PSI - Issue 64

Jiantao Li et al. / Procedia Structural Integrity 64 (2024) 500–506 Author name / Structural Integrity Procedia 00 (2019) 000–000

501

2

1. Introduction In recent years, the construction of railway bridges has shown a vigorous development trend. With the increasing volume of train traffic, the railway bridges are subjected to increasingly large and repeated operational loads. Moreover, the bridge structures are inevitably deteriorated due to the aggressive environmental conditions. Therefore, the implementation of effective structural health monitoring (SHM) is essential to ensure the safe operation of railway Bridges. The traditional online SHM method, using the dynamic response measured by the sensors installed at specific locations of the bridge, has good performance in structural health monitoring and condition assessment (Wang et al., 2024). However, for the large number of bridges, traditional SHM methods are expensive and time consuming for the installation and maintenance of the sensory system on the structures (Yang et al., 2004; Wang et al., 2018). In order to meet the huge monitoring demand on all existing railway bridges, the indirect bridge monitoring using instrumented moving vehicles has attracted much attention (Zhang et al., 2023; Wang et al., 2023). The sensors are installed on the vehicle to work as a moving sensory system to measure the dynamic responses and extract the bridge related dynamic information during its passage over the bridge. This indirect method is a convenient and economic alternative to the direct. The feasibility of using vehicle responses for indirect bridge modal identification has been investigated and verified by many studies (Wang et al., 2022; Li et al., 2022). The responses measured by the vehicle-based moving sensory system also contain the components related to the vehicle dynamic mode and the driving frequency that can mask bridge related dynamic components (Yang and Yang, 2018). In order to improve the feasibility and accuracy of bridge modal identification,Yang et al. (2018) utilized contact-point (CP) response to replace the vehicle response to identify the modal frequency and mode of the bridge. The CP response was defined as the response at the contact point between the vehicle and the bridge, and the vehicle was regarded as an undamped linear single degree of freedom (SDOF) system with known modal frequency. It was verified that CP response could exclude vehicle frequency to enhance the bridge modal identification compared to that directly from vehicle responses. Nayek and Narasimhan (2020) used the dynamic responses measured from a moving 2-DOF test vehicle to extract CP response. Li et al. (2020) used Dual Kalman filter to identify the vehicle states and CP responses from the acceleration responses of vehicle axles. Xue et al. (2020) combined the augmented Kalman filter, Robbins-Monro algorithm (RM), and Rauch-Tung-Striebel (RTS) smoothing to estimate road profile. However, the performance of the above methods for the CP response identification are sensitive to the measurement noise when only the vehicle acceleration responses were used. This study proposed a method by integrating the augmented Kalman filter (AKF) with the Bayesian Expectation Maximization (BEM) strategy to identify the CP displacement responses. The CP responses associated with the road surface roughness and the bridge displacement at the contact points are included as the states variables that can be estimated directly along with the vehicle states. The BEM strategy is adopted to estimate the most probable process and observation noise values of the state space model (Teymouri et al., 2023) to enhance the robustness to measurement noise and improve the accuracy of the identification. 2. Vehicle-bridge interaction model A simplified vehicle-bridge interaction model as shown in Fig. 1 is adopted for the analysis of drive-by bridge model identification using vehicle responses.

Fig. 1. Vehicle-bridge interaction model