Seminar by Pedro Casado

Array signal processing techniques applied to leak pinpointing in buried pipes

Pedro Henrique de Melo Casado Matos
PhD candidate at the University of Trento, Supervisor: Dr. Emiliano Rustighi

Abstract
Leakages in buried pipelines are a significant cause of water wastage in distribution systems, resulting in water losses ranging from 30% to 50% in most countries. To address this issue, techniques have been developed to detect leaks in buried pipes over the last few decades. The leak detection procedure typically involves three steps: (1) leak detection, which involves analysis of water pressure measurements along the pipelines; (2) estimation of the approximate region where the leak occurred through local pressure variations along the pipeline; and (3) pinpointing the exact location of the leak to perform maintenance procedures. Acoustic pinpointing techniques are among the most effective. The latter tool exploits the time delay of arrival between different sensors placed around the suspected leak. By calculating the cross-spectral densities between sensors and analyzing their phase difference behavior over frequency, it is possible to infer the location of the radiating source. Existing methods rely on access points to the pipeline through correlators or external excitation on the free-soil surface. However, in case the buried pipe acts as a radiating source, its location can be estimated through the ground vibration signals only. The seminar presents preliminary results on the application of array signal processing formulations on the data measured by an array of sensors and shows that this is as a potential complementary approach to existing ones. Although array signal processing techniques applied to source localization are well-established in the aeroacoustics field, their adaptations to vibroacoustic fields are still to be investigated. The main differences between acoustic and seismic waves and how they affect signal processing formulation will be discussed. Among the many challenges, the characterization of wave velocity is one of the most troublesome. The uncertainty of the wave velocity effect on the steering vector will be presented and tested against numerical and experimental data. Results shows that the estimation of the leak position is highly sensitive to wave velocity uncertainties, although deviations along the depth coordinate are highly more significant than the ones along the ground surface. A pre-processing analysis of the wavenumber frequency spectrum allows to have a good initial estimate of the actual wave velocity and is proposed as a solution to address the issue.