An invisible leak shows up in pressure and flow
On a drinking water network, most losses are never visible from the surface. They are deduced from two perfectly ordinary measurements.
Photo: Crystal Kwok / Unsplash, Licence Unsplash
A leak on a drinking water network leaves no puddle. Most of them drain into the ground, leave no trace at the surface, and last for as long as nobody goes looking. That is what makes the loss considerable: not its flow rate, but its duration.
Searching by hand, or finding out afterwards
Traditional methods rely on manual inspection or on delayed analysis. Both share the same flaw: they come late. An inspection covers one section at a time on a network that has many; a deferred analysis flags an anomaly once the water is already gone. The cost is twofold, in repairs and in wasted resource. Between two search campaigns the network is not being watched: it is being assumed watertight.
Two quantities that give a leak away
A leak does not hide completely: it leaves a signature in two ordinary measurements, pressure and flow. Sensors installed on the network record those variations in real time, and they are the typical indications of a pierced pipe. None of these readings requires opening up the road. They are quantities operators already know; what changes is how often they are taken and the fact that they are read together.
From the reading to the crew
The value of a detection is measured by the delay between it and the intervention. Data reaches the control centre as soon as it is recorded, which makes it possible to locate the leak precisely and send a crew to a point rather than to a section. An exact location turns a search campaign into a repair. Without that precision, detection would merely move the search work from one desk to another.
What the network gains
- Losses fall, because the time between a leak appearing and being repaired gets shorter.
- Repair costs drop, prevention taking the place of emergency work.
- Water resources are managed on the real state of the network, with a direct environmental effect.
That leaves the sizing question: how many measurement points, and where. A network is not instrumented in one go, and the real question is not the number of sensors but the order: which sections deserve to be listened to first.