Bayesian inference of time-dependent corrosion-induced deterioration using multi-time observations of corrosion-induced crack width

Corrosion of steel reinforcement is a major cause of serviceability and safety degradation in reinforced concrete structures. Because environmental exposure is uncertain, prior models of time-dependent corrosion-induced deterioration may deviate from the actual deterioration process of field structures, affecting corrosion assessment and subsequent reliability evaluation. To address this issue, this paper develops a parameter-driven Bayesian updating framework to infer time-dependent steel weight loss from corrosion-induced crack width measurements collected at multiple inspection times. Temporal uncertainty is represented through uncertain corrosion-evolution parameters in the state transition equation, while a data-driven forward model links latent steel weight loss with observed crack width. A Gibbs-type cyclic sampler is developed to sample from the joint posterior distribution of corrosion states and the hyperparameters governing corrosion evolution. The proposed framework extracts information on time-dependent deterioration from multi-time observations to update the prior deterioration model. Validation using experimental data demonstrates that the method can identify time-dependent corrosion characteristics and improve the reliability of corrosion-state inference.