Time delay estimation using cascaded LMS filters fused by correlation coefficient for pipeline leak localization

Abstract

Water leaks are commonly detected using acoustic/vibration sensors, which are sensitive to sounds and vibrations caused by leakage. In practical surveys, the degree of success for leak detection and localization depends on the operating conditions and construction materials of the pipeline in question. In particular, under the condition of low signal-to-noise ratios (SNRs) and small leaks, the traditional localization method based on time delay estimation (TDE) tends to give pseudo peaks, which may lead to appreciable errors in the estimate of the leak position. In this paper, TDE is carried out based on cascaded least mean square (LMS) filters fused by correlation coefficient. Combined with the propagation theory of leak signals in the pipeline, the TDE algorithms using the LMS adaptive filters are introduced for leak localization. It is found that the single LMS filter is replaced by cascaded architectures, further improving its performance for TDE in a low SNR environment. Numerical simulations are presented to demonstrate the best performance for TDE by using the cascaded LMS filters in comparison with the basic cross-correlation, the phase transform generalized cross-correlation and the single LMS filter. This is further verified using some measurements made in the laboratory and field tests.