Stochastic modelling of fatigue crack growth

Abstract

The article deals with the influence of the skewness of quasi-constant stress range on the fatigue resistance of a steel bridge member subjected to multiple repeated loading. It is shown that it is necessary to know the statistical characteristics and type of probability density function relatively accurately in the case of quasi-constant stress range, otherwise the estimation of the fatigue resistance may be inaccurate. The fatigue resistance is studied using linear fracture mechanics. The random characteristics of the steel member are simulated using the Latin Hypercube Sampling method. The results of experimental research on calibration functions used in linear fracture mechanics are applied in this study. The calibration function for pure bending, which is presented on an extended domain of relative crack length, is applied. Non-parametric rank-order correlation identifies a strong influence of the quasi-constant stress range on the fatigue resistance. A detailed comparison of three variants of fatigue resistance as output random variables is performed. The results show that the probabilistic assessment of reliability may be burdened with greater or lesser epistemic uncertainty, which can distort the estimates of the remaining service life of the load bearing bridge member.