Experimental identification of beams with localized nonlinearities

Abstract

The idea underlying time-frequency identification techniques is that, for certain classes of structural response signals, the availability of a limited number of experimental data can be partially mitigated by taking into account the localization in time of the frequency components of the signals. This paper aims to assess the efficacy of time-frequency and time-scale estimators in the identification of weakly nonlinear systems. The example described refers to a beam characterized by a concentrated nonlinearity, whose first mode and related super-harmonics were seen to simulate a Duffing oscillator. A parametric time-frequency identification was conducted under the assumption that the beam’s input/output relationship could be approximated by a certain number of terms of the Volterra series representation, this resulting in a set of diagrams of instantaneous estimators. Though a substantial stability over time was observed only for the estimates associated with linear parameters, the identified model showed a good predictive capacity. Experimental data used in this research come from tests performed on a steel beam tested within the framework of the European research project COST Action F3 on Structural Dynamics.