矿井突水水源的Bayes逐步判别分析研究

对Bayes逐步判别法在矿井突水水源判别中的应用进行研究分析。选用六大常规离子（ Ca2＋、Mg2＋、K＋＋Na＋、SO42-、Cl-、HCO3-）作为判别因子,建立Bayes逐步判别分析模型,以内蒙唐家汇矿区突水水源判别为例,在建立的判别模型回判检验准确率仅60%,分析原因可能与选定的特征判别因子对该矿区水样分类影响能力较弱有关。增加总硬度、碱度、PH值和矿化度作为判别因子,重新建立Bayes判别分析模型,使回判准确率提高至90%,证明适当增加特征判别因子对改善Bayes逐步判别模型的可靠性和稳定性有利。经对唐家会矿区的3个未知样本进行了判别分析,并与距离判别法和模糊综合评判法判别结果对比,结果表明Bayes逐步判别模型准确性较好,判别准确率与距离判别结果完全相同,而优于模糊综合评判方法。在合理选取特征判别因子的情况下,Bayes逐步判别法是目前矿井突水水源判别的有效方法。

Application of Bayes Stepwise Discrimination Analysis on Identification of Mine Well Inrush-water Resource

Taking the six regular indexes （ Ca2＋, Na＋ ＋K＋, Mg2＋, HCO3 -, Cl-, SO4 2-, PH and TDS） as discrimination factors, the paper builds a stepwise discrimination model to identify mine well inrush-water resource in diggings in Tangjiahui, In-ner Mongolia. Actually, the primary research accuracy is only 60%, which may be affected by the indexes＆#39; weak effect on water sample. Then, the paper uses total hardness, alkalinity and PH value as discrimination factors, and the accuracy increases to 90%. The change indicates that appropriate addition of discrimination factors will improve research accuracy and stability. Moreover, the paper analyzes three unidentified water samples by using Bayes discrimination method, distance discrimination method and fuzzy comprehensive evaluation respectively, and compares their results, which shows that result of first method is identical with that of the second one, which are more accurate than that of the third one. To conclude, with proper discrimination factors, Bayes discrimi-nation method is the most efficient method to identify mine well inrush-water resource.