Monitoring-based Identification of Corrosion in Steel Structures

Over the service lifetime of a structure, corrosion in steel components progressively deteriorates their load-bearing capacity and compromises overall structural safety. Despite its substantial impact on structural performance, research on corrosion progression and its longterm effects remains limited. Most existing studies focus on corrosion incidents in reinforced concrete bars and steel cables, while effective methods for corrosion detection in steel members, such as girders and other structural elements are scarce. This study presents a methodology for detection and quantification of corrosion-induced damage in large-span steel structural components. The proposed approach leverages measured strain data from the monitored structure and employs a probabilistic framework to model and assess the progression of corrosion over time. Corrosion localization is achieved via use of a data-driven indicator, relying on transmittance functions, which are inferred from pairwise strain measurements at locations within the structure. To address the challenge of limited sensor measurements, a Gaussian Process Regression (GPR) model is employed to predict strains at unmeasured locations. The method is validated through numerical simulations employing a high-fidelity finite element model of a steel bridge.