PCA-BP神经网络在矿山岩溶突水水源判别中的应用研究  

Study on application of PCA-BP neural network in discrimination of karst water inrush source in mine

作　　者：韩忠 王晓丽 施龙青 HAN Zhong;WANG Xiaoli;SHI Longqing(No.6 Institution of Geology and Mineral Resources of Shandong Province,Weihai 264209,Shandong,China;College of Geosciences and Engi-neering,Shandong University of Science and Technology,Qingdao 266590,Shandong,China)

机构地区：[1]山东省第六地质矿产勘查院,山东威海264209 [2]山东科技大学地球科学与工程学院,山东青岛266590

出　　处：《河南理工大学学报（自然科学版）》2023年第1期46-53,共8页Journal of Henan Polytechnic University（Natural Science）

基　　金：国家自然科学基金资助项目(51804184);山东省自然科学基金资助项目(ZR2020KE023)。

摘　　要：肥城煤田奥陶系灰岩水和徐家庄灰岩水水质十分相近,因此,导致该煤田矿井突水水源难以判别,为了解决这一问题,选取突水水源中F,Br,I,Rn和H3BO3等微量元素以及化合物质量浓度作为判别指标,利用SPSS软件进行主成分分析,并将所得主成分代入MATLAB软件,建立PCA-BP神经网络判别模型,对比PCA-BP神经网络模型与BP神经网络模型、系统聚类分析判别模型、Fisher判别分析模型的收敛速度和输出精度。结果表明:PCA-BP神经网络模型判别准确率为100%,具有输出结果精度高、误差小、收敛速度快、训练次数少等优点。该模型对于识别水质相近的灰岩突水水源具有一定应用价值。The water quality of Ordovician limestone and Xujiazhuang limestone in Feicheng Coalfield is very similar,so that it is difficult to discriminate the source of water inrush. In order to solve this problem,five trace elements including F,Br,I,H3BO3and Rn were selected as discriminant indexes. SPSS software was used to model principal component analysis,the PCA-BP neural network model was established by substituting the principal components into MATLAB software. In terms of convergence process and output accuracy,PCA-BP neural network model was compared with BP neural network model,system clustering analysis discriminant model and Fisher discriminant analysis model. The results showed that the accuracy of PCA-BP neural network model was 100%,it hasd the advantages of the highest output accuracy,the smallest error,the fast convergence speed and the few iterations. Therefore,the model had a certain application value for discriminating similar limestone water inrush sources.

关 键 词：肥城煤田 突水水源 PCA-BP神经网络 奥陶系灰岩 徐家庄灰岩 判别精度 

分 类 号：TD745[矿业工程—矿井通风与安全]