Eigen RadarAI
Analysis

NTT tests water-leak localization with AI and road-surface vibration

NTT and a Fukuoka plumbing cooperative have tested a system that locates buried water-pipe leaks from vibrations measured at the road surface. A fiber sensor collects the signal and AI separates leak patterns from surrounding noise. The cooperative confirmed leak positions through excavation. NTT plans to improve analysis and field usability before wider deployment.

Artificial Intelligence··Morning
A thin sensor lies on a neighborhood road, connected by a cable to a measurement box farther along the street.

NTT publishes results from the Fukuoka leak trial

NTT and a plumbing cooperative in Fukuoka, Japan, have tested AI that locates leaks in buried water pipes using vibration measured at the road surface. The field demonstration combines fiber sensing with automated analysis. Initial measurements do not require digging down to the pipe. The cooperative supplied pipe information and confirmed actual leak locations through excavation.[1], [2]

A road sensor separates leak signals from noise

The sensor head holds fiber arranged densely over a flat surface and sits on the road above the pipe. Distributed acoustic sensing tracks light scattered back through the fiber to measure vibration along it. AI extracts leak-related vibration from environmental noise and looks for patterns unlike those at nearby measurement points. The operator positions the sensor and starts the measurement.[1]

Further work targets accuracy and field use

The demonstration began after an announcement in September; October 21 brought its results. NTT links the method to making leak detection accessible to workers without lengthy specialist training. It plans to use collected data to improve analysis and practical operation, with wider municipal use an objective. The company also plans to exhibit the work at its research forum in November.[1]

References

  1. News sourceNTTNTT field-tests AI that locates water leaks from fiber vibration↩1↩2↩3
  2. News sourceImpress AI WatchNTT demonstrates AI water-leak localization in about 15 minutes↩