Finite element model updating of a long span suspension bridge

Abstract

Errors and uncertainties in numerical models of structures affects the ability of these models to accurately predict the dynamic behaviour. However, model updating techniques can be used to calibrate the models based on experimental data. This paper presents a case study of sensitivity-based model updating applied to the Hardanger Bridge, a long span suspension bridge. Thirteen stiffness and mass parameters are chosen to represent the system uncertainties in a finite element (FE) model. Thirty vibration modes from system identification based on acceleration data is used to calibrate the FE model, using identified natural frequencies and mode shapes as objectives. In the updated model the average error in natural frequencies is reduced from 3.65% to 1.28%. The MAC numbers for the updated modes range from 0.678 to 0.999. The study indicates FE models of large suspension bridges can be significantly improved, but many uncertainties related to modelling simplifications are still present.