Vibration-based anomaly detection method for structural health monitoring of in-service structures under ambient vibration

This paper presents a vibration-based monitoring system that is able to identify hazard events like earthquakes or typhoons, and to detect anomalies in the structures after these events using ambient vibrations (i. e. vibrations induced by random excitation such as traffic, wind or white noise). The statistic distributions of modal parameters such as natural frequencies, mode shapes are indicators for anomaly. Change of distribution model of modal properties indicates that the structure may be damaged or anomaly. An original algorithm was developed to identify modal parameters effectively and detect anomaly in sub real- time. Long term monitoring data sets in two in-service bridges confirmed that a very small change in frequencies due to asphalt work during construction can be detected by proposed system. The results show that the variation of natural frequencies due to seasonal change can be filtered in order to detect very small changes in modal characteristics. The proposed system could be applicable for nearly real- time monitoring of structures such as bridges and buildings.