Water software company Trinnex has announced the launch of its leadCAST Predict, a new predictive modeling tool that the company says improves water utilities’ ability to optimize resources to achieve compliance with U.S. EPA’s Lead and Copper Rule Revisions (LCRR).
The tool is designed to predict and classify unknown service line materials to reduce the presence of lead in community water systems.
leadCAST Predict is part of Trinnex’s larger end-to-end LCRR compliance solution called leadCAST. After more than 1,000 development and testing hours and collaboration between data scientists, machine learning (ML) engineers, clients, and water domain experts, the ML models were built to seamlessly integrate with a water utility’s existing inventory data to help generate predictions of service line material, including an optimization model for water systems that may not have lead service lines.
leadCAST Predict supports two types of water utilities:
- Those with lead service lines that need to prioritize getting the lead out
- Those with no lead service lines that need to prove the existence of acceptable materials
ML model performance improves as it continuously gets trained with new data obtained from additional data acquisition methods such as construction and plumbing records, inspections, and field verifications. leadCAST Predict users get their results in an interactive dashboard and map views, along with an easy-to-follow report with an overview of how the ML models were built, deployed, and tested.
“We are committed to supporting water utilities in their mission to achieve LCRR compliance and continue delivering safe water services, and we believe leadCAST Predict will help them get there more quickly and accurately,” says Amy Corriveau, Trinnex president .
Trinnex estimates it would cost $4.7 million to field verify at least 5,000 service lines without predictive modeling, based on other similar program costs.
As Katie Deheer, one of the senior data scientists behind the tool explains, “LeadCAST Predict is different from existing tools in the market in that you can perform the analysis on communities with no lead. Our clients don’t have to bear the burden of verifying every single service line with unknown material without some help with prioritization.”
She adds, in some cases, utilities might have a biased set of existing data that can skew the ML model. As such, leadCAST Predict includes a process to evaluate whether existing field verifications are truly representative of an entire water system or if additional data is needed. Learn more about leadCAST Predict.
