Worldwide, an estimated 30 percent of treated drinking water is lost through leaks, breaks, and other shortcomings in water delivery systems. According to the American Society of Civil Engineers, the nation’s drinking water systems are not highly resilient and their present capabilities to prevent failure and properly maintain or reconstitute services are inadequate. In its 2009 Report Card for America’s Infrastructure, the U.S. drinking water systems received a grade of D-, along with a warning that the systems face “staggering public investment needs over the next 20 years.”
In a future in which clean water is expected to become increasingly scarce, the regular loss of nearly a third of treated water to leaks and pipe failures seems especially egregious. But what if water main breaks could be predicted in advance? Not only would this allow municipalities to greatly reduce costly emergency repairs (and the sometimes catastrophic consequences of major water main failures), it could also go a long way toward conserving an increasingly precious natural resource.
One project currently proposed through the Whiting School’s infrastructure center would focus on developing systems-level optimization methods for determining likely leak locations. The goal: to enable water utilities to perform preventive maintenance in advance of catastrophic failure. This would represent an enormous advance over current system maintenance practices. “People are still putting rods to the ground and putting their ears up to it to listen for leaks,” says Buddy Cleveland, senior vice president of Bentley Systems Inc. and chair of the Civil Engineering Visiting Committee. The Pennsylvania-based Bentley Systems is one of the leading providers of software for the design, construction, and operation of infrastructure systems and is a corporate partner in the infrastructure center proposal.
“If you can measure pressure and flow in various parts of a water delivery system, then you can apply genetic algorithms to predict where the leaks will be,” Cleveland says. “This represents the next step in information modeling, where we start to marry the digital sensors to the physical assets and begin to monitor how these things are performing in real time.”