Stochasticity study of fatigue crack growth model parameters in probabilistic damage tolerance analysis

Purpose The primary objective of this study is to investigate the influence of the stochastic variability of fatigue crack growth (FCG) model parameters on the structural FCG life and reliability. Design/methodology/approach This study focused on a high-temperature steam turbine rotor involved in a fatigue fracture accident at a power plant. From the perspective of the randomness of FCG model parameters (C, m) and the initial crack length a0, the probabilistic characteristics of the random variable a0 and the bivariate random vector (C, m) under different distribution types were investigated. A method based on the Nataf model was proposed to construct the bivariate Weibull distribution. On this basis, the probabilistic distribution of FCG life and the prediction of failure probability in probabilistic damage tolerance analysis were studied. Findings The results reveal that the distribution type and coefficient of variation significantly affect the probabilistic distribution of FCG life. In terms of failure probability prediction, the variation patterns of failure probability curves under different distribution types and coefficients of variation were obtained. Furthermore, sensitivity analysis shows that for the bivariate random vector (C, m), the failure probability is consistently more sensitive to m than to C, even under different distribution types. Therefore, particular attention should be paid to accurately quantifying the randomness of m and addressing its correlation with C, in order to achieve more accurate reliability predictions. Originality/value This work provides a reference for the development of more comprehensive probabilistic damage tolerance analysis methods for structural applications.