Digital twin for intelligent maintenance towards sustainable bridges

Aging steel bridges constitute a significant portion of transportation infrastructure, requiring more efficient and intelligent maintenance to enhance their sustainability. This work presents a digital twin (DT) framework for the intelligent maintenance of aging steel bridges subject to time-dependent deterioration, by fully exploiting the prediction model and the data from monitoring and inspection. A tied-arch bridge, subjected to prominent corrosion fatigue (CF) in its hanger system, is selected as a case study. The DT framework comprises three modules: digital data perception, deterioration prognosis, and condition-based decision making. First, a finite element (FE) model of the bridge is developed as the mechanical backbone of the DT, based on the design document and updated with the latest inspection result. Then, the traffic data collected by the structural health monitoring (SHM) system are converted into fatigue stress spectra. Next, the stress spectra and site-specific environmental data are integrated to perform a hybrid deterioration prognosis using a time-variant probability-stress-life model. Finally, optimal maintenance actions are determined based on the prognosis results, by constructing a stochastic environment and adopting a customized deep learning-based solution algorithm. The research outcome demonstrates the value of incorporating SHM data in deterioration prognosis and maintenance planning, and provides a promising framework for the interoperable DT of aging structures at the mechanistic level. The result reveals the nonlinear evolution of C-F deterioration in steel bridge hangers, with increasing failure probability and dispersity. Consequently, different strategies are recommended by the proposed DT framework, such as regular repairs in the early life and frequent inspections and timely replacement in the late life. Furthermore, the present DT framework suggests that the appropriate use of SHM systems can effectively reduce the life-cycle costs and risks, with a value of information that exceeds the initial investment.