Research On Parameter Identification Of The Structural Nonlinear Vibration By Wavelet Transform

Structural parameter identification plays a very important role in structural vibration control, health monitoring and condition assessment. There are lots of studies on parameter identification of linear structures. However, practical structures are nonlinear. Although many researchers have made abundant achievements about the structural parameter identification of nonlinear vibration, a unified theory is not yet formed. Regarding the existing identification algorithms of structural nonlinear vibration there are still some problems need to be solved, especially to verify the proposed algorithm based on experiment. In this paper, the complex Morlet wavelet is used for parameter identification of structural nonlinear vibration. Since there are some problems of complex Morlet wavelet transform on the identification accuracy of nonlinear model parameters under the different cases of edge-effect of wavelet transform and less sampling points, a BP neural network is proposed for the prediction of signals, and identification algorithm was verified by nonlinear vibration experiment. The following is studied in this paper:(1) Aiming at the complex Morlet wavelet transform method which has problems of edge-effect and less sampling points, a method of signals prediction based on BP neural network is presented. Then the principle of parameter identification of structure nonlinear vibration model based on the complex Morlet wavelet transform and the BP neural network is introduced.(2) Numerical simulation of the proposed algorithm based on the two kinds of nonlinear vibration model is conducted. The proposed method of BP neural network is used for the prediction extension of the signal. The complex Morlet wavelet transform is used for the original signal and the predicted signal. At the same time, by adding different levels of noise in the signal, the results show that the method has certain level of noise robustness. Finally, the BP neural network is used for the prediction extension of signals under the case of less sampling points of the two models. As well, the complex Morlet wavelet transform is used for the original signal and the predicted signal to identify the parameters.(3) The proposed algorithm is verified on a structural nonlinear vibration model, in which two rotation dampers are used to simulate the plastic hinge of the structure. By measuring the displacement response of the structure under free vibration, the complex Morlet wavelet transform is used to identify the parameters of the structural nonlinear vibration model under different damping. And then seismic response of the model under different damping cases is measured. The results of numerical simulation are compared with the experimental results which verify the accuracy of the nonlinear vibration model of the steel frame.