Abstract

A methodology is developed and applied that determines the sensitivities of the probability-of-fracture of a gas turbine disk fatigue analysis with respect to the parameters of the probability distributions describing the random variables. The disk material is subject to initial anomalies, in either low- or high-frequency quantities, such that commonly used materials (titanium, nickel, powder nickel) and common damage mechanisms (inherent defects or surface damage) can be considered. The derivation is developed for Monte Carlo sampling such that the existing failure samples are used and the sensitivities are obtained with minimal additional computational time. Variance estimates and confidence bounds of the sensitivity estimates are developed. The methodology is demonstrated and verified using a multizone probabilistic fatigue analysis of a gas turbine compressor disk analysis considering stress scatter, crack growth propagation scatter, and initial crack size as random variables.