Research On Inversion Method Of Underground Coal Goaf Characteristics Based On SAR/InSAR

Coal is one of the most important energy sources in the world and plays an important role today.The surface subsidence caused by coal mining can cause irreparable damage to the ecological environment of the mining area for a long time,which has a great impact on various industrial and agricultural production activities of mankind.In serious cases,it can even cause damage to the safety of people’s lives and property.On the other hand,the phenomena of indiscriminate coal mining,illegal mining,and cross-border mining have been repeatedly prohibited,which wastes a lot of coal resources and leads to geological disasters easily like surface collapse and crack.However,the existing supervision methods mostly use the methods of mass reporting,on-site inspection,and geological exploration,which are difficult to meet the timeliness and are not suitable for the detection of underground goaf without target areas in a large range,which brings difficulties to the supervision of the government and enterprises.Therefore,to monitor the underground coal mining in advance,extract the dynamic deformation information of the surface,and prevent the geological disasters caused by coal mining,it is vital to obtain accurate information on the location of underground goaf.As a relatively advanced earth observation technology,Interferometric Synthetic Aperture Radar(In SAR)has been widely used in the fields of geological hazard monitoring and surface deformation information extraction because of its characteristics like short revisit periods,strong penetration capability,wide coverage,and high monitoring accuracy.However,there are relatively few studies on the inversion of underground goaf location information using surface deformation obtained by In SAR.Due to the large magnitude of surface deformation in mining areas and the large amount of surface vegetation coverage,the SAR image pair is decoherent,which will cause errors in the actual inversion process.At the same time,the variability of geological mining conditions in the mine area also brings difficulties for the inversion of the underground goaf.Therefore,to address the above shortcomings,this thesis selects the Daliuta and Zhangshuanglou coal mines with different geological mining conditions as the research objects,and takes SAR / In SAR surface deformation as the constraint,to deeply study the method of inversion of underground goaf by using swarm intelligence optimization algorithm and deep learning network.The main research results are as follows:(1)An inversion method of underground goaf location parameters in a large gradient surface deformation area integrating SAR Offset-Tracking(OT)algorithm and PIM is constructed.The correlation model of underground goaf and surface settlement is established by using the probability integral method(PIM),the parameters of underground goaf are regarded as unknown parameters,and the surface settlement calculated by Oppositition-based Learning Chaotic Bat Algorithm(OLCBA)joint correlation model is combined with the real surface settlement obtained by SAR/In SAR method to obtain the real parameters of underground goaf.The simulation experiment shows that the error between the underground goaf parameters obtained by inversion and the real values is small.The maximum relative error and mean square error of coal seam mining depth are 3.70% and 6.24 m respectively,at which time,the absolute error is 0.37 m and the goaf inversion model can better approximate the real goaf parameters.Applying the method to the 52304 working face in the Daliuta mining area,the maximum relative error is 81.50% and the maximum absolute error is 34.56 m.(2)An inversion method of location parameter of the deep goaf area integrating DS-In SAR and PIM is constructed.To address the problems of relatively small surface deformation in deep mining areas and the fact that the surface of mining areas is often covered by more vegetation,which leads to decoherence,as well as the PIM parameters in the inversion model need to be inversely measured or determined empirically,this thesis investigates the inversion method of the goaf parameters and the main PIM parameters based on obtaining DS-In SAR time-series deformation.The simulation experiment shows that the error between the parameters of underground goaf obtained by inversion and the real value is small,the maximum relative error is 6.28%,and the maximum absolute error is 7.50 m.Applying the method to the 93604 working face in Zhangshuanglou mine,the maximum relative error in the inversion of the goaf parameter is 14.79%,the maximum absolute error is 42.67 m,and the parameter error of the PIM model is no more than 34%.(3)An inversion method of underground goaf location parameters based on deep learning is proposed.In order to obtain the surface deformation directly from SAR/In SAR to invert the goaf parameters,this thesis firstly generates the mining subsidence deformation field by combining the known goaf characteristic parameters with PIM,and converts it into a winding phase diagram to construct the training data set;Then a convolutional neural network deep learning model is designed and the model parameters with the dataset are trained;Finally,the surface deformation acquired by In SAR is brought into the trained deep learning network model to directly invert the location parameters of underground goaf.In this thesis,the training data set is constructed,and the inversion experiments are completed with the Zhangshuanglou Coal Mine as the research object.The experimental results show that the trained deep learning framework can obtain the real goaf parameters quickly and with high accuracy,with a maximum relative error of 33.17% and a maximum absolute error of 38.50 m.There are 38 figure,13 tables and 91 references.