Digital Twin-Empowered Dynamic Inspection and Maintenance of Steel Bridge Hangers under Deterioration

The corrosion-fatigue (C-F) coupled deterioration of steel bridge hangers has resulted in substantial operational and structural risks, demanding proper and timely inspection and maintenance actions. Especially, a better understanding of C-F deterioration state in hangers is urgently required in light of available monitoring data. Moreover, optimal planning of management actions is still lacking with the deterioration accounted for. This work presents a digital twin-empowered dynamic inspection and maintenance (DIM) method for steel bridge hangers and applies to a case-study bridge equipped with a structural health monitoring (SHM) system. Initially, the finite element (FE) model of the bridge is established after the archived design document, and then tuned with the latest inspection result. Based on the FE model, the traffic data monitored by the SHM system are then transformed into fatigue stress spectra of hangers, assisted by Bayesian dynamic linear models (BDLMs) for future prediction. By integrating the derived spectra and site-specific ambient data, the hybrid prognosis is conducted on the hanger deterioration using a time variant probability-stress-life (t-PSN) model. Finally, optimal maintenance actions are decided after the prognosis results, by constructing a partially observable Markov decision processes (POMDP) environment and tailored deep learning-based solution algorithm. The research output not only highlights the incorporation of SHM data in deterioration prognosis and maintenance planning, but also offers a promising backbone for the interoperable digital twin of ageing structures at the mechanism level.