Ordering Information:
ORDER NUMBER:  91219
DATE AVAILABLE: Summer 2008

Printed Report		PDF
Subscribers	Order Report
Non-Subscriber	AWWA Bookstore (Sept. 2008)	N/A

PRINCIPAL INVESTIGATORS:
Wayne Einfeld, Sean A. McKenna, and Mark P. Wilson

OBJECTIVES:
The purpose of this project was to develop a simulation tool to assess overall contaminant warning system (CWS) performance under various contamination event scenarios and to incorporate specific sensor performance characteristics in the tool to more realistically assess how sensor operational parameters can influence overall system performance.

BACKGROUND:
Water utilities are faced with the ongoing need to provide safe, contaminant-free water to their customers. Contamination sources can include compromised source water that penetrates treatment barriers or intentional contamination attacks from disaffected groups or terrorist organizations. In nearly all contamination event scenarios, early detection and response can help reduce public health risk and impact. To that end, contaminant warning systems offer advantages over more traditional and slower means of contaminant detection and the ability to evaluate various contaminant warning system designs using simulation methods prior to procurement and installation is a desirable capability.

HIGHLIGHTS:
This study has shown that simulation methods can be successfully used to evaluate contaminant warning system designs for user-specified chemical contamination events. Specific sensor performance attributes can be configured within the simulation to allow comparison of various commercial sensor options in a utility-specific contaminant warning system design. Initial testing results reveal that most commercial sensors perform equally well in chemical contamination events of moderate intensity with sensor sampling frequency being one of the most important sensors attributes that influences overall system performance.

APPROACH:
A sensor simulator was configured to mimic such sensor attributes as sensitivity, sampling rate, and drift using various mathematical and statistical algorithms. The spatial and temporal features of chemical contamination events were modeled using a pipe model of a real water distribution system and EPANET. The resulting contaminant pulse shapes at specific locations within the distribution system were combined with normal water quality background signals and processed through the sensor simulator. The sensor-altered signal was then further processed through a second event detection algorithm. Contaminant event detection statistics were then compiled and analyzed to assess overall system performance.

RESULTS/FINDINGS:
Study results showed that sensor/detection module performance degraded with decreasing contaminant concentration levels. At the highest contaminant concentration levels, performance of all combinations was relatively good; whereas, at the lowest contaminant concentration levels sensor/change detection combinations provided essentially no value. Several different change detection algorithms were tested. In this study with single sensor testing, the more advanced statistical change detection algorithms performed on par with the simple threshold method. They are expected to outperform the threshold approach when used with a system design that includes multiple sensor types and locations. The limited scope of this study did not include the evaluation of more complex contaminant warning system designs involving multiple sensor types and locations within a water distribution system.

This study illustrates the potential advantages of simulation tools that can be used during the design of a contaminant warning system for a water utility. Through the use of such tools, inter-comparisons of candidate sensor performance can be made under a variety of realistic contamination scenarios that will enable system designers to fully optimize the warning system design prior to component acquisition, installation, and deployment.

IMPACT:
The development of these types of simulation tools will ultimately prove useful in contaminant warning system design evaluation and optimization by utilities or their consultants. The simulation methods illustrated in this study can be used to evaluate the relative merits of various sensor types and locations within a water distribution system under realistic contamination scenarios using utility-specific pipe layouts. Further work is needed to make the tool more user-friendly and to minimize data manipulation that is presently required to carry out the simulations.

RESEARCH PARTNER:
Sandia National Laboratories

ISBN 978-1-60573-021-9