Damage Detection for Expansion Joints of a Combined Highway and Railway Bridge Based on Long-Term Monitoring Data

The operation performance of expansion joints is crucial to the driving safety of high-speed trains and the structural integrity of long-span bridges. However, the fact that displacement in expansion joints is influenced by multiple related thermal variables in a nonlinear way poses great challenges in evaluating the performance of expansion joints. In this paper, a method originating from the least squares support vector machine (LSSVM) technique is developed to establish a temperature-displacement model and detect damage in expansion joints. The principal component analysis (PCA) is first introduced to extract inputs for the LSSVM-based temperature-displacement model with the aim of removing correlations among thermal variables. The hybrid movement firefly algorithm (HMFA), which integrates directional movement and nondirectional movement to enhance the global searching ability of the original firefly algorithm, is then proposed to optimize the parameters in the LSSVM-based temperature-displacement model and improve the model accuracy. Finally, the Pauta criterion is adopted to deduce damage thresholds from residual errors between the monitored displacement and the predicted results. The proposed method is verified by data recorded in a sophisticated structural health monitoring system deployed on the Tongling Yangtze River Bridge, which is a combined railway and highway bridge. The results demonstrate that after improvement by PCA and HMFA, the prediction accuracy of the LSSVM-based temperature-displacement model is dramatically improved. The threshold can reliably indicate damage in expansion joints.