Prediction of remaining useful life for fatigue-damaged structures using Bayesian inference

Structural health monitoring enables fatigue damage for in-service structures to be evaluated and the remaining useful life to predicted. In this paper, Bayesian inference using a random walk method was implemented to predict the remaining useful life of an aircraft fuselage panel subjected to repeated pressurization cycles. The Paris' law parameters, m and C, were treated as uncertain along with the initial crack size, a₀. Random samples from the joint distribution of m, C, and a₀ were used to generate the fatigue crack growth curve using Paris' law. Using simulated crack size data, the probability that a selected fatigue crack growth curve represented the true fatigue crack growth curve was updated. Crack sizes were calculated using Paris' law with uncertain parameters and random noise and bias were added to the simulated crack sizes. With this approach, fatigue crack size was characterized by a probability distribution at each loading cycle. A detailed explanation of Bayesian updating using the random walk method is provided. crack size. The effect of the likelihood on the remaining useful life predictions was also evaluated. The proposed method takes into account model uncertainties as well as the presence of noise and bias in the measurement data.