Bridge Dynamic Deflection WSST Denoising And Instantaneous Total Energy Integral Abnormal Detection Using GB-SAR Data

Bridges play an important part of public transport infrastructure,and their operation status monitoring is the key to ensuring urban traffic safety.Ground-based Synthetic Aperture Radar(GB-SAR)has been widely used for non-contact vibration and displacement monitoring of various bridges due to the advantages of non-contact measurement,wide frequency response range,high displacement sensitivity,and simple setup.Dynamic deflection is an important indicator for judging the stiffness and stability of bridges.The dynamic deflection collected by GB-SAR can be used to obtain the position and trend of bridge abnormal through TimeFrequency Analysis(TFA)methods.However,in the process of dynamic deflection acquisition,the instrument error,the surface reflectivity of the measured object and the surrounding environment will all affect the progress,which makes the dynamic deflection data contain noise information,which needs to be denoised.However,traditional noise reduction methods cannot satisfy the characteristics of low initial signal-to-noise ratio and short-term high-frequency variation of GB-SAR dynamic deflection.After noise reduction,there will be large effective information loss of dynamic deflection occurred.In bridge abnormal identification,the previous multi-period comparative measurement cannot effectively meet the engineering needs.At the same time,the general TFA energy spectrum discrimination method cannot accurately obtain the projection of time-varying energy in the fixed frequency spectrum.To solve the above problems and obtain more accurate bridge abnormal information through a single measurement,this paper expands from two perspectives in signal noise reduction and bridge abnormal detection:(1)Systematically introduces the theoretical basis and decomposition methods of the mode decomposition method,and explains the Empirical Mode Decomposition(EMD),ExtremePoint Symmetric Mode Decomposition Method(ESMD)methods,The basic principles of Mutual Information Entropy(MIE)algorithm and Wavelet-based Synchro-squeezing Transform(WSST)method are discussed in detail.Discusses the ESMD method for the definition of the total energy of the signal.(2)Aiming at the problem that GB-SAR dynamic deflection data is susceptible to noise caused by environmental factors,this paper proposes an improved ESMD-WSST highfrequency denoising algorithm for noise removal of original data.By introducing ESMD decomposition and mutual information entropy extraction,frequency classification is performed on the original dynamic deflection signal,and WSST denoising progress is performed on the high-frequency part of the signal,which solves the problem that the original WSST denoising method is not sensitive to the identification of high-frequency fluctuations in a short time window.The improved ESMD-WSST high-frequency denoising method provide effective data support for bridge mutation identification.(3)To solve the problem that ESMD time-varying energy cannot be directly projected on the fixed frequency spectrum,this paper constructs a bridge abnormally detection model of ESMD energy integration,calculates the instantaneous total energy of the dynamic deflection signal through the kinetic energy formula,and obtains the function of energy with respect to time.Using the relationship between bridge variation and bridge energy accumulation,the energy-time function is integrated and calculated through the energy integral model,and the bridge variation position and variation degree(trend)are judged according to the integral size.(4)The ESMD-WSST high-frequency denoising method proposed in this paper and the bridge abnormally identification model of ESMD instantaneous energy integration were verified by experiments.Through the analog signal experiment with a signal to noise ratio(SNR)of 5d B,compared with the ESMD,WSST,EMD-WSST noise reduction methods,the SNR increased by 11.5%,56.5%,and 13.4%.The root mean square error(RMSE)were raised respectively.By comparing the results of the actual measurement experiment of Wanning Bridge in Beijing with the results of the other three methods,it is further verified that the ESMD-WSST high-frequency noise reduction method proposed in this paper has a better noise reduction effect on the actual acquisition of GB-SAR dynamic deflection.The denoised signal is smoother and has a higher noise rejection ratio(NRR).The NRR is 5.48 d B,and ratio of the variance root(RVR)is 0.0501 mm.Through the variation identification experiment of Beishatan Bridge in Beijing,the ESMD instantaneous energy integral abnormal identification results proposed in this paper were compared with the bridge crack positions obtained earlier,and the effectiveness of the abnormal identification model proposed in this paper was preliminarily verified.Comparing this result with the Digital Surface Model(DSM)of the lower surface of the bridge constructed by 3D laser of Beishatan Bridge in the same period,it was found that the potential variation area of the left bridge was consistent with the DSM model.The existing crack position and settlement area of the right bridge also coincide with the DSM model,which further verifies the effectiveness of the variation identification model in this paper.