Bayesian-based Finite Element Model Updating, Damage Detection, and Uncertainty Quantification for Cable-stayed Bridges

Abstract

Long-span bridges are important components of civil infrastructure systems because they are vital links in transportation systems. Therefore, as bridge systems age, understanding the safety and serviceability performance of structural components of these systems through structural health monitoring (SHM) techniques is necessary to achieve economically sustainable maintenance. Application of Bayesian inference in SHM techniques provides a reliable platform to deal with different sources of uncertainty in the process and also to obtain probabilistic results which are more meaningful for decision-making. This research seeks to address some of the key challenges in SHM of large-scale civil infrastructures such as analyzing a huge quantity of measured data for system identification, dealing with uncertainty in measurements and analytical models of structures, performing a real-world application of Bayesian Finite Element (FE) model updating, and Bayesian-based damage detection. The proposed research focuses on the following tasks: 1) development of an autonomous data pre-processing and system identification to analyze a large amount of response measurements, and extraction of statistical features of dynamic properties of a large-scale cable-stayed bridge, 2) recommendation of an effective way to systematically deal with different sources of uncertainty in Bayesian FE model updating, and implementation of a real-world application of Bayesian FE model updating on a large-scale bridge to achieve a more accurate FE model for response predictions, and finally 3) proposing a new Bayesian-based structural damage identification technique applicable for bridge structures based on the measurements of their healthy and unhealthy states.