Probabilistic sensitivity analysis in life-prediction of an α + β titanium alloy

Probabilistic sensitivities using the score function method are developed for a lifing analysis of an α + β titanium alloy in a round bar under axial fatigue load. Sensitivities with respect to the statistical inputs of the crack initiation size (a), and Paris crack growth intercept (C) and exponent (m) are developed with consideration of the correlation between C and m. The sensitivities are obtained using a single Monte Carlo sampling analysis and do not involve finite difference approximations. The sensitivities indicate the importance of the random variable input parameters on the mean life and standard deviation of life and can be used as a basis for determining constructive data collection efforts. For this example, the crack growth intercept (C) is the dominant variable that affects mean-life and standard deviation of life, indicating that improved confidence in the results can be obtained most efficiently by improving the statistical characterization of C.