Analysis of Fault Classification Method for SDH Optical Fiber Communication Networks Based on SVR

DOI: https://doi.org/10.62517/jhve.202516505

Author(s)

Zhong YunYu, Hu ChunHua

Affiliation(s)

The 34th Research Institute of CETC, Guilin, Guangxi, China

Abstract

This study aims to address the problems of low accuracy, poor adaptability to complex faults, and slow response in traditional fault classification methods for Synchronous Digital Hierarchy (SDH) optical fiber communication networks, and to improve the efficiency and reliability of SDH network fault diagnosis. The research adopts the Support Vector Regression (SVR) algorithm as the core classification tool, combined with fault feature extraction technology and parameter optimization methods. First, a comprehensive SDH network fault dataset is constructed, which includes feature parameters of common fault types such as signal loss (LOS), bit error rate (BER) exceeding the standard, frame alignment error (FAE), and path mismatch, and the dataset is preprocessed through normalization and outlier removal to eliminate interference factors. Second, the SVR model’s key parameters (including kernel function type, penalty factor C, and gamma coefficient) are optimized using the grid search method combined with 5-fold cross-validation to determine the optimal parameter combination that balances classification accuracy and generalization ability. Finally, the preprocessed fault feature data are input into the optimized SVR model for training and testing, and the model’s performance is compared with traditional fault classification methods such as BP neural network and decision tree based on evaluation indicators including accuracy, recall, F1-score, and processing time. The results show that the SVR-based fault classification method achieves an average accuracy of over 96.5%, which is 8.2% and 11.7% higher than that of BP neural network and decision tree respectively; its recall rate for complex faults reaches 95.3%, and the average processing time per sample is reduced by 0.32s compared with traditional methods. This method can effectively identify various faults in SDH optical fiber communication networks, providing a reliable technical support for rapid fault location and maintenance of the network.

Keywords

SVR; SDH Optical Fiber Communication Network; Fault Classification; Parameter Optimization; Classification Performance

References

[1] Li, F. J., & Zhou, X. (2011). SDH Optical Transmission Equipment Commissioning and Maintenance. Beijing: Science Press. [2] Liu, Z. G. (2021). Analysis and processing of SDH equipment failures. Railway Signaling & Communication, (1), 31. [3] Sun, J. K. (2019). SDH digital microwave transmission system in radio and television and its fault analysis and processing suggestions. Science and Technology Innovation Herald, 16(28), 132-133. [4] Lan, H. (2019). Analysis on the reliability of SDH digital microwave transmission in radio and television. China Electronics, (12), 77-78. [5] Qin, B., & Wang, X. Y. (2020). Fault analysis and processing of smart radio and television bidirectional services. Radio & Television Network, 2(12), 99-100. [6] Li, M. T., Wang, Y., & Liu, J. (2021). Core ideas of SDH optical fiber communication equipment maintenance and fault handling. Popular Science, (5), 66-67. [7] Liu, X. M., & Ye, M. (2018). Fault analysis and processing method of SDH transmission network. Scientific Research, (5), 47. [8] Huang, J. H. (2017). Flexible support vector regression algorithm and its application in fault diagnosis. Journal of Zhejiang University of Technology, 45(6), 629-634. [9] Wen, C. L., & Lü, F. Y. (2020). Review of deep learning-based fault diagnosis methods. Journal of Electronics & Information Technology, 42(1), 234-248. [10] Zhang, Z. P., & Zhao, J. S. (2017). Fault diagnosis of chemical processes based on deep belief network. Computers & Chemical Engineering, 107, 39-407. [11] Li, Y. H., & Wang, H. B. (2021). Research on fault prediction and location of optical transport network in high-speed railway. Journal of Beijing Jiaotong University, 45(3), 321-328. [12] Chen, X. Y., & Wang, Y. (2024). Machine learning-based alarm classification and correlation in SDH/WDM optical networks. Optical Engineering, 63(2), 026105. [13] Wang, Y. Q., & Li, X. D. (2025). AI-embedded optical module with millisecond-granularity analysis for autonomous metro transport network. Journal of Lightwave Technology, 43(5), 1234-1242. [14] Ma, S. Q., & Zhang, Y. L. (2023). Research on fault classification method of SDH optical fiber communication network based on improved SVR. Optical Fiber and Optical Communication Technology, 41(4), 56-62. [15] Liu, Y. F., & Zhao, G. Q. (2024). Risk early warning and evaluation method for electric power SDH networks based on BP neural network. IEEE Transactions on Power Delivery, 39(2), 876-883.