Water pipe failure prediction tool takes out Eureka Prize
A tool that uses artificial intelligence (AI) and machine learning to make intelligent predictions about water pipe failures has taken out the University of Technology Sydney Eureka Prize for Excellence in Data Science.
Developed by the CSIRO’s Data61 Smart Infrastructure Team, the tool can help utilities prioritise which pipes need maintenance, reducing the overall cost of asset management and minimising disruption to water supplies.
“We use AI and machine learning to collect various data – and there’s a lot of it – to make smart predictions. We then use these smart predictions to guide a decision or action,” Data61 Research Group Leader – Enterprise Analytics Dr Fang Chen said.
“It covers a lot of areas and applications, from predicting pipe failures to estimating sewer corrosion and if you’re dosing chemicals smartly.”
Chen and her team have been working on the tool for about eight years. She said it is constantly improving and incorporating different data streams, which has led to a more comprehensive view of a utility’s assets.
For example, while information about hydraulics, energy, chemicals and sewage would previously have been collected and used in silos, Chen said this data is now being combined to form a larger picture.
She compared the tool to a human brain, which is able to take in and process many different streams of data.
“All the data we look at sounds very separate, but it’s like a brain. If a brain has the thinking methodology then it can take in different data and process it using the same framework. For example, if a human is trying to navigate somewhere, he or she will formulate decisions based on what they see, what they read and what the map is telling them,” she said.
Data61, a member of the Australian Water Association, has analysed the networks of more than 30 utilities, and Chen said the tool has helped businesses in Australia and overseas realise the data they collect has practical applications, including pinpointing the most at-risk pipes in their network.
“Of course, you still need a lot of expertise to interpret the data, but it gives you insight into where your priority areas are,” she said.
“We can help utilities focus on the top risk areas when it comes to pipe failure prediction … This means you can focus on the top 20 or top five risk areas – whatever is suitable for each utility.”
This allows utilities to approach asset management in a more cost-effective way.
“It saves money as you can replace your high-risk pipes, which then saves the cost of those being broken and having to repair them. You save water, save the congestion – there are social and economic benefits and less disruption to customers’ daily lives.”
Chen said the water industry has become a lot more comfortable with using and understanding data in recent years.
“When we started you would hardly see anyone using words like ‘data science’ or ‘machine learning’. Now these are so popular they’re written into nearly every strategy document,” she said.
“This is a good thing; it means people are shifting their mindset around data and working towards the right direction. I think that’s the future – data will be a lot more powerful.”