Effect of accuracy of the initial defect distribution on successful prognosis of aircraft structures

We utilize fatigue-crack-growth and load-spectra data published in the open literature to estimate aircraft probability of failure (PoF). We show that because single flight PoF values should not exceed 10^-7, the behavior of the tail of the initial defect distribution significantly affects the predicted safe life. We demonstrate that even a slight variation of initial defect distribution parameters considerably changes the number of allowable flight hours. A similar effect is observed when the aircraft fatigue-damage assessment is based on the existence of a defined size crack with a specified (small) probability. We consider on-line damage-monitoring sensors and show that use of such sensor data leads to improved accuracy of the safe-life predictions. Finally, we outline an approach for a more accurate assessment of the initial defect distribution that is based on applying recently developed analytical and computational tools to extensive metallographic examination data.