Research On Identification Of Structure Parameteres By Wavelet Analysis

Structural parameter identification techniques are widely applied in civil engineering, most of which are in frequency domain or in time domain. Engineering structures are usually excited by non-stationary ambient loads, the ratio of signal to noise is low in dynamic detection. Most of structures in civil engineering are nonlinear. Once a structural element is damaged, the stiffness of that element is reduced. Hence the structure damage may be reflected by the changes of parametric values of the damaged element. For non-stationary signals and responses of nonlinear or time-variant systems, the local analysis in time and frequency simultaneously may not be satisfied by the techniques in frequency domain or in time domain. According to the consideration mentioned above, the thesis focuses on the research of structural parameter identification by wavelet analysis in order to reduce the influence of measurement noise to identification and recognize nonlinear or time-varying systems:1 An efficient least squares estimation with frequency band weighting is proposed for parameter identification of linear system. Based on wavelet multi-resolution analysis theory, differential equation of linear structural dynamic system are decomposed and dynamic time series are described on different scales. The loads and structural responses are transformed by orthogonal wavelet and the parameter identification equation is gained by wavelet coefficients on different frequency bands. The physical parameters are obtained by weighting least-square method and the weighing values are derived by the ratios of noise to signal on different frequency band. The identification accuracy is greatly improved by the proposed method.2 An efficient multi-scale parameter Kalman filter method is proposed for identification of the physical parameters. The measurement data is decomposed by orthogonal wavelet function and the state and measurement equations can be resolved into different scales. The physical parameters of dynamic system are estimated by parameter Kalman filter method in different scales. The theory analysis and numerical simulations indicate that the identification of system parameters in multi-scale is more efficient than in single scale.3 An efficient multi-scale nonlinear least-square method is proposed for identification of the physical parameters. The condensed story model of frame structure is gained by static agglomeration and the complete output information is obtained by integral operator. The elastic modulus of each story is identified by nonlinear least square method in multi-scale. The numerical simulations show that more information is gained by identification in multi-scale than in single-scale and the identification accuracy is improved. A four-story frame structure, with one span in x-coordinate and two-span in y-coordinate, is excited by white noise on top floor and the response data on every floor is analyzed. The free vibration signal can be get by the random decrement technique and the vibration frequency and damping ratios are detected. The elastic modulus of each story of the concrete frame model was identified by the multi-scale nonlinear least square method.4 The identification of weak nonlinear system by Morlet wavelet transform is studied. The free vibration response of weak nonlinear system is analyzed by Morlet transform. The instantaneous frequency and amplitude are recognized by the scale and the wavelet coefficient of the ridge, so the inherence frequency, damping ratios and nonlinear coefficient are estimated. The instantaneous frequency identified by short time Fourier transform, Hilbert transform and wavelet transform are compared. The vibration equation of simple supported reinforced concrete beam is deduced considering the material nonlinearity. It is testified that free small-amplitude bending vibration of reinforced concrete beam contains weak square nonlinear term by theory deduction and a series of impact excitation vibration tests on a reinforced concrete beam under different damage degrees. The inherence frequency, damping ratios and nonlinear parameter of the system are identified by compound Morlet wavelet transform. The tests also demonstrate that the inherence frequency decreases with damage while the nonlinear parameter and damping ratios increase with damage.5 The identification of time-variant modal parameters by modified L-P wavelet transformation is studied. The floor accelerations of a shearing structure which is damaged subjected to earthquake ground motion are transformed by modified L-P wavelet. The bands of frequencies in which the natural frequency lie is determined by the energy corresponding to each band, so weather the structure is damaged and the time when it is damaged can be detected. The mode shape is estimated by the wavelet coefficients of each floor at an instant time corresponding to the band containing the natural frequency. The damage location can be detected by the change of relative displacement of each floor calculated by the mode shape from the undamaged structure to the damaged one.The recorded acceleration of a reinforced concrete building observed by California strong-motion instrumentation program is analyzed by modified L-P wavelet. It is show that damage can be detected by the wavelet coefficients without the model information, the excitation information. This method is also efficient when there is noise in recorded response and it can be used for on-line detection.6 The identification of time-varying physical parameters by wavelet function is studied. By expanding the time-varying stiffness and damping into linear combination of a series scaling functions by Mallat algorithm, the time-varying parameters identification problems is turned into the problem of identifying invariant coefficients, which can be identified by least-square method when the input, output of the structure and the scaling function are known. The time-varying parameters are then reconstructed by the coefficients. The theory analysis and numerical simulations show that the time-varying parameters of shearing structure are identified accurately. The influence of damping to the method is analyzed. The Tikhonov regularization method can obtain fairly accurate result when there is noise.