Investigation into an adaptive leak detection method in water distribution pipes

Pipeline leakage is a subject of increasing concern in China and across the world for transporting fluids and gases. Acoustic leak detection systems based on cross-correlation have been successfully applied to locate and detect leaks, in particular in water distribution pipes. Conventional methods for time delay estimation (TDE) involve some pre-filtering processes prior to performing the cross-correlation based on a priori knowledge of the leak signal and noise spectra to achieve the desired performance. Difficulties are often encountered in practical situations, since the leak signals in water pipes are narrow-band and of low frequency, which inevitably will be contaminated by noise. To overcome this, this paper presents an adaptive leak detection method based on an LMS algorithm for TDE between two acoustic/vibration signals measured using accelerometers/hydrophones bracketing a suspected leak. It builds on an analytical model of wave propagation in a fluid-filled pipe and combines the adaptive algorithm with the propagating characteristics of the leak signal in the pipe system. Compared to the basic crosscorrelation method, the adaptive method shows some promise in the presence of ambient noise, which may offer improvement of TDE for locating a leak having a low signal-to-noise ratio. Experimental results confirm the improved performance of the LMS algorithm for TDE in water distribution pipes.