RNL used for the verification of the efficiency of a localization model in a real WDNs

Water distribution systems (WDNs) suffer from leakage problems. The latter can cause damage to infrastructure and also act on public health by the penetration of microbes of waterborne diseases transmissible through the orifices of leaks as soon as a drop in pressure occurs. For this purpose, the network managers always look for the best devices to detect and locate in time the anomaly of the networks as soon as it appears. To realize a system of detection and location of leakage a prototype circuit was realized equipped with pressure sensors and a system of acquisition based on a Digital Signal Processor (DSP). Pressure measurements for specific distances were carried out. A digital filter was applied for denoising. Pairs of pressure values before and after leakage are applied to a mathematical model for localization. We opted to use nonlinear regression (RNL) for the determination of non-measurable physical parameters based on pre-localized leak positions. Validation tests are performed to demonstrate the effectiveness of the model. In our study, a divergence of the model used was found.