Abstract: Bridge health monitoring system has produced a huge amount of data in long-term service periods, and how to reasonably deal with these data to predict the structural reliability is a key issue in the field of structural health monitoring (SHM). Considering the coupling and randomness of SHM extreme stress data, firstly, the coupled extreme stress signal was decoupled with single moving average method. Then, considering the distribution characteristics of decoupled data, Bayesian dynamic linear models (BDLM) with unknown variance of observation error were established to predict the decoupled extreme stress. Finally, the time-dependent reliability of bridge members was analyzed with first-order second moment (FOSM) reliability method. The research results will provide the theoretical basis for structural reliability prediction.

Key words: bridge engineering, health monitoring, coupling, no-prior distribution information, Bayesian dynamic models, reliability analysis

摘要： 在役桥梁健康监测系统在长期运营中积累了大量数据，如何有效处理这些数据来预测结构的可靠性是结构健康监测(SHM)领域的一个关键问 题。鉴于SHM极值应力信号的耦合性和随机性，首先采用一次移动平均法解耦极值应力耦合信号；然后，针对解耦数据的分布特性，建立观测误差 方差未知的贝叶斯动态线性模型(BDLM)，预测解耦极值应力；最后，结合一次二阶矩(FOSM)可靠度方法进行桥梁构件的时变可靠性分析。研究成 果将为结构的可靠性预测提供理论基础。

关键词: 桥梁工程, 健康监测, 耦合性, 无信息先验分布, 贝叶斯动态模型, 可靠度分析