Damage detection of large-scale steel truss bridge based on Gaussian Bayesian network

To ensure the safety performance of bridge structures during operation, the work on damage detection of steel truss bridges based on Gaussian Bayesian network (GBN) was carried out, and a large-scale steel truss bridge for highway and railway was used as the research background. In this study, a finite element model of the steel truss bridge was first developed, and the coupling vibration analysis of the bridge under the loads of vehicles and trains were performed. The influence of different load factors was also considered. Then, a GBN model was constructed with the moving load as the first-level nodes, bridge deflection as the second-level nodes, and the peak stress of truss elements as the third-level nodes. By performing the finite element analysis, the training data under the different load factors was obtained, and the constructed GBN network was trained to obtain a network model that can accurately represent the complex mapping relationship of nodes at all levels. Combining the trained GBN network model and the damage index based on the peak stress residual of the structural elements, the damage detection of the steel truss bridge under single-point damage, multi-point damage, and different damage levels were studied respectively. The effects of different moving loads on the results of damage detection of the bridge structure were discussed. The investigation results show that the trained GBN network can be accurately conducted for damage detection of the bridge structure, and damage degree of the structure can be further quantified based on the calculated damage index. In addition, by comparing the damage detection results under the vehicle and train loads, it is concluded that the damage detection based on the vehicle loads have a higher accuracy. © 2022, Central South University Press. All rights reserved.