COMPARISON OF A STOCHASTIC-MODEL OF FATIGUE-CRACK GROWTH WITH EXPERIMENTS

A stochastic model of fatigue crack growth under constant amplitude loading, derived from a global/local energy balance, is compared with experiment data. Four random variables and a Gaussian white noise random field were assumed to model, respectively the statistical scatter of the crack length versus N curves and their stochastic fluctuations reflecting the effect of stochastic material non-homogeneity within a specimen. The statistical analysis of the so called Virkler and Ghonem-Dore data sets is extensively used to identify the probabilistic characteristics of the random variables and random field and to verify the stochastic model. Based on the importance sampling integration technique the fatigue lifetime probability distributions are effectively calculated. The influence of the particular assumptions accounting for the random material properties is investigated and the quality of the model prediction is verified by comparison with empirical statistical characteristics of the lifetime.