岩层移动角预测的粗糙集-BP神经网络模型

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。

Rough Set-BP Neural Network Prediction Model of Strata Movement Angle

Rock movement angle is one of the important parameters to characterize the surface movement regularity in metal mine. In order to overcome the difficulties of mechanical model matching,parameters selection,boundary conditions set-ting of the theoretical calculation method and numerical analysis method, and improve the strata movement angle prediction method,firstly,the basic principle of BP neural network and rough set are analyzed,organic combination of them is conducted to establish the rough set-BP neural network model to predict strata movement angle,the rough set is taken as the the front-end processor pretreatment of fuzziness,uncertainty and incomplete information,the BP neural network is used as the core to estab-lish the mapping relationship between input and output;then,a rough set-BP neural network prediction model is established by testing and learning of the 34 sets of actual measured data of rock strata movement samples including 6 factors ( footwall lithol-ogy,upside rock coefficient,ore-body inclination angle,ore-body thickness,mining depth and mining method) ,and it is used to predict the strata movement angle of open-pit to underground mining of Yongping copper mine. The results show that the overall strata movement angles of the mine are 62° (hanging side),68° (heading side),73° (strike) respectively,the prediction re-sults is basically consistent with the ones of the classical prediction methods ( engineering analogy method,numerical simula-tion method,etc. ) ,because of the rough set-BP neural network model with the characteristics of selection the variables scien-tifically,simplify the network prediction model and improving the anti-interference ability of fault tolerance and classification, therefore,it is superior to the classical prediction methods,it is good to improve the prediction precise of mine strata movement angle.