An investigation into the effects of frequency response function estimators on model updating

Model updating is a very active research field, in which significant effort has been invested in recent years. Model updating methodologies are invari ably successful when used on noise-free simulated data, but tend to be unpr edictable when presented with real experimental data that are-unavoidably-c orrupted with uncorrelated noise content. In the development and validation of model-updating strategies, a random zero-mean Gaussian variable is adde d to simulated test data to tax the updating routines more fully. This pape r proposes a more sophisticated model for experimental measurement noise, a nd this is used in conjunction with several different frequency response fu nction estimators, from the classical H-1 and H-2 to more refined estimator s that purport to be unbiased. Finite-element model case studies, in conjun ction with a genuine experimental test, suggest that the proposed noise mod el is a more realistic representation of experimental noise phenomena. The choice of estimator is shown to have a significant influence on the viabili ty of the FRF sensitivity method. These test cases find that the use of the H-2 estimator for model updating purposes is contraindicated, and that the re is no advantage to be gained by using the sophisticated estimators over the classical H-1 estimator. (C) 1999 Academic Press.