The Coal Floor Water-inrush Risk Prediction Research Based On The PSO-SVM

The problem of north China type coal floor water inrush in our country is generally serious, and the accidents will cause significant casualties and property losses. However, risk prediction and evaluation is an important link in mine flood prevention and control, is also the basic premise and important basis to realize safety mining. This article selection of support vector machine (SVM) for coal mine water inrush risk prediction, through researching to the present situation of coal mine water inrush prediction, combining the nonlinear characteristics of coal seam floor water-irruption.Support vector machine (SVM) is a new type of machine learning algorithms based on the statistical learning theory, and it has stronger generalization ability, and it is suitable for solving the problem of nonlinear and small sample. However, the generalization ability and prediction accuracy of support vector machine (SVM) affected by the punishment parameters and kernel function. The parameters in using support vector machine forecasting model is difficult to determine,particle swarm optimization (PSO) algorithm is presented in this paper,through the random initialization of population, fitness function set,particle update Settings, termination conditions, to optimize the related parameters of support vector machine (SVM), then get the improved support vector machine forecasting model by particle swarm optimization.In the PSO - SVM forecasting model of water inrush applications,according to the hydrogeological conditions of coal mine to select the master influence factors of coal mine water inrush at first (that is, the thickness of water-resisting layer, water pressure, bottom damage depth,aquifer and fault throw), and then to collect coal mine water inrush history data, and these data can be divided into training set and test set;with MATLAB experiment platform, combining with the Microsoft Visual C + + compiler, using LIB SVM toolbox programming for simulation training , the data of the training set of support vector machine (SVM) optimal penalty factor C and kernel function parameters are 694.8591 and 317.1063 respectively; and then testing whether the water inrush in coal mines predictions, and the prediction results were analyzed the last. The experimental results show that the PSO - SVM model has higher accuracy and engineering application promotion value in water inrush prediction, has the vital significance to guarantee coal mine safety production.