Research And Application Of Bridge Health Detection Based On Time-frequency Analysis

Bridge is an important part of highway transportation system.However,with the increase of bridge service time,the continuous erosion of natural environment,the rapid increase of traffic volume and traffic load,the bridge has natural accumulation or accidental structural damage in local structures.Once a bridge is damaged in structure and not repaired and maintained in time,it will not only shorten the service life of the bridge,but also affect the safety of driving,and even cause a bridge safety accident when the situation is serious.Therefore,how to accurately complete the bridge health inspection is of great significance.This thesis mainly studies and analyzes the signals obtained by the acceleration sensor sampling.The vibration signal of bridge often presents non-stationary characteristics.In this thesis,Fourier transform,wavelet packet analysis and Wigner Ville distribution are used to study and analyze the health of the bridge.Through analyzing the mechanism of bridge fault signal,it is found that the signal energy is mainly distributed in the high frequency region of time-frequency diagram.Aiming at the measured vibration signals of more than 2000 Bridges,wavelet packet analysis was firstly used to denoise the vibration signals,among which db5 wavelet was used for the wavelet function,the number of decomposition layers was 3,SURE was selected as the threshold selection criterion,and then the denoised signals were processed by Hilbert transformation to obtain their analytical signals.Finally,Fourier transform and WVD time-frequency analysis are carried out for the processed signal,and its spectrum diagram and time-frequency distribution map are analyzed.Through the analysis of the spectrum diagram and time-frequency diagram,the health condition of the bridge is judged.In the experiment,The health condition of the bridge was divided into three conditions:no disease,slight disease and disease.In order to facilitate the calculation,the minor diseases and diseases were classified as one category,while no diseases were classified as another category.Finally,the omission rate and false alarm rate were calculated to be 9.67% and5.19%.Compared with the pure frequency domain analysis method using Fourier transform alone,the accuracy of the method used in this thesis is greatly improved.The experimental results show that the accuracy of this method can meet the engineering needs of bridge detection.