基于PCA-Bayes综合判别方法的祁东矿煤层顶板突水水源判别研究

Research on the Source of Water Inrush from Coal Roof in Qidong Mine Based on PCA-Bayes Comprehensive Discrimination Method

摘要: 祁东煤矿煤层顶板压架突水事故频发，准确、快速地识别突水水源是防止突水再次发生的关键。为此，从祁东煤矿二类含水层中提取了28个训练水样以及潘二矿三类含水层的11个水样作为验证样本，以Ca²⁺、Mg²⁺、Na⁺+ K²⁺、HCO₃^-、CI^-、SO₄^2-为评价变量。采用主成分分析法消除训练样本中的冗余离子变量，然后结合Bayes法建立模型，判别结果表明：新生界松散含水层(I)水样判别正确率为100%，煤系砂岩水(II)水样判别正确率为66.67%，太原组灰岩水(III)水样判别正确率为100%，模型的综合精度为90.91%。与单一Bayes判别方法相比，该方法具有准确率高、速度快等优点，基于该模型判别结果，可在煤矿突水事故发生后快速查明突水原因，有效预防矿井水害。

Abstract: Because water inrush accidents occurred frequently in Qidong Coal Mine’s roof pressing support, it’s the key to identify the source of water inrush accurately and rapidly to prevent water inrush from happening again. Based on the hydrogeological data collected in Qidong Coal Mine, according to the hydrochemical characteristics of aquifer, Ca²⁺, Mg²⁺, Na⁺+ K²⁺, HCO₃^-、CI^-、SO₄^2- were selected to establish a water inrush source discrimination model based on the combination of Principal Component Analysis and Bayes discrimination. Taking 28 water samples from Qidong Coal Mine as training samples and 11 water samples from Pan NO.2 Coal Mine as verification samples, the model is tested and applied. The Principal Component Analysis method is used to eliminate redundant ion variables in the training samples, and then the Bayes method is used to establish a model. The discriminant results show that the discriminant accuracy of Cenozoic loose aquifer (I) water samples is 100 %, the discriminant accuracy of coal measure sandstone water (II) water samples is 66.67 %, the discriminant accuracy of Taiyuan Formation limestone water (III) water samples is 100 %, and the comprehensive accuracy of the model is 90.91 %. Compared with the direct Bayes discrimination, the model reduces the calculation error caused by a large number of redundant original data and improves the calculation accuracy, makes the discrimination more accurate.