Water leak location based on improved dual-tree complex wavelet transform with soft thresholding de-noising

Abstract

Water leakage control emerges as a prime concern among researchers and water utility companies due to ever-increasing water loss level. Acoustic leak detection technique is a promising and widely used approach. Unfavorably, the deficiency of this method is that acoustic waves are interfered by undesirable noise and presence of multi-modal dispersive wave. The conventional correlation-based method assumes leak noise propagates as a single non-dispersive wave and this leads to unreliable detection method. This paper concerns on noise reduction through an improved de-noising method and leak localization by considering wave dispersion. Wavelet de-noising is a common de-noising method used in past leak detection works. But it has shortcomings of frequency aliasing and shift-variant due to DWT decomposition. Therefore, a shift-invariant Dual Tree Complex Wavelet Transform (DTCWT) is introduced here to substitute DWT for wavelet de-noising. DTCWT will decompose the signal into several bandwidths then soft thresholding is applied to remove noise by eliminating the irrelevant signals. Multilevel DTCWT decomposition aids to tackle the problem encountered by basic localization method because the wave velocity is evaluated based on the dominant frequency and dispersion curve. Experimental results show that the proposed de-noising method outperforms ordinary wavelet de-noising. Besides, the method effectively removes noise and makes the peak of cross-correlation function more pronounced and hence it is achievable to increase leak localization accuracy. The proposed method outweighs other methods by offering the lowest false trips occurrence and minimal localization error of 3.33%. A field test result proved that the method is able to locate the leak in a 235-meter-long, underground buried pipe with a minimal error of 2.63% which is 57% lower than the conventional method.