Remaining useful life (RUL) prediction is a crucial component of prognostics and health management (PHM). In structural health monitoring, accurate degradation models are essential for reliable RUL prediction. Conventional exponential degradation models rely on the Gaussian distribution, which may result in negative degradation rates. This paper improves the degradation model by assigning Gamma distribution to the degradation coefficients. Nonconjugacy in Bayesian inference is addressed through moment matching, and particle filtering is used to approximate the posterior distribution. Finally, the probability distributions of degradation trajectories and RUL are derived using characteristic functions. Experiments are conducted on the Virkler fatigue crack-growth data set. The results demonstrate that the proposed framework can provide both RUL estimates and  uncertainty quantification, supporting predictive maintenance decisions for fatigue-critical structures.

Remaining useful life (RUL),degradation model,Bayesian inference,uncertainty quantification