Statistical size effect in the fatigue limit of steel

There is always a large quantity of initiated cracks in a specimen subjected to cyclic loading. This kind of specimen forms a sample of initial cracks: the larger the specimen, the larger the sample. A method for the prediction of the statistical size effect due to the sample size is introduced in this paper. The statistics of extremes can be used to calculate an estimate of the largest expectable crack in specimens with varying surface area. Connection between the initiated crack size and the fatigue limit is obtained with the help of the linear elastic fracture mechanics. The presented method is tested to eight sets of experimental data for cylindrical specimens made of steel 34CrNiMo8. The mean fatigue limit tends to decrease and the scatter increase in proportion to increasing material thickness. If these effects can be accounted for, as done in this paper, the constant amplitude fatigue limit can be very accurately predicted. The error of the calculated estimates is within −3% to +6% compared to experimental results.