首页|基于微振动宽带相位运动放大与深度学习的输电线路索张力测量方法

基于微振动宽带相位运动放大与深度学习的输电线路索张力测量方法

扫码查看
索类构件在输电线路中广泛分布,其张力值及变化情况是影响输电线路本质安全的关键因素,因此也是输电线路工程施工及运维期间状态监测的重点.传统的索张力测量方法存在精度低、环境要求高、难以带电监测等问题,在输电线路中不具备普适性.提出宽带相位运动放大(broadband phase-based motion magnification,BPMM)与深度学习语义分割结合的图像张力测量方法,通过增强图像振动幅度,实现环境激励下输电线路索类构件微振动图像的放大.为去除BPMM算法对于振动视频处理后出现的噪音伪影问题同时提升识别精度,提出基于深度学习U-Net网络与水平集损失熵的联合分割方法来提取索类构件形心,实现了微振动像素变化量的准确拾取,进而通过频域分析得到自振频率并计算索张力.试验及工程应用表明:基于微振动放大的输电线路索类构件张力测量方法能有效识别环境激励下索微小振动变化,测得的索张力值与传感器测量值相比,误差在6%以内,实现了输电线路索类构件张力的高精度、非接触测量,解决了输电线路张力带电测量困难的问题.
Measurement method of cable tension in transmission lines based on micro vibration broadband phase-based motion amplification and deep learning
Cable type components are widely distributed in transmission lines,and their tension values and changes are key factors affecting intrinsic safety of transmission lines.Therefore,they are the focus of state monitoring during engineering construction,operation and maintenance of transmission lines.The traditional method of measuring cable tension has problems of low accuracy,high environmental requirements and being difficult to monitor with electricity,and is not universally applicable in transmission lines.Here,an image tension measurement method combining broadband phase-based motion magnification(BPMM)method with deep learning semantic segmentation was proposed.By enhancing image vibration amplitude,micro vibration image of transmission line cable component under environmental excitation could be magnified.To eliminate noise artifacts caused by BPMM algorithm in vibration video processing and improve recognition accuracy,a joint segmentation method based on deep learning U-Net network and level set loss entropy was proposed to extract the centroid of cable components,and realize accurate picking of micro vibration pixel changes.Then,natural frequency was obtained with frequency domain analysis and cable tension was calculated.Tests and engineering applications showed that the tension measurement method for cable component of transmission line based on micro vibration amplification can effectively identify small vibration changes of cables under environmental excitation;the measured cable tension value has an error of less than 6%compared to the measured value with sensor to realize high-precision,non-contact measurement of cable component tension of transmission line and solve the difficult problem of tension with electricity measurement of transmission line.

transmission linemicro vibrationtension measurementdeep learningimage recognitionvibration frequency

姜岚、叶卿辰、唐波、程若恒、陶文心、黄荥

展开 >

三峡大学 电气与新能源学院,湖北宜昌 443002

湖北省输电线路工程技术研究中心,湖北宜昌 443002

国网河南省电力公司平顶山供电公司,河南平顶山 467000

输电线路 微振动 张力测量 深度学习 图像识别 振动频率

2025

振动与冲击
中国振动工程学会 上海交通大学 上海市振动工程学会

振动与冲击

北大核心
影响因子:0.898
ISSN:1000-3835
年,卷(期):2025.44(1)