In this study, we investigated the feasibility of using bridge-weigh-in-motion (BWIM) as a feature for bridge anomaly detection in machine learning. To balance additional influence line due to bridge damage, the proposed BWIM approach evaluates actual and virtual wheel loads. The displacements used in BWIM were estimated from video footage using a deep learning method. The weight assigned to the virtual wheel has been considered as a feature in machine learning based anomaly detection. Damaged model bridge experiments showed the proposed method’s ability to detect bridge anomalies and its sensitivity to a damaged position.
