Improvement of acoustic water leak detection based on dual tree complex wavelet transform-correlation method

Water leakage is one of the important agendas across the globe thus several effectual systems for leakage detection had been developed with the aim to improve sustainable use of water. Among the methods, acoustic leak detection technique had been proven as a promising approach to detect and localize leaks in water or gas pipeline. However, existence of noise in acoustic signals complicated the leak detection practices. Traditional de-noising methods like filtering and wavelet de-noising are not suitable for non-stationary and broadband acoustic signals. Therefore, Dual Tree Complex Wavelet Transform (DTCWT) is applied in this paper to reduce acoustic noise and decompose signals into several frequency bands. DTCWT decomposition is intended to resolve the problem encountered by the typical correlation-based leak localization method. Due to dispersive and frequency-dependent nature of wave propagation, correlation-based method normally assumes constant wave velocity which results in inaccurate leak source localization. In this paper, signal will be de-noised and decomposed by employing DTCWT. Wave velocity is evaluated based on the dominant frequency and dispersion curve. Then, time delay can be estimated via comparison study among cross correlation, CWT localization and convolution. CWT localization and convolution are proposed as new time-delay estimation method attributable to enhance localization accuracy. Experimental results validated that the proposed method, DTCWT-correlation outperforms other methods with a localization error of 4.67 %. Both CWT localization and convolution are also capable to pinpoint the location of leaks. Besides, leaking and non-leaking condition can be differentiated after multilevel decomposition of DTCWT.