Water Divining With AI
Water Divining With AI
A startup in Brazil is using pattern recognition AI in a smartphone app to find the source of water leaks throughout the water supply network.
Join the DZone community and get the full member experience.Join For Free
Last month, I reported on a novel use of AI in a diagnostic smartphone app, after spotting it in The Economist. A research team in the U.S. has built an artificial neural network that uses pattern recognition to listen to the sounds of bees in a hive and determine whether the colony are suffering from particular illnesses that are characterized by specific buzzes.
This month, I spotted another great article in The Economist about the use of AI to trace leaking pipes, using a similar approach to the diagnostic app for bees.
According to the Economist article, over 35% of the water that enters the pipes from Brazil’s suppliers is never billed for because it either leaks away or is stolen via pipe- or meter-tampering. The article goes on to say:
“[The loss]...amounts to 6.5bn cubic metres in wasted water a year, worth some 8bn Brazilian real ($2.3bn) in forgone revenue, according to some estimates.”
Scientists in water companies sometimes use acoustic equipment to listen for leaks in the pipe network. The difference in sound between a leak and water passing normally through a pipe is something that AI can be taught to spot through pattern recognition, and a startup in Brazil has done just that.
The solution, they say, is to place a handheld sensor against the pipework and to record vibrations in the water flowing through the local network. The sample, which can be as short as 15 seconds, is then analyzed by a smartphone app that compares it against a database of pipe sounds stored in the cloud, matching across a range of frequencies, including those not audible to humans. For example, water diverted illegally from the pipes may be exposed by matching unexpected vibrations detected within the flow.
At present, ten water companies are testing the system and uploading 800 new sound samples a day into its database. Eventually, the tool is anticipated to help them find the most likely sites where water is disappearing from the system, speeding up detection and resolution, to reduce water wastage and theft.
Published at DZone with permission of Jo Stichbury , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.