LAUSR

Abstract This paper proposes an approach based on a k-Nearest Neighbour classification algorithm (k-NN) to identify regions in a water distribution network (WDN) that are affected under presence of leaks. The classification algorithm is trained with numerical data coming from a MATLAB simulator based on a dynamic model of the WDN that involve leaks in its formulation. Concretely, the training is done by using the numerical solutions of a dynamic model of the WDN under several leak cases. The dynamic model is formulated by taking into account typical assumptions of the rigid water column (RWC) theory and using the graph theory. The proposed approach was evaluated in a hydraulic pilot plant.