Time-varying structural identification for shear-beam buildings using wavelet synthesis

Most structures could exhibit intrinsically time-varying behavior when subjected to strong excitation or when damaged. However, most existing structural identification techniques assume that the structure is linear time-invariant. It is necessary to develop techniques that can quantify the time-varying behavior for structural damage identification, reliability and safety evaluation. In this thesis, time-varying parametric identification of structures using wavelet synthesis is investigated. It is well known that arbitrary transient functions can be effectively and accurately approximated using wavelet multi-resolution approximation (WMRA) due to wavelet's good time-frequency localization property. By representing the time-varying structural parameters with WMRA, the time-varying parameter estimation problem is transformed into a time-invariant nonparametric one.;Three parts for time-varying identification are addressed in the thesis. First, a substructural time-varying parameter estimation of large structures is presented. Civil structures usually contain a large number of degrees of freedom. Using a substructural identification, only responses surrounding the substructure are needed. In this approach, it is assumed that the substructural displacement and velocity responses are noise-free so that the information-theoretic criterion can be used for resolution level selection in each WMRA model. Next, the issue of incomplete response measurements is considered when performing time-varying identification. The displacement and velocity used for estimation are obtained from the fusion of displacement and acceleration through the multi-rate Kalman data fusion technique. The different sampling rates of the displacement and acceleration can be properly treated. The resolution level is empirically determined according to the fitting performance due to the unknown statistical properties as needed by the information-theoretic criterion. Finally, time-varying hysteretic structure is considered for identification using WMRA. With a given Bouc-Wen hysteretic model, the time-varying estimation problem is formulated. The wavelet denoising technique is used to reduce the noise in the measured velocity and acceleration responses. Both smooth and abrupt changes in the parameters can be accurately estimated.