A probabilistic total fatigue life model incorporating material inhomogeneities, stress level and fracture mechanics

Abstract Testing on notched specimens from thin sheet aluminum 2024-T3 was carried out to investigate the formation of fatigue cracks at constituent particles and to quantify the critical distributions and their stress level dependence. The distributions of fatigue lives, nucleation lives, and crack nucleating (CN) particle sizes were determined for each specimen and exhibited a significant stress level dependence. The measured distributions provided the foundation for the total fatigue life model, which uses a probabilistic Monte Carlo method in conjunction with Newman's fastran ii crack closure model. The total life model closely predicted the cumulative distribution function (CDF) of fatigue lives for the three stress levels examined and specifically predicted the shortest fatigue lives, critical from a design for reliability standpoint, and their variability. The total life model accounted for both nucleation and propagation lives; however, the results based on modeling the total life entirely as crack propagation were accurate and slightly conservative. Additionally, a probability of crack nucleation (POCN) concept to relate the distribution of all particles to the distribution of CN particles was developed based on the experimental observations and provides a better representation of the data than traditional threshold approaches.