Investigation on fatigue life prediction approach of welded joints via integrated data-driven method

In this study, a novel intelligent fatigue life prediction approach was established via the integrated data-driven method (Borderline-Synthetic Minority Over-Sampling Technique, eXtreme Gradient Boosting, Deep Convolutional Neural Network). Among them, the Borderline-Synthetic Minority Over-Sampling Technique was used to enhance the data quality of the fatigue performance dataset, the eXtreme Gradient Boosting was used to realize the weight analysis of the influencing factors of fatigue life, and the Deep Convolutional Neural Network was used as the model framework to understand the multiple nonlinear relationships between fatigue life and its influencing factors. Based on the analysis of different technology combinations, it was found that weight analysis and data augmentation were both beneficial for improving prediction accuracy, with the former having better results than the latter. And by comparing with other novel prediction models, the accuracy and stability of the proposed method were verified.