Unsupervised Identification of Arbitrarily-damped Structures using Time-Scale ICA: Experimental Investigations

To identify the dynamic parameters of the structures with sufficient accuracy, a new method is developed and used in this study. In this method, the wavelet-transformed (WT) representation of system responses is conducted on the measured responses, and then the independent component analysis (ICA) is implemented to obtain the modal features. Effectiveness of the proposed method under ap-plied loading condition is shown by applying the white noise, and extracting the simulation results of a multi-degree-of-freedom sys-tem to illustrate the applicability of the proposed methodology for both lightly- and highly-damped structures. In this study, it is deter-mined that continuous wavelet transform (CWT) is indicated to have a better efficiency due to its higher adaptive resolution in time-frequency to be incorporated into independent component analysis compared to other conventional methodologies. The applicability of the proposed method for assessing the natural modal frequencies and mode shapes of the existent structures is investigated by study-ing the IASC-ASCE structural health monitoring benchmark. It is shown that in all the cases the modal properties along with the mod-al assurance criterion (MAC) values are in satisfactory agreement with their exact values and the proposed method is sufficiently ro-bust in accurate extraction of higher modes of vibration.