Parameter selection for model updating based on the global sensitivity method

Abstract

A new model-updating parameter selection method based on global sensitivity analysis is presented in this work. A specifically designed evaluation function is used for the probability that the sample fits the distribution of test data. In contrast to other parameter selection methods the test-data information is introduced to the parameter selection procedure. Global sensitivity analysis is performed and a set of composite indices for parameter selection is calculated. The parameters are selected based on the values of these composite indices. The method is validated using simulation data from a pin-jointed truss structure model. The cases of independent and correlated parameters are studied and the presented method is shown to be effective for both.