Improving prediction capability of finite element models of bridges using static and dynamic data

Abstract

Although finite-element models are commonly used in designing civil infrastructure, they may fail to represent the true operational behaviour. Measurements from monitoring and inspection can provide valuable information to improve the prediction capability of models. A significant amount of research has focused on system identification using either dynamic or static measurements separately. However, few research includes the systematic nature of many sources of uncertainties. In this paper, the methodology of error-domain model falsification recently proposed by Goulet and Smith is adopted with emphasis on the combination of both dynamic and static measurements. This methodology is most useful when uncertainties are systematic such as those originating from epistemic sources. Two case studies are presented to demonstrate this methodology. The first involves a simply supported beam in which the static and dynamic responses can be derived analytically. In the second case, a field test on an in-service prestressed concrete highway bridge in Singapore was conducted. The values of deflection and inclination as well as natural frequencies were measured. The results of model falsification using both static and dynamic measurements show higher falsification capacity compared with using only dynamic measurements. It is also shown that updating models using a model-falsification approach is more robust than using a traditional implementation of Bayesian inference.