Structural finite element model updating based on response surface optimization

In order to solve the low accuracy problem in structural finite element analysis, a model updating method of surface response optimization was proposed. The sample points acquired by optimal Latin hypercube experimental design and the F-test evaluation utilized for parameter selection were used to construct the polynomial response surface model, and then the validity of response surface was tested. The residual error of the results between response surface and experiment was taken as the objective function. The genetic algorithm was employed to optimize the updating parameters. The updated model was obtained by substituting updated parameters into initial finite element model (FEM). The ability of the model updating methodology was proved under three working conditions, accurate parameters, no parameters and inaccurate parameters by the three degree-of-freedom numerical model. The model of GARTEUR was taken as an example. Significant parameters were selected out to implement the model updating. The results show that it not only proves the reappearance and prediction ability of the updated FEM, but also shows the ability to predict the modal frequency of locally modified structures, and verifies the validity of the model updating methodology.