Greeley pilots AI to prioritize pipeline replacements; consultants show mixed first‑year accuracy

The Water and Sewer Board heard May 21 about a condition assessment program that combines asset inventory, noninvasive pipe condition testing and two machine‑learning pilots to prioritize pipe rehabilitation and replacement.

Cort Nickel, a water utilities staff member, reviewed the system: about 164 miles of transmission mains (replacement value estimated $3–$5 million per mile) and more than 500 miles of distribution piping with roughly 13,000 valves and 4,000 fire hydrants. “The oldest pipe is 135 years,” Cort said, and staff listed multiple risk factors that drive replacement decisions including pipe age, material, under‑sized mains for fire flow and regulatory requirements.

Patrick (asset analytics staff) summarized a three‑year pilot with two vendors. ResiTech uses asset data combined with satellite, elevation and vegetation inputs; its first‑year output identified about 66 miles as high likelihood of failure. VOTA/piperankvota.ai ingested a broader training set and ran EPulse acoustic condition assessments on sample mains. Patrick compared both vendors against 2024 leak data: VOTA’s high‑risk list captured 20 of the sampled leaks (about 86% accuracy for that sample), while ResiTech identified 11 high‑likelihood leaks and 11 low‑likelihood leaks (roughly 50% accuracy for the sample). Patrick cautioned this is year one of three and staff will not rely solely on AI outputs when planning CIP.

Staff also reported results of an EPulse acoustic pilot: no active leaks were found on tested segments, but the tool flagged approximately 360 feet in poor condition (about 34% wall thickness loss) and roughly 2,592 feet in moderate condition. Cort and Patrick said using predictive analytics aims to reduce costly emergency repairs (staff estimated emergency repair costs near $2,000 per foot vs. proactive replacement averages around $850 per foot) and to inform defensible CIP prioritization.

Why this matters: Greeley’s water system is aging and expensive to replace by emergency repair; staff seek to use data‑driven methods to focus limited capital dollars and reduce reactive costs.

The board asked questions about how AI fits within master planning and emphasized the tool is a predictive component of a larger asset management toolbox; staff said pipeline replacements and selected transmission projects (for example the Cash pipeline replacement) are moving forward under conventional planning while the pilot continues.