Multi-objective based updating of finite element model of bridge using modal properties

Abstract

Construction of new bridges steadily increases, as the national road networks are continuously expanding. Considering the traffic volume and their weights are getting increasingly larger, The bridge performance evaluation is getting more important. Bridge structural health monitoring (SHM) is concerning performance and safety of bridges. Reliability of the Finite element model for performance predictions gets deteriorated, because of the discrepancy between the analytical model and the real structural condition. The way to reduce this discrepancy to the model updating is to correct the geometrical and physical parameters and boundary conditions. By updating, the FE model can predict responses of the structural more accurately. The success of model updating depends on the optimization methods, experimental data, and selection of parameters for updating. In this study, FE model updating has been carried out on Palgok-3 Bridge using the modal information obtained from the acceleration measurement data. The modal properties identified from the acceleration measurement data have been found to be fairly far from those obtained the initial FE model. It might be because the initial FE model does not properly reflect the current structural condition. FE model updating has been carried out using parametric optimization. The study has been concentrated on optimization methods, grouping of updating parameters, and objective function. Two optimization schemes are used: Nelder-Mead downhill simplex method as a point-to-point approach and genetic algorithm as a population-based approach. GA (population-based) gives the better solution than Nelder-Mead method (point-to-point). The main drawback of point-to-point approach is being trapped in the local optimal solution. Nelder-Mead is very sensitive to the initial values, while GA is not sensitive to the initial values. GA requires relatively more computational effort than Nelder-Mead method. Two parameter grouping methods are...