A PCA-MSA Based Arithmetic Research On Water-bursting Source Discrimination Of Mines

Mine water inrush is one of the largest natural disasters, and it threats the safety production in mines. Finding out the water inrush source can provide an important basis for taking measures to avoid mine water disasters. Therefore, how to determine the mine water inrush source quickly and effectively is very important. However, presently, most of the water source discrimination models do not take into account the influence about the information superimposition of the water chemistry indexes. This paper chosen two mines hydrogeological data and water chemistry data as study object. And Introduced the principal component analysis combined with two method of multivariate statistical analysis create water discriminant model.After Compared with the tradition multivariate statistical analysis model (Fisher discriminant modeh、Bayes discriminant mode、Stepwise discriminant model), a optimal source discrimination model was founded. In order to make the mine water inrush discrimination work automation and intelligence, a mine water inrush discrimination system was established by C++language. The results are as follows:(1) Water chemistry data was selected in A mine, which constitute of Ca2-、Mg2-、K++Na-、HCO3-、SO42-、C1、TDS. By using water chemistry analysis software (Aquachem) to draw piper trilinear nomograph、 Box plots of Conventional ion、the chart of Conventional ion relationship with TDS, and analyzing the hydrochemistry characteristics of different water sources. The results show that water has different chemistry characteristics, it is hard to distinguish water bursting source with single Ion distribution law. The accuracy of the water inrush model can be improved by using SAS date analysis software.(2) Fisher discriminant model, Bayes discriminant model, stepwise discriminant model were established on the basis of water chemistry of A,B mines, above two discriminant models with PCA were also established respectively.The results show that the discriminant models with PCA have more high correct rate, the back discrimination ratio and the the text ratio were increased by6%,18%. Eliminate the influence caused by the superposition of the index information, Improve the accuracy of the water inrush discriminant model.(3) The mine water inrush discrimination system was established (using C++language) on the basis of water chemistry differences as theoretical foundation. The core of the system consists of Fisher discriminant method, Bayes discriminant and combined with principal component analysis disciminant modes, and the system was established by C++language.The system has the features of easy to operate, practicality and reliability, furthermore it can discriminant the inrush water source quickly.