摘要
以某大型铅锌矿为研究对象,针对绿色矿山建设中存在的矿体回采损失、贫化率高的问题,提出了采用上向进路充填采矿法的回采方案,采用矿业软件建立了上向进路充填采矿法三维可视化模型,对采矿工艺进行了分析。设计了36组充填材料配比试验,获得了不同配比的试验数据。采用响应面法(RSM)研究了不同养护龄期下砂浆浓度、灰砂比对充填体强度形成的影响规律,建了多因素非线性回归模型。研究结果表明多因素的交互作用对充填体强度形成具有较大的影响,确定当输送砂浆浓度介于70%~72%、灰砂比介于1∶6~1∶10时,矿山可以取得最大的经济及安全效益。采用小波神经网络将36组试验数据作为训练样本,通过多次迭代,获得了3d、7d、14d、28d、60d充填强度预测模型,分析发现建立的预测模型可靠,可用于指导生产实践中的充填配比,为充填配比优化提供了一种新的思路。
Taking a large lead-zinc mine as the research object,aiming at the problems of high stoping loss and dilution rate of orebody in green mine construction,a stoping scheme using the upward approach backfill mining method was put forward,and a three-dimensional visualization model of the upward approach backfill mining method was established with mining software,and the mining process was analyzed.The experiment of filling material ratio of 36 was designed and the test data of different ratio were obtained.Using the response surface method(RSM)was studied under different curing ages,slurry concentration,contrast ratio on the strength of filling body formation rule,built a multi-factor nonlinear regression model,the results showed that many factors interaction has great influence on the formation strength of filling body,determine when transporting slurry concentration between 70%~72%,the contrast between 1∶6~1∶10,The mine can achieve the greatest economic and safety benefits.Using the wavelet neural network,36 groups of test data were taken as training samples.Through several iterations,the 3 d,7 d,14 d,28 d and 60 d filling strength prediction models were obtained.The analysis showed that the established prediction model was reliable and could be used to guide the filling ratio in production practice,providing a new idea for the optimization of filling ratio.
作者
张国胜
张雄天
ZHANG Guosheng;ZHANG Xiongtian(Lanzhou Engineering&Research Institute of Nonferrous Metallurgy Co.,Ltd.,Lanzhou 730000,China)
出处
《金属矿山》
CAS
北大核心
2021年第12期112-117,共6页
Metal Mine
关键词
上向进路充填采矿法
充填试验
响应面分析
小波神经网络
upward drift filling mining method
filling test
response surface analysis
wavelet neural network
