Assignment 5

Berry, J., Fleisher, L, Hart, W. Phillips, C. and Watson, J.-P. (2005) “Sensor Placement in Municipal Water Networks”, Journal of Water Resources Planning and Management, 131(3) pg. 237-243.

Summary

This paper was quite similar to the last paper. They used a mixed integer programming to optimize placements of sensors in municipal water systems. The objective function was to minimize the fraction of population exposed to contaminants, in which each node of the model was weighed based on population that will consume the water. Currently, the early warning system (EWS) is used, which identifies the contamination incident while allowing time for response. The EPA used these EWS systems by placing sensors at multiple location in which the coverage of flow in terms of detection of contamination is maximized. The way in which a contamination event was to occur was modeled by a fixed probability distribution.

The assumptions that the authors used was: An attack occurs at one point in the network; The total population is considered exposed without looking at health impacts; Downstream populations are protected by nodes with sensors, meaning that a population is considered exposed if its node is reached by a flow path that doesn’t pass a sensor; Time periods are treated independently. These assumptions allowed the researched to ignore both the temporal effects of the contaminant, along with the concentration effects. The authors said that this enabled them to say that their model can be re-used for situations where large volumes of contaminants flow quickly through a network.

The network was modeled in which the pipes were “edges”, and the nodes were vertices. The authors’ input data was: probability of an attack at each node, population density at each node, the network layout. The constraints also included were flow direction constraints along with maximum number of sensors. The authors tested the model with example data sets using the program EPANet. The results showed that the population at risk from contamination is reduced as the number of sensors utilized is increased. When the authors realized that the results were showing little sensitivity to the objection function, they then discussed areas in which their model can be more generalized for every day use: using a model that incorporates temporal effects, placement of sensors other than nodes, incorporating cost and maintenance cost of sensors, modified objective function to relate to multiple objectives.

Discussion

This paper was significant, because this is still quite a large issue in water resources these days. However, there are some faults in this paper. It seems that the model isn’t taking any consideration of the water demands of each node, just the population. The authors put large constraints on the direction of flow, but the velocity of the flow was not incorporated at all. If I were to expand on the research, I would start with the four future goals of the model that the authors discussed at the end of the paper.