Bayesian Dynamic Coupled Linear Prediction of Bridge Extreme Stresses

Abstract:

Based on the monitoring extreme stress time series data with trend and randomness, the Bayesian dynamic coupled linear prediction for bridge extreme stresses is studied in this paper. Firstly, with the decoupled historical extreme stress data, the dynamic coupled linear model(DCLM) and the corresponding finite dynamic linear models(DLMs） are respectively built. Then, based on the dynamic monitoring decoupled extreme stress data, the probability recursive processes of the built dynamic coupled linear model are proposed with Bayesian method, which can be used to dynamically predict bridge decoupled extreme stresses. Further, the sums of the predicted decoupled extreme stresses can be considered as the predicted results of the bridge extreme stresses. Finally, based on the monitoring extreme stress data with trend and randomness from an existing bridge, the feasibility and application of the proposed method is verified through comparing with auto-regressive integrated moving average (ARIMA) model.