Feature Extraction of Leaks Signals in Plastic Water Distribution Pipes Using the Wavelet Transform

Abstract

The location of leaks in water distribution pipes are traditionally identified by cross correlating two vibro-acoustic signals measured with accelerometers either side of the leak, but this method provides little information about the features of the signal (such as the leak flow rate) and does not differentiate between the leak signal and background noise. Due to its capability to provide time and frequency information, wavelet analysis can be used to extract features about the leak. However one of the greatest challenges is the selection of the mother wavelet, and there has been limited studies investigating how varying the mother wavelet can produce alternate results in the context of leaks in water distribution systems. This study demonstrates the potential of the wavelet transform to distinguish between the different features composed in the signal, including the leak noise, ambient noise, pump noise and leak type on plastic pipe when applied to data from a test rig. Variables such as the influence of pressure or distance from the leak are also included in order to demonstrate the efficacy of the wavelet transform in providing information about the signal. Moreover, this research gains an understanding of how the choice of different mother wavelet can influence the information provided by the leak signal. The results demonstrate that an appropriate mother wavelet can be identified and that the wavelet transform can localise features of the signal, including pump noise, background noise and the leaks characteristics.