A fully automated Event Recognition System (ERS) for the near real-time detection of pipe bursts and other network events such as boundary valve status changes and pressure management valve faults has been recently developed by the authors. This paper focuses on the further development of this system. The aim is to enhance the ERS approximate event location and alarm handling capabilities by developing and testing a new methodology that, in the case of cascading District Metered Areas (DMAs), automatically determines in which DMA an event occurred. The newly developed methodology makes use of a set of heuristic rules based on engineering knowledge, the Water Distribution System (WDS) schematic and the ERS outputs. The results of applying the new methodology to the historical pressure/flow data from several groups of cascading DMAs in the United Kingdom (UK) with real-life burst events are reported in this paper. The results obtained illustrate that the developed methodology not only enabled detecting the burst events occurred in a timely (i.e., within 30 minutes) and reliable (i.e., without any false alarm) manner but also allowed to always successfully determine in which DMA the event happened. The latter capability enables water companies to target the resources for the identification of the exact burst location to the greatest effect. Additionally, it enables reducing the potential of false alarms and the overall number of detection alarms, thereby facilitating interpretation of the ERS results.
Romano, Michele; Woodward, Kevin; Kapelan, Zoran; and Savić, Dragan A., "Near Real-Time Detection Of Pipe Burst Events In Cascading District Metered Areas" (2014). CUNY Academic Works.