Case
The top 1% of the prioritised mains list contained 24% of the breaks that effectively took place
A large Texas public utility validated the reliability of using data analytics software for failure predictions through a blind test: the data from the most recent two years were withheld, and the predictions made on the previous history were compared with the failures that actually took place in that period.
Had the utility addressed the top x% of the pipes prioritised by the software, what % of the failures would have been avoided?
The results were unequivocal:
- The top 1% contained 24% of the breaks
- The top 5% contained 53% of the breaks
The utility sought our data analytics software to help them proactively prioritise the pipes most likely to fail, with benefits for both O&M and CIP planning. The data available for the initial blind test were the pipe break records since 2005, from two separate maintenance record systems, and the GIS repository. These were connected to the software and fed to the failure prediction module, part of a suite of apps dedicated to reliability, risk and planning.
This module uses advanced predictive models to correlate failure data with GIS, environmental, hydraulic and other contextual factors, in order to predict pipe break behaviour in the entire network. Our proprietary data syncing technology automatically reconciles the data continuously received from the multiple IT systems, georeferencing all features and filling in data gaps such as install dates or pipe materials.
This utility benefits from permanently updated:
- Asset probability of failure (PoF), predicted break rates and number of breaks/year; by pipe, pipe cohort (material, year/batch, diameter), project, zone, etc;
- Deterioration curves specific to each pipe cohort and the utility’s own data (not theoretical);
- Total Cost of Failure (including repair cost, direct and indirect costs, service & liability costs, other);
- An unlimited range of user-defined Consequence of Failure (CoF) dimensions, including the ability to calculate hydraulic criticality using the software’s built-in network model; and
- The resulting ability to assess Business Exposure Risk in a quantified, objective, and defensible way.
As the behaviour of different materials and pipe cohorts is increasingly understood, and the quality of work order data improved, updated data is fed back to the GIS and the quality of predictions increases.
Among the outcomes derived from this capability:
- optimised renewal planning through the software’s dedicated module to generate renewal projects, using real-life project-size units;
- joint planning for water and sewer;
- the ability to compare set projects’ performance on avoided breaks, avoided cost or avoided risk;
- increased O&M efficiency in active leakage detection, directing teams to the most likely pipes or service lines to fail.
Utility profile
Population served: 1,908,000
# clients: 511,300
Total network length (water supply): 11,230 km
Total network length (sewer system): 9,599 km
Available data systems
- GIS
- Maintenance/work order records history available since 2004
- Billing / AMI /AMR history available since 2010
Focus of software implementation
- Network renewal planning
Data leveraged in this case
Our software is designed to take full advantage of the data that already exists in the utility, depending on each application objective.
The software connects to the available data systems in an automated, non-intrusive and completely secure manner.
In the case of this specific utility and the application described here, data from the systems highlighted below were used.
GRUNDFOS CONNECT NETWORK ANALYTICS
Grundfos has entered into a strategic partnership with Baseform to bring powerful digital services to water utilities. The Grundfos global value proposition is being up-scaled to serve the water digital market with Grundfos Connect Network Analytics, a state-of-the-art Artificial Intelligence (AI), machine-learning asset management technology provided by Baseform.