FIDO AI eliminated human error from leak identification. Harnessing the power of the cloud does the same for accurate leak location.

The accurate first-time location of leaks avoids dry digs, cuts leak run time and reduces water loss. It’s done using a water leak’s audio signature. The low-tech way, still widely used, is a listening stick. The high-tech way is correlation.
At its simplest, correlation is a function of distance and time. You use the difference in the time it takes a single noise to reach two points to work out how far away the noise source is.
But there are variables that affect accuracy. The biggest is whether the noise is actually even a leak. Then there is the pipe material. Different materials work better with different processing algorithms.
Without specialist on-site equipment and trained operators, which is costly and time-consuming, getting accurate results from correlation is challenging.
With FIDO Cloud Correlation, FIDO wanted to match or better the accuracy of first-time leak location at much lower cost by simplifying it and reducing human error.
Challenge 1: Synchronicity
To correlate with any degree of accuracy you need to gather two data samples from the same noise at exactly the same time. Even a millisecond of drift can throw leak location out by metres.
For FIDO Cloud Correlation, acoustic sampling is done by FIDO Bugs which provide better data than many proprietary sensors. Many of our clients already use them as smart acoustic samplers. Using them for correlation too avoids introducing additional devices into the process.
The challenge was synchronising any two Bugs precisely enough to enable simultaneous sampling. Bugs don’t use radio signals like many traditional correlators, and 4G and IoT networks have an inherent latency which threw timings out. We discovered that even expensive super high accuracy clocks had a tendency to drift over relatively short periods of time.
Because we don’t want to charge clients for Bugs, we opted for a low-cost quartz crystal clock chip in every FIDO Bug. Synchronising them is as simple as tapping any two Bugs together. This achieves an accuracy of better than a quarter of a millisecond and the clock ensures the Bugs stay together with an accuracy of around ten parts per million.
That means that if you do the recording within one minute your drift is less than half a millisecond – well within our design brief of better than one millisecond concurrency between the recordings.
Challenge 2: Computing accuracy
Working out leak location based on all the variables FIDO AI spots in leak signatures needs some very advanced maths. And that takes a lot more computing power than is possible with an independent hand-held device.
Using the cloud gives us the power to use many more functions, filters and combinations of algorithms as well as the ability to learn from every correlation.
In our case, FIDO AI uses up to 18 different algorithms to determine each leak’s unique features just from listening to the samples from the Bugs.
Its machine-learning brain picks out the precise frequency of the leak from other extraneous noises. Then, based on the other information it has gleaned, like pipe material and leak size, it automatically selects the best combination of processes to pinpoint leak location exactly.
Access to network GIS information further eliminates the need for human input, as accurate asset location and sensor distance is automatically built-in.
Is there a straight-to-dig future?
FIDO Cloud Correlation is now live as part of our service and proving to be incredibly successful.
As we build more and more successful cases it is becoming clear that smart correlation is best done using AI in the cloud.
FIDO AI already accurately identifies leaks and their size. By identifying their location even more room for error is removed from the leak detection process.
There’s no need for any additional equipment or special training. And for utilities with GIS -mapped assets the process is even simpler.
For the future, we are already looking at ways to synchronise data collection offline or in the cloud itself, opening up the tantalising prospect of a truly straight-to-dig leak detection solution.