.png)
Acoustic intelligence that finds developing pipe faults before they become failures
Adelitics Sentinel ingests data from existing acoustic logger fleets, classifies what it hears, tracks how that changes over time, and delivers prioritised, decision-ready alerts — often weeks or months before visible failure.
It does not simply flag elevated noise between adjacent sensors. It models the acoustic environment at each monitoring point and separates genuine fault signatures from the background noise of an urban network.
Most acoustic monitoring tells you something is there.
Adelitics Sentinel tells you what it is, whether it is growing, and what to do about it.
Conventional acoustic monitoring relies on correlation between adjacent sensor pairs. When noise appears above a threshold, an alert is generated. The problem is that correlation cannot distinguish a developing crack from a water meter, a pump, traffic, or any other persistent noise source. The result: high alert volumes, high false positive rates, and field crews investigating non-events.
We've seen utilities operating conventional platforms reporting false positive rates above 50%. Across Australian utility deployments, Adelitics Sentinel has reduced this to 12%.
Correlation is also a point-in-time analysis. It cannot detect slow-developing faults that fall below threshold in any individual recording, or track whether a detected noise source is stable or escalating. These are precisely the faults most likely to cause catastrophic failure.
Adelitics Sentinel addresses both.
Five complementary analytics.
One decision.
Adelitics Sentinel applies five analytical methods to each sensor's data. No single method is sufficient on its own: together, they enable classification, tracking and prioritisation that correlation cannot achieve.
Rate of Change Analysis
Tracks acoustic power at key frequencies over 30 to 100 days. As a crack develops, acoustic energy increases and frequency content shifts, often weeks before any threshold-based system would trigger. Sentinel detects the growth trajectory, not just the presence of noise.
Pattern Recognition & Noise Suppression
Learns the acoustic fingerprint of each sensor's environment and suppresses persistent environmental and operational noise: traffic, pumps, irrigation, HVAC, customer activity. Genuine anomalies become visible against a cleaner baseline.
Event Classification
Trained machine learning models classify each recording by source type: leak, machinery, customer-side noise, or background. Leak signatures are broadband and persistent; machinery noise is frequency-periodic and intermittent. Classification accuracy exceeds 90% in peer-reviewed field validation.
Signal Correction & Normalisation
Corrects for sensor variability, automatic gain control differences between logger vendors, and network condition changes to ensure consistent, comparable data across the full fleet.
Localisation & Prioritisation
Uses coherent power analysis to produce a spatial map of where acoustic energy is concentrating. More robust than standard cross-correlation in the presence of multiple noise sources. In field validation, faults have been localised to within approximately 1% of pipeline length across kilometre-scale sections.
From overnight recordings to morning decision
Acoustic data is ingested each morning from the logger fleet. By the time the operations team starts the day, Sentinel has processed overnight recordings, updated the acoustic history for each sensor, and generated a prioritised alert list.
Alerts are assigned one of three action levels:
MONITOR
Strong evidence of an active leak or developing fault. Field investigation recommended within 24 hours.
.png)
LEVEL 1
INVESTIGATE
LEVEL 2
Probable leak signal, consistent pattern. Schedule field investigation.
DISPATCH
LEVEL 3
Possible signal, insufficient confirmation. Continue tracking.
Each alert includes sensor location, fault classification, confidence level, acoustic history, and a localisation estimate. Confirmed field outcomes feed back into the platform, continuously improving classifier performance for each utility's specific network environment.
Weekly summary reports are provided to operations and asset management teams, documenting alert status, fleet health, and any escalating conditions across the monitored network.
Works with the sensors already in your network
Sentinel connects to existing acoustic logger fleets through vendor APIs. No hardware replacement required to get started.
The platform has been configured for most leading logger types. Additional logger types can be integrated subject to configuration and testing.
Where sensing is not yet in place, Sentinel can be deployed as part of a turnkey engagement including sensor specification and installation.
The platform is built on the world's largest event-validated IoT acoustic dataset: millions of labelled recordings across thousands of sensors in eight cities, accumulated since 2016.
Leak-before-break is demonstrated as a structured, measurable operational process.
Deployment options
Adelitics Sentinel is available through multiple delivery models. The right starting point depends on what sensing is already in place, the scale of monitoring required, and the operational outcomes being targeted.
Data as a Service
Ongoing analytics for an existing logger fleet. Adelitics ingests data through vendor APIs, runs the full Sentinel analytics suite, and delivers prioritised alerts and weekly reporting. No change to existing hardware or workflows. The preferred starting point for utilities with deployed fleets.
Turnkey
End-to-end deployment for new monitoring programs, existing fleet expansions, or where multiple sensor types need to be brought onto a single analytics platform. Includes sensor specification, installation coordination, platform configuration, and operational workflow integration.
Advisory
Targeted support to embed leak-before-break approaches within existing asset management programs. Includes network prioritisation, monitoring framework design, pilot scoping and operational capability development.
Already have acoustic loggers?
Want to start monitoring your network?
If your network has a deployed logger fleet but alert volumes are difficult to action, or false positive rates are eroding confidence in the monitoring system, Sentinel is designed for exactly that situation.
If permanent acoustic monitoring isn't yet in place, we can help you define the right approach — sensing, analytics, or both.