92 Percent Accurate Pipeline Assessment Using Artificial Intelligence and IoT

Traditionally, water utility districts managed underground infrastructure on a haphazard basis. Managers relied on visual occurrences and reports from the public about problems and sent field crew to fix pipeline leaks and ruptures as they happened. The unpredictability of pipeline leaks and failures was costly, and water utility managers had no way of conducting pipeline assessments to prevent future breaches.

Today, Artificial Intelligence (AI), in conjunction with the Internet of Things (IoT), helps water utility district officials to predict the likelihood of pipeline leaks. With predictive analytics and smart data in hand, city officials can implement maintenance plans to prevent costly ruptures and plan new infrastructure.

The following is an example of how Nobel Systems’ GeoViewer enables water utilities to conduct pipeline assessment using AI and IoT. Armed with smart data, a utility can analyze the business risk index, or BRI, of its water system’s pipeline conditions.

GeoViewer AI / IoT Pipeline Assessment

During the pipeline assessment, the following data collected includes:

• Pipe material
• Pipe age
• Pipe diameter
• Pipe length
• Elevation of the pipeline.
• Soil information, obtained from the United States Geological Survey.
• Road ratings.
• Previous leak history, obtained using Nobel’s GeoViewer to record leaks over time.
• Nearby pipe leak history.
• Pressure from IoT devices: GeoViewer IoT pressure monitoring devices are placed across a district’s water lines. GeoViewer records any spikes in pressure in real-time, and the resulting data impacts the pipeline grading assessment.
• SCADA pressure and flow are also used to grade pipelines.

Nobel Systems’ AI calculates a Business Risk Index (BRI) based on the overall probability of failure (POF) and the consequence of failure (COF).

Nobel calculates POF based on current data collected via GeoViewer, historical data from the water district, and information from government records and other sources.

With predictive analytics and smart data in hand, city officials can implement maintenance plans to prevent costly ruptures and plan new infrastructure. #smartutilities #iot #AI #watermanagement Click To Tweet

The COF is determined in meetings with water district and city officials by how a rupture will affect the area in which it occurs. For example, if a water main break occurs near a school, leaving children and instructors without water for several hours, the COF is rated higher than if it occurred near unoccupied land.

Forecasting Leaks

Nobel Systems uses the Autoregressive Integrated Moving Average (ARIMA) method to forecast pipe leaks. The ability to predict leaks helps district managers with budget and infrastructure planning.

ARIMA is a regression analysis that gauges the strength of one dependent variable relative to other changing variables. Regression analysis is a statistical method that allows us to examine the relationship between two or more variables of interest, such as pipe length, age, location, etcetera. In testing, this method monitored errors between forecasted leaks and actual collected leaks.

Collecting Data via IoT

Nobel staff installs IoT real-time monitoring devices at strategic points across a water utility’s system. The prediction algorithm feeds real-time data from transients spikes and pipeline leaks in the systems using these devices.

The results are pushed back to GeoViewer as a geospatial rating, indicating the POF for each analyzed pipe segment. The data will show current leaks and, combined with historical and other conditional data, gives a utility a POF rating.

In testing the algorithm, we collected data via GeoViewer from an existing water utility client. We split the data 80% to train the algorithm and 20% for testing. The trained model was simulated to predict future pipeline failure probabilities. In comparing the expected test data to the existing test data, it was observed that the predictive test data was 92% accurate compared to the test data. The initial error of 30% was reduced by 8%.

Pipeline Prediction of Failure

Using the resulting data, we can assign pipelines the following POF grades:

• Grade 1 – [ 0%-50%, less Likely to fail]
• Grade 2- [ 50%-80%]
• Grade 3- [ 80%-100%, most likely to fail]

 

The Consequence of Failure Determinations

In a hydraulic modeling simulation, each pipeline is turned off and simulated to understand the impact on neighboring valves and operation. Interviews with field operators help determine the effects of each pipe leak and the complete operational cycle. The cost of pipeline replacement is then calculated.

Leak points that resulted from a damaged hydrant, stolen water, or a repair failure are confirmed with field operators and excluded from the data. Pipelines are then categorized on a project basis, depending on construction quality information obtained from the district.

In Summary

Smart technologies that incorporate AI and IOT, coupled with dedicated water utility management software such as Nobel Systems’ GeoViewer, provide accurate pipeline assessment and predictions for pipeline failures. Such technology can help water utility operators and managers accurately assess the health of existing infrastructure and avoid costly repairs and disruption to services. Furthermore, cities and local stakeholders can create preventative maintenance pipeline replacement plans based on POF modeling.