Publications

Minimize impact or maximize benefit: The role of objective function in approximately optimizing sensor placement for municipal water distribution networks

Hart, William E.; Murray, Regan; Phillips, Cynthia A.

We consider the design of a sensor network to serve as an early warning system against a potential suite of contamination incidents. Given any measure for evaluating the quality of a sensor placement, there are two ways to model the objective. One is to minimize the impact or damage to the network, the other is to maximize the reduction in impact compared to the network without sensors. These objectives are the same when the problem is solved optimally. But when given equally-good approximation algorithms for each of this pair of complementary objectives, either one might be a better choice. The choice generally depends upon the quality of the approximation algorithms, the impact when there are no sensors, and the number of sensors available. We examine when each objective is better than the other by examining multiple real world networks. When assuming perfect sensors, minimizing impact is frequently superior for virulent contaminants. But when there are long response delays, or it is very difficult to reduce impact, maximizing impact reduction may be better. © 2011 ASCE.