This paper proposes a leakage detection method based on detecting significant discrepancies between pressure measurements and their estimations obtained from the simulation of a calibrated water distribution network model. Every sensor in the network will allow to detect a discrepancy in pressure due to leakage depending on its location. Then, a set of well distributed pressure sensors will generate a leakage signature that allows leakage localisation. This paper presents the methodology used in the Barcelona network for distributing properly the sensors for a good discrimination in the leakage localisation process. The methodology for sensor placement uses the pressure sensitivity matrix to the leakage presence. This matrix is normalised and binarised in order to be used as a leakage signature matrix using the standard model based fault diagnosis approach. Sensors may be installed in any node and leakages are simulated as a constant demand that can appear in any node too. The problem of deciding which are the best localisations for a small number of sensors in order to detect and localise leakages is an inverse problem that should be solved using optimisation. The resulting optimisation problem is of discrete nature and very huge for a real network. This type of problem is, in general, hard to solve and very time consuming. The use of GA (Genetic Algorithms) has been proved adequate according to the formulation of the signatures in the sensitivity matrix.

You do not currently have access to this content.