A Study Of Key Technology For Dynamic Deformation Monitoring Using GNSS

With the progress of human society and the development economic level, construction scalehas been continuously enlarged in recent years. So, Structural health monitoring of large scaleengineering construction became one of the important issues concerned by engineers. This thesisis focused on the crucial questions that how to improve the accuracy and the continuity of GlobalNavigation Satellite System (GNSS) in dynamic deformation monitoring. To this end, theinstantaneous ambiguity resolution method, systematic error especially multipath modeling andcorrection, de-noising method of dynamic measurement are discussed systemically in thisdissertation. The main works and contributions are summarized as follows:1. The observation model, stochastic model and error model for kinematic relativepositioning using GNSS on short baseline are summarized, and then, the influence of residualtropospheric delay on kinematic positioning accuracy is studied. The experimental resultindicates that residual tropospheric delay will affect the result of baseline significantly when theheight difference between base and rover stations is too large, for vertical component theinfluence may reach a level of several centimeters. The residual zenith tropospheric delayestimation method is effective to improve the accuracy of baseline.2. The influence of priori coordinate error, residual tropospheric delay and multipath effecton real ambiguity is analyzed in detail, and then the partial search method based on prioricoordinate constraint is put forward to resolve integer ambiguity instantaneously. This method issuitable for single-frequency receiver and gives attention to efficiency and reliability. Theoreticalanalysis shows that the tolerance of priori coordinate error for this method is5cm. Experimentalresults illustrate the effectiveness.3. An instantaneous ambiguity resolution method suitable for dual-frequency receiver isestablished for high rate deformation. This method involves two steps. Firstly, two combinationobservations with longer wavelength are selected to reduce the ambiguity search space, and then,the inner correlation between dual carrier phase data, L1and L2, is used to determinate thecorrect integer ambiguity. Theoretical analysis shows that the tolerance of priori coordinatecomponent error for this method is exceeded to30cm. To decrease the influence of errors such asmultipath, iterative search based on robust adaptive Kalman filter is adopted in the algorithm,and the critical value of weighting factor in robust estimation is studied.4. The character of multipath was analyzed, and then the effect of several stochastic modelson mitigating multipath error is compared. The experimental data collected on a fixed baseline in two consecutive days demonstrated the repeatability of multipath either in coordinate space or inobservational space. The true repeat time of multipath was studied for high-rate GNSSmonitoring system using the GPS broadcast ephemeris from IGS. The results show that therepeat time varies for each satellite and differs significantly from sidereal period. The meanrepeat time is about10s less than the sidereal.5. The principle and drawbacks of Empirical Mode Decomposition (EMD) do-noisingmethod is discussed, and then a modified Ensemble EMD (EEMD) method is introduced. Boththeoretical analysis and experimental data demonstrate that multipath observable is a sort ofcomplex and intermittent signal and EEMD is more suitable to de-noise and extract multipathmodel than EMD.6. A scale threshold de-noising method is put forward based on the characteristics ofIntrinsic Mode Function of white noise decomposed by EEMD. With simulated data andobserved data from a settlement monitoring campaign, EEMD forced de-noising, EEMDthreshold de-noising and several kinds of wavelet de-noising methods are compared, and theresults show that EEMD threshold de-noising method perform better than the forced method dueto some significant details within high frequency band are reserved. Compare to wavelet method,EEMD is adaptive and free from the choice of wavelet base and the determination of the numberof decomposition order.7. The principle of modeling multipath error based on single difference residuals used forhigh-rate GNSS structural deformation monitoring is analyzed. Using real GPS observable, it isindicated that the precondition of this method is satisfied approximately on short baselines.Multipath models are extracted using EEMD soft threshold de-noising method on the referenceday collected on a fixed baseline both in coordinate and single difference domain respectively,and are used to remove multipath errors on subsequent days. The results show that both incoordinate and single difference domain, the accuracy of GNSS positioning is improved,especially for the vertical component. Due to the fact that each satellite has a different multipathrepeat period, and in sidereal filter using single difference residuals, the multipath model of eachsatellite is shifted according to its true repeat period, so, more precise results are gotten than thatusing coordinate series.8. The function, structure and flow of the algorithms of our GNSS data processing softwaresuitable for dynamic deformation monitoring are discussed. Then real data collected on a testplatform was processed and then the ability of our software is verified through comparing theresults of our software to those of GrafNave and IMU. At last, a prototype deformationmonitoring system in a coal mine and the test result are introduced. The prototype system iscomprised of single-frequency GPS receivers, wireless communication network and our software. The real experimental result indicates that the root mean square of coordinate series reach0.8mmfor horizontal component and1.1mm for vertical component after removing multipath andde-noising. And the time consumption of data processing for every epoch is less than0.1s. It isdemonstrated that this prototype system can satisfy high-rate dynamic deformation monitoringwith1Hz.