Finding where a water network loses its water

What's on the screen

• The 3D district. Its streets, its buildings and the mains under the streets; the meter chamber at the district's inlet, the noise loggers on the valves, the pressure loggers on the hydrants. By night and by day. "Hydraulic model" turns the ground to glass; every main takes the colour of the water the model estimates it carries beyond normal, and the water's direction runs inside it.
• Inlet meter. The district's flow now against its normal, the last seven nights' night flow, and today's loss.
• Noise loggers. The quietest level each of the eight valves hears at night, against the leak threshold.
• The advisor. It sets the night flow against its normal and the noise loggers against each other. Here, at 01:40, a main bursts under Lale Sokak. The burst doesn't surface and the pressure drops by only 0.04 bar: by day the loss would vanish in the district's demand. At night, though, the inlet meter shows twice the normal flow. The hydraulic model, fitted to the pressure loggers, narrows the loss to the district's south-east, and the noise loggers either side of the burst narrow it to the stretch between two valves; the crew finds the exact spot with a ground microphone. The advisor gives the likely cause, the risk and where to send the crew, together.

How to try it

1. Use the camera buttons to move from the whole district to the view from above, the inlet meter and Lale Sokak, or drag to orbit.
2. Scrub through the day in the bar below. The burst starts at 01:40 and the advisor gives its finding at 02:30. By day the loss vanishes in the demand; the chart keeps showing it.
3. "Hydraulic model" colours the mains by the water the model estimates they carry beyond normal. "Show on the network" on the advisor card shows Lale Sokak and the stretch where the burst most likely is.
4. "Watch the film" plays the same scene full screen as a 26-second walkthrough.

In a real installation

This page is a demo: the district, the network and the data are simulated and representative; it does not show a system installed by a water utility. In a real installation, inlet flow and pressure come from SCADA, the loggers' data from their own platforms, and the network from the GIS and the hydraulic model. Flows in the mains aren't measured: the hydraulic model estimates them, fitted to the loggers. The people figure assumes 150 litres a person a day. The advisor's rules are set up for that network's districts.

If you would like to talk about your network, write to us with the districts, the data sources you have and the problem you want to solve.