Acoustic monitoring is the go-to solution for locating a leak in a large urban pipe network, as the sounds from leaks are unique and travel far in water, but even this method struggles in complex systems. To tackle the problem, Pranav Agrawal and Sriram Narasimhan from UCLA developed algorithms that operate on acoustic signals collected via hydrophones mounted on fire hydrants. In doing so, the team can avoid costly excavation and reposition the devices as needed. Combined with novel probabilistic and machine-learning techniques to analyze the signals and pinpoint leaks, this technology could support water conservation efforts.
