Advanced methodology for assessing distribution characteristics of Paris equation coefficients to improve fatigue life prediction

This paper offers a methodology for coping with information loss following consolidation of data on fatigue crack propagation rates derived from different experiments. It is customary, both in the literature and in standardization, to consolidate results of several experiments conducted under similar conditions, using identical materials. This reduces the ability to implement a probabilistic fracture mechanics approach in order to reliably calculate the distribution of the number of cycles needed to reach a critical value (CV; onset of instability or failure). Such reliable calculation requires, among other things, an estimation of the distribution characteristics of the crack progression curves coefficients represented by models such as Paris or NASGRO, and an estimation of joint distributions of equation coefficients representing such models. Consolidated data reduce the ability to estimate these required distribution characteristics. This work suggests an analytical approach that uses consolidated data, but enables the information to be treated as if it were possible to attribute the data to the various experimental specimens from which they were obtained. Consequently, information required for the evaluation of the distribution of the number of cycles needed to reach a CV can be obtained. The proposed approach is generic and can be applied in additional scientific fields that can benefit from separation of data obtained from different experiments.