Fatigue crack identification for steel structures based on distributed fiber optic sensing in a hybrid Digital Twin framework

This study presents a hybrid Digital Twin framework for fatigue crack identification in steel structures using strain measurements obtained from distributed fiber optic sensing (DFOS). The framework integrates established data-driven and physics-based methods to provide a comprehensive workflow from damage detection and localization to crack assessment and fatigue-life prognosis. In contrast to conventional vibration-based structural health monitoring (SHM) approaches, the framework exploits quasi-continuous distributed strain measurements to identify localized fatigue-induced strain redistributions on a global structural scale, enabling the detection of cracks that do not directly intersect the sensing fiber. In the first stage, damage-sensitive strain features are extracted from DFOS measurements and evaluated for anomalies relative to a reference structural condition. Subsequently, a parametric finite element model incorporating predefined crack paths is updated using the measured strain fields to estimate crack locations and sizes. The identified crack parameters are then used to predict the remaining fatigue life based on linear elastic fracture mechanics, providing information for inspection and maintenance planning. The framework is experimentally validated through uniaxial tensile fatigue tests on welded steel specimens containing single and multiple fatigue cracks that initiated and propagated during testing until failure. The results demonstrate reliable detection and localization of fatigue damage using distributed strain measurements, while physics-based model updating enables the estimation of physically interpretable crack parameters. The remaining fatigue life predictions are conservative, with accuracy improving with progressive crack growth. Overall, the results demonstrate the potential of combining DFOS with data-driven and physics-based modeling within a hybrid framework as a first step toward scalable, condition-based SHM and service life prognosis of fatigue-critical steel structures such as bridges.