Acoustic leak localization method based on signal segmentation and statistical analysis

One of the most serious problems occurring in a pipeline network is the appearance of leaks. The process of detecting and localizing leaks in pipeline systems concerns a very extensive field of signal processing methods employed for this matter. In this paper a leak localization method combining the segmentation of acoustic leak signals, both in the time and in the frequency domain, with a statistical algorithm needed for dealing with the non-deterministic (stochastic) nature of these signals is proposed. This algorithm involves the use of cross-correlation techniques along with the grouping of the time-delay data in a histogram and selecting the bin with the largest number of elements as the one that provides the correct answer. The successful detection of the leak position requires the knowledge of the acoustic wave velocity in the pipe. In the present paper the calculation of the acoustic velocity is performed by the use of a PCB hammer to cover more realistic situations. The proposed leak localization method is tested experimentally in a laboratory setup containing a 67-meter steel pipeline and the results show that the presented method can localize leaks efficiently, since the average localization error is around 3%.