Derivative-free approach for time and frequency domain finite element model updating

Abstract

Finite element (FE) models are essential for accurately predicting structural behavior under various loading conditions in structural engineering. This research investigates a derivative-free method for updating FE models in both time and frequency domains. The FE model updating problem is formulated as a stochastic dynamic system with embedded parameter-to-data mapping, which allows for the estimation of unknown model parameters. The unscented Kalman method is utilized to solve these systems and update the parameters effectively. This approach also tackles specific challenges in FE model updating, such as constraints implementation and sparsity regularization. Constraints are integrated to ensure that estimated parameters remain within predefined limits. Additionally, sparsity regularization is implemented to enhance interpretability and accuracy, particularly in applications such as damage identification. Numerical investigations validate the proposed approach, confirming its effectiveness and reliability in precisely estimating unknown parameters for structural engineering FE models.