A Method of Leakage Location in Water Distribution Networks using Artificial Neuro-Fuzzy System

The paper deals with a method to locate uncontrolled leaks in water distribution networks. The idea of the method is based on discretization of the water supply system to the predefined areas and then identifying an approximate location where a leakage can occur. In the proposed method, the location of leakage is determined by means of the group of neuro-fuzzy classifiers. The number of classifiers corresponds to the number of areas in which the network is divided. The task of each classifier is to change the state of its output in the event of a leak in the network associated with this classifier. The input signals of classifiers are residues. They are computed using flow measurements and output signals of predictive models that describe the observed changes in the nominal flow conditions of the water supply network. In this paper the verification case study is conducted for the water distribution network covering industrial and individual consumers in the selected district of the town in the southern Poland.