Effect of limited DOFs and noise in structural updating

In the field of structural updating, several methods have been introduced by many authors. Most of them attempt first to evaluate differences betweeen the numerical model and the actual experimental structure, then to compute a new updated numerical model by the use of the previous results. Those methods are all based on the idea that the dynamic behaviour of a structure is fully represented by dynamic parameters such as eigenvalues, eigenvectors and FRFs. From a theoretical point of view, all of the proposed methods have shown paramount capabilities to update the numerical representation of a structure, with respect to the number of mismodelled regions and with respect to the gap between the numerical and experimental model. Indeed, if the noise and actual experimental degrees of freedom are introduced, a lack of efficiency occours leading to incorrect results and sometimes the experimental identification of the structure, with conseguent numerical updating, is not possible. Furthermore, some methods, when dealing with experimental data, are able to identify structural modifications only if those induce quite big variations of the dynamic parameters while other methods show opposite behaviour. In this paper the effect of limited DOFs available from experimental tests and the effect of noise are both analyzed and results gained from the Minimum Rank Perturbation Method, from Stubbs sensitivity-based method and from Predictor-Corrector method used in FEMTools commercial code for updating are compared. Numerical tests have been performed on structural elements via F.E. analysis.