Abstract
This study introduces a digital twin framework combining finite element (FE) modelling with feedforward neural networks (FNNs) for structural health monitoring (SHM) of reinforced concrete bridges via a case study. A detailed FE model of a post-tensioned bridge was developed using data from the implementation plan and structural conditions. The FE model was verified and validated through extensive load testing, achieving 95% accuracy. Displacements, strains, stresses and crack widths in critical points of the structure, and virtual sensor data, were extracted from the FE model for various load scenarios, to simulate the behaviour and measurements on the real bridge. An FNN was then trained using longitudinal displacements and strain data as inputs to predict key performance indicators, such as deflections, stresses and crack width. The overall digital twin achieved prediction accuracy between 90-93%. The established connection was applied on the real bridge, where sensor data were fed into the FNN model, allowing prediction of structural performance for SLS once appropriate limits are set. This approach offers an automated solution for real-time SHM and proactive bridge maintenance.