Reconstruction of Bridge Lateral and Longitudinal Displacements Based on a Corrective Time Decomposition and Splicing Method

Abstract

Bridge displacement is one of the most important parameters for assessing the health of bridges. Current bridge displacement monitoring is mainly based on direct measurement and displacement reconstruction methods. Direct measurement methods can only guarantee accuracy up front, so they cannot be applied over the long term. Displacement reconstruction methods are primarily used for vertical displacement. They are challenging to apply directly to transverse and longitudinal displacement. Therefore, this study proposed a new displacement reconstruction method, corrective time decomposition and splicing (CTDaS), for long-term monitoring of lateral and longitudinal bridge displacements. It utilizes a variety of environmental data and displacement measurements from the early stages of monitoring to reconstruct bridge displacements. The proposed method consists of a time decomposition-splicing networks (Dec-SpcNets) model of displacement reconstruction and output optimization. The Dec-SpcNet extracts the features of the final time step in displacement and improves the accuracy of reconstructing displacement. Furthermore, a sliding weighted average was used to correct the output of the method. The method performance is validated based on the collected data of a continuous girder bridge. The results showed the average errors of 0.22 mm in lateral displacement and 1.85 mm in longitudinal displacement. The proposed method is also compared with the state-of-the-art methods to demonstrate its superiority. Further analysis based on Dec-SpcNet compares the criticality of each factor. The proposed method served as an effective application for monitoring bridge lateral and longitudinal displacements in the long term, which will further contribute to the health condition assessment of the bridge.