Research On Recognition And Prediction Of Coal Mining Subsidence Characteristics Based On Deep Learning

The basic characteristics of China’s energy structure of "rich in coal,poor in oil and less in gas" determine the important supporting role of coal in safeguarding national security and promoting high-quality development of social economy.However,the exploitation of a large number of mineral resources is easy to break the original stress balance of the overlying strata,resulting in the movement and deformation of the strata and the ground surface,thus causing a series of mine geological environmental disasters,such as mining area subsidence,ground water accumulation,damage of buildings and structures,and roads,and even serious geological disasters such as landslides and debris flows in mountainous areas.Therefore,timely and effective identification of deformation areas caused by coal mining and prediction of subsidence characteristics are of great significance to prevent the occurrence of geological disasters in mining areas.Traditional coal mining recognition methods are small in scope,low in efficiency and inaccurate in prediction of subsidence characteristics.However,with the development of computer hardware and large-scale data collection,in-depth learning stands out in the field of target detection and related prediction.In view of this,this paper carries out the research on coal mining recognition and subsidence feature prediction based on deep learning.The main research contents and work are as follows:(1)Proposed an intelligent detection method for surface movement basins in coal mining by integrating In SAR and CNN.Based on the theory of deep learning and the Support Vector Machine(SVM)model,this paper establishes an automatic detection model of mining subsidence basin under Sentinel-1A data.The Huainan mining area is selected as the experimental area to verify the model.First,the Differential Interferometric SAR(D-In SAR)method is used to process the Sentinel-1A radar data to obtain the interferogram,and the mining subsidence basin and other targets are manually extracted as training samples;Secondly,Alex Net,VGG19 and Res Net50 convolution neural networks are used to extract the feature vectors of mining subsidence basins for SVM classifier,and the Artificial Fish School Algorithm(AFSA)with strong optimization ability and good global convergence is introduced into the optimization of SVM parameters,and an improved Res Net50-AFSA-SVM model is constructed to detect mining subsidence basins in large area In SAR interferograms;Non-maximum suppression is used to remove the repeated search box and improve the detection accuracy of mining subsidence basin.The Huainan mining area with many subsidence basins and obvious subsidence is selected as the experimental area to verify the accuracy of the model.(2)An integrated system of mining subsidence movement deformation data processing and surface subsidence prediction has been developed.On the basis of drawing lessons from and summarizing the advantages and disadvantages of systems developed in different computer languages,this paper develops an integrated system of mining subsidence mobile deformation data processing and surface subsidence prediction by using C# programming language and combining Word and CAD.The system is divided into mobile deformation data processing module and parameter inversion and prediction module.The mobile deformation data processing module mainly has the functions of mobile deformation data management,mobile deformation calculation,output report,mobile deformation curve drawing,etc;The parameter inversion and prediction module has the function of AFSA-based probability integration parameter inversion and prediction,and AFSA-based Logistic single-point settlement parameter inversion and prediction.The system was applied to the 1414(1)working face of Guqiao Mine for verification.(3)Based on AFSA-MLP,a prediction model for the height of waterconducting fracture zone in fully mechanized top-coal caving mining is proposed.Based on the comprehensive analysis of the factors affecting the height of the waterconducting fracture zone of the overlying rock in the fully mechanized top-coal caving mining,and starting from the collectability of the data and the importance of the impact,this paper selects four important factors as the important parameters for the calculation of the water-conducting fracture zone: mining thickness,mining depth,slope length of the working face,and the ratio coefficient of hard rock lithology,The AFSA intelligent algorithm is used to optimize the defects of the MLP neural network in the process of error back-propagation,and a prediction model for the height of the water-conducting fracture zone based on AFSA-MLP is constructed.The accuracy of the model is verified by typical mining areas and engineering examples,and this model is applied to the parameter calculation of probability integration method.Figure [27] Table [9] Reference [121]...