Probabilistic Contaminant Source Identification in Water Distribution Infrastructure Systems

Abstract

Large water distribution systems can be highly vulnerable to penetration of contaminant factors caused by different means including deliberate contamination injections. As contaminants quickly spread into a water distribution network, rapid characterization of the pollution source has a high measure of importance for early warning assessment and disaster management. In this paper, a methodology based on Probabilistic Support Vector Machines (PSVMs) is proposed for identifying the contamination source location in drinking water distribution systems. To obtain the required data for training the PSVMs, several computer simulations have been performed over multiple combinations of possible contamination source locations and initial mass injections for a conservative solute. Then the trained probabilistic SVMs have been effectively utilized to identify the upstream zones that are more likely to have the positive detection results. In addition, the results of this method were compared and contrasted with Bayesian Networks (BNs) and Probabilistic Neural Networks (PNNs). The efficiency and versatility of the proposed methodology were examined using the available data and information from water distribution network of the City of Arak in the western part of Iran.