Method for monitoring an optical communications system
Data di pubblicazione
08-09-2021Codice
DEIB.18.004.AStato
DisponibileData di priorità
08-03-2018Fase
Italian and International (PCT) Patent GrantedTitolare
Politecnico di MilanoDipartimento
Department of Electronics, Information and BioengineeringAutori
Francesco Musumeci, Massimo Tornatore, Achille Pattavina, Shahkarami ShahinDescrizione
The invention consists of a method for detecting failures in optical communication systems and identifying the source cause of the failure.
The method exploits the monitoring of optical transmission quality parameters, such as Bit Error Rate (BER) and/or Optical Signal to Noise Ratio (OSNR), directly at the receiver of an optical communication system. Failure detection and source cause identification are based on Machine Learning and data analytics techniques.
Continuous BER (and/or OSNR) monitoring is performed at the receiver of an optical communication system, by collecting BER traces at a given sampling interval (TBER). Every W seconds a BER window is formed (see figure). This window is then analyzed to extract several features (e.g., minimum/maximum/average BER values in the window) and is given in input to the ML detection-identification module. Note that the ML-based module is pre-trained and optimized for the specific communication system under consideration. The ML module classifies the BER window and decides if it includes an anomaly in the BER behaviour and, if this is the case, the source cause of the anomaly is also outputed. Several consecutive sliding windows are observed and, after a number (to be decided by the system manager) of consecutive «anomaly alarms», equipment reparation can be triggered, after which BER monitoring is restarted.