Acoustic techniques have been used for many years to find and locate leaks in buried water distribution systems. Hydrophones and accelerometers are typically used as sensors. Although geophones could be used as well, they are not generally used for leak detection. A simple acoustic model of the pipe and the sensors has been proposed previously by some of the authors of this paper, and their model was used to explain some of the features observed in measurements. However, simultaneous measurements of a leak using all three sensor-types in controlled conditions for plastic pipes has not been reported to-date and hence they have not yet been compared directly. This paper fills that gap in knowledge. A set of measurements was made on a bespoke buried plastic water distribution pipe test rig to validate the previously reported analytical model. There is qualitative agreement between the experimental results and the model predictions in terms of the differing filtering properties of the pipe-sensor systems. A quality measure for the data is also presented, which is the ratio of the bandwidth over which the analysis is carried out divided by the centre frequency of this bandwidth. Based on this metric, the accelerometer was found to be the best sensor to use for the test rig described in this paper. However, for a system in which the distance between the sensors is large or the attenuation factor of the system is high, then it would be advantageous to use hydrophones, even though they are invasive sensors.

In leak noise correlation surveys, time delay estimation (TDE) is of great importance in pinpointing a suspected leak. Conventional TDE methods involve pre-filtering processes prior to performing cross-correlation, based on a priori knowledge of the leak and background noise spectra, to achieve the desired performance. Despite advances in recent decades, they have not proven to be capable of tracking changes in sensor signals as yet. This paper presents an adaptive phase transform method based on least mean square (LMS) algorithms for the determination of the leak location to overcome this limitation. Simulation results on plastic water pipes show that, compared to the conventional LMS method, the proposed adaptive method is more robust to a low signal-to-noise ratio. To further verify the effectiveness of the proposed adaptive method, an analysis is carried out on field tests of real networks. Moreover, it has been shown that by using the actual measured data, improved performance of the proposed method for pipeline leakage detection is achieved. Hence, this paper presents a promising method, which has the advantages of simple implementation and ability to track changes in practice, as an alternative technique to the existing correlation-based leak detection methods.