摘要
为研究全尾砂粒径级配特征对充填料性能的影响,提出以分维数和分维数相关系数表征全尾砂的几何特征,并选取灰砂配比、料浆浓度、分维数和分维数相关系数作为BP神经网络输入因子,抗压强度、坍落度和泌水率作为输出因子,建立了充填性能预测的分形—BP神经网络模型。对7个矿山实测数据展开分维数和分维数相关系数计算,并采用BP神经网络进行训练和预测。结果表明:(1)尾砂越细,则粒径级配分维数越大,孔隙分维数就越小,且全尾砂的分维数稍大于分级尾砂的分维数;(2)全尾砂的相关系数在0.71~0.97之间,较分级尾砂离散;(3)分形—BP神经网络模型对充填料性能指标预测的相对误差在8%以内。综上可知:分形理论—BP神经网络相结合的充填性能预测模型具有较好的精度,为充填料性能预测提供了一种新途径。
In order to investigate the impact of whole tailing characteristics of size grading on the performance of backfilling material,this study selected fractional dimension number and correlation coefficient of fractional dimension number to characterize the geometric features of whole tailing.By using cement-sand ratio,slurry concentration, fractional dimension number and correlation coefficient of fractional dimension number as the input factors, compressive strength, slump and bleeding rate as the output factors, a fractal-BP neural network model was constructed to predict the properties of backfilling material.Then data of 7 mines were calculated by the fractal dimension and correlation coefficient of fractal dimension, and the BP neural network was used for the training and prediction.The results showed that the finer the tailing, the bigger the size grading fractional dimension, but contrary to the pore fractal dimension.Furthermore,the fractional dimension of whole tailing is a little higher than grading tailing. The correlation coefficient of the grading tailing is between 0.71 to 0.97,which is more dispersed than that of whole tailing.The relative error is under 8% using fractal-BP neural network model to predict the properties of backfilling material.In a conclusion,the fractal-BP neural network model had a fine precision,which provides a new approach to predict the properties of filling material.
出处
《黄金科学技术》
CSCD
2017年第2期38-44,共7页
Gold Science and Technology
基金
国家自然科学基金项目"金属矿海底基岩开采裂隙分形演化与突水混沌孕育机制"(编号:51674288)资助
关键词
粒径级配
分维数
神经网络
充填料性能
size grading
fractional dimension number
neural network
properties of backfilling material