尾矿库坝面安全隐患智能巡检机器人研发及应用  

Development and Application of Intelligent Inspection Robot for Safety Hazards on Tailings Dam Surfaces

作　　者：孟庆胤 王浩宇 聂明哲 余小军 陈荣健 傅金阳[2,3] MENG Qingyin;WANG Haoyu;NIE Mingzhe;YU Xiaojun;CHEN Rongjian;FU Jinyang(Yichun Luming Mining Co.,Ltd.,Tieli,Heilongjiang 152500,China;School of Civil Engineering,Central South University,Changsha,Hunan 410075,China;National Engineering Research Center of High-Speed Railway Construction Technology,Central South University,Changsha,Hunan 410075,China)

机构地区：[1]伊春鹿鸣矿业有限公司,黑龙江铁力市152500 [2]中南大学土木工程学院,湖南长沙410075 [3]中南大学高速铁路建造技术国家工程研究中心,湖南长沙410075

出　　处：《矿业研究与开发》2024年第7期230-238,共9页Mining Research and Development

基　　金：中国国家铁路集团有限公司科技研究开发计划项目(L2022G003);中南大学研究生创新项目资助项目(2022037)。

摘　　要：拉沟、冲沟等病害的及时检测是确保尾矿坝安全性的关键。以尾矿坝为工程背景,基于履带式底盘系统、自动化控制等技术研制了一款用于坝面拉沟与冲沟检测的智能巡检机器人。针对坝面病害检测目标,提出了一种在YOLOv5m主干网络和颈部网络插入卷积块注意力模块的改进网络模型YOLOv5m-ECA,并在尾矿库坝面无人巡检作业中开展了应用研究。结果表明,改进YOLOv5m-ECA算法使模型的准确率、平均精度均值、F_(1)分数相较于改进前分别提高12个百分点,6.1%和3.6个百分点,与目前4种主流目标检测算法的性能相比,YOLOv5m-ECA的综合性能更强,且易于部署在移动检测装备,更适用于坝面拉沟病害的检测。现场应用表明,该方法能够代替人工实现尾矿库坝面的无人化作业,能够为坝面病害的快速检测提供智能化方案,所检测病害位置与实际坝面位置相符,具有很好的现实意义和应用价值。Timely detection of diseases such as slotting and gulch is the key to ensure the safety of tailings dam.Taking the tailings dam as the engineering background,an intelligent inspection robot for slotting and gulch on the tailings dam surface was developed by utilizing technologies such as tracked chassis systems and automation control.An improved network model YOLOv5m-ECA with convolutional block attention modules inserted into the YOLOv5m backbone network and neck network was proposed for the detection of dam surface diseases.Application research was conducted in unmanned inspection operations on tailings dam surfaces.The results show that the improved YOLOv5m-ECA algorithm improves the model's accuracy,mean average precision,and F_(1) score by 12 percentage points,6.1%,and 3.6 percentage points,respectively,compared to those before improvement.Compared with the performance of four mainstream object detection algorithms,YOLOv5m-ECA demonstrates stronger overall performance and is easily deployable on mobile detection equipment,making it more suitable for slotting and gulch detection on dam surfaces.Field applications have shown that this method can replace manual operations for unmanned inspection of tailings dam surfaces,providing an intelligent solution for rapid disease detection on dam surfaces.The detected disease positions correspond to the actual dam surface positions,demonstrating practical significance and application value.

关 键 词：尾矿库坝面安全 智能巡检 机器人研发 拉沟识别 注意力机制 

分 类 号：TD926.4[矿业工程—选矿]