Reliability sensitivity algorithm and its application in creep/fatigue failure

For non-linear limit state equation, two algorithms, i.e., approximate analytics based on first-order and second-moment and numerical simulation based on Monte-Carlo method, are presented for reliability sensitivity analysis with non-normal basic random variables. In the approximate analytic algorithm, the non-normal basic random variable is transformed to normal one equivalently. Then, the reliability sensitivity algorithm for the normal random variables and the chain derivative rule are employed to calculate the sensitivity of failure probability with respect to the distribution parameters (such as mean values and variances) of the non-normal random variables. In the numerical simulation algorithm, the appropriate sampling points are selected from those of Monte-Carlo method for the regression analysis around the vicinity of the design point, afterwards, the chain derivative rule is used to obtain the reliability sensitivity. The presented algorithms are applied to the non-linear creep/fatigue failure model. The feasibility and the rationality of the present algorithms are illustrated by the good agreement between the analytics and the simulation. The reliability sensitivity of the creep/fatigue failure model varying with the distribution parameters of the random variables supplies helpful guidance for the engineering design.