基于主成分分析和支持向量机的矿井水源判别  

Recognition method of mine water sources based on principal component analysis and support vector machine

作　　者：邱梅[1] 施龙青[1,2] 韩进[3] 滕超[1] 景行[1] 冯阳[1] 

机构地区：[1]山东科技大学地球科学与工程学院,山东青岛266590 [2]山东科技大学山东省沉积成矿作用与沉积矿产重点实验室,山东青岛266590 [3]山东科技大学信息科学与工程学院,山东青岛266590

出　　处：《中国科技论文》2015年第3期251-255,共5页China Sciencepaper

基　　金：高等学校博士学科点专项科研基金资助项目(20133718110004);青岛市经济技术开发区重点科技发展计划资助项目(2013-1-62);山东省自然科学基金重点资助项目(ZR2011EEZ002);山东科技大学科研创新团队支持计划资助项目(2012KYTD101);山东科技大学研究生科技创新基金资助项目(YC140103)

摘　　要：针对传统支持向量机法判别矿井水源准确率较低的问题,选取7种水化学成分指标作为矿井水源判别的指标,采用主成分分析(PCA)和支持向量机(SVM)相结合的方法建立PCA-SVM矿井水源判别模型。以山东盛泉矿业有限公司矿井太原组灰岩含水层和奥灰含水层的水化学特征中的27个样本进行训练和检验,并与传统的支持向量机模型结果进行比较,利用PCASVM模型对矿井水源进行判别,分类准确率由66.67%提高到83.40%。研究结果表明,利用PCA-SVM矿井水源判别模型能有效消除判别指标间的相关影响,判别率较高,引进主成分分析算法是必要的。Considering the low accuracy of traditional support vector machine（SVM）method in mine water source recognition,seven kinds of water chemical composition were selected as the sample variables,and a mine water source recognition model was built by combining principal component analysis（PCA）and support vector machine.27 samples obtained from Taiyuan limestone aquifer and Ordovician limestone aquifer of Shandong Shengquan coal mine were trained and the results were compared with those measured by traditional SVM,which indicates that the PCA-SVM model can improve the measuring accuracy from 66.67% to83.40%.The results show that using PCA-SVM recognition model can effectively eliminate relative effects between variables and realize higher accuracy,which proves the necessity of introducing the PCA.

关 键 词：矿井突水 水源判别 主成分分析 支持向量机 

分 类 号：TD745[矿业工程—矿井通风与安全]