature are the dominant features, each exhibiting a physically consistent inverse correlation with fatigue life. Univariate and multivariate analyses showed that strain amplitude is the primary driver of fatigue degradation in both alloy classes, while irradiation dose and temperature introduce alloy-specific sensitivities. The PINN successfully captured key mechanistic trends, including the comparatively stable irradiation response and dose saturation behaviour of F/M steels, as well as a pronounced reduction in fatigue life at elevated temperatures exceeding the tempering threshold. Overall, the proposed PINN framework offers a reliable and interpretable tool for predicting fatigue life in irradiated structural alloys, thereby supporting informed materials selection and performance assessment for advanced nuclear reactor applications.