Study on deep learning methods for coal burst risk prediction based on mining-induced seismicity quantification

Abstract The assessment of Coal burst risk (CBR) is the premise of bump disaster prevention and control. It is the implementation criterion to guide various rock burst prevention and control measures. The existing static prediction and evaluation methods for CBR cannot be effectively combined with the results of underground dynamic monitoring. This study proposed a mining-induced seismicity information quantification method based on the fractal theory. Deep learning methods were used to construct a deep learning framework of coal burst risk (DLFR) based on the fractal dimension of microseismic information. Gray correlation analysis (GRA), information gain ratio (IGR), and Pearson correlation coefficient are used to screen and compare factors. Statistical evaluation indicators such as macro-F1, accuracy rate, and fitness curve were used to evaluate model performance. Taking the Gaojiapu coal mine as a case study, the performance of deep learning models such as BP Neural Network (BP), Support Vector Machine (SVM) and its optimized model based on particle swarm optimization (PSO) algorithm under this framework is discussed. The research results' reliability and validity are verified by comparing the predicted results with the actual results. The research results show that the prediction results of CBR in DLFR are consistent with the actual results, and the model is reliable and effective. The mining-induced seismicity quantification can solve the problem of insufficient training samples for the CBR. With this, different pressure relief measures can be formulated based on the results of the CBR predictions to achieve "graded" precise prevention and control.