| In the field of structural engineering, system identification has increasingly become an important tool for research topics such as structural control, health monitoring and damage detection. Compared with traditional frequency-domain identification methods, various time-domain methods have been formulated in recent years. These methods successfully overcome some limits of frequency-domain ones, such as signal leakage.In this thesis, a Kalman filter based time-domain method is discussed and applied to structural damage detection. The main points are as follows.As a commonly used parametric estimation method, Kalman filter estimation, in fact, is a kind of unbiased estimation with the least variance. In order to avoid filter divergence in its applications, three different modified algorithms are put forward. They are analyzed and compared by numerical examples of the same one degree-of -freedom system.In this thesis, the application of this Kalman filter based time-domain approach in physical parameter identification of multi-story plane frame structures is discussed. The difference between shear-type and shear-bend-type frames is pointed out in the discussion. Based on this, structural parameter identifications with output information at all stories and with output information at part of stories are accomplished. In order to ensure its application in structural systems with large numbers of degrees of freedom, subsystem method is proposed to decompose the initial structures into several subsystems to be identified one by one. The computation results have proved the feasibility of the method in numerical examples.To apply the Kalman filter based time-domain approach to structural damage detections, the method is combined with least-squares estimation. As a result, physical parameters of structural systems can be identified without input information. Changes of physical parameters identified can be used to indicate the location and degree of possible damages. |
PDF Full Text Request