| Coal is an important energy source for China ’s economic development and has made great contributions to the process of urbanization and the vigorous development of the national economy.The problems such as ground subsidence,ground cracks,landslides and building damage caused by coal transition mining pose a major hidden danger to the personal safety of residents in the mining area.In order to reduce the negative impact of geological disasters and environmental damage caused by mining subsidence,it is necessary to carry out land subsidence monitoring and study the law of mining subsidence.Interferometry Synthetic Aperture Radar(InSAR)technology is a new earth observation method,which can realize all-day,all-weather,wide coverage and high-precision surface deformation monitoring.Compared with the traditional surface subsidence monitoring method,it has obvious advantages.However,InSAR technology can only monitor the surface deformation and analyze the subsidence law,and cannot predict the unknown deformation.How to realize the integration of monitoring and prediction of mining subsidence disaster early warning is the current research hotspot of disaster prevention and mitigation.This paper takes Huainan mining area as the research area,based on time series InSAR technology and improved particle swarm optimization(IPSO)to optimize support vector regression(SVR)model to monitor and predict the surface subsidence of mining area,so as to provide new research ideas for dynamic subsidence monitoring and disaster warning in large area of coal mine.The main work and achievements are as follows :(1)In this paper,Pansidong Mine is taken as the research area,and the small baseline set technology(SBAS-InSAR)is used to monitor the surface subsidence of mining subsidence.Because this method can effectively avoid the influence of timespace decoherence and atmospheric phase,the high-precision surface subsidence information is obtained.Finally,the measured leveling data are compared with the SBAS-InSAR monitoring results.The results show that the maximum difference between the two monitoring results is 46 mm,and the correlation coefficient is 0.92,which fully shows the reliability of the time-series InSAR technology for mining subsidence monitoring.The results can be used for subsequent subsidence prediction in mining areas.(2)According to the relevant mining data and leveling monitoring data of the mining area,the surface deformation and deformation characteristics are analyzed at multiple levels.From the settlement curve of the strike and dip section line of the working face,it can be seen that the deformation characteristics of the mining area are funnel-shaped,which is in line with the subsidence law of mining subsidence;at the same time,according to the analysis of the time and settlement of the characteristic points of the working face,the time and settlement value are roughly nonlinear.By using the statistical analysis method of pixel interval,the percentage of pixel points with different interval deformation rates is calculated.The results show that the percentage of deformation rate distribution in the range of-35 mm / a ～ 0 is the highest,about 65.38 %,and the percentage of surface subsidence area is about 71.73 %.Most of the area has subsidence,and the subsidence area is basically the same as the distribution position of the working face,which also verifies the reliability of SBAS technology monitoring results.(3)Based on the monitoring results of SBAS-InSAR technology and support vector regression(SVR),particle swarm optimization support vector regression(PSO-SVR)and improved particle swarm optimization support vector regression(IPSO-SVR)algorithm,a single model and a combined model are established to predict the surface deformation of the mining area.The prediction results of the three feature points are compared and analyzed.The results show that the prediction accuracy of the three models meets the monitoring requirements of the mining area,but the IPSO-SVR combined model established by the improved particle swarm optimization(IPSO)optimized support vector regression(SVR)proposed in this paper.Due to the advantages of strong learning ability,generalization and highdimensional space nonlinear processing ability,the prediction results have the highest accuracy and stable performance,and are more suitable for mining subsidence prediction.Figure [27] Table [5] Reference [81]... |
