Wavelets for adaptive system identification

New schemes of Wavelets based system identification for structures with multi-coupled modes are developed and presented in this dissertation. System identification techniques have been developed in both time and frequency domains. System identification including modal analysis for different modes requires different resolution of data in the time and frequency domains. However, existing techniques usually use the same resolution data to identify different modes of structures. This is time consuming and may cause errors in the parameters to be identified. The advantages of using orthonormal Wavelets for system identification are that good localization and hierarchical multi-resolution can be achieved in both time and frequency domains. Non-orthogonal Wavelets, obtained from multi-resolution with irregular sampling (adaptive frame), possess the same good localization properties of orthonormal wavelets and also provide more flexibility of basis selection. Consequently, the system model with multiple-coupled modes can be identified (by wavelet based identification techniques) more efficiently and accurately, especially for wideband (transient) and nonstationary signals.; Results from the simulation case studies and modeling of a 10-story reinforced concrete building under earthquake loading show excellent promise for the system identification in the Wavelet domain. Comparisons with conventional identification techniques: ARX (auto-regressive exogenous) in time and FRF (frequency response function) in frequency are made. Results show that wavelets based identification techniques performed much better than ARX and FRF techniques, especially for multiple-coupled mode system with wideband (transient), nonstationary and noise-corrupted input and output.