Time-dependent reliability prediction of bridge components with the fusion of BO-LSTM and Copula theory

There exists the nonlinear failure correlation among the multiple monitoring points of bridge components. Considering the influence of this factor on the reliability indices of the bridge, this paper adopts the Bayesian optimized long short-term memory (BO-LSTM) network model in machine learning to dynamically predict the monitoring data of the bridge, and establishes a three-dimensional Gaussian Copula model based on Copula theory to calculate the time-varying reliability indices and failure probability of the bridge construction. The rationality of the model and method is verified by applying the monitoring data of Fumin Bridge in Tianjin.