Research On Noise Reduction And Damage Idetification In Bridge Health Monitoring Based On Signal Theory

Bridge health monitoring system is interfered by noise inevitably because of the complicated environment. A large number of structure behavior information is submerged in noise, which results in relatively low damage identification rate if sampling signals are used directly. Consequently, noise reduction should be carried out firstly in signal analysis, which can eliminate noise influence and highlight structure information furthest, and then the best effect of structure damage identification will be achieved.Three keys of noise reduction in bridge health monitoring which are the feature difference of useful signal and noise, the establishment of noise reduction algorithm and the application of structure damage identification are researched in the thesis. Several signal process methods such as correlation detection, Fourier transform, filter, wavelet transform, Hilbert-Huang transform (HHT) and natural excitation technique (NExT) are used in sampling signal noise reduction and structure damage identification. According to the feature difference of useful signal and noise, reduction algorithms of different types of noise are proposed, structure damage alarming and localization are studied, and relatively integrated bridge damage identification system is established. The main researches include as below:1. Review and summary of research ideas and methods are carried out in bridge health monitoring and related area by reading massive domestic and overseas literatures. Three contents which are the feature difference of useful signal and noise, the establishment of noise reduction method and the application of structure damage identification are proposed in the research on bridge health monitoring noise reduction.2. The feature difference of useful signal and noise in bridge health monitoring system is analyzed and summarized, and the corresponding suggestions are presented. With the concept comparison of noise and measure error in bridge health monitoring, the procedure of noise reduction is proposed and analysis of noise features especially the difference to useful signal is carried out in detail.3. Research on whitening process of non-white noise in bridge health monitoring is proceeded. On the basis of various types of non-white noise features and modern signal theories such as HHT, wavelet transform and Chebyshev filter, industrial noise whitening process based on EMD parameter detection and1/f noise wavelet threshold whitening method based on frequency range are proposed, whose applicability and effectiveness are verified by numerical simulation experiments.4. Research on white noise reduction in bridge health monitoring is carried out. In allusion to the features of bridge health monitoring sampling signal and shortages of traditional algorithms, a new algorithm named EMD wavelet correlation noise reduction algorithm is proposed based on the advantages of EMD, wavelet transform and correlation detection. Simulation experiments demonstrate that the proposed algorithm has weak influence by each parameter, and it is very suitable for noise reduction of sampling signal with low frequency, weak signal to noise ratio and abundant details.5. The effect and rate improvement of structure damage identification on different conditions is proved by the research on structure damage identification using integrated noise reduction algorithm, which is established by combination of non-white noise and white noise reduction method. Noise reduction in bridge health monitoring was carried out by the proposed algorithm and proved by the finite element simulation experiment. Effect and rate comparison of damage identification before and after integrated noise reduction is proceeded using traditional damage alarming algorithm based on wavelet packet energy and proposed damage localization algorithm based on NExT wavelet packet energy, the results of finite element simulation experiment confirm the applicability and effectiveness of integrated noise reduction algorithm in damage identification, and the effect of structure damage alarming and localization using de-noised signals increases significantly.