Abstract:Taking 31 representative bridges in a district and a county of Shanghai (including the structural characteristics of simple-supported fabricated bridges, prestressed concrete continuous box girders, bowstring arch bridges, reinforced concrete multi-arch bridges and suspended-span bridges) as the practical engineering cases, a systematic research is conducted on the setting of early warning thresholds for monitoring data in the bridge structure health monitoring system and the abnormal diagnosis of data from single sensor signals. Based on the theoretical analysis and engineering parameter settings, it is proposed to set the initial values for monitoring and early warning of different bridge types by integrating the standard threshold method. For the problem of abnormal signals from a single sensor, the methods such as median filtering, differential detection, dynamic baseline correction, and Gaussian mixture (GMM) are integrated to keep the false alarm rate within 4% while effectively retaining the key feature values. The research results provide the reliable technical ideas for the threshold optimization, abnormal diagnosis and maintenance decision-making of the bridge group monitoring system, which has certain reference value for the safety monitoring of similar urban infrastructure clusters.
