Prediction Of Gas Emission Quantity Based On Manifold Learning And Support Vector Machine

Coal production is a mining industry with high risk. Potential safety problems, forexample, gas, floods and landslides always occur in the production process. Coal mine gasaccident has been one of the most serious disasters in coal mines. Mine gas emission is one ofimportant factors affecting mine gas disaster. In order to reduce gas disasters of mine, it hasimportant practical significance to predict mine gas emission accurately. Through studyingthe related literature about predicting the gas emission at home and abroad, this paperpresents a method based on manifold learning and support vector machine to predict the minegas emission.Based on the analysis of gas emission, the paper analyses influencing factors of theamount of gas emission. The gas monitoring data of a coal mine was studied. The Matlabsoftware is the modeling platform. The data of mine gas emission was pretreated by usingmanifold learning algorithm. Through this way, the lack of support vector machine is madeup. And the best kernel function parameter g and penalty factor c were selected by ParticleSwarm Optimization (PSO), Genetic Algorithm (GA) or Cross Validation (CV). It took thedata as input and use SVM for regression prediction of the gas emission quantity. Theprediction model was established and predicted the mine gas emission accurately. Manifoldlearning algorithm can keep the manifold structure of the data in the process of dataprocessing. It reduces the dimensionality of the original data sets. The convergence speed andaccuracy is improved. The support vector machine can keep a good trade-off between thelearning accuracy and the learning ability. It makes the model have a good ability ofgeneralization. Combined with the characteristics of mixed explosives, the paper predictedthe detonation parameters based on the manifold learning and the support vector machine. Itproves that the prediction method in this paper also can be applied to other field.Through the research, The result shows that the predicting model based on the manifold learning and the support vector machine can predict the mine gas emission accurately and fast.The model has a high precision and a good correlation. Comparing with the traditionalpredicting method, the fault tolerant ability of the model is stronger and the precision ishigher. This model can provide a reference for the later gas emission and has certain directivesignificance for the forecast method of gas emission. The applicability of the forecast methodbased on the manifold learning and support vector machine is higher, and the method couldbe promoted and practiced.