Research On Time-varying Parameter Identification Based On Waveletanalysis For Engineering Structure

With the rapid development of modern engineering technology, complicated structure forms are constantly emerging. Time-varying and non-linear characteristics of engineering structures should not be ignored under different external loads, working environments and conditions, etc. For example, for the structural system composed of the stadium with a large-span spatial structure and the bleacherite, its mass has time-varying characteristics with the flow of crowd, forming a typical time-varying structural system; super high-rise buildings show the dynamic time-varying characteristics when subject to strong wind load; for isolation buildings widely used in civil engineering field, the particularity of their structure forms make them show obvious time-varying and non-linear characteristics throughout the whole life cycle from construction to use. Thus, it puts forward not only new requirements for design theory and construction technology, but also higher requirements for structural health monitoring technology. As an important component of structural health monitoring study, the parameter identification of time-varying structural systems and the non-linear structural systems has become a hot topic in the field of structural dynamics research and holds a significant position for its research value. This paper aims at researching on time-varying structural systems and non-linear structural systems, and uses wavelet theory to conduct in-depth study on their parameter identification. The main research work of the dissertation is as follows:In a short time-invariant condition, a time-varying modal parameter identification method based on continuous wavelet transform is proposed for solve the identification of time-varying structural modal parameter. A series of random impact excitation is defined as the input excitation, wavelet transform is made only with output response, and the frequency, damping and mode shape at each moment can be acquired by extracting the wavelet ridge. Analysis shows that the method is suitable for slow time-varying structural systems with large damping. A time-varying structure dynamic test with a two-layer steel framework and variable mass is designed and set up to verify its validity.Based on the classical FRF, the concept of wavelet frequency response function(W-FRF) is introduced using wavelet time-frequency analysis technique, expanding the time-invariant frequency response function to time-varying wavelet frequency response function. The characteristics analysis of structural systems based on W-FRF has been proven to accurately reveal the dynamic characteristics of the system. On this basis, a new physical parameter identification method for linear time-varying structural systems is proposed by combining W-FRF with least-squares iteration method. The analysis results show that the proposed method has good identification capacity for time-varying frequency and mode shape. The influence factors of W-FRF estimation precision are analyzed.For the time-varying structure physical parameter identification, a time-varying physical parameter identification method based on WMRA is established. The time-varying parameters were expressed approximately using wavelet multi-resolution analysis from the perspective of the function space subdivision, and its identification can be transformed into time-invariant wavelet reconstruction coefficient estimation in multiple linear regression model. This part aims at further intensive study on robustness of the algorithm, such as noise resistance, decomposition scale, model selection, basis function selection and sampling frequency, etc. The selection of decomposition scale can be considered as model order and contribution of regression factor, and the accuracy can be improved effectively through the combination of orthogonal forward regression and regularized least squares algorithm(OFR-RLS). The frequency-domain characteristics of time-varying parameters to be identified have great influence on the selection of decomposition scale. Where conditions allow, the parameter identification accuracy can be improved by increasing sampling frequency.The Bouc-wen model used to describe non-linear structural systems is studied systematically. A single-parameter proportional change method is adopted to analyze the sensitivity, deriving the general law of the influence of parameters on hysteretic curve. With the classical Bouc-wen as the object of study, different parameter values show the soft or hard characteristics. The relationship between the maximum amplitude of steady-state response and frequency of input excitation of the system under the harmonic excitation is analyzed according to research idea of linear system dynamic characteristics. The harmonic distortion and FRF distortion methods are adopted to judge the non-linear state of the model.The paper takes the typical base isolation structures with time-varying and non-linear characteristics as the object of study. A wavelet coefficient energy ratio(WCER) indicator is proposed to judge the non-linear state of isolation bearings qualitatively. Then, the identification method with/without model is established based on wavelet multi-resolution analysis theory and a three-order valid difference method, which can effectively identify the parameters and non-linear characteristics of the upper structures and isolation layers. Finally, the shaking table model test of series isolation system is used for the experimental validation. The time-varying characteristics of series isolation systems are revealed. The proposed WCER indicator is used to judge the non-linear state of the series isolation systems. The hysteretic non-linear model of series isolation systems is established based on shaking table test data. Its application scope is given and validity is verified.