According to the statistics of XX operator, 25,000 complaints were received from July to December in 2016. The ratio of faults in optical links and optical modules (O/E conversion modules) on optical line terminals (OLTs) and optical network terminals (ONTs) accounts for 34%. Specifically, the proportion of faults on devices using drop cables caused by urban reconstruction, rat bites, and rain and snow exceeds 20%, and the number of faults on outdoor connectors (such as mechanical splicers) exceeds 6%. If a fault occurs on devices in the drop cable segment, passive components cannot send any information. Therefore, it is difficult to locate the fault.
Because PON is a passive network, passive components cannot send any information. Besides, optical cables are buried in pipelines and hard to detect. Therefore, after a user makes a complaint, the only solution is to go to the user home and use an optical time domain reflectometer (OTDR), a professional test instrument to detect the entire link on an ONT and determine the fault cause.
In 2017, the annual fee for operator XX's home visit is CNY 100 million.
The entire process is divided into 2 phases. The first phase is training. A fault decision tree model is trained using feature data. The second phase is to determine the root cause after a fault occurs. The root cause of the fault is inferred and determined using real-time data and the AI technology.
A. The remote fault locating rate increases from 30% to 80%.
B. The fault locating time on average is shortened from 4 hours to less than 2 hours.