Artificial intelligence for water main infrastructure

The Fracta condition assessment solution calculates and visualizes the likelihood of failure (LoF) for every water main across a utility’s infrastructure.

A more accurate, objective tool can align maintenance and capital repair and replacement strategies.

By Doug Hatler

In the next 25 to 30 years, most water distribution mains in the United States will need to be replaced. The American Water Works Association (AWWA) forecasts it will cost nearly $1 trillion to address the problem.

The task of addressing water main replacement needs, including funding, is daunting, but according to the Water Research Foundation (WRF), 75 percent of water utilities cited pipe breaks as a key criterion in pipe replacement decisions. In 2007, the U.S. Conference of Mayors noted that 86.2 percent of cities use the number of water main breaks per unit length to evaluate drinking water pipe performance.

Literature reviews indicate that between 250,000 and 300,000 breaks occur every year in the U.S., which corresponds to a rate of 25 to 30 breaks per 100 miles of pipeline per year (breaks/100 miles/year). According to another WRF publication, the average pipe break rate (regardless of cause) for water utilities is between 21 and 27 breaks/100 miles/year. The AWWA Partnership for Safe Water Distribution System Optimization Program goal for a fully optimized distribution system is 15 breaks/100 miles/year.

Figure 1: AI-Machine Learning process

Recent studies

In 2018, the Utah State University Buried Structures Laboratory released a study on water main breaks in the United States and Canada. It surveyed more than 300 utilities and focused on water main breaks related to pipe material type, separating out main breaks caused by third-party and maintenance damage. The results answer many important questions. Overall water main pipe break rates have increased 27 percent during the last six years. In aggregate, it was estimated the pipe break rate is 14 breaks/100 miles/year.

Another 2018 report from WRF based on research by Purdue University and Louisiana Tech — Practical Condition Assessment and Failure Probability Analysis of Small Diameter Ductile Iron Pipe — focused on the break rates of ductile iron (DI) pipe less than 12 inches in diameter, taking into consideration that many newer pipes, which are thinner, may be failing at a higher rate than older pipes or larger-diameter DI pipes. The research found that, on average, DI pipes are breaking at a rate of 15.1 breaks/100 miles/year. During the last 10 years on average, the direct cost of failure of small-diameter DI pipe was $12,600 per occurrence, while the indirect cost was estimated at $5,600 per occurrence. On average, the total cost was $18,200 per break or $274,820/100 miles/year.

Condition assessments

The water industry has seen many types of academic surveys and studies on water main replacement programs and the benefits of asset

management, condition assessment, and prioritization. During the last 20 years, utilities have begun to track all aspects of their infrastructure in a GIS-centric platform, and have collected records on the types, sizes, and repair histories of their pipes. In addition, asset inventory, condition assessment, and asset management planning practices provide valuable information to enable utilities to more efficiently determine which pipes to repair and replace, taking into consideration relevant variables unique to the water utility.

Condition assessments of buried water mains typically fall into two categories: physical and desktop. Physical condition assessments are accurate for the pipe tested but tend to be slow, expensive, and labor intensive. Multiple physical measurements are required for correlation and confirmation. The results are difficult to extrapolate to system-wide recommendations.

Desktop methods are more straightforward, but many of these methods are based on arbitrary assumptions and weights (i.e., older pipes are more in need of replacement than newer pipes). More advanced statistical modeling may help decipher differences between various variables, although many of these approaches may not have the ability to consider the importance of some adjacent details such as proximity to light rails or the contribution of elevation or pipe material, therefore impacting accuracy.

A new paradigm

A more robust approach would be a large-scale comparison of these various factors to generate a more refined and accurate prediction based on the disparate interactions between component variables. Artificial intelligence, specifically machine learning (AI-Machine Learning), has emerged as a technology to make a significant impact in buried water infrastructure asset management. AI-Machine Learning consumes large, complex data sets containing more variables than humans can process with current tools. This objective, data-driven method overcomes inherent subjectivity and biases and provides results that help utilities make better replacement decisions.

Due to the large amount of historical and geospatial data needed to run AI-Machine Learning algorithms, water main condition assessments contain all the necessary components of an ideal application for water utilities:

• years of historical data covering installation year, pipe material, and break history;
• categorical data including pressure class, geographical location, elevation, and pipe diameter; and
• contingent data including proximity to rail systems and soil composition.

The volume of data is a unique opportunity for water utilities. Analyzing this data consistently can uncover trends, gain insight on pipeline health, and offer data-driven assessments.

Fracta (www.fracta.ai), a Redwood City, Calif., technology company, is using machine learning to help water utilities make pipe replacement decisions. Fracta’s machine-learning algorithms use vast amounts of historical data to quickly solve complex pipe problems using the likelihood of failure (LoF). In the water main industry, the LoF, also known as condition assessment, provides the most valuable actionable predictions.

The Fracta condition assessment solution calculates and visualizes the LoF for every water main across infrastructure. A water main’s LoF score is the result of more than 1,000 data variables for every pipe segment. Fracta uses the following types of data:

• Asset data — pipe ID, location, diameter, length, material, installation date
• Historical data — break history
• Geographical information — location, elevation, slope
• Environmental data — soil, climate, water bodies, structures, population density, etc.

Analyzing this data consistently uncovers trends, gains insight on pipeline health, and offers data-driven assessments. Coupling the LoF with consequence of failure analysis can then accurately pinpoint areas that are most in need of replacement.

AI-Machine Learning process

Data acquisition, assessment, and cleaning for any AI-Machine Learning process is roughly 60 to 80 percent of the work — also known as pre-processing or data wrangling — with the remaining percentage being the machine learning itself. Once the data is assessed, cleaned, and imputed where needed, it is ready to be fed into a machine-learning algorithm where it is subsequently “trained” to learn the patterns that predict breakage events. Figure 1 illustrates the process.

Data quality is critically important. The data used in the analysis must be collected, organized, and normalized. Main data sources Fracta uses are the utility water main asset data and information about historical breaks. A software-led approach to data cleaning and normalizing aids in the challenge of using real-life data from city- and community-specific utilities that varies in its quality.

The more data a model contains, the more robust the model. As utilities collect new data over time, recording new activity, data is continually fed into a machine-learning mode. This subsequently enhances the model by either strengthening previously learned rules around break predictions or from encountering additional circumstances around which new rules can be built.

Financial benefits

Analyzing water main condition data consistently uncovers trends, gains insight on pipeline health, and offers data-driven assessments. Incorporating a machine learning condition assessment solution like Fracta into a proper infrastructure asset management program will lower condition assessment costs while helping reduce water main breaks and more efficiently allocate pipe replacement capital investment.

Coupling LoF with consequence of failure analysis can then accurately pinpoint the pipes and areas of highest risk. A risk-based asset management program relies on accurate predictions of the most vulnerable water mains. An inflated or deflated LoF score could dramatically over or under estimate risk and lead to replacing low-risk pipe or ignoring high-risk pipe. Use of machine learning to determine LoF is a fast, accurate, and affordable way to improve confidence that risk ratings are accurate and reliable.

Industry data says that as much as 30 to 40 percent of replaced pipe has useful remaining life. If capital spending on pipe replacement is $10 million per year, a utility could be wasting $3 million to $4 million per year replacing the wrong pipe.

Conclusion

Incorporating AI-Machine Learning condition assessments into a proper infrastructure and asset management program will contribute to reduction of the economic impacts incurred from water main breaks and more efficient allocation of capital by water utilities. Use of best practices and a more accurate, objective tool will align maintenance and capital repair and replacement strategies to more efficiently leverage scarce financial and human resources. They also inject financial integrity to the planning process and refine the investment strategy so a utility will be in a better position to defend planning efforts and fund needed capital pipe replacement projects.

Doug Hatler is environmental engineer and chief revenue officer at Fracta (https://fracta.ai), a Redwood City, Calif., technology company “Bringing Artificial Intelligence to Infrastructure.” He has more than 30 years of experience managing water, waste, and EHS compliance.