Research On Data-driven Overburden Deformation Prediction Method In Mining

In underground coal mines,the deformation of the overlying strata,the movement of the strata and the breaking of the coal seam caused by the mining of the working face are one of the main causes of mine dynamic disasters such as water inrush from the mine roof and impact ground pressure.However,the deformation of mining overburden is a black-box problem,especially in the deep mining of coal seams,new monitoring techniques and prediction methods for overburden deformation are needed.Distributed optical fiber sensing technology is a new method for monitoring overlying rock deformation,and physical model testing of similar materials is a commonly used scientific research method in the mining field.Therefore,this paper takes the overburden deformation physical model test as the background and the distributed optical fiber monitoring as the data drive to study the overburden deformation prediction and deformation quantitative analysis method during the coal seam mining process.The main work is reflected in the following aspects:(1)In order to predict the deformation state and deformation trend of the overlying strata during the coal mining process,the frequency shift value data obtained by distributed optical fiber monitoring are used as the basis.Combining convolutional neural network(CNN)and gated neural unit(GRU),a CNN-GRU-BP optical fiber full-measurement point frequency shift value prediction model is proposed.Firstly,the integrated empirical mode decomposition is used to decompose the frequency shift value sequence of optical fiber measurement points into three component sequences with clear physical meanings:randomness,periodicity and trend.The input sample is constructed by fusing multi-attribute features such as the height of the measuring point.Then,model training is performed on the three components separately,and the output predicted value is superimposed to form a predicted frequency shift value.Finally,the strong correlation between the frequency shift curve of the optical fiber measuring point and the overburden deformation is used to infer the state and trend of overburden deformation.The experimental results show that the frequency shift value curve of the optical fiber measuring point can accurately fit the actual frequency shift value curve,and the average errors of the RMSE and MAE of the frequency shift value of the measuring point are 5.07 and 3.03.It is proved that the deformation state of the overlying rock can be effectively estimated by using the change state of the optical fiber frequency shift value curve in the future.(2)Based on the frequency shift value data sensed by distributed optical fibers,the spatial entropy of rock strain,the spatial entropy of overlying rock strain,and the information entropy of overlying rock structure strain are defined.A quantitative analysis method for overlying rock deformation law based on overlying rock strain information entropy is proposed.First,from the local point of view,the change process of the rock strain space entropy in the sub-key layer is analyzed.It is found that when the rock’s tensile strain and compressive strain space entropy change in a zigzag shape,it indicates that the rock has large cracks or fractures.From the perspective of the space above the working face,the change process of the overlying rock strain space entropy is analyzed,and it is found that the overlying rock compressive strain and tensile strain space entropy changes sharply before the overlying rock undergoes large deformation.It can be used as the precursor information of overlying rock failure and instability,and on this basis,the similarity function is used to complete the identification of the mining pressure precursor information of the working face.Finally,through the analysis of the information entropy of the overlying stratum structure strain of the overlying stratum system,it is shown that the evolution process of the coal overlying stratum fracture and instability is a nonlinear dynamic process from disorder to order with periodic unbalance.The analysis results prove that the information entropy of overlying rock strain can be used as a more accurate method to characterize the deformation law of overlying rock during the mining process.The research work of this subject has certain theoretical research significance for grasping the evolution process of overlying rock deformation caused by coal seam mining disturbance and disaster warning of strong rock pressure in working face.