Leaks in Gippsland Water (GW) have historically been complicated to detect using existing flow data models. GW was looking to trial a different, real-time method of detecting common failure modes, utilizing their existing remote telemetry system, data stored in the PI System and time-series analytics provided by Seeq Workbench.

While exploring this use case on defined assets, leak events were classified by Nukon into two main categories:
• Short term leaks with high step-change in flows due to main breaks
• Long-term leaks, which were slow leaks due to cracks in pipes

This session will cover the approach and challenges of reliable, early detection of anomalies to create a real-time, asset-specific leak detection model which can be applied at scale.