A water distribution network is a vital component of a water supply system. It cost represents about 70% of the total cost. Therefore, a rigorous planning and analysis is necessary for a water distribution network. A pipe network is a grid formed by nodes, pipes and loops. In practice, it may also contain reservoirs, pumps, valves and other accessories. Several parameters and variables characterize the flow in a pipe network. Some of them are nodal heads, nodal demands, leakage rates and roughness etc. In general, a water supply system is designed to cater the demand at the end of the design period. The water authority encounters some uncertainty in estimating future water supplies as in estimating future demands. The future demands are difficult to predict exactly because of two major uncertainties: future population growth and future per capita consumption. Water is lost from a water distribution system at many places due to different reasons ranging from the natural ageing of pipes and facilities to the simple theft. The most common uses are: leakage from pipes, leakage from connection pipes, unmetered consumption, illegal connection and theft etc. Out of these uses, there is a significant amount of leakages from pipes, which was taken into account in calculating leakage rates. A methodology was developed to incorporate leakage rates into the model of simulation so that the total leakage rates in the system should not exceed the total Unaccounted For Water (UFW) in the system which was calculated based on the measured values. Roughness values subject to change due to encrustation and corrosion in the pipes. Because of all these factors, the nodal heads in a pipe network is subject to uncertainty due to uncertainty in demands, leakage rates and roughnesses. Several software products are used for simulation modelling in network analysis. The software EPANET was used in this particular study. In order to quantify uncertainty in flow variables the sensitivity approach was used in this study. This methodology was tested on a simple hypothetical network with four nodes and five links. Also this methodology was applied to a real network obtained from Sri Lanka which consist of more than one hundred nodes and links. It was seen that there is a significant amount of uncertainties in nodal heads which cannot be neglected. These results justify the necessity of carrying out the uncertainty analysis in order to quantify uncertainties. In the real network the uncertainties of nodal heads along the selected branches which originate from one node were computed. It was investigated how uncertainties vary along branches of the network. By doing this it will be possible to identify in which parts of the network the uncertainty associated with input parameters are significant, particularly where the result are sensitive.
Back to the list of MSc abstracts