Research On Methods Of Structural Parameter Identification Under Ambient Excitation

The parameter identification is an important content of structure damage detection. It can be divided into determinate excitation method and ambient one according to the source of excitation. The parameter identification based on the ambient excitation is practical, since it doesn't need excitation equipment and causes no damage to structures. On the base of summarization of the correlative research about ambient excitation method at home and abroad, the covariance-driven stochastic subspace identification, the data-driven stochastic subspace identification and the extended kalman filter method are studied in this paper. The main contents are as follows:1. Firstly, the state-space model under discrete time is introduced on the base of structural motion equation. Further more, how to get the modal parameters from system matrices in the model is analyzed. This model is the base of the two subspace methods and the extended kalman filter method.2. The principle of the covariance-driven stochastic subspace identification method and the data-driven stochastic subspace identification method is deduced in detail. The two subspace methods all use the singular value decomposition technique. How to get the modal parameters from the matrices obtained from the singular value decomposition is a key problem, which hasn't been proved strictly by now. This thesis gives the process in a systematic way.3. By numerical experimentation, the identification precision and efficiency of the two subspace methods is analyzed under different noise level. The results show that both subspace methods can eliminate noise effectively. The modal parameters can be identified if any one of the displacement, the velocity and the acceleration is measured, while the identification precision is lower if the acceleration is observed. Compared with the data-driven stochastic subspace identification, the data-driven stochastic subspace identification method has higher precision, but lower efficiency.4. Lastly, the extended kalman filter method is introduced, and the identification precision of the stiffness and damp is analyzed under different noise level.