基于人工神经网络模型的岩石特性预测  

Prediction of rock characteristics based on artificial neural network model

作　　者：陈晓君[1] 陈小根 宋刚[1] 陈根龙[1] CHEN Xiaojun;CHEN Xiaogen;SONG Gang;CHEN Genlong(The Institute of Exploration Technology,CAGS,Langfang Hebei 065000,China;Civil and Resource Engineering School,University of Science and Technology Beijing,Beijing 100083,China)

机构地区：[1]中国地质科学院勘探技术研究所,河北廊坊065000 [2]北京科技大学土木与资源工程学院,北京100083

出　　处：《探矿工程（岩土钻掘工程）》2019年第1期34-38,共5页Exploration Engineering:Rock & Soil Drilling and Tunneling

基　　金：国家重点研发计划资助"多金属矿岩心钻探关键技术装备联合研发及示范"(编号:2016YFE0202200);国家自然科学基金"金刚石钻进过程中岩体基本力学参数实时确定方法研究"(编号:51574015)

摘　　要：近年来,软计算技术被用作替代的统计工具。如人工神经网络(ANN)被用于开发预测模型来估计所需的参数。在本研究中,通过利用冲击钻进过程中的一些钻进参数(气压、推力、钻头直径、穿透率)和所产生的声级,建立了预测岩石性质的神经网络模型。在实验室中所产生的数据,用于开发预测岩石特性(如单轴抗压强度、耐磨性、抗拉强度和施密特回弹数)的神经网络模型,并使用各种预测性能指标对所建模型进行检验,结果表明人工神经网络模型适用于岩石性质的预测。In recent years,soft computing technology has been used as an alternative statistical tool.The artificial neural network(ANN)is used to develop predictive models to estimate the required parameters.In this study,a neural network model for predicting rock properties was established by using some of the drilling parameters(pressure,thrust,bit diameter,penetration)and the resulting sound level during the impact drilling process.Data generated in the laboratory was used for the development of neural network models that predict rock properties such as uniaxial compressive strength,wear resistance,tensile strength,and Schmidt rebound number.The models were tested with various predictive performance indicators and the results show that the artificial neural network model is suitable for the prediction of rock properties.

关 键 词：声级 钻头参数 人工神经网络 岩石特性 

分 类 号：P634.1[天文地球—地质矿产勘探] TU452[天文地球—地质学]