With the development of coal mine production towards deep mining,accidents caused by tectonics(such as faults,collapse columns,etc.)occur frequently,which seriously threaten the safety of coal mine production.Therefore,it is urgent to improve the reliability and accuracy of tectonics identification to provide strong support for coal mine safety.3D seismic exploration technology is an effective tectonics detection method,which can extract many seismic attributes through a series of processing of 3D seismic exploration result data;Seismic attributes can be used to predict tectonics after being extracted and analyzed.In coal mining,the structure only accounts for a very small part of the mining area,and most of the areas are tectonic free areas.However,most of the current research on tectonics identification has not considered this factor,so it is of great practical significance to identify tectonics under unbalanced data.In the study,the east wing mining area of Shanxi Xinyuan Coal Mine was taken as the research area,and structural recognition research was mainly conducted from the perspectives of classification algorithms,data,and features.The specific work is as follows:(1)Classification algorithm analysis: By comparing the performance of several typical non integrated and integrated algorithms on imbalanced data,e Xtreme Gradient Boosting(XGBoost)is selected as the basic classification algorithm in this article.(2)Data pre-processing: The idea based on the Edited Nearest Neighbor(ENN)is adopted to improve the Synthetic Minority oversampling Technology(SMOTE)to overcome the problem of deleting useful sample information or generating noise samples caused by the traditional resampling method.The experimental results show that the improved resampling method can generate high-quality samples.(3)Feature selection: A composite feature selection method has been proposed to improve the classification accuracy of minority classes;This method combines the advantages of filtered and encapsulated feature selection,and experimental results show that the feature subset filtered by this method is more conducive to classification algorithms.(4)Research on the application of tectonics identification: taking the east wing mining area of Xinyuan Coal Mine in Shanxi Province as the research area,the above methods are comprehensively used to identify the tectonics in the research area;The experimental results show that the improved XGBoost model can overcome the problem of uneven distribution of coal mine data,and can accurately identify the type of coal mine tectonics. |