摘要
与传统天线相比,泄漏同轴电缆在狭长的传播空间(例如隧道)中具有均匀的电磁信号覆盖特性。泄漏电缆的长期使用以及人为因素不可避免地会导致故障,而不同故障需要不同的处理方法,因此对于漏缆故障类型的识别就极为重要。该文章提出了一种基于标签平滑半监督生成对抗网络(Semi-supervised Generative Adversarial Networks,SSGAN)的漏缆故障类型识别算法。该算法可以利用少量标记数据与大量非标记数据实现较强的分类能力,解决了数据获取困难、标记成本高等问题,并使用了标签平滑技术,使得无监督学习过程更加稳定。实验结果表明,该模型在故障的类型识别下表现良好,使用相同标记数据集的情况下,分类正确率比其他模型更高;相同正确率的情况下,该SSGAN模型使用的标记数据比其他模型更少。
Compared with traditional antennas,leaky coaxial cables have uniform electromagnetic signal coverage in long and narrow propagation Spaces,such as tunnels.Long-term use of leaking cables and human factors will inevitably lead to faults,and different faults need different treatment methods,so it is very important to identify the types of leaking cable faults.In this paper,an algorithm based on tag smooth semi-supervised Generative Adversarial Networks(SSGAN)is proposed for leak cable fault type recognition.The algorithm can use a small amount of labeled data and a large amount of unlabeled data to achieve strong classification ability,solve the problems of difficult data acquisition and high labeled cost,and use label smoothing technology to make the unsupervised learning process more stable.Experimental results show that the model performs well in fault type recognition,and the classification accuracy is higher than other models when the same labeled data set is used.Under the same accuracy,the SSGAN model uses less marker data than other models.
作者
高文笃
刘太君
GAO Wen-du;LIU Tai-jun(Faculty of Electrical Engineering and Computer Science,Ningbo University,Ningbo 315211,China)
出处
《无线通信技术》
2022年第1期11-14,19,共5页
Wireless Communication Technology
基金
国家自然科学基金(U1809203,62071264)。
关键词
半监督生成对抗网络
输入阻抗谱法
故障识别
漏缆监测
semi-supervised Generative Adversarial Networks(SSGAN)
input impedance spectroscopy
fault identification
leak cable monitoring