Assessment on the crash risk factors of a typical long-span bridge using oversampling-based classification method and considering bridge structure movement

In comparison to ordinary highways, traffic accidents on long-span bridges have unique characteristics due to large percentages of large trucks, inclement weather conditions, and dynamically moving bridge structures. Presently, there is a research significant gap in the existing literature about the traffic safety of long-span bridges due to those difficulties. Meanwhile, the structure movement data of bridge is difficult to obtain. In this paper, real-time data related to the bridge crashes including surrounding environment, traffic status and especially structural movement were obtained from monitoring system of a long-span bridge. An oversampling-based classification method was utilized to explore the risk factors of the long-span bridge crashes. The results indicate that higher maximum wind speed and volume prior to a crash tend to increase the likelihood of the occurrences of the crash, while higher temperature, humidity, average vehicle speed and truck percentage are found to decrease the likelihood. Moreover, the structure movement indicators including horizontal vibration acceleration and deformation are found to have significant adverse effects on the traffic safety of the long-span bridge, and we recommend that those factors should be considered at the design stage.