Hybrid particle prediction of coupling extreme strain data of long-span concrete bridges

It's difficult to accurately predict coupling extreme strain data of long-span concrete bridges. To address this problem, this paper adopts the moving weighted average method and the least square method for decoupling. By separating the segmented coupled signal, the extreme sequences of high/low-frequency extreme strain are obtained through the interval selection method. Then the dynamic linear models with two time-variant state variables of the high/low-frequency extreme strain are established to achieve the high-precision predictions of the corresponding extreme strain by adopting Bayesian particle filter theory with the dynamically monitored and decoupled high/low-frequency extreme strain data, whose addition results in the hybrid particle prediction of the coupling extreme strain. The proposed method is verified by the monitoring strain data of the Xijiang Bridge in Zhaoqin City, Guangdong Province. Compared with ordinary dynamic linear models, the models built in this paper have a good effect on the prediction of coupling extreme strain data. © 2021, Editorial Department of Journal of HEU. All right reserved.