Investigation Of Signal Filtering And Denoising Algorithms For Deformation Monitoring Of Large-scale Civil Engineering Structures Based On GNSS

The implementation of deformation monitoring of large-scale civil engineering structures(e.g.long-span bridges,super high-rise buildings)can grasp the structural operation state in real time and find the structural abnormalities in time,which has important practical significance for ensuring the safe operation of the structures and disaster prevention and mitigation.Global Navigation Satellite System(GNSS)technology has many advantages and broad prospects for monitoring the deformation of large-scale engineering structures.However,there are still some problems in GNSS structural deformation monitoring.1)GNSS monitoring accuracy is affected by background noise;2)The existing GNSS signal filtering and noise reduction algorithms lack the applicability research for different scenarios and types of structures;3)The currently commonly used GNSS real-time kinematic(RTK)technology has monitoring bottlenecks due to the poor quality of differential signal and the difficulty of deploying base station.In view of the above problems,it is urgent to further improve the GNSS signal filtering and noise reduction algorithm and develop new technology as a supplementary replacement of RTK,so as to promote the application and development of GNSS technology in the field of structural deformation monitoring.This paper has carried out the research of GNSS structural deformation monitoring signal filtering and noise reduction algorithms based on RTK and precise point positioning(PPP).The main work and innovations of this thesis are given as following:(1)The GNSS monitoring accuracy and time-frequency characteristics of background noise are analyzed in order to provide a basis for the application and improvement of GNSS signal filtering and noise reduction algorithms.GNSS-RTK stability tests were carried out in five environments,including concrete ground,grass,water,high-voltage electricity and shaded balconies.Besides,GNSS-PPP reciprocating vibration tests with different amplitudes and frequencies were carried out.The results show that the GNSS-RTK background noise is mainly composed of low-frequency and high-energy multipath errors and internal white noise of the instrument that is evenly distributed in the frequency domain,and the multipath errors are mainly distributed in the frequency domain of 0.04 Hz;The spectral distribution of background noise of GNSS-PPP is same with GNSS-RTK,while the low-frequency noise of GNSS-PPP is larger.(2)The multipath error is the main error of both GNSS-RTK and GNSS-PPP,which has the characteristic of periodic repetition.For the application scenario of GNSS continuous multi-day fixed-point monitoring,the joint algorithm combined complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)with standard deviation modified principal component analysis(PCA)is proposed and referred to as CEEMDAN-s PCA.The proposed algorithm is successfully used to extract the common multipath errors from the simulation signal and the GNSS-RTK monitoring signal.The results show that the CEEMDAN-s PCA can effectively separate the white noise components and correct the amplitude difference between the principal component and the real component.Compared with the single algorithm,CEEMDANs PCA joint algorithm can extract multipath error model more effectively.(3)Aiming at the background noise interference when monitoring the deformation of bridge structures using GNSS-RTK,a joint denoising algorithm combined the improved CEEMDAN with wavelet packet threshold is proposed,referred to as improved CEEMDAN-WPT.The improvement of the algorithm includes:1)The correlation coefficient,effective coefficient and power spectral density(PSD)are used to select the real IMF components;2)The optimal choices of wavelet basis function and decomposition layers for WPT is determined through simulation signals.The improved CEEMDAN-WPT is used to process the GNSS-RTK signal for a bridge structure.The results show that the improved CEEMDAN-WPT can effectively weaken the background noise of GNSS-RTK.The natural vibration fundamental frequency of Haihe Bridge structure extracted from the GNSS-RTK signal is 0.369 Hz,which is consistent with the monitoring results from the accelerometer.The above studies show that GNSS-RTK technology combined with improved CEEMDAN-WPT can be successfully used to extract the dynamic characteristics of bridge structures under environmental excitation.(4)In view of the GNSS-RTK monitoring bottleneck caused by the poor quality of differential signal and difficult deployment of base station,GNSS-PPP is applied to the horizontal dynamic deformation monitoring of super high-rise structures.Aimed at the background noise interference of GNSS-PPP,the improved ButterworthCEEMDAN algorithm is proposed for filtering.Among them,the signal before and after the Butterworth low-pass filtering is aligned based on the maximum correlation coefficient to eliminate the time lag effect.The improved Butterworth-CEEMDAN is used to process the GNSS-PPP signal to obtain the horizontal dynamic deformation components of a super high-rise building.The results indicate that the improved Butterworth-CEEMDAN can effectively weaken the low-frequency noise of GNSSPPP.Moreover,the natural vibration fundamental frequency of Tianjin Tower is 0.1586 Hz,which is basically consistent with the GNSS-RTK,finite element analysis results and past monitoring results.The above researches confirm that GNSS-PPP in combination with improved Butterworth-CEEMDAN can be effectively used to extract the horizontal dynamic deformation of super high-rise buildings.