Computer aided leak location and sizing in pipe networks

[摘要] South East Queensland (SEQ) experienced marked reductions in seasonal rainfall over the past decade and as a result water storages in the region had reached unsustainably low levels. In response, water conservation measures were adopted across the region, augmented with programs for water recycling and stormwater harvesting. In this environment, losses from the potable water distribution system were brought into sharper focus and scrutiny especially as audit information indicated application of leak detection and rehabilitation technologies in SEQ had the potential to save 36,000 ML or $27 M worth of water annually. The economic benefit of applying such technologies would depend substantially on the cost of leak detection and location technologies. With a view to reducing these costs, this report describes work that is being done towards the development of a novel technique for leak detection and location involving the analysis of pressure sensor readings by a support vector machine (SVM), through which the leak location can be predicted approximately, at an accuracy that is amplified with increasing distance from the actual leak location. Therefore, the technique would be useful to a water authority in restricting the range of locations that have to be searched to find a given leak.The report is organised into 4 main sections. The first section provides an overview of the leakage issue in Australia, the method of auditing leakage, and the factors contributing to errors. The second section is a review of the currently practiced leakage detection methods and developments in emerging leak detection technologies based on hydraulic analysis. Section 3 introduces the incorporation of models that utilise Artificial Neural Networks for data analysis including the use of the SVM. Section 4 describes the development of a model using the SVM and forms the basis of the technique presented in this report.

[发布日期] 2010-07-10 [发布机构] CSIRO

[学科分类] 地球科学(综合)