基于智能自学习模型识别岩芯的RQD标定方法  

RQD Calibration Method for Core Recognition Based on Intelligent Self-Learning Model

作　　者：赵兴东[1] 王宏宇 王小兵 王立君[4] ZHAO Xingdong;WANG Hongyu;WANG Xiaobing;WANG Lijun(School of Resources and Civil Engineering,Northeastern University,Shenyang,Liaoning l10819,China;Sinosteel Maanshan General Institute of Mining Research Co.,Ltd.,Maanshan,Anhui 243000,China;State Key Laboratory of Safety and Health for Metal Mines,Maanshan,Anhui 243000,China;Shandong Gold Group Co.,Ltd.,Jinan,Shandong 250100,China)

机构地区：[1]东北大学资源与土木工程学院,辽宁沈阳110819 [2]中钢集团马鞍山矿山研究总院股份有限公司,安徽马鞍山市243000 [3]金属矿山安全与健康国家重点实验室,安徽马鞍山市243000 [4]山东黄金集团有限公司,山东济南250100

出　　处：《矿业研究与开发》2023年第12期159-165,共7页Mining Research and Development

基　　金：国家自然科学基金重点项目(52130403);辽宁省2023年第一批中央引导地方科技发展资金项目(2023JH6/10010050)。

摘　　要：RQD分级是分析岩体工程地质条件和评价岩体完整性的重要手段。针对传统人工测量并编录RQD工作量较大且效率低下的问题,基于智能自学习(Inception-v3卷积神经网络迁移学习)模型,通过对地质钻探中拍摄的大量岩芯图片进行特征提取和迁移学习,建立岩芯-岩芯盒识别模型,实现对长度大于10 cm岩芯的自动识别和RQD获取,进而帮助钻孔岩芯编录和岩体质量评价。三山岛西岭矿区的应用结果表明,智能自学习模型识别岩芯的RQD标定结果与传统人工方法所得结果差距仅为1.78%,方便快捷的同时准确性较高,适合应用于实际矿山工程中。RQD classification is an important means to analyze the engineering geological conditions of rock mass and evaluate the integrity of rock mass.According to the large workload and low efficiency of traditional method of manually measuring and cataloging RQD,based on the intelligent self-learning(Inception-v3 convolutional neural network transfer learning)model,a core-core box recognition model was established by feature extraction and transfer learning from a large number of core images taken in geological drilling.It can realize the automatic recognition and RQD acquisition of cores longer than 10o cm in length,and then help drilling core logging and rock mass quality evaluation.The application results of the Xiling mining area in Sanshan Island show that the difference between the calculation results of the intelligent self-learning model and the results obtained by the traditional manual method is only 1.78%in RQD calibration for core recognition.The proposed method is convenient,fast,and has a high accuracy,making it suitable for practical mining engineering.

关 键 词：智能自学习 岩芯识别 RQD 卷积神经网络 

分 类 号：TD313[矿业工程—矿井建设] P642[天文地球—工程地质学]