Finite element model updating in extradosed cable-stayed bridges using the RSM-MPSO method

Accurate finite element (FE) models serve as the foundation for assessing the structural condition of bridges. To establish a high-fidelity FE model capable of precisely characterising the static and dynamic behaviours of an extradosed cable-stayed bridge, this study proposes a collaborative optimisation framework integrating response surface methodology and a modified particle swarm optimisation algorithm, achieving efficient and high-precision model parameter calibration. Through experimental design and sensitivity analysis, critical parameters influencing structural responses were identified, and a response surface surrogate model was developed for iterative computation. A comprehensive objective function was formulated by incorporating field-measured data, and the improved particle swarm optimisation algorithm was employed to iteratively optimise the objective function, thereby obtaining precise FE model parameters. Results demonstrate that the response surface surrogate model, based on sensitivity analysis, achieved correlation coefficients exceeding 0.96 with root mean square error approaching zero. The maximum relative error on the displacement at critical girder cross-sections decreased from 31.40% to 11.62%, and the average relative error reduced from 28.97% to 9.89%. In addition, the mean relative error of the first five natural frequencies decreased from 16.21% to 5.80%, confirming that the updated FE model accurately simulates the actual structural response.