Statistical estimation of fatigue design curves from datasets involving failures from defects

Abstract

In the present paper, two methodologies for the estimation of the design curves of datasets with failures originating from defects are proposed. With the first methodology, the Likelihood Ratio Confidence Bound of a specific quantile P-S-N curve is considered. The second method is based on the bootstrap approach, with a large number of datasets simulated starting from the stress life and the defect size distributions estimated from the experimental data. The two approaches have been validated on literature datasets covering also the Very High Cycle Fatigue (VHCF) life region, proving their effectiveness.