Making use of continuous measurements for change detection purposes: an application to water distribution networks

The monitoring and control of industrial processes often requires the capability to identify particular patterns in a set of acquired measurements. One of the most recurrent applications consists in the detection of changes and the related minimization of false alarms. This requirement is true also when dealing with natural systems. The monitoring of a natural resource usually involves the identification of a set of parameters, which are considered as representative of its underlining processes, in order to extract useful information about its current status and its expected behavior. This work is focused on the water resources destined to the drinkable water distribution, paying attention to two particular aspects: i) the need for a suitable metric to detect anomalous values in the assessment of water quality indicators; ii) the experimentation of a simplified data-driven strategy to estimate natural variations of one or more indicators, in order to mitigate false alarms. This paper proposes a preliminary investigation and a selected case study, in order to exemplify one practical implementation of the proposed approach. The possible application to a context of low-cost distributed sensors is also briefly discussed.