岩层移动角预测的粗糙集-BP神经网络模型

发布时间：2018-05-31 11:54

本文选题：金属矿山 + 地表移动规律　；参考：《金属矿山》2017年02期

【摘要】：岩层移动角是金属矿山开采表征地表移动规律的重要参数之一。为克服理论计算法和数值分析法在力学模型匹配、参数选择、边界条件设置等方面存在的问题并改进岩层移动角预测方法,首先在分析BP神经网络、粗糙集基本原理的基础上,将两者进行有机结合,构建了粗糙集-BP神经网络模型,该模型将粗糙集作为前端处理器对具有模糊性、不确定性和不完整性的信息进行预处理,将BP神经网络作为核心建立输入、输出间的映射关系;然后通过对34组实测岩层移动样本数据进行学习训练和测试,构建了包含下盘岩性、上盘岩石普氏系数、矿体倾角、矿体厚度、开采深度、采矿方法等6个因素的粗糙集-BP神经网络岩层移动角预测模型,并对永平铜矿露天转地下开采岩层移动角进行了预测。结果表明:该矿山总体岩层移动角的预测值分别为上盘62°,下盘68°,走向73°,预测结果与工程类比法、数值模拟法等传统方法接近。所提方法由于可科学选择变量、简化网络结构以及具有提高容错抗干扰和分类的能力,相对于传统预测方法而言,具有一定的优势,有助于提高矿山开采岩层移动角的预测精度。
[Abstract]:Strata moving angle is one of the important parameters of metal mining to characterize the law of surface movement. In order to overcome the problems in mechanical model matching, parameter selection and boundary condition setting, and to improve the prediction method of strata moving angle, the BP neural network and the basic principle of rough set are analyzed firstly in order to overcome the problems of theoretical calculation and numerical analysis. A rough set BP neural network model is constructed by combining the two methods. The rough set is used as a front-end processor to preprocess the fuzzy, uncertain and incomplete information. BP neural network is used as the core to establish the mapping relationship between input and output, and then through the study training and testing of 34 groups of measured rock strata moving sample data, the paper constructs a method that includes the lithology of the lower disc, the Pushmann coefficient of the upper disc rock, and the dip angle of the orebody. The prediction model of strata movement angle based on rough set BP neural network for six factors, such as orebody thickness, mining depth and mining method, is presented. The results show that the predicted values of the total strata moving angle of the mine are 62 掳for the upper side, 68 掳for the lower side and 73 掳for the strike. The predicted results are close to those of the traditional methods such as the engineering analogy method and numerical simulation method. Because the proposed method can select variables scientifically, simplify the network structure and improve the ability of fault-tolerant, anti-interference and classification, the proposed method has some advantages over the traditional prediction methods. It is helpful to improve the prediction accuracy of strata moving angle in mining.
【作者单位】：江西理工大学应用科学学院;紫金矿业集团股份有限公司;
【基金】：国家自然科学基金项目(编号:51464015) 江西省教育厅科学技术研究重点项目(编号:GJJ151519)
【分类号】：TD325

【相似文献】