Performance evaluation of fault-branch detection in TDM passive optical networks by spectral analysis of interferometric signals

Abstract

A new fault-branch detection scheme is proposed to troubleshoot the breaks of any distribution fibers in a timedivision multiplexing (TDM) passive optical network. We employ a continuous optical frequency sweeper at the optical line terminal (OLT) and an interferometric (IF) device at each optical network unit (ONU). By analyzing the spectrum of the returned combined signals at the OLT, we can obtain the status of all branches. This detection method not only uses a small optical frequency band for surveillance monitoring, but is also simple to operate. Furthermore, a modified architecture is proposed to relax the specifications of IF devices. The tolerance of the IF device length was analyzed using the Monte-Carlo simulation method.

Application of machine learning in automatic fault detection for passive optical network

“…Due to the fact that the data with identified fault point characteristics may be much smaller than the network operation indicator data with timestamps that can be collected by network management, semi supervised learning methods can be considered when creating the model. For sample points with fault points and fault cause identification, support vector machines (SVM), decision trees, random forests, neural network and other methods can be used for modelling, the unlabeled data is classified by feature aggregation using clustering and other methods [12] . After the model is generated, it can be validated using partially annotated fault point data.…”