摘要
提高巷道掘进效率、减少冒顶事故是实现煤矿安全高效和智能化开采的重要内容,其关键是对巷道围岩的合理支护设计,而煤矿顶板岩层强度的实时智能感知对巷道支护设计至关重要。利用自主搭建的微型钻进实验平台和制作的砂浆试样,开展钻进实验以获取随钻参数,并测定砂浆试样的单轴抗压强度。采用小波阈值法对随钻参数去噪后,分析钻速、转速和砂浆试样强度对推力和扭矩的影响。基于随钻参数构建预测岩石强度的多元线性回归(MLR)模型,利用径向基函数(RBF)神经网络对MLR模型得到的强度预测残差修正,建立MLR-RBF岩石强度组合预测模型,对MLR和MLR-RBF模型进行验证,并利用MLR-RBF模型对粉砂岩、细粒砂岩和粗粒砂岩3种岩石强度进行预测。研究表明:钻速和转速均与钻进推力呈负相关,但随转速增加旋转扭矩也增加,且扭矩值随钻进深度增加而缓慢线性增加;构建的MLR模型的预测相对误差均值为8.58%,MLR-RBF模型的预测相对误差均值为1.75%,证明了MLR-RBF模型的有效性;MLR-RBF模型对岩石强度的预测误差均值为6.67%,该模型对岩石强度的预测效果较砂浆试样差,主要是因为岩石与砂浆的均质性不同。
Improving mine tunneling efficiency and reducing roof caving accident are important contents to realize safe,efficient and intelligent mining in coal mines.The key is to design the reasonable support of roadway surrounding rock,where the intellisense of strength and structure of roof strata in real time is very crucial.Using the self-building micro-drilling experimental platform and the mortar samples,the drilling experiments of mortar samples were carried out to obtain drilling parameters,and the uniaxial compressive strength of mortar samples was measured.After denoising the drilling parameters by wavelet threshold method,the effects of drilling velocity,rotating speed and rock strength on thrust and torque are analyzed,and the rock interface is identified by thrust and torque.The multiple linear regression(MLR)model for predicting rock strength is constructed based on drilling parameters,and the prediction residual error is corrected by radial basis function(RBF)neural network.The combined prediction model of MLR-RBF was established and verified,and the MLR-RBF model was used to predict the strength of siltstone,fine sandstone and coarse sandstone.The results show that drilling velocity and rotating speed are negatively correlated with thrust,but torque raise with the increase of rotating speed,and torque increases slowly and linearly with drilling depth.The average prediction relative error of MLR model is 8.58%,and the average prediction relative error of MLR-RBF model is 1.75%,which proves the effectiveness of MLR-RBF model.The average prediction error of MLR-RBF model for rock strength is 6.67%,and the prediction effect of MLR-RBF model for rock strength is worse than that of mortar sample,mainly due to the difference of homogeneity between rock and mortar.
作者
孙鑫
张少华
程敬义
王东
葛颂
李想
万志军
SUN Xin;ZHANG Shaohua;CHENG Jingyi;WANG Dong;GE Song;LI Xiang;WAN Zhijun(School of Mines,China University of Mining and Technology,Xuzhou,Jiangsu 221116,China;State Key Laboratory of Coal Resources and Safe Mining,China University of Mining and Technology,Xuzhou,Jiangsu221116,China;Deere Group Limited Company,Jining,Shandongg272000,China)
出处
《采矿与安全工程学报》
EI
CSCD
北大核心
2022年第5期981-991,共11页
Journal of Mining & Safety Engineering
基金
国家自然科学基金项目(52274102)。
关键词
钻进参数
岩石强度
小波阈值去噪
MLR-RBF模型
随钻探测
drilling parameters
rock strength
wavelet threshold denoising
MLR-RBF model
measurement while drilling
