Fisher multivariate statistical method based on PCA for identifying water filling source in mine

The multi-source uncertainty of water filling source in the process of coal winning in working face is the research emphasis of mine water disaster prevention. It is often difficult to define the characteristic water quality threshold of each water filling source when the water chemical information is used to identify the water source, and most of the existing analysis and discrimination methods results in low accuracy of discrimination. Based on the samples of potential water filling source from mine inflow in 8210 working face of Majiliang Coal Mine collected on site, the database of samples was established. The water quality types and the training sample database of each water filling source were analyzed by Piper three-line graph method and multi-factor method. Fisher discriminant water filling source model based on principal component analysis was established, and the types of water filling source in goaf at the sampling stage were analyzed and identified on the basis of Euclidean distance discrimination. The results show that the primary source of water filling is Jurassic goaf water, followed by floor limestone water and roof sandstone water. The discriminant precision of this model based on principal component analysis can reach 99.9%. The model is of great guiding significance for the field identification of water filling source in working face.