Study On Structural Modal Identification Using Blind Source Separation Techniques

With the rapid progress of the city construction, the civil construction engineering has been developed significantly. Simultaneously, some damage occur in the built and building engineering structure due to the effect of design, construction and aging, which leads to a serious threat to the security of people's life and property. Thus, how to identify the health condition of the current engineering structure quickly and accurately has been the difficult subject in engineering field. Based on the summary and analysis of classical modal parameter identification method, this thesis deeply studied the output only modal parameter identification method. This thesis introduced the blind source separation technique of blind signal process field into the modal parameter identification of the engineering structure, proposed an non-parametric time domain method----the modal parameter identification method based on blind source separation technique. The following work and results had been achieved:1. Based on the complex modal theory, it used Hilbert transformation enrich the virtual measuring points, the original vibration system had been expanded effectively. Proposed the improved modal parameter identification method (AMUSE-C) which has less computation task. It leaded to the result that the modal parameter identification method based on the BSS can not only identify the modal parameter of general damping system, but also improve the capacity of identifying the close modes. Furthermore, it can identify the modal frequency and damping ratios directly through extracting the complex frequencies of vibration system. This approach made the computation cost less expensive and the robustness to noise of BSS was extended to the identification of modal frequencies and damping ratios, from which it leaded a totally new way in the application of the BSS in engineering structural modal parameter identification field.2. Based on the improved modal parameter identification method AMUUSE-C developed in this thesis, it introduced the joint time-delay covariance matrix, established the improved modal parameter identification method (SOBI-C) which has high stability, overcame the instability of the identification algorithms caused from the non-positive of the single covariance matrix and improved the robust of the identification algorithms, from which the modal parameters identification based on the BSS could be applied more generally.3. Through the optimal of establishing the improved method SOBI-C, The extended method SOBI-C decomposed the eigenvalue of the joint time-delay covariance matrix so that the main computation cost pressure was transferred from the joint approximate diagonalization of many time-delay covariance matrices to compute the two groups of the joint time-delay covariance matrix. It effectively decreased the cost of computational analysis, which was more valuable and meaningful. The analysis of engineering application showed that the result of the proposed method in this thesis converged to the actual measured value.