Ground Pressure Hazard Prediction Based on Random Forest and MLP Models

The aim of this study is to predict the risk of impact ground pressure based on electromagnetic radiation (EMR) and acoustic emission (AE) signals during deep coal mining by using Random Forest and Multi-Layer Perceptron (MLP) models. First, for the interference components in the signals, this paper proposes an interference signal detection method based on threshold and time interval, and identifies the time interval of the interference signals by using Random Forest and MLP classifiers. Secondly, the model for identifying the time interval of shock ground pressure occurrence is constructed by analyzing the trend characteristics of the precursor signals and applying statistical description, autocorrelation and mutual correlation analysis methods, as well as ADF test and MK test. Finally, in order to achieve the probability prediction of the occurrence of precursor signals at each acquisition moment, MLP neural network is employed and the generalization ability of the model is optimized through cross-validation. The method proposed in this paper can effectively identify the interference signal and the precursor signal, and improve the accuracy and timeliness of the impact ground pressure early warning, while the research provides a scientific and effective technical means for the safe production of coal mines.