Water source determination of mine inflow based on non-linear method

The mathematical principles and models of gray relational analysis and BP neural network used to determine the source of water were introduced. In order to determine the water source of mine inflow, the K++Na+, Ca2+, Mg2+, Cl_, SO42+, HCO3- were selected as evaluation index because of the importance of 6 water hydrochemical element factors. The clustering analysis was performed for water samples from Wutongzhuang mine of Fengfeng mining areas in North China and the determination standards of aquifers were proposed. Based on the training of BP network for 12 selected water samples, two methods mentioned above were used to determine the water sources of 3 typical samples and compared the calculated sources with the actual water source. The results show that both the grey relational analysis (GRA) and BP neural network are appropriate to source determination of mine inflow, but both have the superiority and limitation, which lies on the status of hydrochemical data.