Structural Monitoring Data Analysis,Parameter Identification And Dynamic Reliability Prediction Based On Hilbert Transform

Structural monitoring data contains important structural information,such as load effect,performance evolution and safety state,which can provide references for maintenance management and decision-making of structures.Therefore,how to process and analyze the monitoring data efficiently and accurately,as well as extract the deep feature information of the data and apply the information to the bridge safety assessment,are the current research hotspots and cutting-edge areas in the field of civil structural health monitoring.Due to the complex loads and the degradation phenomena of bridges in their long-term service lives,the vibration responses of civil engineering structures are fundamentally non-stationary.Thus,it is of great theoretical value to study the timefrequency analysis method for processing structural non-stationary data.The Hilbert transform-based time-frequency analysis method,a common non-stationary data processing method with the advantage of high time-frequency resolution,has been widely used in engineering structures.However,the Hilbert transform-based method has difficulties in dealing with the response of spectrum overlapping,and it deserves further study on the decomposition of structural dynamic response as well as the demodulation of the monocomponent data.Therefore,the Hilbert transform-based time-frequency analysis theory of structural non-stationary vibration response,and its application in structural parameter identification as well as dynamic reliability prediction are sophisticated studied in this dissertation.The main contents and achievements are presented as follows:(1)The Hilbert transform-based discrete analytical mode decomposition(DAMD)is proposed to decompose the multicomponent non-stationary vibration response of structures.Then,to address the shortcoming of the classic Fourier spectrum in selecting the cut-off frequency,an automatic cut-off frequency selection algorithm based on the auto-regressive power spectrum is proposed.Furthermore,based on the wavelet ridge,the selection of time-varying cut-off frequency and the decomposition of vibration response are realized.The effectiveness of the proposed method is verified by decomposing a simulated dynamic signal with amplitude decays and a non-stationary data with spectrum overlapping.(2)The causes of demodulation error in Hilbert Square Demodulation(HSD)processing are analyzed by the spectrum expansion technology.The error indexes are derived to quantify the demodulation error of the amplitude envelope function and the instantaneous frequency estimation error.Then,two error mitigation techniques,the amplitude redistribution method and the Recursive Hilbert Transform(RHT)method are presented to mitigate the demodulation error.An amplitude modulation data and an amplitude and frequency modulation data with various sampling frequencies and noise levels are used to verify the efficiency of the error quantifying method and mitigation techniques.(3)The applications of DAMD and HSD in structural parameter identification are then studied.Combining DAMD and Stochastic Subspace Identification(SSI),a DAMDSSI method for structural modal parameter identification is proposed,which can identify closely spaced modes.Based on the simulated response of a 36-story frame structure with closely spaced modes and the measured response of a steel-concrete composite girder bridge,the effectiveness of the proposed DAMD-SSI method are validated.Then,a DAMD-HSD method is put forward to identify the instantaneous frequency of nonlinear structural vibration response.By demodulating the simulated response and the measured dynamic response of the nonlinear structures,it is demonstrated that the DAMD-HSD method can process the response of the nonlinear structure.The HSD method combined with the recursive Hilbert transform can improve the identification accuracy of the instantaneous frequency of structural response.(4)Based on Bayesian Hilbert dynamic linear model(BHDLM)and regular-vine Copula method,a dynamic reliability prediction approach using the monitoring stress data for bridges is proposed.First,considering that the bridge stress data is a multicomponent data,the BHDLM is established by using DAMD and HSD.Then,the dynamic prediction of bridge monocomponent stress data is achieved through the recursive probabilistic predictions and modifications of the model parameters.Subsequently,to consider the correlation between data at multiple measurement points,a dimension reduction prediction method of bridge dynamic reliability based on BHDLM and regular-vine Copula is proposed.To validate the effectiveness of the proposed method,the dynamic reliability is predicted based on the simulating vehicle-bridge coupling vibration response of a steel plate composite girder bridge and the monitoring data of an existing bridge.