A study of the volume effect and scatter at the fatigue limit – experiments and computations for a new specimen with separated notches

In this paper, a new fatigue specimen is presented that has been developed in order to conduct detailed investigations of the volume effect and the scatter of fatigue data in the fatigue limit regime. The specimen has two separated notches, A and B, with different size of the loaded volume associated with them, respectively. By changing the location of the applied load, the fatigue failure in the experiments can be controlled to occur in either notch A or B. Comparisons were made with simulations based on statistical weakest link (WL) theory. In the weakest link statistics, the three parameter Weibull distribution is used for estimation of the fatigue failure probability. It is shown that the predictive capability of WL-models is poor for high and low failure probabilities when the two separated notches, A and B, are considered. If only one notch (either A or B) is considered, and the other one is disregarded, the predictive capability of WL is drastically improved. Experiments with almost equal failure WL-probability in A and B (28% and 27%) did not result in equal experimental failure probability. WL-statistics can not predict the experimental outcome when separated notches are present.