AI System for Routing Water Network Leak Signals for Review

The problem

When flow, pressure, meter, fault history and location data are held in separate sources, their fields must be matched by hand before a potential leak area and field inspection record can be prepared.

The water network operations engineer must compare every suggestion in the potential leak area and field inspection record with the source record; missing or conflicting evidence must not replace the final decision.

If signal distortion caused by a planned discharge or sensor fault is not reviewed separately, the wrong valve could be closed or a genuine leak could be delayed; the record must not progress without a human decision.

What we set out to improve

• Time taken for the draft potential leak area and field inspection record to reach the water network operations engineer with its supporting evidence
• Outcomes of records referred for human review because a planned discharge or sensor fault distorted the signal
• Number of suggestions corrected after identifying the risk of closing the wrong valve or delaying a genuine leak

How the system works

1. 01 Retrieve flow, pressure, meter, fault history and location data from authorised sources
2. 02 Verify field and document versions
3. 03 Prepare the potential leak area and field inspection record with its supporting evidence
4. 04 Exception check: signal distortion caused by a planned discharge or sensor fault
5. 05 Have the water network operations engineer accept, correct or reject the suggestion
6. 06 Decision gate: critical valve, site closure and intervention decisions
7. 07 Record the decision and source separately

Methods we used

• A source-identified data schema for flow, pressure, meter, fault history and location data
• A structure that separates deterministic rules from AI suggestions
• A screen showing the potential leak area and field inspection record beside its supporting evidence
• A human review queue for signal distortion caused by a planned discharge or sensor fault
• A record identifier that prevents duplicate processing

Where people stay involved

Data and security

Only the flow, pressure, meter, fault history and location fields required for the task should be processed. Role-based access should be applied, and unnecessary identity, health, financial and internal organisational data must not be sent to the model. Source retrieval, AI suggestions and the water network operations engineer's decisions should be recorded separately.

Who this suits

Frequently asked questions