Research on Realization of Coal Mine Impact Ground Pressure Signal Early Warning Based on BP Neural Network and LSTM

Abstract

With the reduction of coal resources makes deep coal seam mining important, but also increases the ground stress and geological complexity, especially impact ground pressure.AE signal monitoring technology plays an important role in preventing impact ground pressure by real-time monitoring of cracks and stress changes in coal rock. In this paper, wavelet analysis and RFE techniques are used to extract key features from signals, and BP neural network is chosen as the classification model after comparing multiple machine learning models. Random Forest is accurate but has low recall, and LSTM suffers from slow classification and poor generalization, neither of which is applicable. Compared with prophet and ARIMA models, LSTM performs better in high-frequency data prediction and dealing with fluctuating data, so it constructs a coal mine impact ground pressure signal early warning system based on BP neural network and LSTM to improve the accuracy of coal mine safety monitoring. The innovation of this paper is to use different network frameworks for classifiers and predictors, and by using feature extraction techniques, data transformation is performed to maintain the accuracy of the results while reducing the model complexity, improving the response speed, and enhancing the model generalization ability.