基于Weibull模型的高台阶抛掷爆破爆堆形态BP神经网络预测  

BP neural network forecast of blasting muck pile form of high bench cast blasting based on Weibull model

作　　者：韩亮[1,2] 刘殿书[1] 李红江 王宇涛[1] 

机构地区：[1]中国矿业大学(北京)力学与建筑工程学院,北京100083 [2]北京天地华泰采矿工程技术有限公司,北京100013 [3]北京宝地益联地质勘查工程技术有限公司,北京100011

出　　处：《煤炭学报》2013年第11期1947-1952,共6页Journal of China Coal Society

摘　　要：为了预测高台阶抛掷爆破爆堆形态,引入Weibull模型作为数学模型对实测爆堆曲线进行模拟,在此基础上,将试验中影响爆堆形态的8个参数与模拟得到的Weibull控制参数及松散系数作为输入输出层学习样本,建立BP神经网络。结果表明:Weibull模型可以较好地模拟真实爆堆。利用训练完成的BP神经网络预测爆堆形态时,各参数的预测误差均未超过5%,预测爆堆曲线接近真实爆堆曲线,其中Weibull模型控制参数的预测精度直接影响预测结果,松散系数ξ的预测精度则在将Weibull概率密度函数反无量纲化的过程中间接影响预测效果。To forecast blasting muck pile form of high bench cast blasting, Weibull model was introduced as mathemati- cal model to simulate actual muck pile curve, and then, used eight parameters that tested and proven to have influence on muck pile form and simulated Weibull control parameters, and bulk factor as studying sample on input and output layers so that to establish BP neural network. The results show that Weibull model can simulate real blasting muck pile comparatively well. When the above simulated BP neural network is used for forecasting the muck pile, the error of each parameter is not more than 5% ,and the forecasted blast curve is close to actual muck pile curve. Of all parame- ters, the Weibull model control parameters have direct impact on the forecasted result, and predicted accuracy of bulk factor has indirect impact on the result during nondimensionalization of Weibull probability densitv function.

关 键 词：高台阶抛掷爆破 爆堆形态模拟 爆堆形态预测 Weibull模型 BP神经网络 

分 类 号：TD235.31[矿业工程—矿井建设]