Pressure Sensors Placement Using Graph Signal Processing and Sampling Theory

A suitable strategy for sensor placement in water distribution systems allows water managers and engineers for better monitoring and controlling of the system. Sensor placement problem is often addressed by optimization techniques able to deal with its combinatorial nature. This paper addresses such a question from the innovative framework of graph signal processing. For the sake of the problem reproducibility, the case-study chosen is the Modena water distribution system. A directed, flow-weighted graph is obtained from an abstraction of the Modena hydraulic model from which graph signal processing provides a minimum number of sensors for a minimum (given) information loss. Particularly, graph signal tools such as random node sampler is applied to identify a set of nodes that better represents the graph structure and flow information. The results show the benefits of graph signal processing as a sensor placement strategy in terms of sensitivity parameter, entropy information and coverage rate in leaks case.