ABSTRACT: As one of the important means to distinguish the type of water quality, the chemical characteristics of groundwater are of great significance for determining the source of water inrush in mines.According to the difference in water chemistry characteristics of different aquifers, the water-rock interaction analysis of strong alkaline mine water is carried out, and a mixed water identification model based on SFLA-BP neural network algorithm (Shuffled Frog Leading Algorithm-Back Propagation neural network) for mine water source is proposed. Taking the Muduchaideng Coal Mine as an example, the water quality of four aquifer systems (loose layer—LA, coal bearing sandstone—CA, and the underlying sandstone—TA) are analyzed.Groundwater samples are further classified by cluster analysis and factor analysis.Parameter optimization of BP neural network model using SFLA. Using 70% of the 67 sets of sample data as training samples and 30% as prediction samples, the simulation experiment was carried out to establish the SFLA-BP neural network model, which was compared with the BP neural network algorithm.The results suggest that major ion concentrations of the aquifer systems were different from each other, and factor analysis indicates that their chemical compositions are mainly originated from two kinds of contributions: dissolution of soluble minerals and weathering of silicate minerals.And the main water source of the mine is TA.Research believes that the water source identification model based on the SFLA-BP neural network algorithm can more effectively eliminate the influence of interference factors and accurately identify the type of mine water inrush.
 

Water source identification SFLA-BP neural network Water chemistry characteristics Mine water hazard