An efficient strategy for multi-objective model updating of a cable-stayed bridge based on Kriging model and global sensitivity analysis

Accurately and efficiently establishing the multi-scale finite element (FE) model of a cable-stayed bridge remains challenging. This study proposes a high-performance multi-objective model updating strategy to improve the accuracy and efficiency in calibrating a cable-stayed bridge's multi-scale FE model. The strategy employs the Kriging model performing global sensitivity analysis and model updating to reduce computational cost, and defines global and local multi-objective functions to improve the model's accuracy. The proposed approach is validated using a high-speed railway cable-stayed bridge. The bridge's global and multi-scale models are established and updated using the field-measured displacement, strain, and frequency. The results demonstrate that the proposed strategy delivers both high accuracy and markedly reduced computational overhead. The defined global and local objective functions avoid the determination of weighting factors and provide accurate updating results. The Kriging model enhances the computational efficiency. Compared to the global FE model, the multi-scale model more accurately predict the critical main girder's behavior. After model updating, the maximum displacement error between the multi-scale model predicted and the experimental measured is under 10%, and that of stress and frequency are under 9% and 6%, respectively.