Evaluating the importance of spatial variability of corrosion initiation parameters for the risk-based maintenance of reinforced concrete marine structures

• Sensitivity Analysis is applied to failure probability at durability limit state. • All hyperparameters are defined from a measurement database on a wharf RC beam. • Logs of apparent diffusion parameters are found to be linearly correlated. • It is essential to characterize apparent diffusion parameters’ correlation for RBM. • An accurate characterization of fluctuation scales is unnecessary for RBM. In risk-based maintenance (RBM) of reinforced concrete (RC) marine structures, modeling the spatial variability of corrosion initiation parameters is crucial for ensuring durability. However, the necessity for an accurate characterization of this spatial variability has not yet been fully investigated, despite the potential increase in measurement costs. This study addresses this gap by focusing specifically on the failure probability at the Durability Limit-State (DLS) due to chloride-induced corrosion initiation. A robust sensitivity analysis (SA) methodology, combined with global quantitative All-At-Time (AAT) methods, is applied to a case study of a wharf beam. The objective is to identify the spatially variable degradation parameters whose fluctuation scales have at least the same impact on failure probability as other statistical hyperparameters (HP). The results highlight that key parameters – namely the correlation coefficient of diffusion parameters and the mean and standard deviation of total chloride apparent diffusivity – significantly impact failure probabilities, ranking as the first, second, and third most sensitive HP, respectively. Among fluctuation scales, only that of chloride diffusivity can affect failure probability, while others rank no higher than fifth in sensitivity. The findings demonstrate that a broad, pre-defined range for fluctuation scales (4%–20% of element dimensions) is sufficient for RBM, minimizing the need for costly updates over time. The study also reveals that incorporating aging and diffusion parameter correlations significantly changes both failure time and failure probabilities, increasing them up to 33% and 40 percentage points, respectively, in some scenarios.