神经网络与遗传算法预测充填配比的研究  

Filling ratio prediction with neural network and genetic algorithm

作　　者：黄永刚[1] 饶运章[1] 刘剑[1] 张学焱 

机构地区：[1]江西理工大学资源与环境工程学院,江西赣州341000

出　　处：《有色金属科学与工程》2016年第5期76-80,共5页Nonferrous Metals Science and Engineering

基　　金：2011年度江西省安全生产重大课题(JXAJ2011002)

摘　　要：为确定最优充填配比方案,在实验的基础上,基于神经网络遗传算法,预测全局最优充填实验条件,最优实验条件为灰砂质量比0.202 4,养护时间5.863 d,溶度67.8%,最大充填体抗压强度0.677 7 MPa,与实际最佳配比方案灰砂质量比1∶4、养护天数28 d、溶度75%、最大抗压强度5.48 MPa相差较大,预测结果不是很满意.说明该方法有较强的适用条件,神经网络的预测精度对遗传算法的极值寻优有影响,建议扩大样本的数量.On the basis of neural network and genetic algorithm, the optimal filling experimental conditions were predicted, which including： gray sand ratio 0.202 4, curing time 5.863 d, solubility 67.8 %, and the maximum compressive strength backfill 0.677 7 MPa. The predicted plan is quite different from the actual optimal matching program： gray sand ratio 1：4, curing time 28 d, solubility 75 %, and the maximum compressive strength 5.48 MPa. It shows that the forecast results are not exact and the prediction method can′t be applied in all conditions. The prediction accuracy of neural networks has effect on the optimization of extreme genetic algorithm. The suggestion of expanding sample size is put forward.

关 键 词：神经网络 充填体 抗压强度 遗传算法 充填配比方案 

分 类 号：TD853.34[矿业工程—金属矿开采] TP389.1[矿业工程—矿山开采]