Model Updating and Damage Detection for Bridge Integrity Management

The integrity management of bridges is crucial for ensuring public safety and economic stability. In practice, Structural Health Monitoring data recorded during bridge operation is increasingly used to guide maintenance decisions. However, incorporating structural damage information more effectively can lead to optimal strategies for integrity management. In this study, we employ Bayesian Model Updating to develop a more reliable structural model. The updated finite element model is then used to train a variational autoencoder-based surrogate model for damage detection, localization, and severity estimation. The variational autoencoder model establishes a link between damage-related features and the modal properties derived from SHM data. Damage information supports maintenance decision-making through a predefined decision rule.