Kitchener-Waterloo

If it ain't broke, fix it anyway — models could predict water main breaks

A K-W engineer who heard CBC's water main break coverage and was inspired to develop machine learning models that can predict where water mains will break.

Engineer says his computer models predicted water main breakage, and city needs to be less reactive

It's hard to detect water main breaks because the pipes are underground, but a local engineer has created a computer program that could help predict what pipes are at risk of breaking before it happens. (Jane van Koeverden/CBC)

A Ktichener engineer has developed a computer model that he says has the ability to predict where water mains will break, before they actually happen. 

Mehrdad Varedi quit his job to work full time on the software after hearing CBC's coverage of water main breaks in the region. That sacrifice seems to have paid off, so far this year the modelling has been accurate. 

Varedi founded Waterlix two years ago after quitting his job at a software company. (Courtesy of Mehrdad Varedi)

"All the water main breaks that happen [sic] from the first quarter of 2017 were forecasted by my models," he told Craig Norris, host of The Morning Edition.

Water main breaks are costly for the City of Kitchener. In 2015 city councillor and finance chair Scott Davey said it costs roughly $15,000 every time a water main breaks.

On top of that, the city is also facing a massive infrastructure deficit. Pipes that were designed to last 70 or 80 years are failing earlier than expected — when they are just 40 or 50 years old. 

Predicting vulnerabilities

Varedi used historical figures provided on water main breaks, provided by the city, and other open source data such as satellite images to develop the computer model.  It evaluates hundreds of variables such as age and length of pipes, salt usage on the roads near a pipe and types of buildings surrounding the pipe to find common factors involved in water main breaks.

The model then looks for similar situations and factors existing elsewhere in the city to predict vulnerability in segments of water mains.

This screenshot of the model shows a map around St. Mary's General Hospital and the vulnerable segments of the water mains. The blue dots are spots of vulnerability, the darker shades have a higher probability of breaking. The red stars are historical break sites. (Courtesy of Mehrdad Varedi)

Varedi said right now the management of water pipes is reactive, meaning cities wait until a pipe breaks to repair it. 

He said his company Waterlix's software could help predict where breakage could occur and service them before they break, which would save money and water.

He has met with the asset management team at the City of Kitchener to show them the model. He said although they were open to the idea, there are no plans for implementation yet.