Aiming at the existing problems of backfill quality in stopping of residual ore in a mine,it was proposed to adopt classified tailings in cemented tailings backfill.The feasibility of considering classified sailings as filling aggregate was established by analyzing physical and chemical properties of tailings.Influencing factors of slurry quality were preliminarily identified by ratio test.To obtain the optimal ratio,the neural network was used to optimize,taking the density of slurry and addition amount in each group as input data,and slump,7 d compressive strength as well as 28 d compressive strength as output data.Data from ratio test were used as samples of training and testing to build the prediction model for BP neural network.The optimal structure of prediction model was selected to be 4-9-3 by comparing the influences of hidden layer nodes on model training process and prediction accuracy.By means of entering the refined ratio parameters into prediction model,optimal samples are searched and the optimal ratio is 1:3:8 for cement-fly ash-tailing.The optimization indicates that adding fly ash can reduce filling cost effectively and increase economic benefit on the premise that the strength is maintained.%针对目前某矿山残矿回采充填体质量所存在的问题,提出采用分级尾砂胶结充填的方案.分析分级尾砂的物化特性,得出分级尾砂作为充填骨料的可行性.通过配比试验,初步确定影响料浆质量的因素.为了得到最优配比,采用神经网络进行优化,以料浆浓度及各组分添加量作为输入因子,塌落度、7d抗压强度及28 d抗压强度作为输出因子,并以配比实验数据为训练和检验样本来建立BP神经网络预测模型.对比隐含层节点数对模型训练过程及预测精度的影响,选取最佳预测模型结构为4-9-3.将配比参数细化输入到预测模型中,从而搜索出优选样本,得到最优配比为所(水泥)∶m(粉煤灰)∶m(尾砂)=1∶3∶8.优化结果表明:在保证强度的前提下,粉煤灰的添加可有效地降低充填成本,经济效益显著.
展开▼
