Research On Deformation Monitoring System With Geo-robot

In the practices such as base pit and side slope deformation monitoring, the new generation of Geo-robot are greatly improving the accuracy and effectiveness of the monitoring work with its high degree of automation, high accuracy and high efficiency. However, the improvement of hardware has raised some questions for traditional data processing. In this paper, on the basis of previous work, the theory and methods of gross error elimination based on PauTa’s Criterion, wavelet filtering method of deformation monitoring data, gray filtering and prediction theory, filtering theory and method about colored noise are discussed respectively. Based on these algorithms, a software system for the Geo-robot was developed.In this paper, a systematic analysis was done for the limitations of " PauTa’s Criterion", which proved theoretically that the gross errors can be eliminated one by one with " PauTa’s Criterion" only certain conditions are met, when the observations contain more than one gross error. Then a new method,"data jump method" of eliminating gross error was proposed.Wavelet threshold de-noising and force de-noising methods were studied. According to the slop subsidence monitoring data of Dagushan Iron Mine, wavelet de-noising research has been done, which shows that both threshold and functions based wavelet algorithm have a significant impact on the results, the effect of force de-noising is obvious. In this procedure, wavelet signals were decomposed and the wavelet coefficients were reconstructed by selecting a suitable approximation signal and detail signal. The method of force de-noising is relatively simple, and the signal after de-noising is smooth, the sequence according to filtering effect is "db5、 db3、db2".Gray prediction method after wavelet filtering was studied and better satisfactory results have been achieved than the prediction before wavelet filtering. The optimal dimension was determined. According to the advantages of wavelet de-noising and gray adaptive prediction, a improved gray adaptive method based on wavelet de-noising was finished on the basis of the initial correction of gray adaptive. This kind of prediction has better effect than single gray model. There was a comparison between gray prediction models based on Kalman filtering and wavelet filtering, the latter has better effect. Kalman filter is a filter dependent on the model. When the monitoring data deviate largely from the model, gray prediction of wavelet filtering can achieve better results.For high precision observation data by Geo-robot, based on the assumption that the noise is limited and colored, and the characteristics of current filtering method, the H∞filtering method was proposed. Compared with Kalman filtering method, the H∞filtering method is relatively conservative, and is hard to lose useful information. When signal to noise ratio is poor and high precision is needed, the H∞filtering method is effective. While its disadvantage is that the method is not designed for white noise, so some of the white noise component is not filtered after filtering, he waveform was not very smooth. Considering these shortcomings, for the advantages and disadvantages of H2and H∞filtering program, a mixed program of both was given, which take into account the advantages and disadvantages of two filtering programs. The mixed program can minimize the impact of white noise components with the influence limitation of non-white noise components, which is an effective program to treat mix noise components.Based on studies of Geo-robot (TCA2003), GEOCOM development platform and the monitoring data processing methods, an automatic monitoring system has been developed with VB6.0as a development tool on the GEOCOM platform. The main functions include creating a new project and opening the project, editing point, editing point group, editing cycle, setting communication, setting parameter, setting up station, automatic collecting data, displaying data and processing data in real time. Data processing includes precision assessment, excluding gross errors, meteorological correction, and filtering and distortion prediction. A three-dimensional deformation monitoring system whose core is measuring robot and the computer was implemented. The system can achieve automatic monitoring data acquisition and efficient data processing, laying a solid foundation for real time landslide warning.