Data fusion and machine learning for bridge damage detection

This study presents a new approach for bridge damage detection using multi-level data fusion and anomaly detection techniques. The approach utilises as input data accelerations, deflections and bending moments, measured at multiple sensor locations on a bridge subjected to a moving vehicle. A damage sensitive feature is constructed, coupling principal component analysis and Mahalanobis distance, allowing for initial data dimensionality reduction and information integration. Anomaly detection using a convolutional autoencoder is performed to identify the presence of damage on the bridge. The proposed approach is independent of the mass and speed of the moving vehicles. The performance of the proposed approach is demonstrated using synthetic data generated from surrogate numerical models, showing applications for a variety of damage scenarios. The accuracy of damage identification via anomaly detection is shown to be consistently greater than 99%.