Research On Early Warning Methods Of Mine Disasters Based On Coal Mining Process Identification

The large amount of gas emission from the coal mining face leads to the accumulation of gas in the coal mining roadway,which is prone to gas accidents.Different coal mining operations have different degrees of damage to the coal body stability of the coal mining face,which results in different gas concentration changes and characteristics,and the laws are still to be discovered.The paper takes the gas time series of coal mining face as the research object,and proposes the classification model and early warning model of coal mining process.The main research contents of the thesis are as follows:(1)Based on the analysis of coal mining technology and coal mining process,the variation law of gas concentration time series is studied,and the feasibility of using coal gas concentration time series to classify coal mining process is discussed.The type of coal mining process is divided by the difference in work intensity.(2)In order to improve the recognition accuracy of coal mining process,a classification method lbased on wavelet packet energy spectrum and lm algorithm for bp neural network rmining process is proposed.Firstly,the wavelet packet transform is used to decompose the gas concentration time series after noise reduction,and the energy spectrum of the gas concentration time series of different coal mining operations is extracted as the feature vector,and then the extracted feature vector is used as the input of the neural network.Layer,train,ing obtained a classification model of the coal mining process.The e.xperimental results show that the classification accuracy of this method is 90.5%,the accuracy is 9.5%higher than the SVM classification model,and 7.2%higher than the KNN classifi,cation model.(3)In order to improve the prediction accuracy of gas concentration time series,a prediction method of gas concentration in coal mining face based on empirical mode decomposition and long-term and short-term memory network is proposed.Firstly,the method uses the empirical mode decomposition to decompose the non-stationary gas sequence,and then predicts each decomposition component.Finally,the prediction results of each component are summed to obtain the final gas prediction result.The experimental results show that the prediction average relative error of traditional long-term and short-term memory networks is 2.02%,and the average relative error of gas concentration prediction using this method is 0.39%.This method improves the prediction accuracy of gas concentration.The rules for the classification of gas hazard degree in the early warning model are given.Combined with the coal mining process classification model,the hazard degree of the forecast data is identified.The experimental results show the reliability of the early warning model.(4)Developed a mine gas disaster warning system identified by the b/s mode coal mining process.In the system development process,the modular design idea is applied,and each function is packaged to facilitate user selection and use.The main functions of the system include management functions,real-time monitoring of gas concentration,analysis of gas concentration series,and prediction and warning functions.The system has friendly interface,simple operation and high expandability,and has good practical value and application prospect.