Identification and updating of uncertain dynamic models of a flat beam for nonlinear forced response predictions

This paper addresses the validation of uncertain dynamic models of flat beam test articles to be used for forced nonlinear geometric response prediction. Following a recent analysis, it is recognized that the test articles will be subjected to a significant uncertain preload induced by the clamping process but this preload cannot account fully for the variations in measured natural frequencies. In this light, the focus of the present effort is to assess whether additional, globally distributed uncertainty also needs to be introduced or whether updating of the reduced order model needs to be carried out. Both options are investigated here using a maximum likelihood framework with the preload modeled as a Gaussian random variable with unknown mean and standard deviation while the global uncertainty is introduced following the stochastic nonparametric modeling approach on the reduced order stiffness matrix. The small number of experimental data samples prevents a clear choice between the two options but favors the updating of the reduced order model.