Abstract: Mine water inrush is one of the greatest natural disasters in the mining production process. Quick and effective identification of mine water inrush source is an important guarantee of safe production in mining engineering. Seven kinds of water chemical composition having been selected as the sample variables in water bursting source recognition，and a forecast model of water inrush source is built by combining distance discriminant analysis with principal component analysis. This model is tested and applied in the different water layer of Xieyi Coal Mine in Huainan with thirty-three training samples and eleven predicting samples，and it is compared with the gray correlation discrimination model and Bayesian discrimination model. The results show that the model shows 95% of accuracy，and the predicting correct rate is up to 91%，which provides a new method for mine water inrush identification.

Key words: identification of water source, mine water inrush, distance discriminant analysis, principal component analysis