Probabilistic Damage Detection of Long-Span Bridges Using Measured Modal Frequencies and Temperature

This paper aims to develop a new probabilistic monitoring-based framework for damage detection of long-span bridges, by eliminating the temperature effects from the measured modal frequencies, probabilistic modeling of modal frequencies using kernel density estimate, and detection damage using the control chart. A methodology is presented to address the issue of modal frequencies' non-normal distribution, which has been neglected in the past studies using the control chart to detect the modal frequencies' abnormality caused by structural damages. The efficiency of the proposed framework is validated through a case study of long-term monitoring data of a long-span suspension bridge. The results show that after elimination of the temperature effects, the selected modal frequencies are not normally distributed, while the Q statistics transferred from the modal frequencies follow the standard normal distribution. The abnormality of modal frequencies can be detected when the data points of the Q statistics exceed the limits of the control chart. Further, the control chart has sufficient sensitivity and thus can be used to detect minor abnormalities of the prototype bridge's modal frequencies. It is concluded that the proposed probabilistic monitoring-based framework offers an effective technique for structural health monitoring of long-span bridges.