Fast Bayesian structural damage localization and quantification using high fidelity FE models and CMS techniques

Abstract

Bayesian estimators are proposed for damage identification (localization and quantification) of civil infrastructure using vibration measurements. The actual damage occurring in the structure is predicted by Bayesian model selection and updating of a family of parameterized, high-fidelity, finite element (FE) model classes with the members in the model class family introduced to monitor the large number of potential damage scenarios covering most critical parts of the structure. Asymptotic approximations as well as efficient stochastic simulation techniques are employed for estimating the posterior distribution of the model parameters and multi-dimensional probability integrals arising in the formulation. The proposed Bayesian estimator requires a large number of FE model simulations to be carried out which imposes severe computational limitations on the application of the damage identification technique. Component mode synthesis (CMS) techniques are effectively used to drastically reduce the computational effort. The methodology is illustrated by applying it to damage identification of a bridge using simulated damage scenarios.