An R&D collaboration between Teredo Analytics and PUB’s water supply network department.
BY WONG LIANG JIE AND DR RAJAT MISHRA
Teredo Analytics is a deep tech start-up incorporated under the NUS Graduate Research Innovation Programme (GRIP). Its co-founders are from NUS Acoustic Research Laboratory (ARL) and draw on their expertise in acoustic signal processing, big data analytics, the Internet of Things (IoT) and artificial intelligence (AI) to develop monitoring solutions for water pipeline networks and water plant machinery.
The developed technology was able to detect leaks along PUB’s water pipeline networks and has since been patented. This technology has also been adapted to provide continuous condition monitoring on water plant machinery, enabling early detection of potential machinery failure.
The R&D collaboration between PUB and Teredo Analytics seeks to develop the next generation of low-powered leak detection system that is capable of detecting near real-time leaks with a low false-positive alarm rate. This is made possible with the technological advancement in low-powered microprocessors and an AI algorithm residing in a cloud-based architecture that analyses and processes a large number of acoustic datasets collected.
Existing battery-operated leak detection systems that are installed on the water pipeline are pre-configured to be alerted at a pre-defined time to record leak-associated data parameters such as acoustics and pressure. Such an approach can only alert the authority upon detecting a leak during its pre-configured time and is thus unable to provide real-time leak detection. As such, the delay in which the authority will be alerted to a leak can potentially exceed 20 hours and given the leak size, the water loss can be substantial. In addition, the authority is also faced with a high false-positive alarm rate from previously installed systems where environmental acoustic signatures are mistakenly interpreted as pipe leaks.
The solution includes a two-fold check based on the frequencies of the acoustic signature and a machine learning (ML) framework which will reduce the number of false-positive alerts arising from noisy non-leak activities such as construction work. This involves a leak-event triggered mechanism with the ability to notify the leak detection system and send the acoustic signature of interest to the cloud platform for further processing. The event-based alert mechanism is based on a low-power circuit that notifies the processor upon detecting a persistent leaking acoustic signature.
Wong Liang Jie and Dr Rajat Mishra are co-founders of Teredo Analytics.
The full article is available in the latest edition of Water & Wastewater Asia Jul/Aug 2022 issue. To continue reading, click here.