Ambient Exciting Modal Parameter Identification Based On Modern Time-frequency Analysis

The modal parameters identification of structure has an important sense to applied huge engineering structure such as health monitoring, structure damage estimate, dynamical character analysis of structure, etc. It is difficult to exert artificial excitation and difficult to measure exciting signal effectively on some large engineering structure because of their huge volume, mixed material, complex bind conditions, and so on. So it is not easy to apply traditional modal parameters identification technologies to test input and output signal of these types of structure. Having a series of advantages such as not needing manual excitation, just requiring analysis output data with mechanics and mathematics theory, identifying structure modal parameters in time domain or frequency domain, the structural modal parameters identification method based on ambient excitation (for instance, wind load, earth and automobile excitation) is widely applied to engineering. Meanwhile, the large structure usually show the character of low-frequency close coupling modal,however , identification accuracy of present method is not high enough. Based on the modern testing modal analysis theory and ambient vibration technology,using modern time-frequency analysis technology, this paper emphasizes on identification of low frequency close coupling modal by employing continuous wavelet transform-based and Hilbert-Huang transform-based modal parameter identification method. The main research work and conclusions are as follows:â‘ The theory and method of CWT (continuous wavelet transform)-based modal parameters identification is discussed in detail, then a method of adaptive modal parameters identification based on complex Morlet wavelet is proposed. The CWT has characteristics of infinite frequency resolution in theory, the automatic decoupling of the low-frequency close modes usually occurred in engineering can be realized easily, and the logarithm curve of the wavelet transform coefficients is a line. The modal parameters of the structure can be obtained easily by curve fitting of the logarithmic curve, the principle is simple and easy to exploit. Choosing the complex Morlet wavelet as mother wavelet, designing the the optimism wavelet base suitable for signal character with the wavelet energy entropy, parameters of close coupling modal can be effective identified, using the optimum decomposition scale extracting method to get wavelet decomposing coefficient which most suitable for analyzing. The endpoint data continuation approach base on SVM (Support Vector Machines) is presented to effectively restrain the edge effects of the CWT which occurred when modifying factor is taken big value, in order to improve the parameter identification precision of the CWT, as a result, the identification accuracy of modal parameters is improved.â‘¡The theory and method of HHT(Hilbert-Huang Transform)-based modal parameters identification is discussed in detail. Making use of the bi-filter characteristics of EMD (Empirical Mode Decomposition), a method of BR-EMD(Bandwidth Restrict Empirical Mode Decomposition) is proposed to overcome the shortcoming of the tranditional EMD such as modal incompletely decoupling , modal mixture and low identification precision. This method would not break the adaptability of EMD, and also don't like Schmidt orthogonalization which considers the first IMF must be orthogonal. The method would improve the ability of identifying close modes by restraining the model combination effectively and improving the frequency resolution of EMD by employing bandwidth restrict signal before every IMF is finally decomposed. The numerical simulation and successfully application on Chaotianmen bridge verified that this method can improve the precision of modal parameters identification effectively.â‘¢Stratified sampling technology-based modal parameter identification from short data sequence is discussed. In condition of short data sequence, satisfied random decrement response signal is difficult to extract easily due to the lack of average times of RDT (Random Decrement Technology). Combining with improved RDT, the identification result can be obtained by applying stratified sampling technology which has the ability to improve precious of sampling and total average times. A novel method was supposed to determine the weight of the layer based on fitting error and sample volume, for the number of trigger points of improved random decrement technology can't be beforehand known, the common weight of the layer determination method is not fit for this situation. Combining with optimism complex Morlet wavelet and BR-EMD, this method can effectively identify structure modal parameters under the condition of short data sequents.â‘£A testing modal analysis system is developed successfully, including identification methods for modal parameters based on modern time-frequency analysis. It contains designs of structure model module, modal shaping module, data acquisition module, data analysis and process module, parameters identification module, results checking and normalizing module. Parameters identification module include frequency-domain method, time-domain method and time-frequency-domain method. Applied the three type of method to engineering applications, the experimental results demonstrate that the structure testing modal analysis system designed in this thesis has the value of engineering application.At the end of the thesis, the summarization of the article and expectation of the feature extraction technology development are presented.