With aging infrastructure, cross connections and backflow, pressure transients, and the increasing possibility of intentional injections by disgruntled people and/or terrorists, the potential for ingress of contamination into water distribution systems is causing increasing concern for municipalities. One approach to identify the locations of ingress is through reliance upon a contaminant warning system (CWS). While potentially effective, locating the CWS and subsequently, locating the points of contaminant ingress given the information assembly made possible from the CWS, are challenging problems. The elapsed time before the location of contaminant ingress is identified is critical to the ability to protect consumer exposures. The earlier the contaminant is identified and emergency response occurs, the lower the exposure risk to consumers.
A multi-stage response procedure is described which is developed by locating and quantifying probabilities of Potential Ingress Nodes or PINs, to address the question of contaminant ingress which may occur from any location within the distribution system. The procedure uses data mining to successively narrow the PINs at each stage, with the intent of narrowing the potential locations to numbers that can be assessed in the field. Query sentences are executed to locate the PINs and a Euclidean distance is employed to quantify probability for the purpose of responding to several locations with the highest probabilities. As demonstrated in one case study, the multi-stage procedure may save three hours waiting for the third sensor alarm to provide the information; except for the first sensor alarm, the distance metric can identify the true ingress nodes; the program run time is less than two minutes. The multi-stage response procedure is shown to be an effective and efficient way to address the contaminant ingress source identification problem.