人工神经网络在矿井多水源识别中的应用  

Application of artificial neural network for recognition of multiple water sources in mine

作　　者：徐星[1] 郭兵兵[1] 王公忠[1,2] 

机构地区：[1]河南工程学院安全工程学院,河南郑州451191 [2]武汉理工大学资源与环境工程学院,湖北武汉430070

出　　处：《中国安全生产科学技术》2016年第1期181-185,共5页Journal of Safety Science and Technology

基　　金：国家"十二五"科技支撑计划重点项目(2011BAB05B03);建设经费项目(200925);2016年度河南省高等学校重点科研项目(16A44001)

摘　　要：以焦作矿区水化数据为例,选用Na++K+、Ca2+、Mg2+、Cl-、SO2-4、HCO-36种水化离子浓度作为识别矿井水源的依据,运用Matlab软件分别建立BP神经网络与Elman神经网络的矿井多水源识别模型。应用结果表明:与静态的BP神经网络相比,在结构上增加承接层的Elman神经网络在训练和仿真中拟合能力更强,识别精度更高和泛化能力更好;矿井地下水随着地下开采与扰动具有动态性,将具有非线性动态特征的Elman神经网络应用于矿井多水源的识别,对准确判断突水来源和分析地下水运移规律具有一定的辅助和指导意义。Taking the hydration data of Jiaozuo mining area as example,and using the concentrations of six hydrated ion including Na^＋＋ K^＋,Ca^2＋,Mg^2＋,Cl^-,SO4^2- and HCO3^- as the basis for recognition of water sources in mine,the recognition models of multiple water sources in mine were established based on BP neural network and Elman neural network respectively by using Matlab software. The application results showed that compared with the static BP neural network,the fitting ability of Elman neural network which added an undertaking layer in structure in training and simulation was stronger,the recognition accuracy was higher and the generalization ability was better. The mine groundwater has dynamic nature with underground mining and disturbance. It has certain auxiliary and guidance to accurately determine water inrush sources and analyze the migration laws of groundwater by applying the Elman neural network with the nonlinear dynamic characteristics in recognition of multiple water sources in mine.

关 键 词：矿井多水源 BP神经网络 ELMAN神经网络 识别 泛化能力 

分 类 号：X936[环境科学与工程—安全科学]