Sensor placement for robust burst identification in water systems: Balancing modeling accuracy, parsimony, and uncertainties

Urban water systems are seeing an uptake in using advanced sensing technology. Incorporating sensors for monitoring water distribution systems (WDSs) provides promising benefits to water utilities by enabling a shift from reactive to proactive pipe failure detection and from delayed water loss management to automatic sense-and-respond capabilities. Despite the opportunities that new sensing technologies create, a budget-constrained utility is challenged with identifying sensing locations in the WDS that will maximize information gain. To address this gap, this paper studies the problem of optimal placement of high-frequency pressure sensors in WDSs for pipe burst identification. This paper proposes a sensor placement strategy to address the challenges of data and modeling uncertainty by incorporating robust representation and tolerance analysis into an optimization framework with the objective of achieving the best detection and identification of burst events. Transient simulations are first used to predict system's response to burst events, demonstrating the importance of modeling accuracy over approximation methods. A robust event representation approach is then presented to summarize system response to pipe bursts using signature matrices. Subsequently, the identification problem is cast as a minimum test cover problem when the number of available sensors is unlimited, and as the maximum covering test problem when the number of available sensors is limited. The optimization problems are then formulated and solved using mixed integer linear programming. Four complementary metrics are suggested to evaluate the performance of the sensor placement designs. Multiple criteria decision analysis is then applied to select the placement design while balancing information gain and cost. The results show that incorporating more information can improve event identification, but sufficient accuracy of the extracted information is required to accrue the benefits.