The PCA-Logistic method for identification of water burst in Nalinhe No.2 coal mine

Nalinhe No.2 coal mine is the first pair of large-scale mines in the Nalinhe mining area. The water inrush events occur from time to time due to its own complex hydrogeological conditions and strong disturbance of excavation in the initial stage of production, which has caused serious threat to the mining activity safety. Finding the source of gushing water quickly and effectively is the key to control the mine water disaster. Based on the water quality analysis of main aquifers and goaf water samples in Nalinhe No.2 coal mine, and drawing the piper trilinear nomograph water samples, the hydrogeochemical characteristics of the groundwater in each aquifer and goaf water were revealed. Then eight indexes, snch as Ca²⁺, Mg²⁺, Na⁺+K⁺, HCO₃^-, Cl^-, SO₄^2-, pH and salinity, were counted as the original data of water source discrimination. After the principal component analysis, four principal components F₁, F₂, F₃ and F₄ were obtained. Taking the values of these four principal components as the discriminant of the Logistic regression model, a discriminant model for gushing water sources in the Nalinhe mining area was established. Using 36 groups of standard water samples as training samples, the resubstitution accuracy was 97.22%. The established model was used to discriminate 4 groups of water samples. The research results showed that the method of principal component analysis and disordered multi-class logistic regression could eliminated redundant information effectively between the original data of the samples, and made the results of water source discrimination more rapid and accurate. It could meet the needs of mine production, and provide decision-making and basis for prevention and controlling of water inrush.