Calculating Business Risk Exposure Using Machine Learning

digital face illustration

By Doug Hatler & Matti Kakkori


Aging drinking water mains are facing an increasing number of failures leading to service interruptions, higher operating costs and non-revenue water loss. Some failures cause much bigger financial or socioeconomic impacts than others. A 2018 study led by Dr. Steven Folkman at Utah State University (USU) revealed that water main breaks across the United States and Canada were up 27 percent over the previous six years.

With fiscally tight budgets, water utilities face an uphill battle discerning which pipes in their system need to be replaced and when. Artificial intelligence, specifically machine learning, can make a significant impact in buried water infrastructure asset management.

Machine Learning, Likelihood of Failure and Risk

Fracta, a Redwood City, Calif.-based technology company, commercialized the application of machine learning to the assessment of water mains in 2017. Fracta uses Machine Learning and GIS to predict and visualize the Likelihood of Failure (LOF) of water mains in a distribution system. There are now dozens of water utilities using LOF predictions to help prioritize water mains for rehabilitation and replacement, target water mains for advanced leak detection work, identify containment valves to exercise and maintain, prepare rate cases and studies, and make changes to operating scenarios to avoid break.

LOF is helpful but it does not tell the whole story of the financial risk of aging water mains. Consequence of Failure (COF), or severity, is a critical part of the equation for understanding risk. LOF and COF are used in the Business Risk Exposure (BRE) formula: LOF (%) x COF ($) = BRE ($). A monetized approach to COF and BRE gives utilities an objective risk assessment of entire water distribution system and translates the results into financial terms water engineers, planners and finance professionals can use to make fast, accurate and affordable decisions about buried water main infrastructure.

When an old 6-in. cast iron main breaks on Franklin Street on a cold February morning, the crews rush in to shut off the water and start assessing damage. A rupture like this is relatively standard, but it happens to be on a water main providing water to a large elementary school. Now the school is out of water, parents are notified and school is closed for the day.

Early morning news crews pick up the story and are on-site filming the repair crew. The city’s water deputy director is asked to discuss the situation – the talking points are ready from another break that made the news some months earlier. “The city has aging infrastructure, but we are actively going through the system and replacing all the old cast iron mains. This is a 15-year program that will modernize the water system. Our replacement program had not reached Franklin Street yet, so this unfortunate break happened,” the utility official says.

This is an imaginary story, but it might ring true for many water agencies across the country. Is it unavoidable to experience breaks like these, or could there be improved ways to prioritize replacement plans? Some breaks in our aging water systems are unavoidable, but by using solid asset management practices and targeting the highest risk parts of the system, utilities can mitigate vulnerabilities.

risk matrix

Consequence of Failure…The Missing Link to Measuring Risk

In water system asset management, the asset risk is typically assessed by calculating asset criticality or BRE. The main components of risk assessment are the LOF and COF.

LOF is a value that can be derived purely mathematically. An objectively calculated LOF should be the foundation of the asset risk assessment. The LOF calculation should be as comprehensive as possible, resulting in LOF scores for every water main segment.

Pipe failures result in various consequences, including economic, social, and environmental impacts. COF calculations are a bit more complicated. COF can’t be derived with purely objective, mathematic rules. Best practice for COF determination is to consider the various factors of consequence in a “triple bottom line” analysis that considers financial, social and environmental impacts of main breaks. Measuring financial impacts, i.e. repair costs of a break, is possible and quite straightforward; calculating social impacts, on the other hand, requires a more subjective, opinion-based approach.

One way to aid in simplifying the complex challenge of defining COF is to build standard methodologies to assess the different aspects by defining typical COF categories:

Direct Costs:

  • Pipe repair cost
  • Repaving and restoration cost

Indirect Costs:

  • Customers without water
  • Impact on critical facilities
  • Property damage
  • Traffic disruption
  • Environmental costs

It is possible to rank every pipe segment in these categories, but there are still challenges – how is this done systematically across the entire distribution system, and how do these categories compare against one another to determine the final COF impact?

consequence and likelihood of failure

Monetizing Consequence of Failure and Business Risk Exposure

In order to systematically assess the entire distribution system and calculate the COF impact, a scalable approach is needed where asset properties such as diameter, material and various geospatial features that include proximity to critical facilities, important traffic patterns, and environmentally sensitive areas are objectively assessed. However, we are still left with the challenge of comparing the COF categories against one another.

One way to do this is to convert all categories to a financial impact value. The Water Research Foundation report, Managing Infrastructure Risk: The Consequence of Failure for Buried Assets, explains how these different COF categories can be assessed monetarily. Monetized COF provides a range of risk values, not just simple scoring from 1-5. Why is a range of values a benefit? Let’s illustrate this by considering a risk matrix with scores 1-5 and compare it to a risk matrix with LOF probabilities and monetized COF values.

The above graph illustrates the benefit of a monetized risk matrix. Using the score-based matrix, the highest COF is only 25 times greater than the lowest COF; however, in using the monetized matrix, the difference is 3,200 times greater. Differentiating between higher risk assets and lower risk assets is beneficial when it comes to rehabilitation and replacement planning.

When using a score-based matrix there is potential for overweighting parts of the system with a higher COF. If a replacement plan is built on these scores, perfectly good water mains could be prematurely replaced because of inaccurate ranking.

For example, a 6-in. cast iron main installed in 1932 in a typical neighborhood might receive a LOF score of 4, but a COF score of 1, resulting in total business risk of 4. A brand new 12-in. PVC main close to a school could have a LOF score of 1, but a COF score of 4, resulting in the same business risk in a score-based matrix. However, using a monetized matrix, the business risk would be $1,125 and $300 respectively, more accurately showing COF impacts.

READ MORE: Can Machine Learning Really Provide A Cost-Effective Desktop Analysis?

Summary

Business risk based on objective, data-driven LOF and systematic, monetized COF sets the strong foundation for balanced, proactive rehabilitation and replacement plans. Fracta added COF and BRE to its market leading machine learning and GIS condition assessment solution enabling water utilities to measure and mitigate aging water main risk in financial terms.

Breaks will happen in water systems even with perfect risk assessment and prioritization, but when a proactive program is implemented, the breaks will be fewer as high LOF mains are proactively replaced, and those that do break will have a lower COF, meaning there will not be news crews on-site on an early February morning.


Doug Hatler is an environmental engineer and the chief revenue officer at Fracta Inc., a Redwood City, Calif.-based technology company using artificial intelligence and machine learning to solve the aging infrastructure problem, starting with water distribution mains. For 32 years, Hatler has worked in numerous industries in roles such as environmental engineer, management consultant and regulatory specialist.

Matti Kakkori is an engineer and the head of product delivery for Fracta, Inc. Kakkori has worked in various aspects of hardware and software product development and management for two decades. Prior to joining Fracta, Kakkori was running his own product management consulting practice. He holds a M.Sc in Electrical Engineering from the Helsinki University of Technology.

Leave a Reply

Your email address will not be published.

*