Across the continent, water mains laid down during the post-WWII housing boom are reaching the end of their life. Out of sight, small cracks and leaks are going undetected until they turn into catastrophic bursts. And all that wasted water can cost cities millions of dollars every year.

Echologics has a solution. The Mississauga company has designed remote sensors that sit inside the caps of fire hydrants. Each night, they come to life, listening to the water network for suspicious noises. When one sensor detects a potential leak, it alerts a central hub, which then uses data from neighbouring sensors to pinpoint the problem.

This isn’t simple science. The EchoShore-DX sensors have to pick out specific noise patterns from all the other surrounding sounds: traffic, rain, barking dogs, backhoes, the normal flow of water through the network, and more. To complicate things further, leaks make different noises depending on the diameter of the pipe, the type of pipe material and the size of the leak.

Fortunately, Echologics brings plenty of expertise to the task. “We really know acoustics, vibration and how to design systems for optimal performance,” said Kevin Laven, Vice-President of Technical Services. “We believe the acoustic performance of our devices is head and shoulders above others in the industry.”

That said, Laven still sees room for improvement. With the help of Advancing Water Technologies (AWT) funding from Southern Ontario Water Consortium (SOWC), he’s teaming up with a pair of engineering professors from the University of Waterloo. Together, they’re using machine learning to help the EchoShore-DX sensors zero in on leaks with even greater accuracy.

“We believe that there’s more information present in the data we’re gathering than what we’re really making use of right now,” said Laven. “And that’s the goal of the project: to really take it to the next level.”

To do that, Professor Kumaraswamy Ponnambalam and Professor Alexander Wong will identify the best algorithms to mine a treasure trove of data from 5,000 EchoShore-DX devices across North America. Because the researchers know which records came from leaking pipes and which didn’t, they can test how well different algorithms tease out subtle differences between the two sets of data.

According to Professor Wong, one of the challenges is to maximize the accuracy of detection without introducing false positives. “We need to make sure that it is as reliable as possible,” he said.

According to Laven, preliminary results have already shown a big jump in performance—and he foresees even bigger gains to come.

“The beauty of machine learning is that once the algorithms are developed, you can retrain it to a larger database over time and further improve the system,” he explains. Better performance should lead to wider adoption of the technology, which in turn will generate more data to keep driving improvements.

Two years after introducing EchoShore-DX, Echologics has sold units to water utilities across Canada and the United States. Although the company is currently focused on the North American region, it ultimately plans to tackle overseas markets, reconfiguring the sensors to work with the different fire hydrant models found in different countries.

It’s a perfect example of the kind of technology the AWT program was designed to accelerate. SOWC established the program, which is funded by the Federal Economic Development Agency for Southern Ontario (FedDev Ontario), to help companies in Ontario leverage the province’s research facilities and academic expertise to develop water technologies with global potential.

The funding has allowed Echologics to maximize its R&D budget. In fact, the company places so much value in the program, it has applied to AWT to advance two other projects as well.

One project, involving a researcher from Queen’s University, is clarifying how the soil surrounding a pipe influences noise within the water network. Once they can tease out that confounding factor, Echologics’ acoustic technology will be able to measure pipe tuberculation.

Another collaboration with Fleming College aims to automate part of the company’s system for measuring pipe wall thickness. “Partnerships with universities and colleges let us draw on expertise that we don’t necessarily have in house,” said Laven.

Of course, their collaborators benefit as well. At the University of Waterloo, the AWT funding has enabled professors Ponnambalam and Wong to hire six graduate students and post-doctoral fellows, training a new generation of water experts.

It’s also a great opportunity to put new machine learning techniques to the test, using academic advances to improve the world’s water systems. “The whole idea of engineering is to apply math, science and other things that we learn to solve real-world problems,” said Professor Ponnambalam.

Professor Wong agrees. “It introduces us to some really cool, practical engineering problems that we might not otherwise encounter,” he said. “And for me, I can see the huge potential impact for this particular project.”