Finding Leaks On Large Water Mains In Singapore Using Advanced Algorithms

A single leak on a large transmission main can mean serious water loss, and replacing the pipe is costly. PUB, Singapore's national water agency, now monitors 200 kilometers of these mains with a newer acoustic system. Data analyst Marcin Kloc argues that anyone can feed data to an AI model; the harder work is using physics and math to transform it first. The results show up in fewer noise events for analysts to review, simpler scaling across sites, and notifications that arrive in 1-2 days instead of 2-3. One leak was caught where correlation was impossible. PUB also tracks classification and sensitivity, which show how well a system separates leaks from network noise and how many leaks it finds. Read how PUB measures whether its leak detection works.