Research on Event Detection Technology for Fiber Optic Cable based on CNN

ZHANG Jianyu; XU Ruiwan; SHEN Li; YU Wenhai; MEI Liang

doi:10.13756/j.gtxyj.2026.240183

您当前的位置：

首页 >

文章列表页 >

Research on Event Detection Technology for Fiber Optic Cable based on CNN

更新时间：2026-03-03

- Research on Event Detection Technology for Fiber Optic Cable based on CNN
- Study on Optical Communications Issue 1, (2026)
- 作者机构：
  
  1.武汉邮电科学研究院，武汉　 430074
  2.烽火通信科技股份有限公司　网络产出线，武汉　 430205
- 作者简介：
- 基金信息：
- DOI：10.13756/j.gtxyj.2026.240183
  CLC： TN929
- Received：26 August 2024，
  
  Revised：2024-09-12，
  
  Published：10 February 2026
- 稿件说明：
移动端阅览
张剑羽，徐瑞晚，沈力，等. 基于CNN的通信光缆事件检测技术研究[J]. 光通信研究，2026(1): 240183.

Zhang J Y, Xu R W, Shen L, et al. Research on Event Detection Technology for Fiber Optic Cable based on CNN[J]. Study on Optical Communications, 2026(1): 240183.
张剑羽，徐瑞晚，沈力，等. 基于CNN的通信光缆事件检测技术研究[J]. 光通信研究，2026(1): 240183. DOI： 10.13756/j.gtxyj.2026.240183.

Zhang J Y, Xu R W, Shen L, et al. Research on Event Detection Technology for Fiber Optic Cable based on CNN[J]. Study on Optical Communications, 2026(1): 240183. DOI： 10.13756/j.gtxyj.2026.240183.

摘要

目的

光纤故障识别是无源光网络中的重要功能。随着信息技术的高速发展与通信容量需求的不断升级，光网络体系结构变得越来越复杂，对于其中光纤链路事件点的定位检测也越发困难。光时域反射仪（OTDR）作为光纤故障检测的重要手段，能够定位光纤中的事件位置和评估光纤的状态。针对目前OTDR事件检测方法存在阈值固定、噪声干扰和定位精度不高等问题，文章提出了一种以卷积神经网络（CNN）为主要框架的OTDR事件检测新方法。

方法

模型采用仿真数据进行训练，以经过预处理后的时域曲线及其对应的方差和差分曲线作为输入，并预测OTDR迹线上事件的精确位置。在CNN模型的基础上构建高效通道注意力（ECA）机制。通过增强通道间的特征表示，自适应地调整不同通道的重要性，学习OTDR迹线的多层次特征，提升模型的性能。

结果

文章中实验以OTDR为基础完成测试数据采集。在采集的数据集上，模型预测事件的召回率达到93.0%，精确率达到94.7%。

结论

相比传统的事件检测方法，文章所提方法无需阈值选择，仅通过归一化预处理就可以自动学习和提取特征进行事件识别。实验结果显示，该方法具有较高的识别准确率与数据适应性，为相关的实践开发提供了有效的理论验证与实例展示。

Abstract

Objective

Fiber fault identification is an important function in passive optical networks. With the rapid development of information technology and the continuous upgrading of communication capacity demand

the optical network architecture has become increasingly complex

and the location and detection of event points in fiber links have thus become more challenging. Optical Time-Domain Reflectometer (OTDR)

as an important means of fiber fault detection

is able to locate the events in the fiber and assess the state of the fiber. Aiming at the problems of fixed threshold

noise interference

and low localization accuracy that exist in the current OTDR event detection methods

this paper proposes a new OTDR event detection method using Convolutional Neural Networks (CNN) as the main framework.

Methods

The model is trained with simulated data

taking the preprocessed time-domain curves as well as their corresponding variances and difference curves as input

and predicts the precise locations of events on the OTDR traces. Efficient Channel Attention (ECA) mechanism is constructed on the basis of CNN model. By enhancing the feature representation between channels

adaptively adjusting the importance of different channels

and learning the multilevel features of OTDR traces

the performance of the model is improved.

Results

The experiments are completed with OTDR as the basis for test data acquisition. On the experimental dataset

the model could allocate events with recall of 93.0% and precision of 94.7%.

Conclusion

Compared with traditional event detection methods

this event detection method can automatically learn and extract the features only using normalized OTDR data that is independent of the chosen threshold. The results show that the method has high detection accuracy and data adaptability

which provides an effective theoretical validation and example demonstration for the practical development.

关键词

Keywords

references

刘中华 , 龚浩敏 , 吴岩 , 等 . 相干光时域反射仪技术与应用 [J ] . 光通信研究 , 2022 ( 5 ): 43 - 48 .

Liu Z H , Gong H M , Wu Y , et al . Principle and Application of the C-OTDR [J ] . Study on Optical Communications , 2022 ( 5 ): 43 - 48 .

陈诚 , 肖逸 , 李爱东 , 等 . 基于RBF神经网络的OTDR小波分析算法 [J ] . 光通信技术 , 2017 , 41 ( 2 ): 21 - 24 .

Chen C , Xiao Y , Li A D , et al . OTDR Events Analysis Algorithm based on Wavelet Transform and RBF Neural Network [J ] . Optical Communication Technology , 2017 , 41 ( 2 ): 21 - 24 .

孔衡 . 光时域反射的事件分析算法研究 [D ] . 上海 : 上海交通大学 , 2015 .

Kong H . Research on Events Analysis Algorithms in Optical Time Domain Reflectometry [D ] . Shanghai, China : Shanghai Jiao Tong University , 2015 .

臧益鹏 , 李现勤 , 吴松桂 , 等 . 基于改进小波变换的OTDR事件检测方法 [J ] . 光通信技术 , 2024 , 48 ( 1 ): 23 - 28 .

Zang Y P , Li X Q , Wu S G , et al . Event Detection Method of OTDR based on Improved Wavelet Transform [J ] . Optical Communication Technology , 2024 , 48 ( 1 ): 23 - 28 .

钟志宏 , 文科 , 王荣 . OTDR事件检测和定位算法 [J ] . 解放军理工大学学报(自然科学版) , 2004 ( 5 ): 22 - 25 .

Zhong Z H , Wen K , Wang R . Event Detection and Location in OTDR Data [J ] . Journal of PLA University of Science and Technology (Natural Science Edition) , 2004 ( 5 ): 22 - 25 .

唐良瑞 , 祁兵 , 陈常洪 , 等 . 一种基于小波变换的光纤故障分析算法 [J ] . 中国电机工程学报 , 2006 , 26 ( 2 ): 101 - 105 .

Tang L R , Qi B , Chen C H , et al . Algorithm of Fiber Fault Analysis based on Wavelet Transform [J ] . Proceedings of the CSEE , 2006 , 26 ( 2 ): 101 - 105 .

Abdelli K , Cho J Y , Azendorf F , et al . Machine-Learning-based Anomaly Detection in Optical Fiber Monitoring [J ] . Journal of Optical Communications and Networking , 2022 , 14 ( 5 ): 365 - 375 .

Abdelli K , Tropschug C , Griesser H , et al . Faulty Branch Identification in Passive Optical Networks Using Machine Learning [J ] . Journal of Optical Communications and Networking , 2023 , 15 ( 4 ): 187 - 196 .

Straub M , Saier T , Reber J , et al . AI-based OTDR Event Detection, Classification and Assignment to ODN Branches in Passive Optical Networks [C ] // 49th European Conference on Optical Communications (ECOC 2023) . Glasgow, UK : IET , 2024 : 10484603 .

吴俊 , 管鲁阳 , 鲍明 , 等 . 基于多尺度一维卷积神经网络的光纤振动事件识别 [J ] . 光电工程 , 2019 , 46 ( 5 ): 180493 .

Wu J , Guan L Y , Bao M , et al . Vibration Events Recognition of Optical Fiber based on Multi-Scale 1D CNN [J ] . Opto-Electronic Engineering , 2019 , 46 ( 5 ): 180493 .

徐杰伊 . OTDR动态范围提高及事件检测研究 [D ] . 哈尔滨 : 哈尔滨工业大学 , 2020 .

Xu J Y . Study on OTDR Dynamic Range Improvement and Event Detection [D ] . Harbin, China : Harbin Institute of Technology , 2020 .

Wang Q , Wu B , Zhu P , et al . ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks [C ] // 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Seattle, WA, USA : IEEE , 2020 : 9156697 .

杨光乔 , 李颖 , 王国程 , 等 . 基于ECA改进1D-CNN的柱塞泵故障诊断 [J ] . 石油机械 , 2023 , 51 ( 11 ): 34 - 40 .

Yang G Q , Li Y , Wang G C , et al . Fault Diagnosis of Plunger Pump based on ECA Improved 1D-CNN [J ] . China Petroleum Machinery , 2023 , 51 ( 11 ): 34 - 40 .

Views

下载量

CSCD

CNKI被引量

Alert me when the article has been cited

提交

Tools

Publicity Resources

Wavelet Denoising Method based on Improved FOX Algorithm in OTDR System

Research Progress on MgF₂ Microcavity Optical Frequency Combs

Photonic Integration and CPO Applications based on TGV Interposer

Visible Light Frequency Comb Generation based on High-Frequency Mixing Wave Driving

Fiber Nonlinear Impairment Compensation based on Multi-source Domain Transfer Learning

Related Author

LI Xixian

LI Changyuan

WANG Pan

WANG Mengyu

GE Chang

DU Jiangbing

SHEN Jiaxin

LEI Ting

Related Institution

Graduate Department, Wuhan Research Institute of Posts and Telecommunications

Network Output Line, FiberHome Telecommunication Technology Co., Ltd.

Key Laboratory of Nondestructive Test (Ministry of Education), Nanchang Hangkong University

State Key Laboratory of Photonics and Communications, Shanghai Jiao Tong University

a. Nanophotonics Research Center, Institute of Microscale Optoelectronics; b. State Key Laboratory of Radio Frequency Heterogeneous Integration, Shenzhen University

Address：3rd floor, Building 4, Fenghuo Science Park, Gaoxinsilu, East Lake New Technology Development Zone, Wuhan, Hubei Province, P.R.China Postal code：430205
Technical support is provided by Beijing Founder electronics co., LTD 鄂ICP备2022010197 鄂公网安备 42011202002092号
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
Total Visits：312958 Visits Today：124

⁰