罗健1 华攸水1 张浩2 曹立超2 蒋晓明2
（1.广东工业大学机电工程学院，广东 广州 510006
2.广东省科学院智能制造研究所，广东 广州 510070）
摘要：针对目前金属壁面焊缝表面缺陷检测主要由人工完成，检测结果依赖检测人员的工作经验，且高空壁面作业较危险的情况，设计一款轮式磁吸附爬行机器人系统。该系统的轮式磁吸附爬行机器人可在金属壁面运动，利用YOLOv5目标检测框架训练神经网络分类模型，并将焊缝表面缺陷分类模型搭载于人机交互端软件，实现金属壁面焊缝表面缺陷的实时检测。
关键词：金属壁面；焊缝表面缺陷；缺陷检测；轮式磁吸附爬行机器人；YOLOv5
中图分类号：TP242.2；TP391.41 文献标志码：A 文章编号：1674-2605(2023)02-0005-05
DOI：10.3969/j.issn.1674-2605.2023.02.005
Crawling Robot System for Detecting Surface Defects of Metal Wall Welds
LUO Jian1 HUA Youshui1 ZHANG Hao2 CAO Lichao2 JIANG Xiaoming2
(1.Guangdong University of Technology, Guangzhou 510006, China
2.Institute of Intelligent Manufacturing, Guangdong Academy of Science, Guangzhou 510070, China)
Abstract: A wheeled magnetic adsorption crawling robot system is designed to address the current situation where the detection of surface defects in metal wall welds is mainly done manually, and the detection results rely on the work experience of the testing personnel. In addition, high-altitude wall operations are more dangerous. The wheeled magnetic adsorption crawling robot can move on the metal wall, and uses the YOLOv5 object detection framework to train a neural network detection model. The detection model is installed in the human-computer interaction software to achieve real-time detection of surface defects in metal wall welds.
Keywords: metal wall surface; weld surface defect; defect detection; wheeled magnetic adsorption crawling robot; YOLOv5
