Contaminant Source Identification for Priority Nodes in Water Distribution Systems

Abstract
A multi-stage response procedure is described to assist in the identifi-cation of contaminant source ingress locations into water distribution systems. In each stage, data mining procedures are used, with query sentences with several conditions, to select possible intrusion nodes (PINs) for priority nodes such as hospitals or schools. A Euclidean dis-tance metric defined in a response versus time system using a 2–D coordinate system is employed to quantify the probability of each PIN being the true injection node. Use is made of time delay as the negative of the first detection time in sensor and important nodes; further, the priority degree of each PIN is defined as the normalized time delay. Ac-cordingly, the determination of the PINs to respond under emergency is formulated as a two-criteria decision process: first, the consideration of the probability, and second, the priority degree. The proposed proce-dure is demonstrated by a case study. The query sentences with conditions I and II identify 48 PINs in the first stage, and 17, 10, 4 and 4 in the subsequent 4 stages; the PINs with conditions I-1 and II-1 identify priority nodes of particular interest due to their positive time delay in the first, second and third stages. In each stage the Euclidean distance metric identifies the true intrusion node to be node 44 with the highest probability; and priority degree supplies the other criteria in deciding the emergency response nodes. The developed algorithm runs in less than one minute, which demonstrates the utility of the procedure for real-time application.
This paper is only available in PDF Format: View full text PDF