Matched-field Processing for Leak Detection with Prior Uncertainty Information: Experiments from Laboratory to Field

Matched-field processing (MFP) has been proposed for transient-based pipeline leakage detection and has proven to be fast, accurate and robust with noise. Essentially, MFP estimates leak parameters by matching the transient model to measured pressure. However, experimental results from real pipeline systems are always submitted to different kinds of uncertainties, such that the theoretical model and experimental data are mismatched and this may result in failure of leakage detection. It is found that a large part of uncertainties is deterministic, which means that they appear in every transient test, and may stem from internal property of a pipe. Therefore, this paper proposes to include these deterministic uncertainty information in the leakage detection scheme, which can be obtained via an initial measurement just after the construction of a pipe system or via a regular monitoring of pipeline condition. As a consequence, a novel MFP approach is proposed which incorporates these information of modeling uncertainties. The deterministic property of uncertainty and the novel MFP leakage detection method are assessed via three experimental scenarios (from simple to complex in terms of uncertainty level): (i) a straight steel pipe system in laboratory; (ii) a looped high-density-polyethylene (HDPE) pipe system in laboratory; (iii) a HDPE pipe system with an upstream pressure reducing valve (PRV) in a field test at Beacon Hill, Hong Kong. It is shown that, in all the three scenarios, the proposed novel MFP approach improves leakage detection accuracy than the conventional MFP by incorporating uncertainty information.