Application of Explainable Machine Learning in Metal Fatigue Life Prediction: Methods, Challenges and Prospects

Fatigue life prediction is crucial for the safety of metallic materials and structures. Machine learning (ML) models have demonstrated strong predictive capabilities in this field, but their "black-box" nature limits their reliability, trustworthiness, and application in engineering practice. Explainable Artificial Intelligence (XAI) provides key techniques to open the ML "black box". This paper aims to systematically review the current applications of XAI in the field of metal fatigue life prediction. Addressing the issue that current research applies various interpretation methods without systematic categorization, this paper proposes classifying existing methods into two main categories: "post-hoc explanations" and "interpretable by-design". This paper outlines the key techniques of these two categories and their specific application examples in fatigue life prediction, discusses the limitations of model interpretation methods and current technical challenges, and prospects for opportunities and challenges in subsequent research.