Non-contact acoustic-visual diagnosis of bridge expansion joint networks: Inspection and health monitoring

Bridge expansion joints are critical yet failure-prone components whose malfunction can lead to traffic-safety risks. Conventional inspections rely heavily on lane closures and manual checks, which are inefficient for large-scale bridge networks and often fail to detect concealed defects. To address these limitations, this study develops a non-contact acoustic-visual diagnostic framework for bridge expansion joint networks, supporting automated, scalable, and traffic-interruption-free inspection and monitoring. The framework integrates vehicle-induced acoustic responses with image-based event information through a compact acoustic-camera sensing configuration and an edge-cloud collaborative architecture. Visual information is used for vehicle-event detection, acoustic-signal synchronization, and vehicle-type routing, while acoustic signatures serve as the primary condition-sensitive input for unsupervised anomaly detection. Within the investigated category of modular H-beam expansion joints with rubber bearings, the portable model reduces dependence on site-specific labeled defect data and bridge-specific retraining, whereas the fixed monitoring mode requires site-specific baseline calibration. Full-scale laboratory experiments established the relationship between representative high-risk defects and acoustic signatures, providing the basis for the diagnostic model design. A Hybrid-Autoencoder was then developed and evaluated, with ablation analysis confirming the contribution of crossed multi-head (CM)-Attention, tokenization/compression, and structured feature fusion. Field validation on 28 in-service bridges demonstrated the system’s ability to identify anomalous joints, including concealed internal defects missed by surface visual inspection. A subsequent 1-year deployment further verified system stability under realistic traffic and environmental conditions. The results demonstrate the practical potential of the proposed acoustic-visual diagnostic framework for preventive maintenance within the investigated expansion-joint category.