Utilizing AI For Predictive Maintenance In GPON

One of the most impactful and insightful ways of ensuring proactive troubleshooting, improved uptime, customer experience and operational efficiency is using AI for predictive maintenance in GPON.
To predict failure AI would require diverse information such as but not limited to: Optical power levels (Tx/Rx), port/device status, temperature, CPU load, Loss of signal (LOS), historical reboots, link flaps, service interruptions, speed tests, related application performance reports, complaints.
USE CASE: Customer Behaviour as a Signal
A sudden rise in speed complaints, repeated slow speeds, speed tests failure from ONTs from a certain cluster.
AI Method: Using Natural Language Processing (NLP) to cluster customer complaints and correlate with the networks KPIs.
Outcome:
1. Early fault detection before alarms are triggered — by identifying rising clusters of similar complaints.
2. Automated fault localization — complaints done on location basis
3. Fewer duplicate tickets/escalations — repeated issues from multiple users but linked by a common network event or equipment issue (BNG/ASG/RSG/Core Equipment)
4. Analysis trend for capacity planning — identify slow speeds during peak times or buffering for certain sites.
5. Improved RCA — this would in turn lead to improved/shortened MTTR