Integration of physics-based model and measurement data for probabilistic condition monitoring of bridges

Abstract

Recently, structural condition monitoring technologies based on sensor data have been actively studied for civil infrastructure. Although these techniques made important contributions, there are still several difficulties resulting from using only sensor data. This study proposes a novel platform for the condition monitoring of bridges. The proposed platform requires three types of input, a finite element model, time-series sensor data, and areliability analysis software package. First, a condition monitoring model is constructed using the Gaussian process regression (GPR), a non-parametric regression method based on Bayesian inference, and a probabilistic model is derived for bridge performance assessment by performing reliability analysis in conjunction with finite element analysis. The proposed platform is applied to an actual bridge in the Republic of Korea, and corresponding analysis results are addressed with the discussion on the applicability of the proposed platform.