Vibration-based water leakage detection system for public open data platforms

Abstract. Water Distribution Networks are known to lose a consistent percentage of drinkable water due to the presence of leakages. In this paper it is proposed a solution to detect water leaks consisting of: i) a new sensing equipment able to acoustically monitor the external surface of a newly laid underground pipe; ii) a training of several machine learning models able to analyse the data collected by the new sensing equipment; iii) a user dashboard to give the final user the possibility to monitor the pipe’s condition. The research process included the generation of artificial leakages capable to produce a suitable dataset necessary to properly train machine learning models onto.