Skip to content

CASE STUDY

San Francisco Public Utilities Comission

Supplementing Drinking Water Pipe Replacement Decisions With Artificial Intelligence

Background

SFPUC is the third largest munipal utility in California and serves more than 2.7 million residential, commercial and industrial companies across the Bay Area. SFPUC has approximatley 1240 miles of distribution pipe delivering clean water to the residents within San Francisco.

A large percentage of pipes in SFPUCs network are over 75 years old and needed to be prioritized for replacement. The SFPUC engineers supported Fracta’s machine learning model with distribution , and pipe break data in order to begin developing a large data driven method to support pipe replacement decisions.

The water engineers at SFPUC knew they needed to prioritize these older mains for replacement but their in house age and break model was not specific enough to give them a confident way to prioritize the replacement of these pipes. Enter Fracta.

Objective

Model and evaluate more than 500 miles of total pipe network which was at least 75 years old for replacement. SFPICs existing model was unable to reliably predict which pipes on the network were the most at risk. Combining SFPUCs data with Fractas data driven objective approach should allow them to better identify the most troublesome pipes

Results

SFPUC used Fracta to pinpoint the Riskiest 1% of 100-year old pipes for priority replacement within 1 year.

Within one calendar year, SFPUC had decided to begin incorporating Fracta’s likelihood of failure metric as an additional assessment criteria when evaluating borderline pipes for replacement. Borderline pipes were determined as those who do not have a clear data supported suggestion for pipe replacement.

While using Fracta’s machine learning algorithm SFPUC was consistently able to meet and exceed their annual pipe replacement goal of 13 miles by using Fracta to prioritize the top 1% of riskiest pipes annually.

Being able to differentiate which aging pipes are the most likely to fail and what the higest predicted reason for failure is is only possible using Fracta’s multi channel

Services

"To leave good pipe in the ground as long as possible is economically important.”

Katie Miller
SFPUC