Probabilistic Time-Variant Linear Finite Element Model Updating for Nonlinear Structural Systems

Abstract

This paper presents a method to update linear time-variant finite element (FE) models of civil structures experiencing nonlinear behavior due to earthquake excitation. First, the time-varying modal parameters of the structure are identified using the input–output dynamic data recorded during a damaging seismic event by employing a short-time windowing approach. Then, the identified modal parameters are used to update a linear FE model of the structure using a Bayesian approach. Global sensitivity analysis based on Sobol’ indices is employed to select the most influential parameters for the model updating stage. The evolution of the equivalent stiffness of different elements of the FE model are tracked and their estimation uncertainties are also quantified. The method is verified using numerically simulated data of a two-dimensional nonlinear FE model of a nine-story steel frame.