APPLICATION OF RESPONSE SURFACE METHODOLOGY IN FINITE ELEMENT MODEL UPDATING

摘要

Response surface methodology (RSM) is an approach that combines mathematical and statistical techniques to approximate multivariate input-output relationships of a physical system. With the aid of the RSM, the effective finite element (FE) model updating analyses becomes a relatively easy task. In this paper, the work focuses on sampling with design of experiments, selecting updating parameters, constructing a quadratic polynomial response surface (RS). The objective functions are built up based on residuals characterizing the natural frequency discrepancies between RS model and the analytical model. The parameters are updated by using a multiobjective optimization technique that is employed to minimize objective functions simultaneously which overcomes the difficulty confronted by single-objective optimization in weighting each objective function. A numerical example of a simply supported beam is given to verify the proposed method.