Switching Bayesian dynamic linear model for condition assessment of bridge expansion joints using structural health monitoring data

Age-related deterioration and premature failure have been primary concerns for bridge expansion joints. It is essential to improve the understanding of their operational performance. The existing approaches mainly formulate the deterministic/probabilistic temperature-displacement relationship (TDR) model to assess the structural condition of bridge expansion joints. Nevertheless, it is not easy to guarantee a strong correlation between representative temperature and displacement. Rather than establishing the TDR model, this work uses the displacement response to evaluate the expansion joint condition by combining the Bayesian dynamic linear model (BDLM) with Markov-switching theory. The external temperature effect on the displacement is modeled by the superposition of harmonic components in BDLM. The expectation–maximization (EM) algorithm initialized by the subspace method is employed to optimize the initial parameters. The presented approach is validated through the simulated data, and then it is applied to the expansion joint of a long-span bridge. Results show that EM with the subspace method involves high computational accuracy and efficiency in estimating unknown parameters compared to the Newton-Raphson approach. The switching BDLM successfully identifies the degradation process of expansion joints and offers the transition probability from the normal to other states.