Identification Of Structural Modal Parameters Based On Wavelet Transformation

Nowadays, more and more attention is focused on structural damage detection and healthy assessment with nondustructive test. The structural modal parameters, which reflect its intrinsic characteristics, are the critical factors in the vibration-based damage detection and healthy assessment. Due to the huge volume of civil engineering structure, the vibration excitation is a difficulty; the modal identification based on environmental excitation is available.Firstly, the basic theory of modal analysis, stochastic process and discrete signal system is introduced in this thesis, which includes: the process to derivate frequency response function, the relation between frequency response function with power spectrum and correlation coefficient, the time-frequency transform and frequency mixing.Then, the basic application of Fourier transformation (FT), Short-Time Fourier transformation (STFT) and Wavelet transformation in modal identification is presented. The basic principle of how to acquire modal parameters and denoise original signals with wavelet method is discussed. Besides, the software aiming at identify the modal parameters of large civil engineering structures based on Matlab platform is accomplished, its critical technology is presented in this thesis, including the method of modal parameters identification, denoise dispose, counterpoise and filter treatment.Finally, the three methods mentioned above are applied into three projects. From the results, it is concluded that, STFT and Wavelet methods can achieve a better identification precision in practical structures application, and the wavelet method can realize a desirable precision both in time and frequency domain because of its variable windows. However, the selection of the wavelet function is significant, which influences the analysis effect and efficiency. The appropriate wavelet function can be chosen by theoretical analysis and repeated attempt. The wavelet method can denoise the original signals effectively, i.e., restrain the high-frequency components, and extrude the low-frequency components, which provide more useful information. The method based on level threshold value can attain a far better denoise outcome than on glabal threshold value, but losing some information of original signals.