A novel nonlinear fatigue cumulative damage model based on machine learning

Abstract

Machine learning serves as a potent tool for predicting fatigue damage. By integrating traditional physics-based models with machine learning techniques, prediction accuracy can be notably enhanced. Drawing from a substantial pool of experimental data, this article introduces a novel nonlinear fatigue cumulative damage model. This model combines the traditional Manson-Halford model with machine learning algorithms, enabling the prediction of residual fatigue damage across various materials. The predictive capabilities of this model are compared with those of traditional approaches. The article subsequently investigates the key parameters influencing the nonlinear fatigue cumulative damage model. The findings demonstrate that the proposed model exhibits significantly higher accuracy in residual fatigue damage prediction compared to traditional methods. Notably, the primary factor impacting the nonlinear cumulative fatigue damage model is identified as the first level loading damage. This study underscores the substantial potential of the proposed nonlinear fatigue cumulative damage model for predicting residual fatigue damage.