Anomaly detection of bridges by Bayesian method with optimized sensor placement

This study aims to investigate the effectiveness of a Bayesian hypothesis-based damage detection using sensor arrangement optimized by the effective independence (EFI) method. The feasibility of Bayesian anomaly detection is verified using vibration and damage experimental data on an actual steel plate girder bridge. It was found that the same vibration frequencies could be identified with the optimally located three sensors as when all sensors were used. Observations demonstrated that it is possible to detect damage and to identify the damage location approximately even with optimally arranged three sensors. In addition, the EFI method was also useful in Bayesian anomaly detection.