Vibration-based Bayesian model updating of an actual steel truss bridge subjected to incremental damage

Abstract

As the finite element (FE) model has become increasingly important in engineering, the model updating method has also received much attention as a means of improving the structural FE model accuracy. However, the objectives of many earlier studies have been laboratory experiments and numerical simulations. Actual structures have not been adequately investigated. This report describes work on a vibration-based Bayesian model updating for an actual steel truss bridge. The transitional Markov chain Monte Carlo (TMCMC) sampling method was used to estimate the posterior distribution. A field experiment assessing this target bridge was conducted under five damage scenarios. Then a fast Bayesian FFT method was used to identify the modal properties. Based on sensitivity analysis, three model classes were proposed for the model updating process of the target bridge. In all cases, the updated modal properties were found to fit well with the experimental data, whereas the updated model parameters cannot. The case with more prior information identified the structural damage. Results demonstrate that making full use of prior information can improve the model updating accuracy. Feasibility of damage detection was observed for an actual steel truss bridge based on the FE model updating method.