Abstract
Fatigue cracking threatens the safety of orthotropic steel decks (OSDs) in long-span bridges, yet current manual inspection lacks predictive depth. We present a closed-loop framework integrating autonomous robotic inspection, vision-based quantification, and finite-element fracture mechanics to enable adaptive fatigue prognosis. In laboratory validation, the robotic platform achieved a mean localization accuracy of 2.7 ± 0.8 cm, meeting structural precision requirements. Field deployment on an in-service cable-stayed bridge demonstrated that automated inspection reduced average time per girder from 124.6 to 50.4 minutes-a 59.6% reduction. Identified cracks were assimilated into a digital twin for adaptive state updating. Analysis of discrepancies between simulated and observed propagation paths-interpreted via stress intensity factor fields-highlighted significant mixed-mode fracture effects, particularly elevated Mode-II (shear) contributions, as a primary source of predictive uncertainty under in-service conditions. This integration of robotics and digital twins provides a scalable solution for automated maintenance. Beyond labor reduction, the framework establishes a data-driven path toward proactive life-cycle management, enhancing the structural resilience and long-term safety of critical transportation infrastructure.