Predicting Water Line Failure Helps Topeka Maximize Limited CIP budget
The City of Topeka
The City of Topeka required a method for determining the failure risk of its water lines so a water line replacement plan could be implemented within its limited CIP budget. Using a machine-learning algorithm to predict pipe failures, Topkeka built a water line replacement plan to systemically replace water lines.
General Fund/Existing Public Funds
Operational since 2021
Director of Utilities Department
The City of Topeka needed a way to identify the condition of buried water lines so replacement could be prioritized.
Topeka, Kansas’ water infrastructure is in critical condition. Of the 890 miles of water line stretched throughout the city, 60 need to be replaced immediately because they are beyond their life expectancy of 100 years.
The consequences of not replacing these water mains can be severe. One water main broke within 15 feet of a Topeka home, releasing four million gallons of water and pushing the home off its foundation. Topeka has seen an average of 65 breaks per 100 miles of water line over the past 10 years, over four times the average found by a study of 300 utilities across the U.S. and Canada. Although necessary, fixing the situation through water line replacement is not simple.
Replacing water lines is expensive and the success of determining which lines to replace varies. According to Topeka’s Utilities Director Braxton Copeley, “The cost to replace one mile of water line is generally about $1 million a year.” Apart from a line breaking, the methods used to identify a pipe’s need for replacement are to dig it up for inspection or to replace it simply due to age. The current methods to determine a pipe’s health are expensive and can be misleading as older pipes are not always in poor condition.
Topeka needed to determine how to prioritize replacing old cast-iron water lines beyond their life expectancy and middle-aged pipes that have shown signs of premature failures. Starting in the 1950s, the city began installing ductile iron pipes that were rated for 100 years, but the way they were installed led to breaks after only 60 years. A particular half-mile length of ductile iron waterline had 55 breaks alone since 2008.
Facing a large number of capital needs armed with a very limited budget, the city needed a standardized method for determining the conditions of pipes in the ground that would enable better resource allocation for line replacement.
The City built out a water line replacement plan using machine learning to determine the likelihood of pipe failure.
Topeka is making data-driven decisions to determine how to allocate resources when fixing water mains.
“By having an AI-based model on the likelihood of failure, it allows us to use real data in terms of what lines are most likely to break,” Topeka Utilities Director Copley says. “I can identify those lines for replacement. Back in the old days, we’d be relying on guys out in the field.”
The city can now predict the risk of failure for each pipe, having incorporated Topeka’s break and pipe asset data for the past 17 years into Fracta’s machine learning algorithm. Parameters that can affect the deterioration of an underground pipe asset include soil characteristics, weather, automobile traffic, and topography, among others. The algorithm uses this data to identify patterns conducive to pipe failure and enable the forecasting of future breaks. Topeka staff can then use that data to accurately determine the likelihood of failure for the city’s cast-iron and ductile iron water lines, as well as the impact of those failures.
Now aware of failure risks and impacts to the community, Topeka staff can make the most of the city’s limited budget by building out a replacement plan to address the most at-risk areas. While the probability of a water main break might be higher in one location than another, access to accurate predictions allows the city to prioritize replacement based on other factors such as where the water main is located (ex. near a school or hospital).
"Given the fact that myself and two engineers are managing $6.5M worth of waterline replacements each year, we find that having a resource like Fracta to help make data-driven decisions on which line to replace, is invaluable,” Copeley stated.
Meanwhile, the model that is predicting pipe health is growing even smarter with new data sources. One year into using the new tool, Topeka was able to correctly predict 65% of pipe failures in a sub-region with pipes of interest that occurred in the following year. This prediction allowed the city to prioritize those replacements while protecting the safety of its residents and maintaining the water supply.
Topeka is able to accurately prioritize aging pipes that need to be replaced based on the pipe’s condition and the estimated budget needed to replace it
The city can maximize its limited CIP budget by building out a replacement plan to address the most at-risk water lines
Staff and officials can now identify weaknesses in water pipe networks and understand the impact of an unplanned failure
With 5-6 projects annually, Topeka saves an average of $80-100K annually on repairs by predicting pipe breakages without the high-cost work of digging them up for inspection
Who Should Consider
Towns, cities, and water utilities needing a standardized method for determining pipe conditions to enable better resource allocation for line replacement.
Government Project Team
- Braxton Copeley, Utilities Director
Last UpdatedJul 8th, 2022
More resources about this case study