Abstract
In this paper, we propose a method to identify leaks and bursts in pipeline networks. The method uses frequency domain analysis of transient signals. As transient signals are less prone to noise, the detectable range is much wider than that of acoustic signals. Continuous transient signals are obtained from pressure sensors deployed across an actual Water Distribution System (WDS) in Singapore. The signals are first filtered with wavelet transform to remove the high frequency noise. The denoised signals are analyzed using spectrograms that reveal the energy of the signals in the Time-Frequency domain. After that, the Gabor transform is used to fine-tune the spectrograms' result. The novelty of this method, compared with conventional time domain methods (such as Inverse Transient Analysis (ITA) [1], Wavelet Transform [2]), is its simplicity and ability to provide more robust anomaly identification. Moreover, its performance is less dependent on the characteristics of the WDS. As a proof of concept, our method has been validated through emulated bursts on the WaterWiSe@SG [3] test bed.