黄艺航1 蔡凯武2 黄晓智1 袁澄1 梁恩源1 林智海1
(1.广东工业大学机电工程学院,广东 广州 510006
2.广东工业大学先进制造学院,广东 揭阳 522000)
摘要:为解决传统的通信电缆故障检测与定位方法存在的灵敏性不足和智能化程度低等问题,提出基于机器学习的通信电缆故障检测与定位方法。首先,基于行波法检测原理搭建通信电缆故障仿真模型来采集实验数据样本;然后,提出基于粒子群优化-支持向量机(PSO-SVM)的通信电缆故障检测模型,其故障识别准确率达99.4%;接着,提出基于卷积神经网络-长短时记忆(CNN-LSTM)的通信电缆故障定位模型,该模型对故障点定位的平均绝对误差为0.334 9,均方根误差为0.320 8;最后,通过对比实验验证CNN-LSTM的网络准确率较单独使用CNN和LSTM模型分别提高了9.47%和6.2%。
关键词:PSO-SVM模型;CNN-LSTM模型;行波法;通信电缆;故障检测;故障定位
中图分类号:TN391.4 文献标志码:A 文章编号:1674-2605(2024)01-0006-08
DOI:10.3969/j.issn.1674-2605.2024.01.006
Fault Detection and Localization Method of Communication Cables Based on Machine Learning
HUANG Yihang1 CAI Kaiwu2 HUANG Xiaozhi1 YUAN Cheng1
LIANG Enyuan1 LIN Zhihai1
(1.School of Electromechanical Engineering, Guangdong University of Technology, Guangzhou 510006, China
2.School of Advanced Manufacturing, Guangdong University of Technology, Jieyang, 522000, China)
Abstract: To address the issues of insufficient sensitivity and low intelligence in traditional communication cable fault detection and localization methods, a fault detection and localization method of communication cables based on machine learning is proposed. Firstly, based on the principle of traveling wave detection, a communication cable fault simulation model is constructed to collect experimental data samples; Then, a communication cable fault detection model based on Particle Swarm Optimization Support Vector Machine (PSO-SVM) is proposed, with a fault recognition accuracy of 99.4%; Next, a communication cable fault location model based on Convolutional Neural Network Long Short Term Memory (CNN-LSTM) is proposed. The average absolute error of the model for fault location is 0.334 9, and the root mean square error is 0.320 8; Finally, through comparative experiments, it was verified that the network accuracy of CNN-LSTM was 9.47% and 6.2% higher than that of using CNN and LSTM models alone, respectively.
Keywords: PSO-SVM model; CNN-LSTM model; traveling wave method; communication cable; fault detection; fault localization