Study On Intelligent Prediction Of Overburden Deformation Caused By Coal Mining Based On Distributed Optical Fiber Sensing Technology

Underground mining of coal has caused the overlying rock and soil mass of coal seam to slip and collapse,and caving zone,fracture zone,and bending zone were formed from bottom to top.In view of the concealment,complexity,time variability,and nonlinearity characteristics of the overburden deformation caused by mining,distributed optical fiber sensing technology was used to obtain the vertical strain distribution of the overburden,and the intelligent prediction model of overburden deformation was proposed by introducing machine learning method.The predictive model realizes the advanced prediction of overburden deformation and promotes the informatization and intellectualization of coal mine overburden deformation monitoring.The central research contents are as follows:(1)The indoor similar-material model test was carried out.The characteristics of overburden deformation under the condition of strike longwall mining were studied by using Brillouin Optical Time Domain Analysis technology(BOTDA).Based on the regression algorithm in machine learning — gradient boosting algorithm,the predictive model of overburden deformation strain was established.Combined with the theory of key strata,the predictive accuracy of the model in multiple mining stages were analyzed.The predictive model was applied to the field test of overburden deformation in coal seam mining,which verified the feasibility of the machine learning method in the prediction of overburden deformation caused by mining.(2)The recognition of the deformation area of the overburden caused by mining was converted into the binary classification problem in machine learning.According to the measured data of the field sensing optical cable and theoretical analysis,the spatial distribution of the overburden deformation area in each mining stage was determined.The collected strain data was segmented and labeled to obtain modeling data sets.Based on the classification algorithm in machine learning — convolutional neural network +residual unit,the recognition model of overburden deformation area was established.The coal seam mining stages were divided,and the trained model of current mining stage was used to identify and predict the distribution of overburden deformation area in the subsequent mining stage.Comparative tests with different classification algorithms were set up to verify the accuracy of the proposed model.(3)The intelligent decision-making system of overburden deformation was established by fusing the predictive model of overburden deformation strain and the recognition model of overburden deformation area to realize the advanced prediction of the overall deformation of the overburden in the study area.According to the predictive results of the system,different treatment measures can be taken to realize the safe,green and intelligent mining of coal.The research results of this paper are of great significance for ensuring the safe mining of coal mine,avoiding and reducing the damage to the ecological environment in the mining area,and providing scientific theoretical basis and technical support for the precise control of mining overburden deformation.