| China is a coal producer and consumer country,a large number of coal resources to make an important contribution to the economic construction,at the same time also bring a series of environmental and geological disasters.Therefore,it is particularly important to monitor the mining subsidence,predict and forewarn the mining subsidence,design reasonable mining methods,and post-mining ecological restoration.For unmanned,automatic and omni-directional mining monitoring and problem of monitoring time-space,it combined on the basis of the SAR data space measurement,on the basis of the technology of UAV aerial survey,on the basis of the leveling and RTK ground monitoring of three kinds of monitoring model.A method of independently using SAR data to obtain large gradient deformation in mining area is proposed.On the one hand,the method of space-sky and earth coordinated monitoring is used to obtain the land types situation and provide basic data for the post-mining ecological restoration work.On the other hand,it extends the mining deformation monitoring technology from monitoring the linear form of discrete points to the exploration of 3D+time scale of the surface area,which provides a new method for the omni-directional study of mining subsidence law.The main work is as follows:1)The applicability of sentinel-1A data in the study area and the influence of data source error on parameter inversion are analyzed.On the one hand,the D-InSAR technology is used to obtain the LOS deformation value of SAR data and decompose it to obtain the vertical subsidence value of working panel,and then the advanced influence Angle of 1613 working panel of Guqiaonan Mine in Huainan is calculated.The applicability of sentinel SAR data in Huainan mining area is verified by comparing the advanced influence Angle with the measured value.On the other hand,combined with the measured data and geological and mining conditions,the probability integral method is used to simulate the subsidence values of strike and dip lines,and different noises are added to them respectively.It is verified that the data source error within 10 cm can meet the requirements of static and dynamic parameter inversion in the mining area.2)A new method combining inverse geocoding and machine learning to construct prediction model to repair "void field" is proposed.The SAR data is affected by ground objects and environment,resulting in partial position incoherence,forming a"void field" at the observation station.As a result,the deformation value cannot be extracted from the SAR data even though there are leveling monitoring values at some points.Firstly,RGC-LSTM,RGC-NARNN and RGC-BP models are constructed for SAR data restoration,which can effectively extract the initial settlement and residual settlement values of observation stations.Then,the fractal model with natural law and the Knothe time function model with subsidence law were used to simulate the surface subsidence,and the accuracy of the two models was verified.Finally,RGC-NARNN and RGC-BP models with higher model accuracy were selected to repair the cavity field of the two differential interferograms in the study area,and the root mean square errors(RMSEs)was obtained within 2 mm.By comparison,the accuracy of RGC-BP model was better than that of RGC-NARNN model.3)InSAR-PEK(Interferometric Synthetic Aperture Radar PIM Exponent Knothe)is constructed by combining the probability integral method(PIM)and the power Exponent Knothe function.The subsidence values of the initial period and the residual period of SAR data are independently used to predict the subsidence values of the whole period,and then the effective subsidence values of the main period are obtained.Compared with the measured leveling value,the average of RMSE is 41.7 mm;when compared with the leveling data in the whole monitoring period,and the average of RMSE is 61.0 mm;when compared with the leveling monitoring data in the main period,the average percentage of the error in the whole settlement period to the maximum subsidence is 2.99%,and the average percentage of the error in the main period to the maximum settlement is 4.13%.Therefore,the InSAR-PEK model can obtain the large gradient deformation value of the mining area by independently using SAR data.4)To solve the unmanned monitoring problem of mining area,DOM,DSM and 3D model of mining area are obtained by using UAV technology.DSM extracted data were compared with leveling and RTK data to verify the precision and reliability of UAV tilt photogrammetry to obtain elevation of surface objects.After proof,the RMSE average of strike line and dip line was 2.85 cm and 3.81 cm,respectively.The elevation error of the verification point of the road strike line is 3.71 cm,and that of the dip line is 2.15 cm.It is preliminarily determined that the elevation value obtained by UAV technology can meet the requirements of subsidence prediction and parameter inversion.5)To solve the spatio-temporal evolution law of mining subsidence,four pairs of differential interference maps were made using 7-scene SAR data to explore the spatio-temporal law of subsidence of deep mining working panel and early mining under the geological condition of huge thick loose layer.The results show that in the initial stage of active subsidence,the surface moving basin expands widely,converges slowly at the boundary,and develops a long slow and slight subsidence zone.The surface deformation value is small,and the subsidence basin is elliptic,the basin trend direction is symmetric,and the basin strike direction is asymmetric;In the middle of active period,the subsidence area gradually draws closer to the center,and the elliptic subsidence basin evolves to nearly round.The main subsidence area of the subsidence basin is nearly round,and the strike and dip direction of the basin develop symmetrically.With the advance of mining,the surface movement and deformation are large,presenting the characteristics of concentrated subsidence.Figure 58 Table 42 Reference 186... |
PDF Full Text Request