Water Leakage Detection System for Underground Pipes by Using Wireless Sensors and Machine Learning

Abstract

Expectations have been raised for a social infrastructure maintenance system that utilizes IoT/M2M technology. We are working to develop a water leakage detection system using wireless sensors set into the water pipes. However, radio propagation is restricted in the constrained underground environment of water pipes so how to collect the data and identify leaks is an issue. We propose a system that combines drive-by data collection and static data collection, in order to collect acoustic sensor data from a wide area effectively [1]. We also propose a random noise removal method by focusing on the constancy of the leak sound, in order to identify leaks accurately. We show the results of evaluating the effectiveness of the proposed system in tests and real fields.