Updating method of bridge finite element model based on real coded genetic algorithm

To overcome the local convergence and improve the corrective accuracy in the modified iterative optimization process of traditional finite element model, an updating method was proposed by combining the real coded genetic algorithm (RCGA) and measured data of static and dynamic characteristics. The quadrilateral isoparametric element theory and Newton iteration method were used to compile the macro command to realize the fast automatic loading of vehicle loads in the FEM. The objective function was constructed by the static and dynamic characteristics of the finite element model of the structure, the RCGA was taken as the optimization strategy, and the modification frame of the model was established by the MATLAB platform. Through the numerical simulation of a frame structure, the convergence efficiencies and updating results of the proposed optimization method and other methods were compared to verify the effectiveness of the proposed method. To determine the modified parameters, the Latin hypercube sampling method was used to analyze the parametric influence of finite element model on the dynamic responses of the bridge, and the proposed method was applied to modify the solid finite element model of a reconstructed hollow slab bridge. Analysis result shows that the zero order algorithm and the first order algorithm are depended on the sensibilities and correction ranges of the parameters. When the parameters have less sensitivities or the correction ranges are greater than 50%, the correction result of the model is erroneous. The RCGA is insensitive to the initial inputs, so the local convergence can be avoided. The main parameters to be corrected by the sensitivity analysis are the elastic modulus of hollow slab, the elastic modulus of cast-in-situ layer and the longitudinal and transversal restraint stiffnesses of the supports. After correction, the elastic modulus of hollow slab increases by about 19.13%, the elastic modulus of cast-in-situ layer increases by about 16.00%, the lateral restraint stiffness increases by about 46.21%, and the longitudinal restraint stiffness increases by about 72.72%. The static and dynamic characteristics of the modified finite element model are in good agreement with the measured responses, the errors of static responses are less than 4%, and the errors of dynamic responses are less than 3%.