Groundwater source identification based on principal component analysis and improved extreme learning machine algorithm using the genetic algorithm: a case study from the Dagushan iron mine, Liaoning Province, China

This study aims to identify the groundwater source of the mine water inrush in the Dagushan iron mine, Liaoning province, in northeast China. The hydrogeochemical characteristics can reflect the process of water–rock interaction. Therefore, it can be used to identify the occurrence type of groundwater. In the face of complex hydrogeological conditions and data in mining area, it is very important to find a fast and accurate method of dealing with these hydrochemical data. The main objective of this study is to use principal component analysis (PCA) with improved extreme learning machine (ELM) algorithm using the genetic algorithm (GA), boxplots diagram, Piper trilinear diagram, and Gibbs diagram to identify the possible source of water inrush in tunnel. The results show that water in the study area is mainly neutral and the ionic species in the water originate from rock weathering, among which most abundant anion and cation are HCO 3− and Ca 2+ . Additionally, a certain hydraulic connection is observed among the surface water, pore water, and fissure water. PCA-GA-ELM is shown to be very effective in identifying the source of mine water inrush. Firstly, the data are processed by PCA and 89.887% of the original information is retained. The PCA results indicate that the total dissolved solids, Na + , Cl − , SO 4 2− , and total hardness have the strongest influence on source identification of mine water inrush. Then, the tenfold cross-validation method is applied to the optimization of ELM parameters, and GA is used to improve the stability and accuracy of the ELM so as to obtain an optimal discriminant model after 100 evolutionary calculations. Through the study, it can be found that PCA-GA-ELM has a strong adaptability and accuracy, and it is effective to use this method to identify mine water inrush source.