Time-history deflection monitoring of railway bridges subjected to high-speed trains

Abstract

In a structural condition monitoring of railway bridges for high-speed trains, the vertical deflection is one of important indicators because it is associated with the train’s running safety. Specifically, as directly monitoring the time-history deflection can give an intuitive sense to check the structural integrity, this study aims to suggest a new method to monitor the time-history deflection. Here, the Bayesian inference is introduced to construct the probabilistic condition monitoring model for considering various uncertain factors, such as train loads. The proposed method consists of three steps: data synchronization, Bayesian inference, and anomaly scoring, and it can be utilized for detecting unusual patterns in time-series deflection data. It was applied to an operating railway bridge for high-speed trains in South Korea, and an applicability of the proposed method to the structural condition monitoring of railway bridges was discussed. It was found that the unusual structural responses can be detected intuitively with the model constructed by the proposed method.