Prediction of random dynamic loads using second-order blind source identification algorithm

Traditional load identification methods are based on the frequency response function matrix. However, in some cases, it is impossible to measure the frequency response functions directly, where only the measured structural dynamic response data are available. In this paper, a novel frequency domain method based on second-order blind source identification (SOBI) algorithm is proposed for identifying the random dynamic loads from some dynamic responses of limited test points. Firstly, the SOBI algorithm is applied to identify the modal parameters from the time histories of the measured displacement responses and then the modal loads are estimated by the identified modal parameters and modal responses in the modal space; finally, the random dynamic loads can be identified in the frequency domain. In order to control the error propagation, the theoretical formulas of the regularization process have been deduced, and the regularization parameters are selected by the generalized cross-validation method. A numerical simulation and an eight-storey spatial frame experimental model are studied to validate the proposed method; the comparison results show a good agreement between the identified random dynamic loads and the actually exerted loads.