Source Discrimination of Mine Water Inrush Using Multiple Combinations of an Improved Support Vector Machine Model

Accurate and timely identification of water inrush sources is important in controlling mine water hazards. We combined various algorithms to offer an improved inrush source identification method. The algorithms include the Fisher identification method, self-organizing correlation method (SOM), improved principal component analysis method (PCSOM), and grey wolf algorithm (GWO) optimized support vector machine (GWOSVM). The model was used to identify the water inrush source of 47 groups of training samples and 20 groups of analysed samples from different water sources in the Zhaogezhuang coal mine. The results show that PCSOM can reduce the information overlap between discriminant indexes, simplify the model structure, and improve the running speed of the algorithm. The penalty factor c and kernel parameter g of the SVM optimized by the grey wolf algorithm are faster and more stable in parameter optimization, and the discrimination result is more accurate. The proposed model can accurately and quickly identify water inrush sources. Thus, it is helpful for rapidly predicting inrush disasters and can serve as a reference for inrush source identification technology.