Abstract: Additive manufacturing features rapid production of complicated shapes and has been widely employed in biomedical, aeronautical and aerospace applications. However, additive manufactured parts generally exhibit deteriorated fatigue resistance due to the presence of random defects and anisotropy, and the prediction of fatigue properties remains challenging. In this paper, recent advances in fatigue life prediction of additive manufactured metallic alloys via machine learning models are reviewed. Based on artificial neural network, support vector machine, random forest, etc., a number of models on various systems were proposed to reveal the relationships between fatigue life/strength and defect/microstructure/parameters. Despite the success, the predictability of the models is limited by the amount and quality of data. Moreover, the supervision of physical models is pivotal, and machine learning models can be well enhanced with appropriate physical knowledge. Lastly, future challenges and directions for the fatigue property prediction of additive manufactured parts are discussed.

Key words: Fatigue, Additive manufacturing, Metallic alloys, Machine learning