Machine learning assisted calibration of a fatigue crack growth model considering temperature and stress ratio conditions

Abstract

Complex working conditions significantly influence fatigue crack growth (FCG), especially the temperature and stress ratio conditions. To describe the influence of temperature and stress ratio conditions on FCG behaviors, the corresponding macroscopic experiments and microscopic observations were performed. Through the introduction of the correction terms of temperature and stress ratio to Paris law, the extended Paris model was built in this study to enhance the predictive ability of FCG behaviors under working conditions. The parameters of FCG model are mainly calibrated by the method of fitting experimental curves, the size and dispersion of data for fitting seriously affects the accuracy of parameter calibration and subsequent FCG prediction. To improve the efficiency and accuracy of parameter calibration, a machine learning assisted calibration method was established to identify the parameters of the proposed model. The good agreement between prediction and experimental results of fatigue crack growth rate curves proves the rationality of the proposed FCG model and the parameter calibration method.