Abstract
The practical modelling of water distribution networks requires a reduction in the number of consumption nodes and connecting pipes in order to save time, computational and data collection efforts. The reduced model called skeletonized network is achieved by aggregating water demands to be normally allocated at pipe intersection nodes while eliminating unimportant pipes from the real network. The proper representation of the real network by the skeletonized model highly depends on the method by which demands are allocated. In this paper, two methods of demand allocation were investigated with respect to the actual situation. The first method uses Thiessen polygons to determine number of consumers to be supplied by each node. The second method uses the principle of insufficient information to distribute demands supplied by each pipe. The hydraulic performance of each method is evaluated using the satisfaction of pressures at the network nodes considering all pipes in service, while measuring the hydraulic performance under pipe failure using the resilience index. The main result of the study is the overestimation of pressures at end nodes of the network when Thiessen method is used. The approximated method outperformed the Thiessen method at all network nodes.