面向抽水蓄能电站的巡检机器人关键技术研究  

Research on key technologies of inspection robot for pumped storage power station

作　　者：李海峰 许德操 LI Haifeng;XU Decao(State Grid Qinghai Electric Power Company,Xining 810000,China)

机构地区：[1]国网青海省电力公司,青海西宁810000

出　　处：《电子设计工程》2025年第1期113-117,共5页Electronic Design Engineering

基　　金：国网青海省电力公司科技项目(SGQHHBOOFZJS200266)。

摘　　要：针对抽水蓄能电站中混凝土墙体现场检测难度大且传统巡检方式效率低的问题,提出了一种基于深度学习的裂缝检测系统。通过FCN网络和CNN网络的相似性匹配,可以有效地判断混凝土裂缝的位置、尺寸及深度等信息,实现对裂缝的自动检测。同时,设计了巡检机器人硬件构架与软件系统,实现巡检路线自主规划、自主导航和检测区域的自主识别等功能。基于CrackForest数据集对所提检测模型的性能进行了验证测试,结果表明所提模型的识别精度可达87.74%,平均误差仅为0.45,综合性能良好。In response to the difficulty of on⁃site inspection of concrete walls in pumped storage power plants and the low efficiency of traditional inspection methods,A crack detection system based on deep learning is proposed.By matching the similarity between FCN network and CNN network,the location,size,and depth of concrete cracks can be effectively determined,and automatic detection of cracks can be achieved.At the same time,the hardware architecture and software system of the inspection robot were designed to achieve functions such as autonomous planning of inspection routes,autonomous navigation,and autonomous identification of detection areas.The performance of the proposed detection model was verified and tested based on the CrackForest dataset,and the results showed that the recognition accuracy of the proposed model can reach 87.74%,with an average error of only 0.45,indicating good overall performance.

关 键 词：抽水蓄能电站 裂缝检测 图像识别 深度学习 智能巡检 

分 类 号：TP391.44[自动化与计算机技术—计算机应用技术] TN929.5[自动化与计算机技术—计算机科学与技术]