Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
770301 | Engineering Fracture Mechanics | 2014 | 27 Pages |
•We present methods for evaluating the fatigue reliability.•We focus on variable amplitude fatigue loading.•We demonstrate that the subset simulation provides a practical assessment tool.•The reliability is influenced by the correlation of the stress range process.
In fatigue reliability assessments, the random load process is commonly represented by its marginal distribution (load spectrum) only. However, as shown in this paper, the correlation characteristics of the load process can have a strong influence on the fatigue reliability and should be accounted for. The paper reviews the modeling of random fatigue crack growth under variable amplitude loading for reliability analysis. Solutions for fatigue crack growth evaluation at different levels of detailing are described and a fatigue crack growth and failure evaluation algorithm, based on a discretization of the random stress process, is presented. As an alternative, a mean approximation is described. Finally, effective computational methods for assessing the fatigue reliability under variable amplitude loading are introduced and applied exemplarily to a case study. The solutions are based on the first-order reliability method FORM and the subset simulation. Using a Markov process model of the loads, the influence of different types of service histories is investigated, by varying the correlation length of the stress cycle process. The results show that the correlation length of the load process has significant influence on the resulting reliability; the resulting probability of failure can vary up to several orders of magnitude for the same marginal probability distribution of stress amplitudes. Based on the results of the case study, the influences of the stress process correlation and of the adopted failure criteria on the reliability are discussed. The mean approximation and the random variable model of the random load process are demonstrated to be applicable under specific conditions.