A hybrid approach of modified bootstrap and physics-based methods for probabilistic fatigue life prediction considering overload effects

Abstract

A hybrid approach is proposed by combining the modified bootstrap and physics-based methods, which aims to achieve a reasonable and accurate probabilistic fatigue life prediction under arbitrary overload conditions based on constant amplitude fatigue data. The fatigue crack growth model and equivalent initial flaw size (EIFS) are employed to calculate the fatigue life. The crack closure mechanism is used to account for the loading interaction effects. Furthermore, the uncertainties of EIFS, coefficient of the fatigue model and overload intervals are considered. Fatigue data are collected by performing testing of aluminum alloy 7075-T6 under constant amplitude loading with/without overloads. The bootstrap method is modified by generating the expanding samples, which is utilized to quantify the dispersion of fatigue data. The resampling is conducted based on the expanding samples, instead of the original data directly obtained from the experiment. Eventually, the fatigue lives with different failure probabilities and confidence levels are calculated. In addition, other testing data sets of Al–Li alloy 2060 are also employed to verify the effectiveness and accuracy. Good agreements are observed.