Prediction and Classification of the Optical Transport Network Faults Using Hybrid Deep Learning Model

Abstract

The Optical Transport Network is a critical constituent of modern telecommunication infrastructure, enabling high-capacity data transmission channels over the same fiber. However, the occurrence of faults in OTNs can lead to service disruptions and network downtime, causing significant financial losses for telecommunication operators. Many types of research have dealt with alarm analysis, fault detection, identification, localization, and diagnosis of OTN, while fault management is important, they are a reactive method that finds faults after they have occurred. Therefore, there is an upward demand to develop efficient and accurate fault prediction techniques to proactively manage faults and ensure network reliability. This research attempts to enhance fault management capabilities, minimize downtime, optimize operational efficiency, and deliver enhanced services by utilizing a hybrid deep learning algorithm. The research methodology involves a comprehensive analysis of historical fault data to identify fault patterns and trends. This analysis serves as the foundation for developing a fault prediction mode. By leveraging advanced monitoring techniques and data in the real world, potential faults can be predicted promptly, and proactive measures to prevent service disruptions. This empowers network administrators to proactively address potential issues, allocate resources efficiently, and improve overall network performance. Hybrid Deep learning models CNN-LSTM, CNN-BiLSTM, and CNN-GRU were evaluated for Optical Transport Network fault prediction using MSE, MAE, RMSE, and R2 metrics. The CNN-LSTM model outperformed others with the lowest MSE (1.5203), MAE (2.7686), RMSE (1.2330), and highest R2 (0.9392), making it the most effective. Both CNN-BiLSTM and CNN GRU models also showed improvements over their standalone versions, with CNN-GRU slightly better in MAE and R2 but not as robust as CNN-LSTM. Generally, the outcomes of this research contribute to the field of fault management in OTNs, providing valuable insights for telecommunication operators to enhance their fault prediction capabilities.