Under the agreement announced today with Sydney Water, NICTA's machine learning capabilities will be used to more accurately identify which pipes are at risk of failure, potentially saving Australia's water utilities and the community $700 million a year in reactive repairs and maintenance.
Australia's critical water mains break on average 7,000 times each year, due to age, material, soil type and other factors, and Rob Fitzpatrick, Director, Infrastructure, Transport and Logistics at NICTA, says NICTA’s technology was trialled in Wollongong and was able to “accurately predict breaks in the following year with twice the precision of the existing technology.”
“NICTA is applying advanced machine learning techniques to pipe failure data from Sydney Water. We have been working together to develop a system that will reduce the inconvenience and expense incurred by water pipe breakages.”
“To do this we need accurate models to identify high-risk pipes which can cope with the differences in age, pipe material, environmental conditions and urbanisation.
“We need smart technology to help provide answers. NICTA’s approach is innovative and has the potential to have worldwide impact in pipe condition assessment. We have introduced the approach to our international colleagues, who are keenly watching the outcomes.”
The project is valued at more than $700,000 in cash and in-kind contributions from NICTA and Sydney Water, and NICTA CEO, Hugh Durrant-Whyte, said the collaboration with Sydney Water was the latest example of NICTA working with industry to “develop the ICT for smart infrastructure that will drive Australia’s future productivity.”
“Our approach draws on NICTA’s specialty in machine learning,” said Dr Fang Chen, NICTA’s Technical Lead on the project.
“We have developed a new computer modelling based approach to estimate the likelihood of pipe failure. Our approach could also be applied to other infrastructure failure prediction, such as bridges.”