Greater Johnstown Water Authority: Machine Learning Targets Non-Revenue Water

A pumping station can signal that water is leaking somewhere in the network, but it cannot say where. For a utility with 303 miles of pipe, cast iron dating to 1867, and limited resources, sending crews to search is slow, costly, and often inconclusive. This account follows the Greater Johnstown Water Authority as Assistant Public Works Director Jim Kukura puts machine learning ahead of the field work, then confirms suspected leaks with correlators and acoustic technology. It also contrasts an age-based view of the network with a failure-likelihood view of the same section, and shows what the authority measured within its first seven to eight months. Explore the results and the utility's plans for what comes next.