Study On Multi-classification Of Tectonic Deformed Coal Base On RBF Neural Networks

Coal and gas outburst is one of the three main dynamic disasters in mine mining,which is mainly manifested in the phenomenon of instantaneous gushing of coal and gas underground.Studies have shown that in coal and gas outburst accidents,different types of tectonically deformed coal generally develop in coal seams,and the risk factors of coal and gas outburst accidents caused by different types of tectonically deformed coal are different.Therefore,in actual coal mining,if we can accurately predict the types of tectonically deformed coal in the mine,and then mark out the gas outburst grade and take corresponding emergency measures to reduce the occurrence of gas outburst accidents,It will play a very important role in coal mine management and safe development and production.Aiming at the problem that the current prediction methods of tectonically deformed coal categories have low prediction accuracy due to various restrictive factors,an improved RBF neural network model based on AGA-DBSCAN optimization is proposed to predict the types of tectonically deformed coals.First,the actual exploration and technical application of the 8# coal seam in Luling Coal Mine of Huaibei Mining Group Co.,Ltd.were carried out to obtain relevant 3D seismic attribute data.Secondly,in order to reduce the dimensionality of the 3D seismic attribute of this coal seam and eliminate the linear correlation between the attribute variables,the principal component analysis(PCA)algorithm is used for preprocessing.Then,use the better optimization ability of genetic algorithm to optimize the key hyperparameters(Eps,min＿samples)in density clustering(DBSCAN)to obtain real and effective core point data.By high density filtering,the best core point data are obtained,and the condensed hierarchical tree is constructed through hierarchical clustering to obtain the initial clustering center of the best k-means clustering.Finally,the cluster center obtained by the k-means clustering operation result is used as the optimal RBF neural network hidden layer center vector parameter,thereby increasing the accuracy and robustness of the model’s prediction.At the same time,there is a problem that genetic algorithm is easy to fall into local optimization.The global and local search ability of genetic algorithm is improved by introducing the adaptive change of crossover rate and mutation rate with the increase of evolution times and increasing the parameters of chaotic variables.In addition,in order to enhance the generalization performance of the tectonic deformed coal classification model,the L2 regularization term is added to the weight parameters of the model,which effectively avoids the influence of noise and outliers on the generalization ability of the model.In order to further explore the classification and prediction performance of the model,the improved model proposed in this paper is compared with the support vector machine(SVM)model that uses K-fold cross-validation to optimize parameters and the traditional BP neural network model.Through the training and testing of the seismic attribute data of the actual coal seam,the better classification and prediction results are obtained,and the results are consistent with the actual geological data.As a result,the improved RBF neural network model proposed in this paper has high prediction accuracy and small error,so it can be applied to the classification of tectonically deformed coal in actual mining area.