New prediction system targets water pipeline failure

Data61 has collaborated with Sydney Water to devise a system that can predict where a water pipeline will fail, which has the potential to minimise maintenance costs. Australian water utilities spend more than $1 billion a year on reactive repairs and maintenance on water pipelines.

The country’s critical water mains break on average 7000 times each year due to a variety of factors, including age, material and soil type. This costs upwards of $1.4 billion a year to repair.

Sydney Water invests about $32 million a year in renewals of critical water pipes. Added to that is $30 million a year on smaller water pipes and another $40 million on large concrete wastewater pipes.

This kind of investment is needed to manage the 21,000 kilometres of water and 25,000 kilometres of wastewater pipes. This includes 5000 kilometres of critical water pipes and 900 kilometres of large concrete wastewater pipes.

In 2011, Sydney Water identified key factors in how it wanted to change the way it maintained its water pipelines. The utility approached Data61 (formerly National Information Communications Technology Australia [NICTA]) asking to assist in finding a better way to identify pipe failure.

Sydney Water and Data61 established a partnership to create a solution to the issue. Data61 provided data analytics research expertise and Sydney Water contributed data and system knowledge.

Using Sydney Water’s information on water pipeline failures for the past 15 years, Data61 commenced a four-year research project with the aim of creating a data-driven tool capable of predicting these failures.

However, the team found that the act of simply identifying the cause of a water pipeline failure is a major challenge. “Some of those pipes have been buried underneath the surface for 50 or 100 years,” Fang Chen, Research Group Manager at Data61, says.

Data61 found a wide range of factors played a role in Sydney Water’s water pipeline failures over the past 15 years. This included the year the pipe was laid, the type of soil surrounding it, its depth, coating, length, lining and more. Dr. Chen explains that this complexity in cause of failure was the first hurdle to overcome.

She gives the example of a lung cancer patient. Not every lung cancer sufferer is a smoker: diet, fitness, health and other factors can contribute to the cause, even if smoking is commonly linked to the illness. “It’s the same thing,” she adds. “For some of the pipes, we found that more than 20 different factors could affect them.”

Dr. Chen explains that the cause of any pipeline failure can be defined as an assumption. “The assumptions made may be very biased for us coming into this field.” Erosion, for instance, can be perceived to be a major cause of water pipeline failure as it physically erodes the asset. Therefore, it could be assumed to be one of the major causes of pipeline failure in general.

Due to the complexity of the cause of pipeline failure, the Data61 team, in collaboration with Sydney Water, took a different approach. “We didn’t have assumptions and pretended we were blind coming into it,” she says.

Dr. Chen and her team started with a clean slate and constructed three sets of data based on Sydney Water’s pipeline failure records. The first set of data was about the pipeline itself – what was it made out of, what lining was used, when it was built. The second set was the failure data, such as when the asset failed and what were identified as the contributing factors. The third was the environmental factors – what kind of soil surrounds the pipeline. “The idea is to try to list the relationship between the [pipe attributes] and the different causes of failure in the past,” says Dr. Chen.

Following the initial research stages, the Data61 team devised a water pipe risk prediction tool, which uses the three sets of data to accurately forecast where and when a water pipeline will fail.

The pilot asset management tool identifies the pipes most likely to have a higher risk of failure.

The tool has now been researched with 25 utilities worldwide, which includes nine million pipe assets and about 525,000 kilometres of pipes.

Data61 is now working in partnership with Sydney Water on implementing the tool into business as usual. “When new technology penetrates an industry, the question is how to integrate this into daily business,” asserts Dr. Chen. She explains that this process is about maximising the benefits of the tool and helping asset managers become familiar with its capabilities.

A comprehensive history of pipeline failures is required for a utility to use the system.

The tool has a variety of applications in other industries too. “We filed a patent of this specification methodology for infrastructure failure,” she says.

Data61 has employed the same methodology on the Sydney Harbour Bridge using sensors to predict structural failures. It also has potential applications in road assets.

“You can predict what sort of road may need to be replaced and that work can be scheduled in the most minimal risk way,” says Dr. Chen. “From a methodology perspective, it’s quite unique.”

Interesting? Share this article