Classification And Prediction Of Coal Structure Based On Recurrent Neural Network

Coal body structure refers to the structural characteristics of coal seam after various geological processes in the process of geological evolution.After deformation and metamorphism,coal body can be divided into primary structure coal and structural coal.At present,many scholars observe the classification of structural coal by macroscopic and microscopic methods.Macroscopic methods are mainly observed by human eyes and hand specimens.Macroscopic methods are divided into om microscope observation and SEM microscope image analysis.The methods are time-consuming and laborious,and the accuracy is not high.In this context,this paper proposes a model based on cyclic neural network to predict coal structure classification.Firstly,LDA method is used to reduce the dimension of 3D seismic attribute data to minimize the correlation between different variables.Secondly,aiming at the problem that the gradient of traditional cyclic neural network is easy to disappear,this paper proposes two improved cyclic neural network structures: LSTM and Gru.According to the high-dimensional characteristics of 3D seismic attribute data,a high-dimensional cyclic neural network is constructed.At the same time,the cyclic neural network model of LSTM as control unit is adjusted.In order to ensure the accuracy of the prediction model,softmax function is used to output the classification results.Finally,the paper applies the improved high-dimensional cyclic neural network prediction model to the 8 coal seam of No.At the same time,the model is compared with LR and SVM,which further shows the reliability of the prediction model.Therefore,the high-dimensional cyclic neural network model proposed in this paper has high prediction accuracy and low error,which can meet the needs of coal structure classification in real mining areas,and has certain research value and promotion significance.