Updating complex structures by a robust multilevel condensation approach

The method of updating proposed in this paper is based on condensed models and the frequency responses. A robust approach of condensation with two levels is presented. The first reduction basis denoted T01 reduces the size of the model to a few hundred of degrees of freedom. The vectors of the basis are sufficient to make the condensed model reliable and robust with respect to the perturbations of the model in the observed frequency band. The second reduction basis denoted T12 condenses the model to the measurement points. It is updated during the updating procedure. The formulation of the updating problem with residues on the inputs leads to a linear relation of the parameters to be identified. This method makes it possible to detect strong errors of mass and stiffness if the Ritz basist T01 is reliable. In practice, the reduced matrices evolve with the modifications of the model. One can update this basis during updating. A strategy of iterative regularization makes it possible to obtain the variations of the parameters. This method is validated on an industrial example. In order to test the robustness of this strategy, the quality of the readjusted model in comparison with the experimental results for various loading configurations of the structure is evaluated.