Research For Dangerous Area Prediction And Prevention Technology Of Coal And Gas Outburst Based On The SVM

Coal and gas outburst is a complex, non-linear problem which is affectedby a variety of factors, it has great defect to predict outburst by using traditionalmethods. With the rapid development of computer and information processingtechnology, a lot of intelligent methods and techniques gradually penetrate into someproblems that is similar to outburst prediction. Support Vector Machine is a MachineLearning method based on Statistical Learning Theory, it is main for settling thepractical problems of the small sample, nonlinearity, high dimension, local minimum,and it has good classification and recognition which has been widely applied to manyareas of pattern recognition and forecasting. To the end, this paper establishedlearning model to achieve the classified prediction of outburst risk by introducingSupport Vector Machine which is based on field and laboratory testing data.Due to the impact of many outburst factors, it is not easy to distinguish thenecessary conditions for the outburst. Therefore, the original data must be pretreatedso as to obtain the key factors of impact outburst, In order to effectively solve theproblem, using the combination of grey relational analysis and entropy method, thekey indicators of characteristics extracted from the original sample.Training and test samples of the prediction model are selected by the keyindicators, and on the basis, established the predicting model of Support VectorMachine, training and testing process of the entire model is completed in theMATLAB platform, and calling the part function model of the LIBSVM library is tobe designed by the simulation program. In addition, further research is for classifiedaccuracy affection of Support Vector Machine model from the perspective of thekernel function selection and parameter optimization, it is verified that it is moresuitable for coal prominent class prediction based on Radial Basis Function(RBF).And on this basis, the punishment parameter C and kernel parameter g of SupportVector Machine are optimized by Cross-Validation method and Genetic Algorithm; itis proved that Genetic Algorithm achieved better test results under the premise of twoparameters preferred.At last, using Support Vector Machine classification method establish area risk prediction model of the Nan Fung expansion zone76,78two regions in WuyangCoal Mine, the test results show that they are in line with the actual outburst dangers,therefore, the Support Vector Machine classification model can be used hazardprediction in the unknown area of the mine area. In addition, the combination of"Four in One" Comprehensive outburst prevention measures, it mainly put forwardto gas pre-pumping for Wuyang coal mine. Finally, it analyzes and inspects theeffectiveness of the outburst prevention measures by advanced drilling. It provides adirection for the future of coal and gas outburst in the mine.