Abstract
Water pipelines are infrastructure essential to our daily life. Due to extensive deterioration, breaks in pipes have become frequent. Measures to prevent pipe leakage rely on highly skilled inspection personnel who use listening sticks or similar devices to detect leaks based on the leak sounds transmitted along the road surface or at valves. It is assumed that these inspectors detect leaks based on the volume, frequency range, and timbre of the sounds. The problem is that their skills are not being passed down, as not enough highly experienced personnel have been trained to meet today's demand. Technology is needed that can substitute for their auditory skills and experience. In this study, leak sound data measured under various conditions is analyzed from the two aspects of time domain and frequency domain. First, peak values measured on a sound recorder are subject to time-series analysis. Second, the features of the frequency spectrum are analyzed using Fourier transform. The objective of this research is to obtain information useful for discrimination of water leakage sounds from other sounds. The results of the discrimination analysis indicate that the frequency domains useful for leakage detection are in the neighborhood of 1000[Hz] to 5000[Hz] for ductile iron pipes and 500 [Hz] to 1500[Hz] in the case of polyethylene pipes. The study also showed that the positioning of the sound recorder in relation to the leak location, as well as the leaking volume, affect the precision of leak detection.
