首页|基于振动信号的岩石单轴抗压强度钻进预测实验研究

基于振动信号的岩石单轴抗压强度钻进预测实验研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
为了研究钻进振动信号与岩体地质力学参数之间的响应关系,准确、快速地感知、预测岩石单轴抗压强度,开展基于钻进振动信号的岩石单轴抗压强度预测研究.以室内钻取花岗岩、石灰岩、砂岩和煤四类原岩(煤)试件实验为基础,结合傅里叶变换和振动信号降噪方法构建GA-BP神经网络模型,并对比分析降噪前后以及不同降噪方法模型的预测性能.结果表明:钻进振动信号与岩石单轴抗压强度之间有响应关系,应用钻进振动信号可预测岩石单轴抗压强度;采用Adobe Audition软件对振动信号进行降噪处理的GA-BP神经网络预测模型决定系数R2为0.838,均方根误差为7.063 MPa,平均绝对误差为5.347 MPa,其结果优于原始预测模型和一般降噪方法预测模型;与原始预测模型相比,最优降噪模型在预测精度上提升了 6.3％,均方根误差减小1.954MPa,平均绝对误差减小1.621 MPa;同一预测模型中不同岩性的预测效果存在一定差异.降噪信号GA-BP神经网络预测模型对单轴抗压强度有较优秀的预测能力,所用方法可为在岩体地质力学参数随钻测量方面提供基础.
Experimental study of rock uniaxial compressive strength prediction with drilling based on vibration signals
In order to study the response relationship between the vibration signal with drilling and the geomechanical parameters of the rock mass,and to perceive and predict the uniaxial compressive strength of the rock accurately and quickly,a research on the prediction of uniaxial compressive strength of the rock based on the vibration signal with drilling was carried out.Based on indoor drilling experiments of four types of raw rock(coal)specimens,namely granite,limestone,sandstone and coal,the GA-BP neural network model was constructed by combining Fourier transform and vibration signal noise reduction methods,and the prediction performance of the model before and after the noise reduction,as well as the models with different noise reduction methods,were compared and analyzed.The results show that there is a responsive relationship between the vibration signal with drilling and the uniaxial compressive strength of rock,and the uniaxial compressive strength of rock can be predicted by using the vibration signal while drilling.The GA-BP neural network prediction model using Adobe Audition software to denoise the vibration signal has a determination coefficient R2 of 0.838,a root mean square error of 7.063 MPa,and an average absolute error of 5.347 MPa.The results are better than the original prediction model and the general noise reduction method prediction model.Compared with the original prediction model,the prediction accuracy of the optimal noise reduction model is improved by 6.3％,the root mean square error is reduced by 1.954 MPa,and the average absolute error is reduced by 1.621 MPa.There are some differences in the prediction effect of different lithology in the same prediction model.The GA-BP neural network prediction model of noise reduction signal has excellent prediction ability for uniaxial compressive strength.The method can provide a basis for the measurement of rock mass geomechanical parameters while drilling.

rock mechanicsvibration signals with the drilluniaxial compressive strengthFourier transformsignal noise reductionartificial neural networks

郝建、刘河清、刘建康、吕家庆、郑义宁、刘建荣

展开 >

山东科技大学能源与矿业工程学院,山东青岛 266590

山东能源集团鲁西矿业有限公司郭屯煤矿,山东菏泽 274700

内蒙古上海庙矿业有限责任公司院士专家工作站,内蒙古鄂尔多斯 016299

岩石力学 钻进振动信号 单轴抗压强度 傅里叶变换 信号降噪 人工神经网络

国家自然科学基金资助项目国家自然科学基金资助项目山东省自然科学基金资助项目

5217412152204099ZR2022QE203

2024

10.13722/j.cnki.jrme.2023.0960

岩石力学与工程学报
中国岩石力学与工程学会

岩石力学与工程学报

CSTPCD北大核心
影响因子：2.589
ISSN：1000-6915
年,卷(期)：2024.43(6)