Application of principal component analysis and Bayes discrimination approach in water source identification

Accurate and efficient determination of water inrush source is the prerequisite for solution of mine water disaster. In order to solve the problem of mine water disaster, based on water quality analysis data of 59 water samples of different aquifers in Yuaner coal mine, principal component analysis in multivariate statistical analysis was used. Principal component analysis was used to calculate the factor score of different water samples, hierarchical clustering was conducted and the erroneous samples were rejected. The remaining water samples were used as learning samples to test the accuracy of determination of Bayes discriminant function. It is concluded that the accuracy was 92.5%. The discriminant results were verified by cross-validation. The discriminant function was used to discriminate the water sample from the water-rich area of a working face floor in the mine, and the results coincided with the actual situation. The results show that the discrimination method based on principal component analysis and Bayes discrimination method was more accurate than single Bayes discrimination method, could eliminate the interaction effect among variables of samples and realize fast and effective discrimination of water inrush source, the discriminant process of water inrush source is simple and the accuracy is high. The interaction between the sample of variables was eliminated, was realized. It provides effective basis for coal mine water prevention and control.