Study On Water Bursting Source Discrimination Model In Fujian Coal Mine

Coal is one of the most important energy sources in Chinese economic development.In the process of coal mine construction and mining,water inrush accidents often occur,resulting in significant casualties and property losses.It has always been one of the main disaster factors affecting the safety of coal mine production.In the 2004～2016 year,14 accidents occurred in Fujian coal mine and 40 people died.Accurate identification of water bursting source is the key to control and prevent water hazards in Fujian coal mine.This study can provide effective theoretical basis and technical guidance for safety production in Fujian coal mine.Taking Fujian coal mine as the research object,this paper systematically analyzed the types of water bursting sources in Fujian coal mine,selected typical coal mines to collect water samples of water bursting source for hydrochemical detection experiments,Based on the analysis of chemical characteristics of water bursting source,a fast discriminant index is put forward,and a discriminant model of Water Inrush Source in Fujian coal mine is constructed.The main research work and achievements in this paper include:（1）According to the analysis of the natural environment,geology and hydrogeology of Fujian coal mine,the basic types of water bursting source in Fujian coal mine,such as goaf water,limestone water,surface water and fissure water,are obtained.Secondly,three typical coalmines in Fujian Province,Changfushan coal mine,Cuipingshan coal mine and Wuci coal mine,are selected as experimental research objects.（2）Thirty-five standard water samples of typical mine water bursting source were collected,including 13 goaf water samples,7 limestone water samples,6 surface water samples and 9 fissure water samples.The contents of Na⁺,K⁺,Ca ²⁺,Mg ²⁺,Cl^-,SO₄^2-,HCO₃^-ion content and p H value of water samples were determined.（3）Analyzing the hydrochemical characteristics of the water bursting source from the measured data,analysis of water quality characteristics of various water sources by using piper trilinear nomograph method,the water quality characteristics of the water source are obtained.The water quality of the goaf water water belongs to SO₄·HCO₃-Ca·Mg type,which is acidic.The water quality of limestone water belongs to HCO₃-Ca type and weakly alkaline.The type of surface water quality belongs to HCO₃-Ca.Mg type,which is neutral.The water quality of fissure water belongs to HCO₃-Ca type and belongs to HCO₃·SO₄-Ca type,which is weakly alkaline.According to the comparison of ionic concentration of different water sources and Box-plot method,it is concluded that Ca²⁺,HCO₃^-and SO₄^2-ions have great differences in different water sources,which can distinguish water source types very well and can be used as a fast discriminant index.（4）Fisher discriminant model,Bayes discriminant model and BP neural network discriminant model of water inrush source are established by using Ca²⁺,HCO₃^-and SO₄^2-as modeling indexes of discriminant models.The results show that the original grouped cases correctly classified rate of Fisher discriminant model is 88.46%,the correct rate of predicted sample discrimination is 77.78%;The results show that the original grouped cases correctly classified rate of Bayes discriminant model is 80.77%,the correct rate of predicted sample discrimination is 77.78%;The results show that the original grouped cases correctly classified rate of BP neural network discriminant model is 100%,the correct rate of predicted sample discrimination is 88.89%.（5）Comprehensive analysis of the discriminant accuracy,advantages and disadvantages of each discriminant model,the selection method of discriminant model for water bursting source in coal mine is put forward.The process first detects the chemical composition of water samples,then preliminarily discriminates the hydrochemical characteristics of the detected data,and finally chooses the appropriate discriminant model to verify according to the preliminary discriminant results.（6）According to the discriminant model selection method,it is applied to Dazhuke coal mine in Fujian.Firstly,the chemical composition of water samples from three water inrush points was detected,and then the hydrochemical characteristics of the detected data were analyzed.The water quality of the three water samples belongs to SO₄·HCO₃-Ca·Mg type,which was acidic,and all of them were primary goaf water.Finally,the BP neural network discriminant model is selected to verify the results,consistent with the preliminary discriminant,the results are all correct,and the ideal application effect is achieved.