金属矿开采岩层移动角预测知识库模型及其工程应用  

Knowledge bank model to predict motion angle of terrane in metal deposit and its application in engineering

作　　者：刘钦[1,2] 刘志祥[3] 李地元[3] 李威[2] 

机构地区：[1]北京科技大学土木与环境工程学院,北京100083 [2]山东黄金集团三山岛金矿,山东莱州261442 [3]中南大学资源与安全工程学院,湖南长沙410083

出　　处：《中南大学学报（自然科学版）》2011年第8期2446-2452,共7页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金与上海宝钢集团公司联合资助项目(51074177);国家重点基础研究计划("973"计划)项目(2010CB732004);教育部博士点基金资助项目(200805331147)

摘　　要：在分析国内外大量充填矿山岩层移动研究成果的基础上,用神经网络建立金属矿充填开采岩层移动角与矿体上下盘围岩性质、地质构造、地下水、开采深度、矿体走向长度、开采厚度及矿体倾角9个影响因素的知识库模型。采用梯度下降法与混沌优化方法相结合,使神经网络知识库模型实现大量样本快速训练的同时,避免陷入局部极小,同时提高了模型计算精度。利用神经网络知识库模型的容错和非线性映射功能,对岩层移动角各影响因素的敏感性进行分析。将知识库模型应用于三山岛金矿新立矿区海下开采岩层移动角预测,分析新立矿区海岸竖井受采动影响的安全范围,并提出保护竖井的安全措施。研究结果表明:矿体上下盘围岩普氏系数是影响岩层移动角的决定因素,地下水和上下盘围岩地质构造是岩层移动角的主要影响因素。Based on data of terrane moving angle in domestic and overseas filling mines,a knowledge bank model,which embodies the relations between moving angle of terrane and characters of wall rock on upside and downside deposit,geologic conformation,underground water,mining depth,length towards deposit,mining thickness and deposit angle,was created with a method of neural network.Coupling grading method with chaotic optimization,the neural network model achieved the merit of rapid training and avoided local minimum,a lot of samples were trained,and the calculating precision of model was improved as well.Using the function of permitting fault and mapping of neural network,the sensitivity of per factor to influence moving angle of terrane was analyzed.The knowledge bank model was applied in predicting motion angle of undersea mining in Xinli zone of Sanshandao Gold Mine,and the safe ranges were analyzed,and the safe measures to protect well were put forward.The results show that the Protodrakonov coefficient of wall rock is the first factor to influence moving angle of terrane,and the underground water and geological conformation are the second and the third influencing factors.

关 键 词：岩层移动角 神经网络 知识库模型 

分 类 号：TD8[矿业工程—矿山开采]