A Novel Multiobjective Function for Finite-Element Model Updating of a Long-Span Cable-Stayed Bridge Using In Situ Static and Dynamic Measurements

The traditional objective function (TOF) of bridge model updating is usually defined on the basis of the structural response residuals between numerically calculated and experimentally measured values. However, this model finds it difficult to determine the weighting factors, and the updated model lacks accuracy. To offset these disadvantages, this study proposes a novel objective function (NOF) for finite-element model updating of a long-span cable-stayed bridge using in situ static and dynamic measurements. The proposed NOF comprises the residual functions of the displacement assurance criterion, the position of the neutral axis, and the natural frequency. The global sensitivity analysis-based model updating uses the Kriging model as the surrogate model to save computational costs. The nondominated sorting genetic algorithm II is employed to find the Pareto solutions. The updating results with the TOF defined directly using displacement, stress, and frequency are listed for the purposes of comparison. The results show that the NOF avoids determining the weighting factors and provides more accurate results than the TOF. The updated displacement curves, the position of the neutral axis, and the frequencies obtained from the NOF are all closer to the experimentally measured ones than those obtained from the TOF. After model updating, the displacement's relative errors are around 5% without considering the displacement values under 10 mm. The relative errors of the first four frequencies are all less than 10%. The updated model can serve as the baseline finite-element model of the bridge.