Remaining fatigue life prediction of additively manufactured Inconel 718 alloy based on in-situ SEM and deep learning

Remaining fatigue life prediction is vital for engineering structures to ensure safety and reliability. The successive and dynamic changes in morphology are valuable informational reminders of the remaining life of a material. This study proposed a novel strategy to monitor and predict the remaining life of materials in a successive and dynamic manner, incorporating the in-situ scanning electron microscope (SEM) and deep learning algorithm. Two in-situ SEM fatigue experiments were conducted to simulate the practical service conditions of additively manufactured Inconel 718 alloy. A series of SEM image sequences reflecting the changing morphology were collected and used for downstream deep learning-based remaining life prediction. Original SEM images were cropped and grouped for training, validation, and test purposes. Results showed that the deep learning-based method exhibited high prediction ability. The proposed method’s predicted life values were highly close to the experimental results. The visualization result showed that the microstructure evolution, such as accumulation of slip bands, surface fluctuation changes resulting from coordinated grain deformation, and fatigue crack growth, provide much information and decision support to the network. These results demonstrated the great promise of monitoring and predicting the remaining fatigue life of materials based on morphology. We believe this image and deep learning-based strategy will inspire current fatigue life prediction and facilitate the move toward successive, automatic, and accurate life monitoring in practical engineering structures.