基于水化学数据的矿山涌水水源识别:主成分分析与残差分析

涌水是矿山安全生产的重要威胁之一,掌握涌水的水源是预防这类灾害的重要依据。基于多期次的水化学监测数据,提出了利用主成分分析(PCA)和残差分析来识别涌水的水源数量和水源类型。主成分分析能够充分提取高维数据的结构信息,消除变量冗余。残差分析利用主成分分析结果,提取不同的主成分个数重构离子浓度,得出重构离子浓度和原始离子浓度之间的残差,利用残差的结构性信息来判断可能的涌水水源数量和水源类型。结果表明:随保留主成分个数的增加,重构离子浓度的精度越高,离子残差的结构性得到消除且表现出随机性。在实例应用中,通过利用5个主成分重构离子浓度消除了所有主要离子残差结构性,保证了信息的最大程度提取,因此,确定研究矿山涌水水源数量为5个; 同时,利用每个主成分对残差结构性的影响程度及离子相关性分析,确定了这5个水源类型为海水、第四系孔隙水、富钙基岩水、富镁基岩水和淡水。

End-members Identification of Mine Water Inrush Based on Hydrochemical Data: Principal Component Analysis and Residual Analysis

Water inrush is a threat to mine safety, and the knowledge of water sources(end-members)identification is the foundation of prediction for potential disasters. Based on the multiple-stage hydrochemical data, a method using principal component analysis(PCA)and residual analysis was developed to identify the numbers and types of water sources. PCA can eliminate the redundancy and extract the structural information of original variables. Residual analysis is based on the results of PCA by extracting different numbers of principal components to reconstruct the ion concentrations, calculating the residual between reconstructed and measured ion concentrations, then, using the structural degree to identify the numbers and types of water sources. The results show that the accuracy is improved, and the structure of residual is eliminated with the increasing number of principal components. Using the 5 principal components can eliminate most of the structural information of main ions, and extract all the useful information in the case; meanwhile, 5 types of water sources, including seawater, Quaternary pore water, Ca-rich water, Mg-rich water and freshwater, are identified based on the different effects of each principal component to the residual and the relationships among ion concentrations.