摘要
在生产和服役过程中线夹表面会产生裂纹缺陷,降低了线夹的力学性能,给电力供应造成威胁。提出了基于涡流检测的线夹缺陷检测方法,运用主分量分析法提取缺陷信号特征,采用支持向量机法自动识别缺陷。搭建了涡流检测实验系统和3D扫描系统,采用电火花技术制作了深度和长度不同的缺陷试件。实验研究了缺陷长度和深度变化对信号的影响规律,分析了缺陷特征信号的分布。实验结果表明:优化后的核参数可使支持向量机的分类精度达到96%以上,提出的线夹缺陷检测和分类方法可为在役线夹缺陷检测和质量控制提供有效的指导。
In the process of manufacturing and service, cracking defects will occur at the surface of wire clamps due to imperfect manufacturing technique or stress concentration, which will reduce the mechanical properties of wire clamps and pose a threat to the safety of power supply. In this work, eddy current testing based method was proposed to evaluate defects in wire clamps. The support vector machine was adopted to automatically identify defects and the principal component analysis method was uti-lized to extract defect signal features. The experimental setup of eddy current testing and 3D scanner were built. Specimens with defects of different lengths and depths were fabricated by Electrical Discharge Machining （EDM）. The influences of defect length and depth on probe signals were experimentally studied and the distribution of extracted defect characteristic signals was ana-lyzed. The results show that kernel parameters after optimization can improve the classification accuracy of the support vector ma-chine beyond 96%. The defect detection and classification method proposed in current work can provide effective guidance for de-fect detection and quality control of the wire clamps in service.
作者
张兴森
边美华
梁庆国
卢展强
ZHANG Xingsen BIAN Meihua LIANG Qingguo LU Zhanqiang(Electric Pozver Research Institute of Guangxi Pozver Grid Co., Ltd. , Nanning 530023, Chin)
出处
《中国科技论文》
北大核心
2017年第4期454-458,共5页
China Sciencepaper
基金
广西电网有限责任公司科技项目(GX2014-2-0022)
关键词
线夹
涡流检测
特征
分类
支持向量机
wire clamp
eddy current testing
feature
classification
support vector machine
