Water loss from water distribution systems (WDS) is an ongoing problem which poses a significant risk to water resources around the world. This paper presents a novel combined sensor placementleak/burst localisation methodology which forms, and analyses by using sc inverse-distance weighted (IDW) interpolation, a sensitivity matrix to determine, on average, how accurately each sensor configuration localises leaks/bursts modelled at all nodes in a WDS. For a given number of sensors, the multi-objective evolutionary algorithm determines the optimal location of sensors to maximise the leak/burst localisation performance using the sc-IDW outputs in its objective function. Once the optimal sensor location is selected, the sc-IDW technique is used when new leaks/bursts occur in the WDS to determine their approximate location. A benchmark WDS was used to compare the leak/burst localisation performance against a baseline sensor placement technique. The comparison indicated that by using the sc-IDW technique for both the sensor placement and leak/burst localisation steps the leak/burst search area was reduced in size by between 9 and 26%. Reducing the leak/burst search area allows field teams to more quickly repair a leak/burst and reduce the impact that it has on water company operational efficiency and customer service.
This paper focusses on the customisation and further enhancement of the recently developed data-driven methodology for the automated near real-time detection of pipe bursts and other (e.g. sensor faults) events at the district metered area (DMA) level. Assuming the availability of pressure/flow data from an increased number of sensors deployed in a DMA, the aim is to: (i) overcome the limitations of the probabilistic inference engine when dealing with the increased data availability; and (ii) exploit the event information resulting from the analysis of the larger number of DMA signals for determining the approximate location of the pipe burst events within the DMA. This is achieved by making use of a multivariate Gaussian mixtures-based graphical model and geostatistical techniques. The novel detection and location methodology is demonstrated and tested on a series of simulated pipe burst events that were performed by opening hydrants in a real-life DMA in the UK. The results obtained illustrate that the new methodology can successfully determine the approximate location of pipe bursts within a DMA (in addition to detecting them in a fast and reliable manner). The performance comparison of several geostatistical techniques shows that the Ordinary Cokriging technique outperforms all other techniques tested.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.