Abstract
The degradation of metallic systems under cyclic loading is subject to significant uncertainty, which affects the reliability of residual lifetime predictions and subsequent decisions on optimum maintenance schedules. This paper focuses two main challenges in developing a reliable framework for the lifecycle management of fatigue-critical components: constructing a stochastic model that captures uncertainties in crack growth histories, and presenting a computationally efficient strategy for solving the stochastic optimization associated with maintenance scheduling. Polynomial chaos (PC) representation is proposed to propagate uncertainty in the fatigue-induced crack growth process, using a database from constant amplitude loading experiments. Additionally, an optimization strategy is implemented based on Gaussian process surrogate modeling to solve the stochastic optimization problem under maximum probability of failure constraints. The sensitivity of the optimum solution to different probability of failure thresholds is examined. The proposed framework offers a decision support tool for informed decision-making under uncertainty, aiming to mitigate fatigue failure.