Sparse Bayesian Identification of Temperature-Displacement Model for Performance Assessment and Early Warning of Bridge Bearings

Abstract

Bearings usually play numerous important functionalities such as deformation regulation, load transfer, and seismic isolation in bridges. A better mastery of their service performance is increasingly desired for bridge owners. In the present study, a novel sparse Bayesian temperature-displacement relationship (TDR) model is proposed to characterize and predict the bearing displacement responses induced by temperature actions in a probabilistic manner, based on the use of long-term structural health monitoring (SHM) data. Compared with the traditional deterministic TDR model, the newly proposed model can deal with two critical problems: (1) most of temperature difference terms barely have effects on bearing displacement responses, leading to the sparsity of model parameters; and (2) uncertainties will inevitably arise from factors such as measurement noise and inherent randomness, resulting in the uncertainty of model parameters. Therefore, it enables to account for the uncertainty associated with the predictions of temperature-induced bearing displacement responses. By combining the probabilistic prediction results with the reliability and anomaly analysis principles, a reliability index is adopted to assess the service performance of bearings subjected to extreme temperature actions. In addition, an anomaly index is defined to determine whether there are performance degradations and then trigger early warnings for the degraded bearings. The long-term SHM data from an in-service long-span railway bridge is employed for effectiveness verifications. The results show that the sparse Bayesian TDR model can achieve effective probabilistic predictions for temperature-induced bearing displacement responses and the reliability and anomaly indices are favorable for bearing performance assessment and early warning.