A two-stage Bayesian algorithm for finite element model updating by using ambient response data from multiple measurement setups

Abstract

This study presents a two-stage Bayesian finite element model updating procedure by using acceleration response measurements obtained from multiple setups. In the presented methodology, parametric uncertainties for the modal parameters are estimated by using the Bayesian Fast Fourier Transform Approach (BFFTA). Different from the previous Bayesian methods, a block diagonal covariance matrix is modeled for prior estimation of measured modal parameters. In addition, the modelling error in the eigenvalue equations is considered as soft constraints to be updated. Numerical and experimental studies are presented to validate the proposed method. The effect of soft constraints on the identification results as well as their posterior uncertainties are investigated. According to the results, it is shown that the proposed methodology can identify the most probable finite element model parameters with high level of accuracy. In addition, the posterior uncertainties obtained by the proposed procedure are significantly small when compared to the methods that consider rigid constraints for prediction and/or modelling error.