Researches On Responce-only Approaches For Identification Of Time-varying Modal Parameters

In the recent years, technology has greatly developed in engineering field. In the field of aerospace, mechanics and civil engineering, engineering structures are becoming high-speed, large, complex and smart, which results in great difficulties of modeling the engineering structures for dynamical analysis. So it is necessary to process the response signals of engineering structures for dynamical analysis as an inverse problem, and mode parameters are the important factors of dynamical properties. So the problem for estimation of modal parameters attracts attentions of researchers and engineeringers all the time. Further, the parameters of structures in operation would change with time flowing or environment changing, such as a great mass loss in rocket lauching and great surface-temperature chang in high-speed missiles, and the effect on the modal parameters cann’t be ignored anymore, thus research on estimation of time-varying modal parameters should be processed.In the view of signal processing, estimation of modal parameters can be grouped in mechanical signal processing. And research achievements in signal processing are often applied to estimate modal parameters in recent years, such as blind source separation (BSS), which attracts wide attentions and is becoming a promising method for estimation of modal parameters. Moreover, estimation of time-varying modal parameters relates to identification of structure systems with time-varying parameters, which corresponds to the identification methods based on subspace tracking in this study. This study reviews the achievements on estimation of modal parameters, and focuses on estimation of time-varying modal parameters. The main works are as follows.1. When the determined BSS is applied to separate mode response signals, the separated mode response signals are not in order and the mode orders are determined by the analysical results of the mode response signals. Introducing the priori knowledge on the sources as constrained conditions, a revised fixed-point method for BSS is obtained using the independent component analysis and is applied to estimate modal parameters. The simulated examples and real-world experiments show the revised method better than the original method;2. The determined BSS used for estimation of modal parameters requires that the number of the active modes is equal (or less than) the number of the observations (the number of sensors). Moreover the signals are non-stationary when the time-varying modal parameters are focused on. So a novel method for identification of modal parameters is proposed by introducing the underdetermined BSS based on sparse component analysis in the time-frequency domain. The simulated examples and real-world experiments show that the proposed method has a good performance on estimation of time-varying modal parameters, a good performance on estimation of time-variant modal parameters too;3. When the identification method for estimation of time-varying modal parameters based on Projection Approximation Subspace Tracking (PAST) is used to process long-length data, recursive least squares (RLS) used in PAST would lose its tracking ability because of data saturation, which would result in the method losing its ability on tracking the modal parameters. By introducing the finite-data window, a revised PAST is proposed and applied in the identification method for estimation of time-varying modal parameters based on subspace tracking. The simulated examples and real-world experiments show the revised method better than the original method;4. By analyzing the algorithm procedure of empirical mode decomposition (EMD) and local mean decomposition (LMD) and the research achievements on the revised EMD and LMD, the relationship between EMD and LMD is revealed in this study. Moreover the potential application of the two methods on estimation of time-varying modal parameters is discussed in this study, too.