Accurate Fault Location Using Deep Belief Network for Optical Fronthaul Networks in 5G and Beyond

In face of staggering traffic growth driven by fifth generation (5G) and beyond, optical fronthaul networks which host such connections require efficient and reliable operational environments. Fault location has become one of the primary factors for post-fault responses. In this paper, we propose a Deep Belief Network (DBN) based fault location (DBN-FL) model to locate single-link fault of optical fronthaul network in 5G and beyond. The DBN-FL model contains two phases including the hybrid pre-training phase and the Levenberg Marquardt (LM) algorithm-based fine-tuning phase. In the hybrid pre-training phase, we combine the supervised and unsupervised learning to reduce the demand for training samples. In the fine-tuning phase, the LM algorithm is adopted to fine-tune the DBN-FL model. The experimental results indicate that the proposed DBN-FL model can realize high-accuracy fault location as a classifier (accuracy over 96%), and outperforms traditional deep learning (DL) approaches both in terms of location accuracy and training efficiency.