Research on Fatigue Life Prediction Method of Spot-Welded Joints Based on Machine Learning

Spot-welding joints are widely used in modern industries, and their fatigue life is crucial for the safety and reliability of structures. This paper proposes a method for predicting the fatigue life of spot-welding joints by integrating traditional structural stress methods and machine learning algorithms. Systematic fatigue tests were conducted on Q&P980 steel spot-welding joints to investigate the influence of the galvanized layer on fatigue life. It was found that the galvanized layer significantly reduces the fatigue life of spot-welding joints. Further predictions of fatigue life using machine learning algorithms, including Random Forest, Artificial Neural Networks, and Gaussian Process Regression, demonstrated superior prediction accuracy and generalization ability compared to traditional structural stress methods. The Random Forest algorithm achieved an R² value of 0.93, with lower error than traditional methods. This study provides an effective tool for the fatigue life assessment of spot-welding joints and highlights the potential application of machine learning in this field.