Abstract: Leaks in water supply pipelines cause waste of precious resources and endanger the environ-ment. Correlator pinpoints leaks depending on the noise from the leaks. The leakage noise meets Gaussian distribution. The fluid-filled pipe is considerably complicated due to fluid-structure interaction. The radial pipe wall vibration is excited by internal water pressure pulsations arising from water-borne sound. The pipeline system is supposed to be a finite impulse response (FIR) system. Adaptive system identification technique employing least-mean-square (LMS) algorithm is applied to modelling the pipeline and analyzing the prop-agation characteristics of leak noise. Signal dispersion is observed corresponding to the significant asymm-etry displayed by the impulse response of the pipeline. Basing on cross-power spectrum phase,the frequ-ency band of leak signal is obtained. The filtered signals are used in the estimation of time delay. It is indicated that the method improves confidence of locating leak.